2015-10-29 07:07:51 +00:00
|
|
|
// Copyright (C) 2014 The Syncthing Authors.
|
|
|
|
//
|
|
|
|
// This Source Code Form is subject to the terms of the Mozilla Public
|
|
|
|
// License, v. 2.0. If a copy of the MPL was not distributed with this file,
|
2017-02-09 06:52:18 +00:00
|
|
|
// You can obtain one at https://mozilla.org/MPL/2.0/.
|
2015-10-29 07:07:51 +00:00
|
|
|
|
|
|
|
package db
|
|
|
|
|
|
|
|
import (
|
2020-01-21 17:23:08 +00:00
|
|
|
"bytes"
|
2020-03-19 13:32:22 +00:00
|
|
|
"errors"
|
2020-08-18 07:20:12 +00:00
|
|
|
"fmt"
|
2020-01-21 17:23:08 +00:00
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
"google.golang.org/protobuf/proto"
|
|
|
|
|
|
|
|
"github.com/syncthing/syncthing/internal/gen/bep"
|
|
|
|
"github.com/syncthing/syncthing/internal/gen/dbproto"
|
2019-11-29 08:11:52 +00:00
|
|
|
"github.com/syncthing/syncthing/lib/db/backend"
|
2020-12-21 11:59:22 +00:00
|
|
|
"github.com/syncthing/syncthing/lib/events"
|
2020-08-18 07:20:12 +00:00
|
|
|
"github.com/syncthing/syncthing/lib/osutil"
|
2015-10-29 07:07:51 +00:00
|
|
|
"github.com/syncthing/syncthing/lib/protocol"
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
"github.com/syncthing/syncthing/lib/sliceutil"
|
2015-10-29 07:07:51 +00:00
|
|
|
)
|
|
|
|
|
2020-05-30 07:50:23 +00:00
|
|
|
var (
|
|
|
|
errEntryFromGlobalMissing = errors.New("device present in global list but missing as device/fileinfo entry")
|
|
|
|
errEmptyGlobal = errors.New("no versions in global list")
|
|
|
|
errEmptyFileVersion = errors.New("no devices in global file version")
|
|
|
|
)
|
2020-03-16 07:45:45 +00:00
|
|
|
|
2015-10-29 07:07:51 +00:00
|
|
|
// A readOnlyTransaction represents a database snapshot.
|
|
|
|
type readOnlyTransaction struct {
|
2019-11-29 08:11:52 +00:00
|
|
|
backend.ReadTransaction
|
2020-12-21 11:59:22 +00:00
|
|
|
keyer keyer
|
|
|
|
evLogger events.Logger
|
2015-10-29 07:07:51 +00:00
|
|
|
}
|
|
|
|
|
2019-12-02 07:18:04 +00:00
|
|
|
func (db *Lowlevel) newReadOnlyTransaction() (readOnlyTransaction, error) {
|
2019-11-29 08:11:52 +00:00
|
|
|
tran, err := db.NewReadTransaction()
|
|
|
|
if err != nil {
|
|
|
|
return readOnlyTransaction{}, err
|
2015-10-29 07:07:51 +00:00
|
|
|
}
|
2021-06-27 06:43:49 +00:00
|
|
|
return db.readOnlyTransactionFromBackendTransaction(tran), nil
|
|
|
|
}
|
|
|
|
|
|
|
|
func (db *Lowlevel) readOnlyTransactionFromBackendTransaction(tran backend.ReadTransaction) readOnlyTransaction {
|
2019-11-29 08:11:52 +00:00
|
|
|
return readOnlyTransaction{
|
|
|
|
ReadTransaction: tran,
|
|
|
|
keyer: db.keyer,
|
2020-12-21 11:59:22 +00:00
|
|
|
evLogger: db.evLogger,
|
2021-06-27 06:43:49 +00:00
|
|
|
}
|
2015-10-29 07:07:51 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
func (t readOnlyTransaction) close() {
|
|
|
|
t.Release()
|
|
|
|
}
|
|
|
|
|
2019-11-29 08:11:52 +00:00
|
|
|
func (t readOnlyTransaction) getFile(folder, device, file []byte) (protocol.FileInfo, bool, error) {
|
|
|
|
key, err := t.keyer.GenerateDeviceFileKey(nil, folder, device, file)
|
|
|
|
if err != nil {
|
|
|
|
return protocol.FileInfo{}, false, err
|
|
|
|
}
|
|
|
|
return t.getFileByKey(key)
|
2019-01-18 10:34:18 +00:00
|
|
|
}
|
|
|
|
|
2019-11-29 08:11:52 +00:00
|
|
|
func (t readOnlyTransaction) getFileByKey(key []byte) (protocol.FileInfo, bool, error) {
|
|
|
|
f, ok, err := t.getFileTrunc(key, false)
|
|
|
|
if err != nil || !ok {
|
|
|
|
return protocol.FileInfo{}, false, err
|
2019-01-18 10:34:18 +00:00
|
|
|
}
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return f, true, nil
|
2019-01-18 10:34:18 +00:00
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
func (t readOnlyTransaction) getFileTrunc(key []byte, trunc bool) (protocol.FileInfo, bool, error) {
|
2019-11-29 08:11:52 +00:00
|
|
|
bs, err := t.Get(key)
|
|
|
|
if backend.IsNotFound(err) {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return protocol.FileInfo{}, false, nil
|
2019-01-18 10:34:18 +00:00
|
|
|
}
|
|
|
|
if err != nil {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return protocol.FileInfo{}, false, err
|
2019-01-18 10:34:18 +00:00
|
|
|
}
|
lib/db: Deduplicate block lists in database (fixes #5898) (#6283)
* lib/db: Deduplicate block lists in database (fixes #5898)
This moves the block list in the database out from being just a field on
the FileInfo to being an object of its own. When putting a FileInfo we
marshal the block list separately and store it keyed by the sha256 of
the marshalled block list. When getting, if we are not doing a
"truncated" get, we do an extra read and unmarshal for the block list.
Old block lists are cleared out by a periodic GC sweep. The alternative
would be to use refcounting, but:
- There is a larger risk of getting that wrong and either dropping a
block list in error or keeping them around forever.
- It's tricky with our current database, as we don't have dirty reads.
This means that if we update two FileInfos with identical block lists in
the same transaction we can't just do read/modify/write for the ref
counters as we wouldn't see our own first update. See above about
tracking this and risks about getting it wrong.
GC uses a bloom filter for keys to avoid heavy RAM usage. GC can't run
concurrently with FileInfo updates so there is a new lock around those
operation at the lowlevel.
The end result is a much more compact database, especially for setups
with many peers where files get duplicated many times.
This is per-key-class stats for a large database I'm currently working
with, under the current schema:
```
0x00: 9138161 items, 870876 KB keys + 7397482 KB data, 95 B + 809 B avg, 1637651 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x08: 1349 items, 12 KB keys + 10 KB data, 9 B + 8 B avg, 17 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 7 B avg, 30 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
Total 10426475 items, 968490 KB keys + 9202925 KB data.
```
Note 7.4 GB of data in class 00, total size 9.2 GB. After running the
migration we get this instead:
```
0x00: 9138161 items, 870876 KB keys + 2611392 KB data, 95 B + 285 B avg, 4788 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 17 B avg, 51 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
0x0d: 44282 items, 1461 KB keys + 61081 KB data, 33 B + 1379 B avg, 1637399 B max
Total 10469408 items, 969939 KB keys + 4477905 KB data.
```
Class 00 is now down to 2.6 GB, with just 61 MB added in class 0d.
There will be some additional reads in some cases which theoretically
hurts performance, but this will be more than compensated for by smaller
writes and better compaction.
On my own home setup which just has three devices and a handful of
folders the difference is smaller in absolute numbers of course, but
still less than half the old size:
```
0x00: 297122 items, 20894 KB keys + 306860 KB data, 70 B + 1032 B avg, 103237 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
Total 1947412 items, 151268 KB keys + 337485 KB data.
```
to:
```
0x00: 297122 items, 20894 KB keys + 37038 KB data, 70 B + 124 B avg, 520 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
0x0d: 18041 items, 595 KB keys + 71964 KB data, 33 B + 3988 B avg, 101109 B max
Total 1965447 items, 151863 KB keys + 139628 KB data.
```
* wip
* wip
* wip
* wip
2020-01-24 07:35:44 +00:00
|
|
|
f, err := t.unmarshalTrunc(bs, trunc)
|
2020-02-11 14:37:22 +00:00
|
|
|
if backend.IsNotFound(err) {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return protocol.FileInfo{}, false, nil
|
2020-02-11 14:37:22 +00:00
|
|
|
}
|
2019-01-18 10:34:18 +00:00
|
|
|
if err != nil {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return protocol.FileInfo{}, false, err
|
2019-01-18 10:34:18 +00:00
|
|
|
}
|
2019-11-29 08:11:52 +00:00
|
|
|
return f, true, nil
|
2015-10-29 07:07:51 +00:00
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
func (t readOnlyTransaction) unmarshalTrunc(bs []byte, trunc bool) (protocol.FileInfo, error) {
|
lib/db: Deduplicate block lists in database (fixes #5898) (#6283)
* lib/db: Deduplicate block lists in database (fixes #5898)
This moves the block list in the database out from being just a field on
the FileInfo to being an object of its own. When putting a FileInfo we
marshal the block list separately and store it keyed by the sha256 of
the marshalled block list. When getting, if we are not doing a
"truncated" get, we do an extra read and unmarshal for the block list.
Old block lists are cleared out by a periodic GC sweep. The alternative
would be to use refcounting, but:
- There is a larger risk of getting that wrong and either dropping a
block list in error or keeping them around forever.
- It's tricky with our current database, as we don't have dirty reads.
This means that if we update two FileInfos with identical block lists in
the same transaction we can't just do read/modify/write for the ref
counters as we wouldn't see our own first update. See above about
tracking this and risks about getting it wrong.
GC uses a bloom filter for keys to avoid heavy RAM usage. GC can't run
concurrently with FileInfo updates so there is a new lock around those
operation at the lowlevel.
The end result is a much more compact database, especially for setups
with many peers where files get duplicated many times.
This is per-key-class stats for a large database I'm currently working
with, under the current schema:
```
0x00: 9138161 items, 870876 KB keys + 7397482 KB data, 95 B + 809 B avg, 1637651 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x08: 1349 items, 12 KB keys + 10 KB data, 9 B + 8 B avg, 17 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 7 B avg, 30 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
Total 10426475 items, 968490 KB keys + 9202925 KB data.
```
Note 7.4 GB of data in class 00, total size 9.2 GB. After running the
migration we get this instead:
```
0x00: 9138161 items, 870876 KB keys + 2611392 KB data, 95 B + 285 B avg, 4788 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 17 B avg, 51 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
0x0d: 44282 items, 1461 KB keys + 61081 KB data, 33 B + 1379 B avg, 1637399 B max
Total 10469408 items, 969939 KB keys + 4477905 KB data.
```
Class 00 is now down to 2.6 GB, with just 61 MB added in class 0d.
There will be some additional reads in some cases which theoretically
hurts performance, but this will be more than compensated for by smaller
writes and better compaction.
On my own home setup which just has three devices and a handful of
folders the difference is smaller in absolute numbers of course, but
still less than half the old size:
```
0x00: 297122 items, 20894 KB keys + 306860 KB data, 70 B + 1032 B avg, 103237 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
Total 1947412 items, 151268 KB keys + 337485 KB data.
```
to:
```
0x00: 297122 items, 20894 KB keys + 37038 KB data, 70 B + 124 B avg, 520 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
0x0d: 18041 items, 595 KB keys + 71964 KB data, 33 B + 3988 B avg, 101109 B max
Total 1965447 items, 151863 KB keys + 139628 KB data.
```
* wip
* wip
* wip
* wip
2020-01-24 07:35:44 +00:00
|
|
|
if trunc {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
var bfi dbproto.FileInfoTruncated
|
|
|
|
err := proto.Unmarshal(bs, &bfi)
|
lib/db: Deduplicate block lists in database (fixes #5898) (#6283)
* lib/db: Deduplicate block lists in database (fixes #5898)
This moves the block list in the database out from being just a field on
the FileInfo to being an object of its own. When putting a FileInfo we
marshal the block list separately and store it keyed by the sha256 of
the marshalled block list. When getting, if we are not doing a
"truncated" get, we do an extra read and unmarshal for the block list.
Old block lists are cleared out by a periodic GC sweep. The alternative
would be to use refcounting, but:
- There is a larger risk of getting that wrong and either dropping a
block list in error or keeping them around forever.
- It's tricky with our current database, as we don't have dirty reads.
