2
2
mirror of https://github.com/octoleo/restic.git synced 2024-12-26 04:17:29 +00:00
restic/internal/restic/find.go
Michael Eischer c4fc5c97f9 prune: Use a single CountedBlobSet to track blobs
The set covers necessary, existing and duplicate blobs. This removes the
duplicate sets used to track whether all necessary blobs also exist.
This reduces the memory usage of prune by about 20-30%.
2022-10-22 18:45:12 +02:00

61 lines
1.4 KiB
Go

package restic
import (
"context"
"sync"
"github.com/restic/restic/internal/ui/progress"
"golang.org/x/sync/errgroup"
)
// Loader loads a blob from a repository.
type Loader interface {
LoadBlob(context.Context, BlobType, ID, []byte) ([]byte, error)
LookupBlobSize(id ID, tpe BlobType) (uint, bool)
Connections() uint
}
type findBlobSet interface {
Has(bh BlobHandle) bool
Insert(bh BlobHandle)
}
// FindUsedBlobs traverses the tree ID and adds all seen blobs (trees and data
// blobs) to the set blobs. Already seen tree blobs will not be visited again.
func FindUsedBlobs(ctx context.Context, repo Loader, treeIDs IDs, blobs findBlobSet, p *progress.Counter) error {
var lock sync.Mutex
wg, ctx := errgroup.WithContext(ctx)
treeStream := StreamTrees(ctx, wg, repo, treeIDs, func(treeID ID) bool {
// locking is necessary the goroutine below concurrently adds data blobs
lock.Lock()
h := BlobHandle{ID: treeID, Type: TreeBlob}
blobReferenced := blobs.Has(h)
// noop if already referenced
blobs.Insert(h)
lock.Unlock()
return blobReferenced
}, p)
wg.Go(func() error {
for tree := range treeStream {
if tree.Error != nil {
return tree.Error
}
lock.Lock()
for _, node := range tree.Nodes {
switch node.Type {
case "file":
for _, blob := range node.Content {
blobs.Insert(BlobHandle{ID: blob, Type: DataBlob})
}
}
}
lock.Unlock()
}
return nil
})
return wg.Wait()
}