syncthing/lib/db/meta.go

472 lines
12 KiB
Go
Raw Permalink Normal View History

// Copyright (C) 2017 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,
// You can obtain one at https://mozilla.org/MPL/2.0/.
package db
import (
"errors"
"fmt"
"math/bits"
"time"
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/dbproto"
2020-07-19 06:55:27 +00:00
"github.com/syncthing/syncthing/lib/db/backend"
"github.com/syncthing/syncthing/lib/events"
"github.com/syncthing/syncthing/lib/protocol"
"github.com/syncthing/syncthing/lib/sync"
)
var errMetaInconsistent = errors.New("inconsistent counts detected")
type countsMap struct {
counts CountsSet
indexes map[metaKey]int // device ID + local flags -> index in counts
}
// metadataTracker keeps metadata on a per device, per local flag basis.
type metadataTracker struct {
2020-07-19 06:55:27 +00:00
keyer keyer
countsMap
mut sync.RWMutex
dirty bool
evLogger events.Logger
}
type metaKey struct {
dev protocol.DeviceID
flag uint32
}
const needFlag uint32 = 1 << 31 // Last bit, as early ones are local flags
func newMetadataTracker(keyer keyer, evLogger events.Logger) *metadataTracker {
return &metadataTracker{
2020-07-19 06:55:27 +00:00
keyer: keyer,
mut: sync.NewRWMutex(),
countsMap: countsMap{
indexes: make(map[metaKey]int),
},
evLogger: evLogger,
}
}
// Unmarshal loads a metadataTracker from the corresponding protobuf
// representation
func (m *metadataTracker) Unmarshal(bs []byte) error {
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 dbc dbproto.CountsSet
if err := proto.Unmarshal(bs, &dbc); 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
m.counts.Created = dbc.Created
m.counts.Counts = make([]Counts, len(dbc.Counts))
for i, c := range dbc.Counts {
m.counts.Counts[i] = countsFromWire(c)
}
// Initialize the index map
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
m.indexes = make(map[metaKey]int)
for i, c := range m.counts.Counts {
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
m.indexes[metaKey{c.DeviceID, c.LocalFlags}] = i
}
return nil
}
// Marshal returns the protobuf representation of the metadataTracker
func (m *metadataTracker) Marshal() ([]byte, error) {
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
dbc := &dbproto.CountsSet{
Counts: make([]*dbproto.Counts, len(m.counts.Counts)),
Created: m.Created().UnixNano(),
}
for i, c := range m.counts.Counts {
dbc.Counts[i] = c.toWire()
}
return proto.Marshal(dbc)
}
2020-07-19 06:55:27 +00:00
func (m *metadataTracker) CommitHook(folder []byte) backend.CommitHook {
return func(t backend.WriteTransaction) error {
return m.toDB(t, folder)
}
}
// toDB saves the marshalled metadataTracker to the given db, under the key
// corresponding to the given folder
2020-07-19 06:55:27 +00:00
func (m *metadataTracker) toDB(t backend.WriteTransaction, folder []byte) error {
key, err := m.keyer.GenerateFolderMetaKey(nil, folder)
if err != nil {
return err
}
m.mut.RLock()
defer m.mut.RUnlock()
if !m.dirty {
return nil
}
bs, err := m.Marshal()
if err != nil {
return err
}
err = t.Put(key, bs)
if err == nil {
m.dirty = false
}
return err
}
// fromDB initializes the metadataTracker from the marshalled data found in
// the database under the key corresponding to the given folder
func (m *metadataTracker) fromDB(db *Lowlevel, folder []byte) error {
key, err := db.keyer.GenerateFolderMetaKey(nil, folder)
if err != nil {
return err
}
bs, err := db.Get(key)
if err != nil {
return err
}
if err = m.Unmarshal(bs); err != nil {
return err
}
if m.counts.Created == 0 {
return errMetaInconsistent
}
return nil
}
// countsPtr returns a pointer to the corresponding Counts struct, if
// necessary allocating one in the process
func (m *metadataTracker) countsPtr(dev protocol.DeviceID, flag uint32) *Counts {
// must be called with the mutex held
if bits.OnesCount32(flag) > 1 {
panic("incorrect usage: set at most one bit in flag")
}
key := metaKey{dev, flag}
idx, ok := m.indexes[key]
if !ok {
idx = len(m.counts.Counts)
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
m.counts.Counts = append(m.counts.Counts, Counts{DeviceID: dev, LocalFlags: flag})
m.indexes[key] = idx
// Need bucket must be initialized when a device first occurs in
// the metadatatracker, even if there's no change to the need
// bucket itself.