This means that if we update two FileInfos with identical block lists in
the same transaction we can't just do read/modify/write for the ref
counters as we wouldn't see our own first update. See above about
tracking this and risks about getting it wrong.
GC uses a bloom filter for keys to avoid heavy RAM usage. GC can't run
concurrently with FileInfo updates so there is a new lock around those
operation at the lowlevel.
The end result is a much more compact database, especially for setups
with many peers where files get duplicated many times.
This is per-key-class stats for a large database I'm currently working
with, under the current schema:
```
0x00: 9138161 items, 870876 KB keys + 7397482 KB data, 95 B + 809 B avg, 1637651 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x08: 1349 items, 12 KB keys + 10 KB data, 9 B + 8 B avg, 17 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 7 B avg, 30 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
Total 10426475 items, 968490 KB keys + 9202925 KB data.
```
Note 7.4 GB of data in class 00, total size 9.2 GB. After running the
migration we get this instead:
```
0x00: 9138161 items, 870876 KB keys + 2611392 KB data, 95 B + 285 B avg, 4788 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 17 B avg, 51 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
0x0d: 44282 items, 1461 KB keys + 61081 KB data, 33 B + 1379 B avg, 1637399 B max
Total 10469408 items, 969939 KB keys + 4477905 KB data.
```
Class 00 is now down to 2.6 GB, with just 61 MB added in class 0d.
There will be some additional reads in some cases which theoretically
hurts performance, but this will be more than compensated for by smaller
writes and better compaction.
On my own home setup which just has three devices and a handful of
folders the difference is smaller in absolute numbers of course, but
still less than half the old size:
```
0x00: 297122 items, 20894 KB keys + 306860 KB data, 70 B + 1032 B avg, 103237 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
Total 1947412 items, 151268 KB keys + 337485 KB data.
```
to:
```
0x00: 297122 items, 20894 KB keys + 37038 KB data, 70 B + 124 B avg, 520 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
0x0d: 18041 items, 595 KB keys + 71964 KB data, 33 B + 3988 B avg, 101109 B max
Total 1965447 items, 151863 KB keys + 139628 KB data.
```
* wip
* wip
* wip
* wip
2020-01-24 07:35:44 +00:00
|
|
|
if err != nil {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return protocol.FileInfo{}, err
|
lib/db: Deduplicate block lists in database (fixes #5898) (#6283)
* lib/db: Deduplicate block lists in database (fixes #5898)
This moves the block list in the database out from being just a field on
the FileInfo to being an object of its own. When putting a FileInfo we
marshal the block list separately and store it keyed by the sha256 of
the marshalled block list. When getting, if we are not doing a
"truncated" get, we do an extra read and unmarshal for the block list.
Old block lists are cleared out by a periodic GC sweep. The alternative
would be to use refcounting, but:
- There is a larger risk of getting that wrong and either dropping a
block list in error or keeping them around forever.
- It's tricky with our current database, as we don't have dirty reads.
This means that if we update two FileInfos with identical block lists in
the same transaction we can't just do read/modify/write for the ref
counters as we wouldn't see our own first update. See above about
tracking this and risks about getting it wrong.
GC uses a bloom filter for keys to avoid heavy RAM usage. GC can't run
concurrently with FileInfo updates so there is a new lock around those
operation at the lowlevel.
The end result is a much more compact database, especially for setups
with many peers where files get duplicated many times.
This is per-key-class stats for a large database I'm currently working
with, under the current schema:
```
0x00: 9138161 items, 870876 KB keys + 7397482 KB data, 95 B + 809 B avg, 1637651 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x08: 1349 items, 12 KB keys + 10 KB data, 9 B + 8 B avg, 17 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 7 B avg, 30 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
Total 10426475 items, 968490 KB keys + 9202925 KB data.
```
Note 7.4 GB of data in class 00, total size 9.2 GB. After running the
migration we get this instead:
```
0x00: 9138161 items, 870876 KB keys + 2611392 KB data, 95 B + 285 B avg, 4788 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 17 B avg, 51 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
0x0d: 44282 items, 1461 KB keys + 61081 KB data, 33 B + 1379 B avg, 1637399 B max
Total 10469408 items, 969939 KB keys + 4477905 KB data.
```
Class 00 is now down to 2.6 GB, with just 61 MB added in class 0d.
There will be some additional reads in some cases which theoretically
hurts performance, but this will be more than compensated for by smaller
writes and better compaction.
On my own home setup which just has three devices and a handful of
folders the difference is smaller in absolute numbers of course, but
still less than half the old size:
```
0x00: 297122 items, 20894 KB keys + 306860 KB data, 70 B + 1032 B avg, 103237 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
Total 1947412 items, 151268 KB keys + 337485 KB data.
```
to:
```
0x00: 297122 items, 20894 KB keys + 37038 KB data, 70 B + 124 B avg, 520 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
0x0d: 18041 items, 595 KB keys + 71964 KB data, 33 B + 3988 B avg, 101109 B max
Total 1965447 items, 151863 KB keys + 139628 KB data.
```
* wip
* wip
* wip
* wip
2020-01-24 07:35:44 +00:00
|
|
|
}
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
if err := t.fillTruncated(&bfi); err != nil {
|
|
|
|
return protocol.FileInfo{}, err
|
2020-05-13 12:28:42 +00:00
|
|
|
}
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return protocol.FileInfoFromDBTruncated(&bfi), nil
|
lib/db: Deduplicate block lists in database (fixes #5898) (#6283)
* lib/db: Deduplicate block lists in database (fixes #5898)
This moves the block list in the database out from being just a field on
the FileInfo to being an object of its own. When putting a FileInfo we
marshal the block list separately and store it keyed by the sha256 of
the marshalled block list. When getting, if we are not doing a
"truncated" get, we do an extra read and unmarshal for the block list.
Old block lists are cleared out by a periodic GC sweep. The alternative
would be to use refcounting, but:
- There is a larger risk of getting that wrong and either dropping a
block list in error or keeping them around forever.
- It's tricky with our current database, as we don't have dirty reads.
This means that if we update two FileInfos with identical block lists in
the same transaction we can't just do read/modify/write for the ref
counters as we wouldn't see our own first update. See above about
tracking this and risks about getting it wrong.
GC uses a bloom filter for keys to avoid heavy RAM usage. GC can't run
concurrently with FileInfo updates so there is a new lock around those
operation at the lowlevel.
The end result is a much more compact database, especially for setups
with many peers where files get duplicated many times.
This is per-key-class stats for a large database I'm currently working
with, under the current schema:
```
0x00: 9138161 items, 870876 KB keys + 7397482 KB data, 95 B + 809 B avg, 1637651 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x08: 1349 items, 12 KB keys + 10 KB data, 9 B + 8 B avg, 17 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 7 B avg, 30 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
Total 10426475 items, 968490 KB keys + 9202925 KB data.
```
Note 7.4 GB of data in class 00, total size 9.2 GB. After running the
migration we get this instead:
```
0x00: 9138161 items, 870876 KB keys + 2611392 KB data, 95 B + 285 B avg, 4788 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 17 B avg, 51 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
0x0d: 44282 items, 1461 KB keys + 61081 KB data, 33 B + 1379 B avg, 1637399 B max
Total 10469408 items, 969939 KB keys + 4477905 KB data.
```
Class 00 is now down to 2.6 GB, with just 61 MB added in class 0d.
There will be some additional reads in some cases which theoretically
hurts performance, but this will be more than compensated for by smaller
writes and better compaction.
On my own home setup which just has three devices and a handful of
folders the difference is smaller in absolute numbers of course, but
still less than half the old size:
```
0x00: 297122 items, 20894 KB keys + 306860 KB data, 70 B + 1032 B avg, 103237 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
Total 1947412 items, 151268 KB keys + 337485 KB data.
```
to:
```
0x00: 297122 items, 20894 KB keys + 37038 KB data, 70 B + 124 B avg, 520 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
0x0d: 18041 items, 595 KB keys + 71964 KB data, 33 B + 3988 B avg, 101109 B max
Total 1965447 items, 151863 KB keys + 139628 KB data.
```
* wip
* wip
* wip
* wip
2020-01-24 07:35:44 +00:00
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
var bfi bep.FileInfo
|
|
|
|
err := proto.Unmarshal(bs, &bfi)
|
|
|
|
if err != nil {
|
|
|
|
return protocol.FileInfo{}, err
|
lib/db: Deduplicate block lists in database (fixes #5898) (#6283)
* lib/db: Deduplicate block lists in database (fixes #5898)
This moves the block list in the database out from being just a field on
the FileInfo to being an object of its own. When putting a FileInfo we
marshal the block list separately and store it keyed by the sha256 of
the marshalled block list. When getting, if we are not doing a
"truncated" get, we do an extra read and unmarshal for the block list.
Old block lists are cleared out by a periodic GC sweep. The alternative
would be to use refcounting, but:
- There is a larger risk of getting that wrong and either dropping a
block list in error or keeping them around forever.
- It's tricky with our current database, as we don't have dirty reads.
This means that if we update two FileInfos with identical block lists in
the same transaction we can't just do read/modify/write for the ref
counters as we wouldn't see our own first update. See above about
tracking this and risks about getting it wrong.
GC uses a bloom filter for keys to avoid heavy RAM usage. GC can't run
concurrently with FileInfo updates so there is a new lock around those
operation at the lowlevel.
The end result is a much more compact database, especially for setups
with many peers where files get duplicated many times.
This is per-key-class stats for a large database I'm currently working
with, under the current schema:
```
0x00: 9138161 items, 870876 KB keys + 7397482 KB data, 95 B + 809 B avg, 1637651 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x08: 1349 items, 12 KB keys + 10 KB data, 9 B + 8 B avg, 17 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 7 B avg, 30 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
Total 10426475 items, 968490 KB keys + 9202925 KB data.
```
Note 7.4 GB of data in class 00, total size 9.2 GB. After running the
migration we get this instead:
```
0x00: 9138161 items, 870876 KB keys + 2611392 KB data, 95 B + 285 B avg, 4788 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 17 B avg, 51 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
0x0d: 44282 items, 1461 KB keys + 61081 KB data, 33 B + 1379 B avg, 1637399 B max
Total 10469408 items, 969939 KB keys + 4477905 KB data.
```
Class 00 is now down to 2.6 GB, with just 61 MB added in class 0d.
There will be some additional reads in some cases which theoretically
hurts performance, but this will be more than compensated for by smaller
writes and better compaction.
On my own home setup which just has three devices and a handful of
folders the difference is smaller in absolute numbers of course, but
still less than half the old size:
```
0x00: 297122 items, 20894 KB keys + 306860 KB data, 70 B + 1032 B avg, 103237 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
Total 1947412 items, 151268 KB keys + 337485 KB data.
```
to:
```
0x00: 297122 items, 20894 KB keys + 37038 KB data, 70 B + 124 B avg, 520 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
0x0d: 18041 items, 595 KB keys + 71964 KB data, 33 B + 3988 B avg, 101109 B max
Total 1965447 items, 151863 KB keys + 139628 KB data.
```
* wip
* wip
* wip
* wip
2020-01-24 07:35:44 +00:00
|
|
|
}
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
if err := t.fillFileInfo(&bfi); err != nil {
|
|
|
|
return protocol.FileInfo{}, err
|
lib/db: Deduplicate block lists in database (fixes #5898) (#6283)
* lib/db: Deduplicate block lists in database (fixes #5898)
This moves the block list in the database out from being just a field on
the FileInfo to being an object of its own. When putting a FileInfo we
marshal the block list separately and store it keyed by the sha256 of
the marshalled block list. When getting, if we are not doing a
"truncated" get, we do an extra read and unmarshal for the block list.
Old block lists are cleared out by a periodic GC sweep. The alternative
would be to use refcounting, but:
- There is a larger risk of getting that wrong and either dropping a
block list in error or keeping them around forever.
- It's tricky with our current database, as we don't have dirty reads.
This means that if we update two FileInfos with identical block lists in
the same transaction we can't just do read/modify/write for the ref
counters as we wouldn't see our own first update. See above about
tracking this and risks about getting it wrong.
GC uses a bloom filter for keys to avoid heavy RAM usage. GC can't run
concurrently with FileInfo updates so there is a new lock around those
operation at the lowlevel.
The end result is a much more compact database, especially for setups
with many peers where files get duplicated many times.
This is per-key-class stats for a large database I'm currently working
with, under the current schema:
```
0x00: 9138161 items, 870876 KB keys + 7397482 KB data, 95 B + 809 B avg, 1637651 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x08: 1349 items, 12 KB keys + 10 KB data, 9 B + 8 B avg, 17 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 7 B avg, 30 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
Total 10426475 items, 968490 KB keys + 9202925 KB data.
```
Note 7.4 GB of data in class 00, total size 9.2 GB. After running the
migration we get this instead:
```
0x00: 9138161 items, 870876 KB keys + 2611392 KB data, 95 B + 285 B avg, 4788 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 17 B avg, 51 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
0x0d: 44282 items, 1461 KB keys + 61081 KB data, 33 B + 1379 B avg, 1637399 B max
Total 10469408 items, 969939 KB keys + 4477905 KB data.
```
Class 00 is now down to 2.6 GB, with just 61 MB added in class 0d.
There will be some additional reads in some cases which theoretically
hurts performance, but this will be more than compensated for by smaller
writes and better compaction.
On my own home setup which just has three devices and a handful of
folders the difference is smaller in absolute numbers of course, but
still less than half the old size:
```
0x00: 297122 items, 20894 KB keys + 306860 KB data, 70 B + 1032 B avg, 103237 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
Total 1947412 items, 151268 KB keys + 337485 KB data.
```
to:
```
0x00: 297122 items, 20894 KB keys + 37038 KB data, 70 B + 124 B avg, 520 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
0x0d: 18041 items, 595 KB keys + 71964 KB data, 33 B + 3988 B avg, 101109 B max
Total 1965447 items, 151863 KB keys + 139628 KB data.
```
* wip
* wip
* wip
* wip
2020-01-24 07:35:44 +00:00
|
|
|
}
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return protocol.FileInfoFromDB(&bfi), nil
|
lib/db: Deduplicate block lists in database (fixes #5898) (#6283)
* lib/db: Deduplicate block lists in database (fixes #5898)
This moves the block list in the database out from being just a field on
the FileInfo to being an object of its own. When putting a FileInfo we
marshal the block list separately and store it keyed by the sha256 of
the marshalled block list. When getting, if we are not doing a
"truncated" get, we do an extra read and unmarshal for the block list.
Old block lists are cleared out by a periodic GC sweep. The alternative
would be to use refcounting, but:
- There is a larger risk of getting that wrong and either dropping a
block list in error or keeping them around forever.
- It's tricky with our current database, as we don't have dirty reads.
This means that if we update two FileInfos with identical block lists in
the same transaction we can't just do read/modify/write for the ref
counters as we wouldn't see our own first update. See above about
tracking this and risks about getting it wrong.
GC uses a bloom filter for keys to avoid heavy RAM usage. GC can't run
concurrently with FileInfo updates so there is a new lock around those
operation at the lowlevel.
The end result is a much more compact database, especially for setups
with many peers where files get duplicated many times.
This is per-key-class stats for a large database I'm currently working
with, under the current schema:
```
0x00: 9138161 items, 870876 KB keys + 7397482 KB data, 95 B + 809 B avg, 1637651 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x08: 1349 items, 12 KB keys + 10 KB data, 9 B + 8 B avg, 17 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 7 B avg, 30 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
Total 10426475 items, 968490 KB keys + 9202925 KB data.
```
Note 7.4 GB of data in class 00, total size 9.2 GB. After running the
migration we get this instead:
```
0x00: 9138161 items, 870876 KB keys + 2611392 KB data, 95 B + 285 B avg, 4788 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 17 B avg, 51 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
0x0d: 44282 items, 1461 KB keys + 61081 KB data, 33 B + 1379 B avg, 1637399 B max
Total 10469408 items, 969939 KB keys + 4477905 KB data.
```
Class 00 is now down to 2.6 GB, with just 61 MB added in class 0d.
There will be some additional reads in some cases which theoretically
hurts performance, but this will be more than compensated for by smaller
writes and better compaction.
On my own home setup which just has three devices and a handful of
folders the difference is smaller in absolute numbers of course, but
still less than half the old size:
```
0x00: 297122 items, 20894 KB keys + 306860 KB data, 70 B + 1032 B avg, 103237 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
Total 1947412 items, 151268 KB keys + 337485 KB data.
```
to:
```
0x00: 297122 items, 20894 KB keys + 37038 KB data, 70 B + 124 B avg, 520 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
0x0d: 18041 items, 595 KB keys + 71964 KB data, 33 B + 3988 B avg, 101109 B max
Total 1965447 items, 151863 KB keys + 139628 KB data.
```
* wip
* wip
* wip
* wip
2020-01-24 07:35:44 +00:00
|
|
|
}
|
|
|
|
|
2021-04-05 08:24:16 +00:00
|
|
|
type blocksIndirectionError struct {
|
|
|
|
err error
|
|
|
|
}
|
|
|
|
|
|
|
|
func (e *blocksIndirectionError) Error() string {
|
|
|
|
return fmt.Sprintf("filling Blocks: %v", e.err)
|
|
|
|
}
|
|
|
|
|
|
|
|
func (e *blocksIndirectionError) Unwrap() error {
|
|
|
|
return e.err
|
|
|
|
}
|
|
|
|
|
2020-05-13 12:28:42 +00:00
|
|
|
// fillFileInfo follows the (possible) indirection of blocks and version
|
|
|
|
// vector and fills it out.
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
func (t readOnlyTransaction) fillFileInfo(fi *bep.FileInfo) error {
|
2020-02-27 10:19:21 +00:00
|
|
|
var key []byte
|
|
|
|
|
|
|
|
if len(fi.Blocks) == 0 && len(fi.BlocksHash) != 0 {
|
|
|
|
// The blocks list is indirected and we need to load it.
|
|
|
|
key = t.keyer.GenerateBlockListKey(key, fi.BlocksHash)
|
|
|
|
bs, err := t.Get(key)
|
|
|
|
if err != nil {
|
2021-04-05 08:24:16 +00:00
|
|
|
return &blocksIndirectionError{err}
|
2020-02-27 10:19:21 +00:00
|
|
|
}
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
var bl dbproto.BlockList
|
|
|
|
if err := proto.Unmarshal(bs, &bl); err != nil {
|
2020-02-27 10:19:21 +00:00
|
|
|
return err
|
|
|
|
}
|
|
|
|
fi.Blocks = bl.Blocks
|
lib/db: Deduplicate block lists in database (fixes #5898) (#6283)
* lib/db: Deduplicate block lists in database (fixes #5898)
This moves the block list in the database out from being just a field on
the FileInfo to being an object of its own. When putting a FileInfo we
marshal the block list separately and store it keyed by the sha256 of
the marshalled block list. When getting, if we are not doing a
"truncated" get, we do an extra read and unmarshal for the block list.
Old block lists are cleared out by a periodic GC sweep. The alternative
would be to use refcounting, but:
- There is a larger risk of getting that wrong and either dropping a
block list in error or keeping them around forever.
- It's tricky with our current database, as we don't have dirty reads.
This means that if we update two FileInfos with identical block lists in
the same transaction we can't just do read/modify/write for the ref
counters as we wouldn't see our own first update. See above about
tracking this and risks about getting it wrong.
GC uses a bloom filter for keys to avoid heavy RAM usage. GC can't run
concurrently with FileInfo updates so there is a new lock around those
operation at the lowlevel.
The end result is a much more compact database, especially for setups
with many peers where files get duplicated many times.
This is per-key-class stats for a large database I'm currently working
with, under the current schema:
```
0x00: 9138161 items, 870876 KB keys + 7397482 KB data, 95 B + 809 B avg, 1637651 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x08: 1349 items, 12 KB keys + 10 KB data, 9 B + 8 B avg, 17 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 7 B avg, 30 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
Total 10426475 items, 968490 KB keys + 9202925 KB data.
```
Note 7.4 GB of data in class 00, total size 9.2 GB. After running the
migration we get this instead:
```
0x00: 9138161 items, 870876 KB keys + 2611392 KB data, 95 B + 285 B avg, 4788 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 17 B avg, 51 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
0x0d: 44282 items, 1461 KB keys + 61081 KB data, 33 B + 1379 B avg, 1637399 B max
Total 10469408 items, 969939 KB keys + 4477905 KB data.
```
Class 00 is now down to 2.6 GB, with just 61 MB added in class 0d.
There will be some additional reads in some cases which theoretically
hurts performance, but this will be more than compensated for by smaller
writes and better compaction.
On my own home setup which just has three devices and a handful of
folders the difference is smaller in absolute numbers of course, but
still less than half the old size:
```
0x00: 297122 items, 20894 KB keys + 306860 KB data, 70 B + 1032 B avg, 103237 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
Total 1947412 items, 151268 KB keys + 337485 KB data.
```
to:
```
0x00: 297122 items, 20894 KB keys + 37038 KB data, 70 B + 124 B avg, 520 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
0x0d: 18041 items, 595 KB keys + 71964 KB data, 33 B + 3988 B avg, 101109 B max
Total 1965447 items, 151863 KB keys + 139628 KB data.
```
* wip
* wip
* wip
* wip
2020-01-24 07:35:44 +00:00
|
|
|
}
|
2020-02-27 10:19:21 +00:00
|
|
|
|
2020-05-13 12:28:42 +00:00
|
|
|
if len(fi.VersionHash) != 0 {
|
|
|
|
key = t.keyer.GenerateVersionKey(key, fi.VersionHash)
|
|
|
|
bs, err := t.Get(key)
|
|
|
|
if err != nil {
|
2020-08-18 07:20:12 +00:00
|
|
|
return fmt.Errorf("filling Version: %w", err)
|
2020-05-13 12:28:42 +00:00
|
|
|
}
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
var v bep.Vector
|
|
|
|
if err := proto.Unmarshal(bs, &v); err != nil {
|
2020-05-13 12:28:42 +00:00
|
|
|
return err
|
|
|
|
}
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
fi.Version = &v
|
2020-05-13 12:28:42 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
|
|
|
|
// fillTruncated follows the (possible) indirection of version vector and
|
|
|
|
// fills it.
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
func (t readOnlyTransaction) fillTruncated(fi *dbproto.FileInfoTruncated) error {
|
2020-05-13 12:28:42 +00:00
|
|
|
var key []byte
|
|
|
|
|
|
|
|
if len(fi.VersionHash) == 0 {
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
|
|
|
|
key = t.keyer.GenerateVersionKey(key, fi.VersionHash)
|
|
|
|
bs, err := t.Get(key)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
var v bep.Vector
|
|
|
|
if err := proto.Unmarshal(bs, &v); err != nil {
|
2020-05-13 12:28:42 +00:00
|
|
|
return err
|
|
|
|
}
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
fi.Version = &v
|
lib/db: Deduplicate block lists in database (fixes #5898) (#6283)
* lib/db: Deduplicate block lists in database (fixes #5898)
This moves the block list in the database out from being just a field on
the FileInfo to being an object of its own. When putting a FileInfo we
marshal the block list separately and store it keyed by the sha256 of
the marshalled block list. When getting, if we are not doing a
"truncated" get, we do an extra read and unmarshal for the block list.
Old block lists are cleared out by a periodic GC sweep. The alternative
would be to use refcounting, but:
- There is a larger risk of getting that wrong and either dropping a
block list in error or keeping them around forever.
- It's tricky with our current database, as we don't have dirty reads.
This means that if we update two FileInfos with identical block lists in
the same transaction we can't just do read/modify/write for the ref
counters as we wouldn't see our own first update. See above about
tracking this and risks about getting it wrong.
GC uses a bloom filter for keys to avoid heavy RAM usage. GC can't run
concurrently with FileInfo updates so there is a new lock around those
operation at the lowlevel.
The end result is a much more compact database, especially for setups
with many peers where files get duplicated many times.
This is per-key-class stats for a large database I'm currently working
with, under the current schema:
```
0x00: 9138161 items, 870876 KB keys + 7397482 KB data, 95 B + 809 B avg, 1637651 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x08: 1349 items, 12 KB keys + 10 KB data, 9 B + 8 B avg, 17 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 7 B avg, 30 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
Total 10426475 items, 968490 KB keys + 9202925 KB data.
```
Note 7.4 GB of data in class 00, total size 9.2 GB. After running the
migration we get this instead:
```
0x00: 9138161 items, 870876 KB keys + 2611392 KB data, 95 B + 285 B avg, 4788 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 17 B avg, 51 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
0x0d: 44282 items, 1461 KB keys + 61081 KB data, 33 B + 1379 B avg, 1637399 B max
Total 10469408 items, 969939 KB keys + 4477905 KB data.
```
Class 00 is now down to 2.6 GB, with just 61 MB added in class 0d.
There will be some additional reads in some cases which theoretically
hurts performance, but this will be more than compensated for by smaller
writes and better compaction.
On my own home setup which just has three devices and a handful of
folders the difference is smaller in absolute numbers of course, but
still less than half the old size:
```
0x00: 297122 items, 20894 KB keys + 306860 KB data, 70 B + 1032 B avg, 103237 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
Total 1947412 items, 151268 KB keys + 337485 KB data.