nkey := metaKey{dev, needFlag}
if _, ok := m.indexes[nkey]; !ok {
// Initially a new device needs everything, except deletes
nidx := len(m.counts.Counts)
m.counts.Counts = append(m.counts.Counts, m.allNeededCounts(dev))
m.indexes[nkey] = nidx
}
}
return &m.counts.Counts[idx]
}
// allNeeded makes sure there is a counts in case the device needs everything.
func (m *countsMap) allNeededCounts(dev protocol.DeviceID) Counts {
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 counts Counts
if idx, ok := m.indexes[metaKey{protocol.GlobalDeviceID, 0}]; ok {
counts = m.counts.Counts[idx]
counts.Deleted = 0 // Don't need deletes if having nothing
}
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
counts.DeviceID = dev
counts.LocalFlags = needFlag
return counts
}
// addFile adds a file to the counts, adjusting the sequence number as
// appropriate
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 (m *metadataTracker) addFile(dev protocol.DeviceID, f protocol.FileInfo) {
m.mut.Lock()
defer m.mut.Unlock()
m.updateSeqLocked(dev, f)
m.updateFileLocked(dev, f, m.addFileLocked)
}
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 (m *metadataTracker) updateFileLocked(dev protocol.DeviceID, f protocol.FileInfo, fn func(protocol.DeviceID, uint32, protocol.FileInfo)) {
m.dirty = true
if f.IsInvalid() && (f.FileLocalFlags() == 0 || dev == protocol.GlobalDeviceID) {
// This is a remote invalid file or concern the global state.
// In either case invalid files are not accounted.
return
}
if flags := f.FileLocalFlags(); flags == 0 {
// Account regular files in the zero-flags bucket.
fn(dev, 0, f)
} else {
// Account in flag specific buckets.
eachFlagBit(flags, func(flag uint32) {
fn(dev, flag, f)
})
}
}
// emptyNeeded ensures that there is a need count for the given device and that it is empty.
func (m *metadataTracker) emptyNeeded(dev protocol.DeviceID) {
m.mut.Lock()
defer m.mut.Unlock()
m.dirty = true
empty := Counts{
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
DeviceID: dev,
LocalFlags: needFlag,
}
key := metaKey{dev, needFlag}
if idx, ok := m.indexes[key]; ok {
m.counts.Counts[idx] = empty
return
}
m.indexes[key] = len(m.counts.Counts)
m.counts.Counts = append(m.counts.Counts, empty)
}
// addNeeded adds a file to the needed counts
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 (m *metadataTracker) addNeeded(dev protocol.DeviceID, f protocol.FileInfo) {
m.mut.Lock()
defer m.mut.Unlock()
m.dirty = true
m.addFileLocked(dev, needFlag, f)
}
func (m *metadataTracker) Sequence(dev protocol.DeviceID) int64 {
m.mut.Lock()
defer m.mut.Unlock()
return m.countsPtr(dev, 0).Sequence
}
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 (m *metadataTracker) updateSeqLocked(dev protocol.DeviceID, f protocol.FileInfo) {
if dev == protocol.GlobalDeviceID {
return
}
if cp := m.countsPtr(dev, 0); f.SequenceNo() > cp.Sequence {
cp.Sequence = f.SequenceNo()
}
}
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 (m *metadataTracker) addFileLocked(dev protocol.DeviceID, flag uint32, f protocol.FileInfo) {
cp := m.countsPtr(dev, flag)
switch {
case f.IsDeleted():
cp.Deleted++
case f.IsDirectory() && !f.IsSymlink():
cp.Directories++
case f.IsSymlink():
cp.Symlinks++
default:
cp.Files++
}
cp.Bytes += f.FileSize()
}
// removeFile removes a file from the counts
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 (m *metadataTracker) removeFile(dev protocol.DeviceID, f protocol.FileInfo) {
m.mut.Lock()
defer m.mut.Unlock()
m.updateFileLocked(dev, f, m.removeFileLocked)
}
// removeNeeded removes a file from the needed counts
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 (m *metadataTracker) removeNeeded(dev protocol.DeviceID, f protocol.FileInfo) {
m.mut.Lock()
defer m.mut.Unlock()
m.dirty = true
m.removeFileLocked(dev, needFlag, f)
}
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 (m *metadataTracker) removeFileLocked(dev protocol.DeviceID, flag uint32, f protocol.FileInfo) {
cp := m.countsPtr(dev, flag)
switch {
case f.IsDeleted():
cp.Deleted--
case f.IsDirectory() && !f.IsSymlink():
cp.Directories--
case f.IsSymlink():
cp.Symlinks--
default:
cp.Files--
}
cp.Bytes -= f.FileSize()
// If we've run into an impossible situation, correct it for now and set
// the created timestamp to zero. Next time we start up the metadata
// will be seen as infinitely old and recalculated from scratch.