```
to:
```
0x00: 297122 items, 20894 KB keys + 37038 KB data, 70 B + 124 B avg, 520 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
0x0d: 18041 items, 595 KB keys + 71964 KB data, 33 B + 3988 B avg, 101109 B max
Total 1965447 items, 151863 KB keys + 139628 KB data.
```
* wip
* wip
* wip
* wip
2020-01-24 07:35:44 +00:00
|
|
|
return nil
|
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
func (t readOnlyTransaction) getGlobalVersions(keyBuf, folder, file []byte) (*dbproto.VersionList, error) {
|
2019-11-29 08:11:52 +00:00
|
|
|
var err error
|
|
|
|
keyBuf, err = t.keyer.GenerateGlobalVersionKey(keyBuf, folder, file)
|
|
|
|
if err != nil {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return nil, err
|
2019-11-29 08:11:52 +00:00
|
|
|
}
|
2020-05-01 07:30:20 +00:00
|
|
|
return t.getGlobalVersionsByKey(keyBuf)
|
|
|
|
}
|
2019-01-18 12:01:39 +00:00
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
func (t readOnlyTransaction) getGlobalVersionsByKey(key []byte) (*dbproto.VersionList, error) {
|
2020-05-01 07:30:20 +00:00
|
|
|
bs, err := t.Get(key)
|
2019-01-18 12:01:39 +00:00
|
|
|
if err != nil {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return nil, err
|
2019-01-18 12:01:39 +00:00
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
var vl dbproto.VersionList
|
|
|
|
if err := proto.Unmarshal(bs, &vl); err != nil {
|
|
|
|
return nil, err
|
2020-05-01 07:30:20 +00:00
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return &vl, nil
|
2020-05-01 07:30:20 +00:00
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
func (t readOnlyTransaction) getGlobal(keyBuf, folder, file []byte, truncate bool) ([]byte, protocol.FileInfo, bool, error) {
|
2020-05-01 07:30:20 +00:00
|
|
|
vl, err := t.getGlobalVersions(keyBuf, folder, file)
|
|
|
|
if backend.IsNotFound(err) {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return keyBuf, protocol.FileInfo{}, false, nil
|
2020-05-01 07:30:20 +00:00
|
|
|
} else if err != nil {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return nil, protocol.FileInfo{}, false, err
|
2019-01-18 12:01:39 +00:00
|
|
|
}
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
keyBuf, fi, err := t.getGlobalFromVersionList(keyBuf, folder, file, truncate, vl)
|
2020-05-30 07:50:23 +00:00
|
|
|
return keyBuf, fi, true, err
|
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
func (t readOnlyTransaction) getGlobalFromVersionList(keyBuf, folder, file []byte, truncate bool, vl *dbproto.VersionList) ([]byte, protocol.FileInfo, error) {
|
|
|
|
fv, ok := vlGetGlobal(vl)
|
2020-05-30 07:50:23 +00:00
|
|
|
if !ok {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return keyBuf, protocol.FileInfo{}, errEmptyGlobal
|
2020-05-11 13:07:06 +00:00
|
|
|
}
|
2020-05-30 07:50:23 +00:00
|
|
|
keyBuf, fi, err := t.getGlobalFromFileVersion(keyBuf, folder, file, truncate, fv)
|
2021-09-07 22:11:16 +00:00
|
|
|
return keyBuf, fi, err
|
2020-05-30 07:50:23 +00:00
|
|
|
}
|
2019-01-18 12:01:39 +00:00
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
func (t readOnlyTransaction) getGlobalFromFileVersion(keyBuf, folder, file []byte, truncate bool, fv *dbproto.FileVersion) ([]byte, protocol.FileInfo, error) {
|
|
|
|
dev, ok := fvFirstDevice(fv)
|
2020-05-30 07:50:23 +00:00
|
|
|
if !ok {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return keyBuf, protocol.FileInfo{}, errEmptyFileVersion
|
2020-05-30 07:50:23 +00:00
|
|
|
}
|
|
|
|
keyBuf, err := t.keyer.GenerateDeviceFileKey(keyBuf, folder, dev, file)
|
2019-11-29 08:11:52 +00:00
|
|
|
if err != nil {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return keyBuf, protocol.FileInfo{}, err
|
2019-01-18 12:01:39 +00:00
|
|
|
}
|
2019-11-29 08:11:52 +00:00
|
|
|
fi, ok, err := t.getFileTrunc(keyBuf, truncate)
|
2020-05-30 07:50:23 +00:00
|
|
|
if err != nil {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return keyBuf, protocol.FileInfo{}, err
|
2019-11-29 08:11:52 +00:00
|
|
|
}
|
2020-05-30 07:50:23 +00:00
|
|
|
if !ok {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return keyBuf, protocol.FileInfo{}, errEntryFromGlobalMissing
|
2020-05-30 07:50:23 +00:00
|
|
|
}
|
|
|
|
return keyBuf, fi, nil
|
2019-01-18 12:01:39 +00:00
|
|
|
}
|
|
|
|
|
2020-01-21 17:23:08 +00:00
|
|
|
func (t *readOnlyTransaction) withHave(folder, device, prefix []byte, truncate bool, fn Iterator) error {
|
|
|
|
if len(prefix) > 0 {
|
|
|
|
unslashedPrefix := prefix
|
|
|
|
if bytes.HasSuffix(prefix, []byte{'/'}) {
|
|
|
|
unslashedPrefix = unslashedPrefix[:len(unslashedPrefix)-1]
|
|
|
|
} else {
|
|
|
|
prefix = append(prefix, '/')
|
|
|
|
}
|
|
|
|
|
|
|
|
key, err := t.keyer.GenerateDeviceFileKey(nil, folder, device, unslashedPrefix)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
2020-05-08 19:31:43 +00:00
|
|
|
if f, ok, err := t.getFileTrunc(key, truncate); err != nil {
|
2020-01-21 17:23:08 +00:00
|
|
|
return err
|
|
|
|
} else if ok && !fn(f) {
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
key, err := t.keyer.GenerateDeviceFileKey(nil, folder, device, prefix)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
dbi, err := t.NewPrefixIterator(key)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
defer dbi.Release()
|
|
|
|
|
|
|
|
for dbi.Next() {
|
|
|
|
name := t.keyer.NameFromDeviceFileKey(dbi.Key())
|
|
|
|
if len(prefix) > 0 && !bytes.HasPrefix(name, prefix) {
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
|
lib/db: Deduplicate block lists in database (fixes #5898) (#6283)
* lib/db: Deduplicate block lists in database (fixes #5898)
This moves the block list in the database out from being just a field on
the FileInfo to being an object of its own. When putting a FileInfo we
marshal the block list separately and store it keyed by the sha256 of
the marshalled block list. When getting, if we are not doing a
"truncated" get, we do an extra read and unmarshal for the block list.
Old block lists are cleared out by a periodic GC sweep. The alternative
would be to use refcounting, but:
- There is a larger risk of getting that wrong and either dropping a
block list in error or keeping them around forever.
- It's tricky with our current database, as we don't have dirty reads.
This means that if we update two FileInfos with identical block lists in
the same transaction we can't just do read/modify/write for the ref
counters as we wouldn't see our own first update. See above about
tracking this and risks about getting it wrong.
GC uses a bloom filter for keys to avoid heavy RAM usage. GC can't run
concurrently with FileInfo updates so there is a new lock around those
operation at the lowlevel.
The end result is a much more compact database, especially for setups
with many peers where files get duplicated many times.
This is per-key-class stats for a large database I'm currently working
with, under the current schema:
```
0x00: 9138161 items, 870876 KB keys + 7397482 KB data, 95 B + 809 B avg, 1637651 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x08: 1349 items, 12 KB keys + 10 KB data, 9 B + 8 B avg, 17 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 7 B avg, 30 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
Total 10426475 items, 968490 KB keys + 9202925 KB data.
```
Note 7.4 GB of data in class 00, total size 9.2 GB. After running the
migration we get this instead:
```
0x00: 9138161 items, 870876 KB keys + 2611392 KB data, 95 B + 285 B avg, 4788 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 17 B avg, 51 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
0x0d: 44282 items, 1461 KB keys + 61081 KB data, 33 B + 1379 B avg, 1637399 B max
Total 10469408 items, 969939 KB keys + 4477905 KB data.
```
Class 00 is now down to 2.6 GB, with just 61 MB added in class 0d.
There will be some additional reads in some cases which theoretically
hurts performance, but this will be more than compensated for by smaller
writes and better compaction.
On my own home setup which just has three devices and a handful of
folders the difference is smaller in absolute numbers of course, but
still less than half the old size:
```
0x00: 297122 items, 20894 KB keys + 306860 KB data, 70 B + 1032 B avg, 103237 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
Total 1947412 items, 151268 KB keys + 337485 KB data.
```
to:
```
0x00: 297122 items, 20894 KB keys + 37038 KB data, 70 B + 124 B avg, 520 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
0x0d: 18041 items, 595 KB keys + 71964 KB data, 33 B + 3988 B avg, 101109 B max
Total 1965447 items, 151863 KB keys + 139628 KB data.
```
* wip
* wip
* wip
* wip
2020-01-24 07:35:44 +00:00
|
|
|
f, err := t.unmarshalTrunc(dbi.Value(), truncate)
|
2020-01-21 17:23:08 +00:00
|
|
|
if err != nil {
|
|
|
|
l.Debugln("unmarshal error:", err)
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
if !fn(f) {
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return dbi.Error()
|
|
|
|
}
|
|
|
|
|
|
|
|
func (t *readOnlyTransaction) withHaveSequence(folder []byte, startSeq int64, fn Iterator) error {
|
|
|
|
first, err := t.keyer.GenerateSequenceKey(nil, folder, startSeq)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
last, err := t.keyer.GenerateSequenceKey(nil, folder, maxInt64)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
dbi, err := t.NewRangeIterator(first, last)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
defer dbi.Release()
|
|
|
|
|
|
|
|
for dbi.Next() {
|
|
|
|
f, ok, err := t.getFileByKey(dbi.Value())
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
if !ok {
|
|
|
|
l.Debugln("missing file for sequence number", t.keyer.SequenceFromSequenceKey(dbi.Key()))
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
|
|
|
|
if shouldDebug() {
|
|
|
|
if seq := t.keyer.SequenceFromSequenceKey(dbi.Key()); f.Sequence != seq {
|
2020-05-18 08:42:51 +00:00
|
|
|
l.Debugf("Sequence index corruption (folder %v, file %v): sequence %d != expected %d", string(folder), f.Name, f.Sequence, seq)
|
2020-01-21 17:23:08 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
if !fn(f) {
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return dbi.Error()
|
|
|
|
}
|
|
|
|
|
|
|
|
func (t *readOnlyTransaction) withGlobal(folder, prefix []byte, truncate bool, fn Iterator) error {
|
|
|
|
if len(prefix) > 0 {
|
|
|
|
unslashedPrefix := prefix
|
|
|
|
if bytes.HasSuffix(prefix, []byte{'/'}) {
|
|
|
|
unslashedPrefix = unslashedPrefix[:len(unslashedPrefix)-1]
|
|
|
|
} else {
|
|
|
|
prefix = append(prefix, '/')
|
|
|
|
}
|
|
|
|
|
|
|
|
if _, f, ok, err := t.getGlobal(nil, folder, unslashedPrefix, truncate); err != nil {
|
|
|
|
return err
|
|
|
|
} else if ok && !fn(f) {
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
key, err := t.keyer.GenerateGlobalVersionKey(nil, folder, prefix)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
dbi, err := t.NewPrefixIterator(key)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
defer dbi.Release()
|
|
|
|
|
|
|
|
var dk []byte
|
|
|
|
for dbi.Next() {
|
|
|
|
name := t.keyer.NameFromGlobalVersionKey(dbi.Key())
|
|
|
|
if len(prefix) > 0 && !bytes.HasPrefix(name, prefix) {
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
var vl dbproto.VersionList
|
|
|
|
if err := proto.Unmarshal(dbi.Value(), &vl); err != nil {
|
2020-03-16 09:09:27 +00:00
|
|
|
return err
|
2020-01-21 17:23:08 +00:00
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
var f protocol.FileInfo
|
|
|
|
dk, f, err = t.getGlobalFromVersionList(dk, folder, name, truncate, &vl)
|
2020-01-21 17:23:08 +00:00
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
|
|
|
|
if !fn(f) {
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
return dbi.Error()
|
|
|
|
}
|
|
|
|
|
2020-05-11 18:15:11 +00:00
|
|
|
func (t *readOnlyTransaction) withBlocksHash(folder, hash []byte, iterator Iterator) error {
|
|
|
|
key, err := t.keyer.GenerateBlockListMapKey(nil, folder, hash, nil)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
|
|
|
|
iter, err := t.NewPrefixIterator(key)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
defer iter.Release()
|
|
|
|
|
|
|
|
for iter.Next() {
|
|
|
|
file := string(t.keyer.NameFromBlockListMapKey(iter.Key()))
|
|
|
|
f, ok, err := t.getFile(folder, protocol.LocalDeviceID[:], []byte(osutil.NormalizedFilename(file)))
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
if !ok {
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
f.Name = osutil.NativeFilename(f.Name)
|
|
|
|
|
|
|
|
if !bytes.Equal(f.BlocksHash, hash) {
|
2021-06-07 21:10:35 +00:00
|
|
|
msg := "Mismatching block map list hashes"
|
|
|
|
t.evLogger.Log(events.Failure, fmt.Sprintln(msg, "in withBlocksHash"))
|
|
|
|
l.Warnf("%v: got %x expected %x", msg, f.BlocksHash, hash)
|
2020-05-11 18:15:11 +00:00
|
|
|
continue
|
|
|
|
}
|
|
|
|
|
|
|
|
if f.IsDeleted() || f.IsInvalid() || f.IsDirectory() || f.IsSymlink() {
|
2021-06-07 21:10:35 +00:00
|
|
|
msg := "Found something of unexpected type in block list map"
|
|
|
|
t.evLogger.Log(events.Failure, fmt.Sprintln(msg, "in withBlocksHash"))
|
|
|
|
l.Warnf("%v: %s", msg, f)
|
2020-05-11 18:15:11 +00:00
|
|
|
continue
|
|
|
|
}
|
|
|
|
|
|
|
|
if !iterator(f) {
|
|
|
|
break
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return iter.Error()
|
|
|
|
}
|
|
|
|
|
2020-01-21 17:23:08 +00:00
|
|
|
func (t *readOnlyTransaction) availability(folder, file []byte) ([]protocol.DeviceID, error) {
|
2020-05-01 07:30:20 +00:00
|
|
|
vl, err := t.getGlobalVersions(nil, folder, file)
|
2020-01-21 17:23:08 +00:00
|
|
|
if backend.IsNotFound(err) {
|
|
|
|
return nil, nil
|
|
|
|
}
|
|
|
|
if err != nil {
|
|
|
|
return nil, err
|
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
fv, ok := vlGetGlobal(vl)
|
2020-05-30 07:50:23 +00:00
|
|
|
if !ok {
|
|
|
|
return nil, nil
|
|
|
|
}
|
|
|
|
devices := make([]protocol.DeviceID, len(fv.Devices))
|
|
|
|
for i, dev := range fv.Devices {
|
2020-06-07 08:31:12 +00:00
|
|
|
n, err := protocol.DeviceIDFromBytes(dev)
|
|
|
|
if err != nil {
|
|
|
|
return nil, err
|
|
|
|
}
|
|
|
|
devices[i] = n
|
2020-01-21 17:23:08 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
return devices, nil
|
|
|
|
}
|
|
|
|
|
|
|
|
func (t *readOnlyTransaction) withNeed(folder, device []byte, truncate bool, fn Iterator) error {
|
|
|
|
if bytes.Equal(device, protocol.LocalDeviceID[:]) {
|
|
|
|
return t.withNeedLocal(folder, truncate, fn)
|
|
|
|
}
|
2020-09-04 12:01:46 +00:00
|
|
|
return t.withNeedIteratingGlobal(folder, device, truncate, fn)
|
|
|
|
}
|
2020-01-21 17:23:08 +00:00
|
|
|
|
2020-09-04 12:01:46 +00:00
|
|
|
func (t *readOnlyTransaction) withNeedIteratingGlobal(folder, device []byte, truncate bool, fn Iterator) error {
|
2020-01-21 17:23:08 +00:00
|
|
|
key, err := t.keyer.GenerateGlobalVersionKey(nil, folder, nil)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
dbi, err := t.NewPrefixIterator(key.WithoutName())
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
defer dbi.Release()
|
|
|
|
|
|
|
|
var dk []byte
|
2020-06-07 08:31:12 +00:00
|
|
|
devID, err := protocol.DeviceIDFromBytes(device)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
2020-01-21 17:23:08 +00:00
|
|
|
for dbi.Next() {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
var vl dbproto.VersionList
|
|
|
|
if err := proto.Unmarshal(dbi.Value(), &vl); err != nil {
|
2020-03-16 09:09:27 +00:00
|
|
|
return err
|
2020-01-21 17:23:08 +00:00
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
globalFV, ok := vlGetGlobal(&vl)
|
2020-05-30 07:50:23 +00:00
|
|
|
if !ok {
|
|
|
|
return errEmptyGlobal
|
|
|
|
}
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
haveFV, have := vlGet(&vl, device)
|
2020-01-21 17:23:08 +00:00
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
if !Need(globalFV, have, protocol.VectorFromWire(haveFV.Version)) {
|
2020-05-11 13:07:06 +00:00
|
|
|
continue
|
|
|
|
}
|
2020-05-30 07:50:23 +00:00
|
|
|
|
2020-01-21 17:23:08 +00:00
|
|
|
name := t.keyer.NameFromGlobalVersionKey(dbi.Key())
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
var gf protocol.FileInfo
|
2020-05-30 07:50:23 +00:00
|
|
|
dk, gf, err = t.getGlobalFromFileVersion(dk, folder, name, truncate, globalFV)
|
2020-03-16 07:45:45 +00:00
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
2020-05-30 07:50:23 +00:00
|
|
|
|
2020-09-04 12:01:46 +00:00
|
|
|
if shouldDebug() {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
if globalDev, ok := fvFirstDevice(globalFV); ok {
|
2020-09-04 12:01:46 +00:00
|
|
|
globalID, _ := protocol.DeviceIDFromBytes(globalDev)
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
l.Debugf("need folder=%q device=%v name=%q have=%v invalid=%v haveV=%v haveDeleted=%v globalV=%v globalDeleted=%v globalDev=%v", folder, devID, name, have, fvIsInvalid(haveFV), haveFV.Version, haveFV.Deleted, gf.FileVersion(), globalFV.Deleted, globalID)
|
2020-09-04 12:01:46 +00:00
|
|
|
}
|
2020-03-16 07:45:45 +00:00
|
|
|
}
|
|
|
|
if !fn(gf) {
|
|
|
|
return dbi.Error()
|
2020-01-21 17:23:08 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
return dbi.Error()
|
|
|
|
}
|
|
|
|
|
|
|
|
func (t *readOnlyTransaction) withNeedLocal(folder []byte, truncate bool, fn Iterator) error {
|
|
|
|
key, err := t.keyer.GenerateNeedFileKey(nil, folder, nil)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
dbi, err := t.NewPrefixIterator(key.WithoutName())
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
defer dbi.Release()
|
|
|
|
|
|
|
|
var keyBuf []byte
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
var f protocol.FileInfo
|
2020-01-21 17:23:08 +00:00
|
|
|
var ok bool
|
|
|
|
for dbi.Next() {
|
|
|
|
keyBuf, f, ok, err = t.getGlobal(keyBuf, folder, t.keyer.NameFromGlobalVersionKey(dbi.Key()), truncate)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
if !ok {
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
if !fn(f) {
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return dbi.Error()
|
|
|
|
}
|
|
|
|
|
2015-10-29 07:07:51 +00:00
|
|
|
// A readWriteTransaction is a readOnlyTransaction plus a batch for writes.
|
|
|
|
// The batch will be committed on close() or by checkFlush() if it exceeds the
|
|
|
|
// batch size.
|
|
|
|
type readWriteTransaction struct {
|
2019-11-29 08:11:52 +00:00
|
|
|
backend.WriteTransaction
|
2015-10-29 07:07:51 +00:00
|
|
|
readOnlyTransaction
|
2021-06-07 08:52:06 +00:00
|
|
|
indirectionTracker
|
|
|
|
}
|
|
|
|
|
|
|
|
type indirectionTracker interface {
|
|
|
|
recordIndirectionHashesForFile(f *protocol.FileInfo)
|
2015-10-29 07:07:51 +00:00
|
|
|
}
|
|
|
|
|
2020-07-19 06:55:27 +00:00
|
|
|
func (db *Lowlevel) newReadWriteTransaction(hooks ...backend.CommitHook) (readWriteTransaction, error) {
|
|
|
|
tran, err := db.NewWriteTransaction(hooks...)
|
2019-11-29 08:11:52 +00:00
|
|
|
if err != nil {
|
|
|
|
return readWriteTransaction{}, err
|
2015-10-29 07:07:51 +00:00
|
|
|
}
|
2019-11-29 08:11:52 +00:00
|
|
|
return readWriteTransaction{
|
2021-06-27 06:43:49 +00:00
|
|
|
WriteTransaction: tran,
|
|
|
|
readOnlyTransaction: db.readOnlyTransactionFromBackendTransaction(tran),
|
|
|
|
indirectionTracker: db,
|
2019-11-29 08:11:52 +00:00
|
|
|
}, nil
|
|
|
|
}
|
|
|
|
|
2020-02-12 10:59:55 +00:00
|
|
|
func (t readWriteTransaction) Commit() error {
|
2020-05-29 11:43:02 +00:00
|
|
|
// The readOnlyTransaction must close after commit, because they may be
|
|
|
|
// backed by the same actual lower level transaction.
|
|
|
|
defer t.readOnlyTransaction.close()
|
2019-11-29 08:11:52 +00:00
|
|
|
return t.WriteTransaction.Commit()
|
2015-10-29 07:07:51 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
func (t readWriteTransaction) close() {
|
|
|
|
t.readOnlyTransaction.close()
|
2019-11-29 08:11:52 +00:00
|
|
|
t.WriteTransaction.Release()
|
2015-10-29 07:07:51 +00:00
|
|
|
}
|
|
|
|
|
2020-03-20 11:07:14 +00:00
|
|
|
// putFile stores a file in the database, taking care of indirected fields.
|
2020-08-18 07:20:12 +00:00
|
|
|
func (t readWriteTransaction) putFile(fkey []byte, fi protocol.FileInfo) error {
|
2020-02-27 10:19:21 +00:00
|
|
|
var bkey []byte
|
|
|
|
|
2020-08-18 07:20:12 +00:00
|
|
|
// Always set the blocks hash when there are blocks.
|
2020-02-27 10:19:21 +00:00
|
|
|
if len(fi.Blocks) > 0 {
|
|
|
|
fi.BlocksHash = protocol.BlocksHash(fi.Blocks)
|
2020-08-18 07:20:12 +00:00
|
|
|
} else {
|
2020-02-27 10:19:21 +00:00
|
|
|
fi.BlocksHash = nil
|
|
|
|
}
|
|
|
|
|
|
|
|
// Indirect the blocks if the block list is large enough.
|
|
|
|
if len(fi.Blocks) > blocksIndirectionCutoff {
|
|
|
|
bkey = t.keyer.GenerateBlockListKey(bkey, fi.BlocksHash)
|
|
|
|
if _, err := t.Get(bkey); backend.IsNotFound(err) {
|
lib/db: Deduplicate block lists in database (fixes #5898) (#6283)
* lib/db: Deduplicate block lists in database (fixes #5898)
This moves the block list in the database out from being just a field on
the FileInfo to being an object of its own. When putting a FileInfo we
marshal the block list separately and store it keyed by the sha256 of
the marshalled block list. When getting, if we are not doing a
"truncated" get, we do an extra read and unmarshal for the block list.
Old block lists are cleared out by a periodic GC sweep. The alternative
would be to use refcounting, but:
- There is a larger risk of getting that wrong and either dropping a
block list in error or keeping them around forever.
- It's tricky with our current database, as we don't have dirty reads.
This means that if we update two FileInfos with identical block lists in
the same transaction we can't just do read/modify/write for the ref
counters as we wouldn't see our own first update. See above about
tracking this and risks about getting it wrong.
GC uses a bloom filter for keys to avoid heavy RAM usage. GC can't run
concurrently with FileInfo updates so there is a new lock around those
operation at the lowlevel.
The end result is a much more compact database, especially for setups
with many peers where files get duplicated many times.
This is per-key-class stats for a large database I'm currently working
with, under the current schema:
```
0x00: 9138161 items, 870876 KB keys + 7397482 KB data, 95 B + 809 B avg, 1637651 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x08: 1349 items, 12 KB keys + 10 KB data, 9 B + 8 B avg, 17 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 7 B avg, 30 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
Total 10426475 items, 968490 KB keys + 9202925 KB data.
```
Note 7.4 GB of data in class 00, total size 9.2 GB. After running the
migration we get this instead:
```
0x00: 9138161 items, 870876 KB keys + 2611392 KB data, 95 B + 285 B avg, 4788 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 17 B avg, 51 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
0x0d: 44282 items, 1461 KB keys + 61081 KB data, 33 B + 1379 B avg, 1637399 B max
Total 10469408 items, 969939 KB keys + 4477905 KB data.