if cp.Deleted < 0 {
m.evLogger.Log(events.Failure, fmt.Sprintf("meta deleted count for flag 0x%x dropped below zero", flag))
cp.Deleted = 0
m.counts.Created = 0
}
if cp.Files < 0 {
m.evLogger.Log(events.Failure, fmt.Sprintf("meta files count for flag 0x%x dropped below zero", flag))
cp.Files = 0
m.counts.Created = 0
}
if cp.Directories < 0 {
m.evLogger.Log(events.Failure, fmt.Sprintf("meta directories count for flag 0x%x dropped below zero", flag))
cp.Directories = 0
m.counts.Created = 0
}
if cp.Symlinks < 0 {
m.evLogger.Log(events.Failure, fmt.Sprintf("meta deleted count for flag 0x%x dropped below zero", flag))
cp.Symlinks = 0
m.counts.Created = 0
}
}
// resetAll resets all metadata for the given device
func (m *metadataTracker) resetAll(dev protocol.DeviceID) {
m.mut.Lock()
m.dirty = true
for i, c := range m.counts.Counts {
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 c.DeviceID == dev {
if c.LocalFlags != needFlag {
m.counts.Counts[i] = Counts{
DeviceID: c.DeviceID,
LocalFlags: c.LocalFlags,
}
} else {
m.counts.Counts[i] = m.allNeededCounts(dev)
}
}
}
m.mut.Unlock()
}
// resetCounts resets the file, dir, etc. counters, while retaining the
// sequence number
func (m *metadataTracker) resetCounts(dev protocol.DeviceID) {
m.mut.Lock()
m.dirty = true
for i, c := range m.counts.Counts {
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 c.DeviceID == dev {
m.counts.Counts[i] = Counts{
DeviceID: c.DeviceID,
Sequence: c.Sequence,
LocalFlags: c.LocalFlags,
}
}
}
m.mut.Unlock()
}
func (m *countsMap) Counts(dev protocol.DeviceID, flag uint32) Counts {
if bits.OnesCount32(flag) > 1 {
panic("incorrect usage: set at most one bit in flag")
}
idx, ok := m.indexes[metaKey{dev, flag}]
if !ok {
if flag == needFlag {
// If there's nothing about a device in the index yet,
// it needs everything.
return m.allNeededCounts(dev)
}
return Counts{}
}
return m.counts.Counts[idx]
}
// Snapshot returns a copy of the metadata for reading.
func (m *metadataTracker) Snapshot() *countsMap {
m.mut.RLock()
defer m.mut.RUnlock()
c := &countsMap{
counts: CountsSet{
Counts: make([]Counts, len(m.counts.Counts)),
Created: m.counts.Created,
},
indexes: make(map[metaKey]int, len(m.indexes)),
}
for k, v := range m.indexes {
c.indexes[k] = v
}
2023-06-19 06:50:53 +00:00
copy(c.counts.Counts, m.counts.Counts)
return c
}
// nextLocalSeq allocates a new local sequence number
func (m *metadataTracker) nextLocalSeq() int64 {
m.mut.Lock()
defer m.mut.Unlock()
c := m.countsPtr(protocol.LocalDeviceID, 0)
c.Sequence++
return c.Sequence
}
// devices returns the list of devices tracked, excluding the local device
// (which we don't know the ID of)
func (m *metadataTracker) devices() []protocol.DeviceID {
m.mut.RLock()
defer m.mut.RUnlock()
return m.countsMap.devices()
}
func (m *countsMap) devices() []protocol.DeviceID {
devs := make([]protocol.DeviceID, 0, len(m.counts.Counts))
for _, dev := range m.counts.Counts {
if dev.Sequence > 0 {
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 dev.DeviceID == protocol.GlobalDeviceID || dev.DeviceID == protocol.LocalDeviceID {
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
devs = append(devs, dev.DeviceID)
}
}
return devs
}
func (m *metadataTracker) Created() time.Time {
m.mut.RLock()
defer m.mut.RUnlock()
return time.Unix(0, m.counts.Created)
}
func (m *metadataTracker) SetCreated() {
m.mut.Lock()
m.counts.Created = time.Now().UnixNano()
m.dirty = true
m.mut.Unlock()
}
// eachFlagBit calls the function once for every bit that is set in flags
func eachFlagBit(flags uint32, fn func(flag uint32)) {
// Test each bit from the right, as long as there are bits left in the
// flag set. Clear any bits found and stop testing as soon as there are
// no more bits set.
currentBit := uint32(1 << 0)
for flags != 0 {
if flags&currentBit != 0 {
fn(currentBit)
flags &^= currentBit
}
currentBit <<= 1
}
}