```
Class 00 is now down to 2.6 GB, with just 61 MB added in class 0d.
There will be some additional reads in some cases which theoretically
hurts performance, but this will be more than compensated for by smaller
writes and better compaction.
On my own home setup which just has three devices and a handful of
folders the difference is smaller in absolute numbers of course, but
still less than half the old size:
```
0x00: 297122 items, 20894 KB keys + 306860 KB data, 70 B + 1032 B avg, 103237 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
Total 1947412 items, 151268 KB keys + 337485 KB data.
```
to:
```
0x00: 297122 items, 20894 KB keys + 37038 KB data, 70 B + 124 B avg, 520 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
0x0d: 18041 items, 595 KB keys + 71964 KB data, 33 B + 3988 B avg, 101109 B max
Total 1965447 items, 151863 KB keys + 139628 KB data.
```
* wip
* wip
* wip
* wip
2020-01-24 07:35:44 +00:00
|
|
|
// Marshal the block list and save it
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
blocks := sliceutil.Map(fi.Blocks, protocol.BlockInfo.ToWire)
|
|
|
|
blocksBs := mustMarshal(&dbproto.BlockList{Blocks: blocks})
|
2020-02-27 10:19:21 +00:00
|
|
|
if err := t.Put(bkey, blocksBs); err != nil {
|
lib/db: Deduplicate block lists in database (fixes #5898) (#6283)
* lib/db: Deduplicate block lists in database (fixes #5898)
This moves the block list in the database out from being just a field on
the FileInfo to being an object of its own. When putting a FileInfo we
marshal the block list separately and store it keyed by the sha256 of
the marshalled block list. When getting, if we are not doing a
"truncated" get, we do an extra read and unmarshal for the block list.
Old block lists are cleared out by a periodic GC sweep. The alternative
would be to use refcounting, but:
- There is a larger risk of getting that wrong and either dropping a
block list in error or keeping them around forever.
- It's tricky with our current database, as we don't have dirty reads.
This means that if we update two FileInfos with identical block lists in
the same transaction we can't just do read/modify/write for the ref
counters as we wouldn't see our own first update. See above about
tracking this and risks about getting it wrong.
GC uses a bloom filter for keys to avoid heavy RAM usage. GC can't run
concurrently with FileInfo updates so there is a new lock around those
operation at the lowlevel.
The end result is a much more compact database, especially for setups
with many peers where files get duplicated many times.
This is per-key-class stats for a large database I'm currently working
with, under the current schema:
```
0x00: 9138161 items, 870876 KB keys + 7397482 KB data, 95 B + 809 B avg, 1637651 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x08: 1349 items, 12 KB keys + 10 KB data, 9 B + 8 B avg, 17 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 7 B avg, 30 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
Total 10426475 items, 968490 KB keys + 9202925 KB data.
```
Note 7.4 GB of data in class 00, total size 9.2 GB. After running the
migration we get this instead:
```
0x00: 9138161 items, 870876 KB keys + 2611392 KB data, 95 B + 285 B avg, 4788 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 17 B avg, 51 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
0x0d: 44282 items, 1461 KB keys + 61081 KB data, 33 B + 1379 B avg, 1637399 B max
Total 10469408 items, 969939 KB keys + 4477905 KB data.
```
Class 00 is now down to 2.6 GB, with just 61 MB added in class 0d.
There will be some additional reads in some cases which theoretically
hurts performance, but this will be more than compensated for by smaller
writes and better compaction.
On my own home setup which just has three devices and a handful of
folders the difference is smaller in absolute numbers of course, but
still less than half the old size:
```
0x00: 297122 items, 20894 KB keys + 306860 KB data, 70 B + 1032 B avg, 103237 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
Total 1947412 items, 151268 KB keys + 337485 KB data.
```
to:
```
0x00: 297122 items, 20894 KB keys + 37038 KB data, 70 B + 124 B avg, 520 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
0x0d: 18041 items, 595 KB keys + 71964 KB data, 33 B + 3988 B avg, 101109 B max
Total 1965447 items, 151863 KB keys + 139628 KB data.
```
* wip
* wip
* wip
* wip
2020-01-24 07:35:44 +00:00
|
|
|
return err
|
|
|
|
}
|
|
|
|
} else if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
2020-02-27 10:19:21 +00:00
|
|
|
fi.Blocks = nil
|
lib/db: Deduplicate block lists in database (fixes #5898) (#6283)
* lib/db: Deduplicate block lists in database (fixes #5898)
This moves the block list in the database out from being just a field on
the FileInfo to being an object of its own. When putting a FileInfo we
marshal the block list separately and store it keyed by the sha256 of
the marshalled block list. When getting, if we are not doing a
"truncated" get, we do an extra read and unmarshal for the block list.
Old block lists are cleared out by a periodic GC sweep. The alternative
would be to use refcounting, but:
- There is a larger risk of getting that wrong and either dropping a
block list in error or keeping them around forever.
- It's tricky with our current database, as we don't have dirty reads.
This means that if we update two FileInfos with identical block lists in
the same transaction we can't just do read/modify/write for the ref
counters as we wouldn't see our own first update. See above about
tracking this and risks about getting it wrong.
GC uses a bloom filter for keys to avoid heavy RAM usage. GC can't run
concurrently with FileInfo updates so there is a new lock around those
operation at the lowlevel.
The end result is a much more compact database, especially for setups
with many peers where files get duplicated many times.
This is per-key-class stats for a large database I'm currently working
with, under the current schema:
```
0x00: 9138161 items, 870876 KB keys + 7397482 KB data, 95 B + 809 B avg, 1637651 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x08: 1349 items, 12 KB keys + 10 KB data, 9 B + 8 B avg, 17 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 7 B avg, 30 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
Total 10426475 items, 968490 KB keys + 9202925 KB data.
```
Note 7.4 GB of data in class 00, total size 9.2 GB. After running the
migration we get this instead:
```
0x00: 9138161 items, 870876 KB keys + 2611392 KB data, 95 B + 285 B avg, 4788 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 17 B avg, 51 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
0x0d: 44282 items, 1461 KB keys + 61081 KB data, 33 B + 1379 B avg, 1637399 B max
Total 10469408 items, 969939 KB keys + 4477905 KB data.
```
Class 00 is now down to 2.6 GB, with just 61 MB added in class 0d.
There will be some additional reads in some cases which theoretically
hurts performance, but this will be more than compensated for by smaller
writes and better compaction.
On my own home setup which just has three devices and a handful of
folders the difference is smaller in absolute numbers of course, but
still less than half the old size:
```
0x00: 297122 items, 20894 KB keys + 306860 KB data, 70 B + 1032 B avg, 103237 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
Total 1947412 items, 151268 KB keys + 337485 KB data.
```
to:
```
0x00: 297122 items, 20894 KB keys + 37038 KB data, 70 B + 124 B avg, 520 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
0x0d: 18041 items, 595 KB keys + 71964 KB data, 33 B + 3988 B avg, 101109 B max
Total 1965447 items, 151863 KB keys + 139628 KB data.
```
* wip
* wip
* wip
* wip
2020-01-24 07:35:44 +00:00
|
|
|
}
|
|
|
|
|
2020-05-13 12:28:42 +00:00
|
|
|
// Indirect the version vector if it's large enough.
|
|
|
|
if len(fi.Version.Counters) > versionIndirectionCutoff {
|
|
|
|
fi.VersionHash = protocol.VectorHash(fi.Version)
|
|
|
|
bkey = t.keyer.GenerateVersionKey(bkey, fi.VersionHash)
|
|
|
|
if _, err := t.Get(bkey); backend.IsNotFound(err) {
|
|
|
|
// Marshal the version vector and save it
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
versionBs := mustMarshal(fi.Version.ToWire())
|
2020-05-13 12:28:42 +00:00
|
|
|
if err := t.Put(bkey, versionBs); err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
} else if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
fi.Version = protocol.Vector{}
|
|
|
|
} else {
|
|
|
|
fi.VersionHash = nil
|
|
|
|
}
|
|
|
|
|
2021-06-07 08:52:06 +00:00
|
|
|
t.indirectionTracker.recordIndirectionHashesForFile(&fi)
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
fiBs := mustMarshal(fi.ToWire(true))
|
2020-02-27 10:19:21 +00:00
|
|
|
return t.Put(fkey, fiBs)
|
lib/db: Deduplicate block lists in database (fixes #5898) (#6283)
* lib/db: Deduplicate block lists in database (fixes #5898)
This moves the block list in the database out from being just a field on
the FileInfo to being an object of its own. When putting a FileInfo we
marshal the block list separately and store it keyed by the sha256 of
the marshalled block list. When getting, if we are not doing a
"truncated" get, we do an extra read and unmarshal for the block list.
Old block lists are cleared out by a periodic GC sweep. The alternative
would be to use refcounting, but:
- There is a larger risk of getting that wrong and either dropping a
block list in error or keeping them around forever.
- It's tricky with our current database, as we don't have dirty reads.
This means that if we update two FileInfos with identical block lists in
the same transaction we can't just do read/modify/write for the ref
counters as we wouldn't see our own first update. See above about
tracking this and risks about getting it wrong.
GC uses a bloom filter for keys to avoid heavy RAM usage. GC can't run
concurrently with FileInfo updates so there is a new lock around those
operation at the lowlevel.
The end result is a much more compact database, especially for setups
with many peers where files get duplicated many times.
This is per-key-class stats for a large database I'm currently working
with, under the current schema:
```
0x00: 9138161 items, 870876 KB keys + 7397482 KB data, 95 B + 809 B avg, 1637651 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x08: 1349 items, 12 KB keys + 10 KB data, 9 B + 8 B avg, 17 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 7 B avg, 30 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
Total 10426475 items, 968490 KB keys + 9202925 KB data.
```
Note 7.4 GB of data in class 00, total size 9.2 GB. After running the
migration we get this instead:
```
0x00: 9138161 items, 870876 KB keys + 2611392 KB data, 95 B + 285 B avg, 4788 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 17 B avg, 51 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
0x0d: 44282 items, 1461 KB keys + 61081 KB data, 33 B + 1379 B avg, 1637399 B max
Total 10469408 items, 969939 KB keys + 4477905 KB data.
```
Class 00 is now down to 2.6 GB, with just 61 MB added in class 0d.
There will be some additional reads in some cases which theoretically
hurts performance, but this will be more than compensated for by smaller
writes and better compaction.
On my own home setup which just has three devices and a handful of
folders the difference is smaller in absolute numbers of course, but
still less than half the old size:
```
0x00: 297122 items, 20894 KB keys + 306860 KB data, 70 B + 1032 B avg, 103237 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
Total 1947412 items, 151268 KB keys + 337485 KB data.
```
to:
```
0x00: 297122 items, 20894 KB keys + 37038 KB data, 70 B + 124 B avg, 520 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
0x0d: 18041 items, 595 KB keys + 71964 KB data, 33 B + 3988 B avg, 101109 B max
Total 1965447 items, 151863 KB keys + 139628 KB data.
```
* wip
* wip
* wip
* wip
2020-01-24 07:35:44 +00:00
|
|
|
}
|
|
|
|
|
2015-10-29 07:07:51 +00:00
|
|
|
// updateGlobal adds this device+version to the version list for the given
|
|
|
|
// file. If the device is already present in the list, the version is updated.
|
|
|
|
// If the file does not have an entry in the global list, it is created.
|
2021-09-07 22:11:16 +00:00
|
|
|
func (t readWriteTransaction) updateGlobal(gk, keyBuf, folder, device []byte, file protocol.FileInfo, meta *metadataTracker) ([]byte, error) {
|
2020-06-07 08:31:12 +00:00
|
|
|
deviceID, err := protocol.DeviceIDFromBytes(device)
|
|
|
|
if err != nil {
|
2021-09-07 22:11:16 +00:00
|
|
|
return nil, err
|
2020-06-07 08:31:12 +00:00
|
|
|
}
|
2020-05-11 13:07:06 +00:00
|
|
|
|
|
|
|
l.Debugf("update global; folder=%q device=%v file=%q version=%v invalid=%v", folder, deviceID, file.Name, file.Version, file.IsInvalid())
|
2015-10-29 07:07:51 +00:00
|
|
|
|
2020-05-01 07:30:20 +00:00
|
|
|
fl, err := t.getGlobalVersionsByKey(gk)
|
|
|
|
if err != nil && !backend.IsNotFound(err) {
|
2021-09-07 22:11:16 +00:00
|
|
|
return nil, err
|
2019-11-29 08:11:52 +00:00
|
|
|
}
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
if fl == nil {
|
|
|
|
fl = &dbproto.VersionList{}
|
|
|
|
}
|
2019-11-29 08:11:52 +00:00
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
globalFV, oldGlobalFV, removedFV, haveOldGlobal, haveRemoved, globalChanged, err := vlUpdate(fl, folder, device, file, t.readOnlyTransaction)
|
2019-11-29 08:11:52 +00:00
|
|
|
if err != nil {
|
2021-09-07 22:11:16 +00:00
|
|
|
return nil, err
|
2015-10-29 07:07:51 +00:00
|
|
|
}
|
|
|
|
|
2018-06-18 06:23:40 +00:00
|
|
|
name := []byte(file.Name)
|
|
|
|
|
2020-05-11 13:07:06 +00:00
|
|
|
l.Debugf(`new global for "%v" after update: %v`, file.Name, fl)
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
if err := t.Put(gk, mustMarshal(fl)); err != nil {
|
2021-09-07 22:11:16 +00:00
|
|
|
return nil, err
|
2020-05-11 13:07:06 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
// Only load those from db if actually needed
|
|
|
|
|
|
|
|
var gotGlobal, gotOldGlobal bool
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
var global, oldGlobal protocol.FileInfo
|
2018-06-02 13:08:32 +00:00
|
|
|
|
2020-05-11 13:07:06 +00:00
|
|
|
// Check the need of the device that was updated
|
|
|
|
// Must happen before updating global meta: If this is the first
|
|
|
|
// item from this device, it will be initialized with the global state.
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
needBefore := haveOldGlobal && Need(oldGlobalFV, haveRemoved, protocol.VectorFromWire(removedFV.GetVersion()))
|
2020-09-04 11:59:04 +00:00
|
|
|
needNow := Need(globalFV, true, file.Version)
|
2020-05-11 13:07:06 +00:00
|
|
|
if needBefore {
|
2020-05-30 07:50:23 +00:00
|
|
|
if keyBuf, oldGlobal, err = t.getGlobalFromFileVersion(keyBuf, folder, name, true, oldGlobalFV); err != nil {
|
2021-09-07 22:11:16 +00:00
|
|
|
return nil, err
|
2020-05-11 13:07:06 +00:00
|
|
|
}
|
2020-05-13 10:56:49 +00:00
|
|
|
gotOldGlobal = true
|
2020-05-11 13:07:06 +00:00
|
|
|
meta.removeNeeded(deviceID, oldGlobal)
|
|
|
|
if !needNow && bytes.Equal(device, protocol.LocalDeviceID[:]) {
|
|
|
|
if keyBuf, err = t.updateLocalNeed(keyBuf, folder, name, false); err != nil {
|
2021-09-07 22:11:16 +00:00
|
|
|
return nil, err
|
2020-05-11 13:07:06 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if needNow {
|
2021-06-17 08:15:11 +00:00
|
|
|
keyBuf, global, err = t.getGlobalFromFileVersion(keyBuf, folder, name, true, globalFV)
|
|
|
|
if err != nil {
|
2021-09-07 22:11:16 +00:00
|
|
|
return nil, err
|
2020-05-11 13:07:06 +00:00
|
|
|
}
|
2020-05-13 10:56:49 +00:00
|
|
|
gotGlobal = true
|
2020-05-11 13:07:06 +00:00
|
|
|
meta.addNeeded(deviceID, global)
|
|
|
|
if !needBefore && bytes.Equal(device, protocol.LocalDeviceID[:]) {
|
|
|
|
if keyBuf, err = t.updateLocalNeed(keyBuf, folder, name, true); err != nil {
|
2021-09-07 22:11:16 +00:00
|
|
|
return nil, err
|
2020-05-11 13:07:06 +00:00
|
|
|
}
|
|
|
|
}
|
2019-11-29 08:11:52 +00:00
|
|
|
}
|
2015-10-29 07:07:51 +00:00
|
|
|
|
2020-07-30 11:49:14 +00:00
|
|
|
// Update global size counter if necessary
|
|
|
|
|
|
|
|
if !globalChanged {
|
|
|
|
// Neither the global state nor the needs of any devices, except
|
|
|
|
// the one updated, changed.
|
2021-09-07 22:11:16 +00:00
|
|
|
return keyBuf, nil
|
2020-07-30 11:49:14 +00:00
|
|
|
}
|
2020-05-13 10:56:49 +00:00
|
|
|
|
|
|
|
// Remove the old global from the global size counter
|
|
|
|
if haveOldGlobal {
|
|
|
|
if !gotOldGlobal {
|
2020-05-30 07:50:23 +00:00
|
|
|
if keyBuf, oldGlobal, err = t.getGlobalFromFileVersion(keyBuf, folder, name, true, oldGlobalFV); err != nil {
|
2021-09-07 22:11:16 +00:00
|
|
|
return nil, err
|
2020-05-11 13:07:06 +00:00
|
|
|
}
|
|
|
|
}
|
2020-05-30 07:50:23 +00:00
|
|
|
// Remove the old global from the global size counter
|
2020-05-13 10:56:49 +00:00
|
|
|
meta.removeFile(protocol.GlobalDeviceID, oldGlobal)
|
2020-05-11 13:07:06 +00:00
|
|
|
}
|
|
|
|
|
2020-05-13 10:56:49 +00:00
|
|
|
// Add the new global to the global size counter
|
|
|
|
if !gotGlobal {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
if protocol.VectorFromWire(globalFV.Version).Equal(file.Version) {
|
2021-06-17 08:15:11 +00:00
|
|
|
// The inserted file is the global file
|
|
|
|
global = file
|
|
|
|
} else {
|
|
|
|
keyBuf, global, err = t.getGlobalFromFileVersion(keyBuf, folder, name, true, globalFV)
|
2021-09-07 22:11:16 +00:00
|
|
|
if err != nil {
|
|
|
|
return nil, err
|
|
|
|
}
|
2020-05-13 10:56:49 +00:00
|
|
|
}
|
2018-07-10 15:32:34 +00:00
|
|
|
}
|
2020-05-13 10:56:49 +00:00
|
|
|
meta.addFile(protocol.GlobalDeviceID, global)
|
2018-07-10 15:32:34 +00:00
|
|
|
|
2020-05-11 13:07:06 +00:00
|
|
|
// check for local (if not already done before)
|
|
|
|
if !bytes.Equal(device, protocol.LocalDeviceID[:]) {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
localFV, haveLocal := vlGet(fl, protocol.LocalDeviceID[:])
|
|
|
|
localVersion := protocol.VectorFromWire(localFV.Version)
|
|
|
|
needBefore := haveOldGlobal && Need(oldGlobalFV, haveLocal, localVersion)
|
|
|
|
needNow := Need(globalFV, haveLocal, localVersion)
|
2020-05-11 13:07:06 +00:00
|
|
|
if needBefore {
|
|
|
|
meta.removeNeeded(protocol.LocalDeviceID, oldGlobal)
|
|
|
|
if !needNow {
|
|
|
|
if keyBuf, err = t.updateLocalNeed(keyBuf, folder, name, false); err != nil {
|
2021-09-07 22:11:16 +00:00
|
|
|
return nil, err
|
2020-05-11 13:07:06 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2020-09-04 12:13:38 +00:00
|
|
|
if needNow {
|
2020-05-11 13:07:06 +00:00
|
|
|
meta.addNeeded(protocol.LocalDeviceID, global)
|
|
|
|
if !needBefore {
|
|
|
|
if keyBuf, err = t.updateLocalNeed(keyBuf, folder, name, true); err != nil {
|
2021-09-07 22:11:16 +00:00
|
|
|
return nil, err
|
2020-05-11 13:07:06 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
for _, dev := range meta.devices() {
|
|
|
|
if bytes.Equal(dev[:], device) {
|
|
|
|
// Already handled above
|
|
|
|
continue
|
|
|
|
}
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
fv, have := vlGet(fl, dev[:])
|
|
|
|
fvVersion := protocol.VectorFromWire(fv.Version)
|
|
|
|
if haveOldGlobal && Need(oldGlobalFV, have, fvVersion) {
|
2020-05-11 13:07:06 +00:00
|
|
|
meta.removeNeeded(dev, oldGlobal)
|
|
|
|
}
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
if Need(globalFV, have, fvVersion) {
|
2020-05-11 13:07:06 +00:00
|
|
|
meta.addNeeded(dev, global)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2021-09-07 22:11:16 +00:00
|
|
|
return keyBuf, nil
|
2020-05-11 13:07:06 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
func (t readWriteTransaction) updateLocalNeed(keyBuf, folder, name []byte, add bool) ([]byte, error) {
|
2019-11-29 08:11:52 +00:00
|
|
|
var err error
|
|
|
|
keyBuf, err = t.keyer.GenerateNeedFileKey(keyBuf, folder, name)
|
|
|
|
if err != nil {
|
|
|
|
return nil, err
|
|
|
|
}
|
2020-05-11 13:07:06 +00:00
|
|
|
if add {
|
|
|
|
l.Debugf("local need insert; folder=%q, name=%q", folder, name)
|
|
|
|
err = t.Put(keyBuf, nil)
|
|
|
|
} else {
|
2018-10-30 04:40:51 +00:00
|
|
|
l.Debugf("local need delete; folder=%q, name=%q", folder, name)
|
2020-05-11 13:07:06 +00:00
|
|
|
err = t.Delete(keyBuf)
|
2018-10-30 04:40:51 +00:00
|
|
|
}
|
2020-05-11 13:07:06 +00:00
|
|
|
return keyBuf, err
|
2018-10-30 04:40:51 +00:00
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
func Need(global *dbproto.FileVersion, haveLocal bool, localVersion protocol.Vector) bool {
|
2021-04-29 20:01:46 +00:00
|
|
|
// We never need an invalid file or a file without a valid version (just
|
|
|
|
// another way of expressing "invalid", really, until we fix that
|
|
|
|
// part...).
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
globalVersion := protocol.VectorFromWire(global.Version)
|
|
|
|
if fvIsInvalid(global) || globalVersion.IsEmpty() {
|
2018-06-02 13:08:32 +00:00
|
|
|
return false
|
|
|
|
}
|
|
|
|
// We don't need a deleted file if we don't have it.
|
2020-05-11 13:07:06 +00:00
|
|
|
if global.Deleted && !haveLocal {
|
2018-06-02 13:08:32 +00:00
|
|
|
return false
|
|
|
|
}
|
|
|
|
// We don't need the global file if we already have the same version.
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
if haveLocal && localVersion.GreaterEqual(globalVersion) {
|
2018-06-02 13:08:32 +00:00
|
|
|
return false
|
|
|
|
}
|
|
|
|
return true
|
|
|
|
}
|
|
|
|
|
2015-10-29 07:07:51 +00:00
|
|
|
// removeFromGlobal removes the device from the global version list for the
|
|
|
|
// given file. If the version list is empty after this, the file entry is
|
|
|
|
// removed entirely.
|
2020-05-30 07:50:23 +00:00
|
|
|
func (t readWriteTransaction) removeFromGlobal(gk, keyBuf, folder, device, file []byte, meta *metadataTracker) ([]byte, error) {
|
2020-06-07 08:31:12 +00:00
|
|
|
deviceID, err := protocol.DeviceIDFromBytes(device)
|
|
|
|
if err != nil {
|
|
|
|
return nil, err
|
|
|
|
}
|
2020-05-11 13:07:06 +00:00
|
|
|
|
|
|
|
l.Debugf("remove from global; folder=%q device=%v file=%q", folder, deviceID, file)
|
2015-10-29 07:07:51 +00:00
|
|
|
|
2020-05-01 07:30:20 +00:00
|
|
|
fl, err := t.getGlobalVersionsByKey(gk)
|
2019-11-29 08:11:52 +00:00
|
|
|
if backend.IsNotFound(err) {
|
2015-10-29 07:07:51 +00:00
|
|
|
// We might be called to "remove" a global version that doesn't exist
|
|
|
|
// if the first update for the file is already marked invalid.
|
2019-11-29 08:11:52 +00:00
|
|
|
return keyBuf, nil
|
|
|
|
} else if err != nil {
|
|
|
|
return nil, err
|
2015-10-29 07:07:51 +00:00
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
oldGlobalFV, haveOldGlobal := vlGetGlobal(fl)
|
|
|
|
oldGlobalFV = fvCopy(oldGlobalFV)
|
2020-05-30 07:50:23 +00:00
|
|
|
|
|
|
|
if !haveOldGlobal {
|
2020-05-13 10:56:49 +00:00
|
|
|
// Shouldn't ever happen, but doesn't hurt to handle.
|
2020-12-21 11:59:22 +00:00
|
|
|
t.evLogger.Log(events.Failure, "encountered empty global while removing item")
|
2020-05-13 10:56:49 +00:00
|
|
|
return keyBuf, t.Delete(gk)
|
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
removedFV, haveRemoved, globalChanged := vlPop(fl, device)
|
2020-05-30 07:50:23 +00:00
|
|
|
if !haveRemoved {
|
2018-10-30 04:40:51 +00:00
|
|
|
// There is no version for the given device
|
2019-11-29 08:11:52 +00:00
|
|
|
return keyBuf, nil
|
2018-10-30 04:40:51 +00:00
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
var global protocol.FileInfo
|
2021-02-08 07:38:41 +00:00
|
|
|
var gotGlobal bool
|
2020-05-13 10:56:49 +00:00
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
globalFV, haveGlobal := vlGetGlobal(fl)
|
2020-05-13 10:56:49 +00:00
|
|
|
// Add potential needs of the removed device
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
if haveGlobal && !fvIsInvalid(globalFV) && Need(globalFV, false, protocol.Vector{}) && !Need(oldGlobalFV, haveRemoved, protocol.VectorFromWire(removedFV.Version)) {
|
2021-09-07 22:11:16 +00:00
|
|
|
keyBuf, global, err = t.getGlobalFromVersionList(keyBuf, folder, file, true, fl)
|
2020-05-13 10:56:49 +00:00
|
|
|
if err != nil {
|
|
|
|
return nil, err
|
|
|
|
}
|
|
|
|
gotGlobal = true
|
|
|
|
meta.addNeeded(deviceID, global)
|
|
|
|
if bytes.Equal(protocol.LocalDeviceID[:], device) {
|
|
|
|
if keyBuf, err = t.updateLocalNeed(keyBuf, folder, file, true); err != nil {
|
|
|
|
return nil, err
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Global hasn't changed, abort early
|
2020-05-30 07:50:23 +00:00
|
|
|
if !globalChanged {
|
2020-05-11 13:07:06 +00:00
|
|
|
l.Debugf("new global after remove: %v", fl)
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
if err := t.Put(gk, mustMarshal(fl)); err != nil {
|
2020-03-19 13:32:22 +00:00
|
|
|
return nil, err
|
2015-10-29 07:07:51 +00:00
|
|
|
}
|
2020-05-13 10:56:49 +00:00
|
|
|
return keyBuf, nil
|
2015-10-29 07:07:51 +00:00
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
var oldGlobal protocol.FileInfo
|
2021-02-08 07:38:41 +00:00
|
|
|
keyBuf, oldGlobal, err = t.getGlobalFromFileVersion(keyBuf, folder, file, true, oldGlobalFV)
|
2020-05-11 13:07:06 +00:00
|
|
|
if err != nil {
|
|
|
|
return nil, err
|
|
|
|
}
|
2021-02-08 07:38:41 +00:00
|
|
|
meta.removeFile(protocol.GlobalDeviceID, oldGlobal)
|
2020-05-11 13:07:06 +00:00
|
|
|
|
2020-05-13 10:56:49 +00:00
|
|
|
// Remove potential device needs
|
2021-02-08 07:38:41 +00:00
|
|
|
shouldRemoveNeed := func(dev protocol.DeviceID) bool {
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
fv, have := vlGet(fl, dev[:])
|
|
|
|
fvVersion := protocol.VectorFromWire(fv.Version)
|
|
|
|
if !Need(oldGlobalFV, have, fvVersion) {
|
2021-02-08 07:38:41 +00:00
|
|
|
return false // Didn't need it before
|
|
|
|
}
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
return !haveGlobal || !Need(globalFV, have, fvVersion)
|
2021-02-08 07:38:41 +00:00
|
|
|
}
|
|
|
|
if shouldRemoveNeed(protocol.LocalDeviceID) {
|
|
|
|
meta.removeNeeded(protocol.LocalDeviceID, oldGlobal)
|
2020-05-11 13:07:06 +00:00
|
|
|
if keyBuf, err = t.updateLocalNeed(keyBuf, folder, file, false); err != nil {
|
2019-11-29 08:11:52 +00:00
|
|
|
return nil, err
|
|
|
|
}
|
2020-05-11 13:07:06 +00:00
|
|
|
}
|
|
|
|
for _, dev := range meta.devices() {
|
2020-05-13 10:56:49 +00:00
|
|
|
if bytes.Equal(dev[:], device) { // Was the previous global
|
2020-05-11 13:07:06 +00:00
|
|
|
continue
|
|
|
|
}
|
2021-02-08 07:38:41 +00:00
|
|
|
if shouldRemoveNeed(dev) {
|
|
|
|
meta.removeNeeded(dev, oldGlobal)
|
2019-11-29 08:11:52 +00:00
|
|
|
}
|
2020-05-11 13:07:06 +00:00
|
|
|
}
|
|
|
|
|
2020-05-13 10:56:49 +00:00
|
|
|
// Nothing left, i.e. nothing to add to the global counter below.
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
if len(fl.Versions) == 0 {
|
2019-11-29 08:11:52 +00:00
|
|
|
if err := t.Delete(gk); err != nil {
|
|
|
|
return nil, err
|
|
|
|
}
|
|
|
|
return keyBuf, nil
|
2017-11-11 19:18:17 +00:00
|
|
|
}
|
2018-10-30 04:40:51 +00:00
|
|
|
|
2020-05-13 10:56:49 +00:00
|
|
|
// Add to global
|
|
|
|
if !gotGlobal {
|
2021-09-07 22:11:16 +00:00
|
|
|
keyBuf, global, err = t.getGlobalFromVersionList(keyBuf, folder, file, true, fl)
|
2020-05-13 10:56:49 +00:00
|
|
|
if err != nil {
|
|
|
|
return nil, err
|
2019-11-29 08:11:52 +00:00
|
|
|
}
|
2015-10-29 07:07:51 +00:00
|
|
|
}
|
2020-05-13 10:56:49 +00:00
|
|
|
meta.addFile(protocol.GlobalDeviceID, global)
|
2018-10-30 04:40:51 +00:00
|
|
|
|
2020-05-30 07:50:23 +00:00
|
|
|
l.Debugf(`new global for "%s" after remove: %v`, file, fl)
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
if err := t.Put(gk, mustMarshal(fl)); err != nil {
|
2019-11-29 08:11:52 +00:00
|
|
|
return nil, err
|
|
|
|
}
|
2019-01-20 07:47:20 +00:00
|
|
|
|
2019-11-29 08:11:52 +00:00
|
|
|
return keyBuf, nil
|
2015-10-29 07:07:51 +00:00
|
|
|
}
|
|
|
|
|
2019-11-29 08:11:52 +00:00
|
|
|
func (t readWriteTransaction) deleteKeyPrefix(prefix []byte) error {
|
2020-12-21 10:10:59 +00:00
|
|
|
return t.deleteKeyPrefixMatching(prefix, func([]byte) bool { return true })
|
|
|
|
}
|
|
|
|
|
|
|
|
func (t readWriteTransaction) deleteKeyPrefixMatching(prefix []byte, match func(key []byte) bool) error {
|
2019-11-29 08:11:52 +00:00
|
|
|
dbi, err := t.NewPrefixIterator(prefix)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
defer dbi.Release()
|
2018-06-02 13:08:32 +00:00
|
|
|
for dbi.Next() {
|
2020-12-21 10:10:59 +00:00
|
|
|
if !match(dbi.Key()) {
|
|
|
|
continue
|
|
|
|
}
|
2019-11-29 08:11:52 +00:00
|
|
|
if err := t.Delete(dbi.Key()); err != nil {
|
|
|
|
return err
|
|
|
|
}
|
2018-06-02 13:08:32 +00:00
|
|
|
}
|
2019-11-29 08:11:52 +00:00
|
|
|
return dbi.Error()
|
2015-10-29 07:07:51 +00:00
|
|
|
}
|
2016-07-04 10:40:29 +00:00
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
func (t *readWriteTransaction) withAllFolderTruncated(folder []byte, fn func(device []byte, f protocol.FileInfo) bool) error {
|
2020-01-21 17:23:08 +00:00
|
|
|
key, err := t.keyer.GenerateDeviceFileKey(nil, folder, nil, nil)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
dbi, err := t.NewPrefixIterator(key.WithoutNameAndDevice())
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
defer dbi.Release()
|
|
|
|
|
|
|
|
var gk, keyBuf []byte
|
|
|
|
for dbi.Next() {
|
|
|
|
device, ok := t.keyer.DeviceFromDeviceFileKey(dbi.Key())
|
|
|
|
if !ok {
|
|
|
|
// Not having the device in the index is bad. Clear it.
|
|
|
|
if err := t.Delete(dbi.Key()); err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
continue
|
|
|
|
}
|
2020-02-27 10:19:21 +00:00
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
f, err := t.unmarshalTrunc(dbi.Value(), true)
|
2020-01-21 17:23:08 +00:00
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
|
|
|
|
switch f.Name {
|
|
|
|
case "", ".", "..", "/": // A few obviously invalid filenames
|
|
|
|
l.Infof("Dropping invalid filename %q from database", f.Name)
|
|
|
|
name := []byte(f.Name)
|
|
|
|
gk, err = t.keyer.GenerateGlobalVersionKey(gk, folder, name)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
keyBuf, err = t.removeFromGlobal(gk, keyBuf, folder, device, name, nil)
|
|
|
|
if err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
if err := t.Delete(dbi.Key()); err != nil {
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
|
|
|
|
if !fn(device, f) {
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
}
|
2020-02-14 07:11:24 +00:00
|
|
|
return dbi.Error()
|
2020-01-21 17:23:08 +00:00
|
|
|
}
|
|
|
|
|
refactor: use modern Protobuf encoder (#9817)
At a high level, this is what I've done and why:
- I'm moving the protobuf generation for the `protocol`, `discovery` and
`db` packages to the modern alternatives, and using `buf` to generate
because it's nice and simple.
- After trying various approaches on how to integrate the new types with
the existing code, I opted for splitting off our own data model types
from the on-the-wire generated types. This means we can have a
`FileInfo` type with nicer ergonomics and lots of methods, while the
protobuf generated type stays clean and close to the wire protocol. It
does mean copying between the two when required, which certainly adds a
small amount of inefficiency. If we want to walk this back in the future
and use the raw generated type throughout, that's possible, this however
makes the refactor smaller (!) as it doesn't change everything about the
type for everyone at the same time.
- I have simply removed in cold blood a significant number of old
database migrations. These depended on previous generations of generated
messages of various kinds and were annoying to support in the new
fashion. The oldest supported database version now is the one from
Syncthing 1.9.0 from Sep 7, 2020.
- I changed config structs to be regular manually defined structs.
For the sake of discussion, some things I tried that turned out not to
work...
### Embedding / wrapping
Embedding the protobuf generated structs in our existing types as a data
container and keeping our methods and stuff:
```
package protocol
type FileInfo struct {
*generated.FileInfo
}
```
This generates a lot of problems because the internal shape of the
generated struct is quite different (different names, different types,
more pointers), because initializing it doesn't work like you'd expect
(i.e., you end up with an embedded nil pointer and a panic), and because
the types of child types don't get wrapped. That is, even if we also
have a similar wrapper around a `Vector`, that's not the type you get
when accessing `someFileInfo.Version`, you get the `*generated.Vector`
that doesn't have methods, etc.
### Aliasing
```
package protocol
type FileInfo = generated.FileInfo
```
Doesn't help because you can't attach methods to it, plus all the above.
### Generating the types into the target package like we do now and
attaching methods
This fails because of the different shape of the generated type (as in
the embedding case above) plus the generated struct already has a bunch
of methods that we can't necessarily override properly (like `String()`
and a bunch of getters).
### Methods to functions
I considered just moving all the methods we attach to functions in a
specific package, so that for example
```
package protocol
func (f FileInfo) Equal(other FileInfo) bool
```
would become
```
package fileinfos
func Equal(a, b *generated.FileInfo) bool
```
and this would mostly work, but becomes quite verbose and cumbersome,
and somewhat limits discoverability (you can't see what methods are
available on the type in auto completions, etc). In the end I did this
in some cases, like in the database layer where a lot of things like
`func (fv *FileVersion) IsEmpty() bool` becomes `func fvIsEmpty(fv
*generated.FileVersion)` because they were anyway just internal methods.
Fixes #8247
2024-12-01 15:50:17 +00:00
|
|
|
func mustMarshal(f proto.Message) []byte {
|
|
|
|
bs, err := proto.Marshal(f)
|
2016-07-04 10:40:29 +00:00
|
|
|
if err != nil {
|
|
|
|
panic(err)
|
|
|
|
}
|
|
|
|
return bs
|
|
|
|
}
|