fix loadChunks by using cache instead of reading file

This way there is only one place where we read chunks and where the
read/write wrapper is used.
This also allows to remove LoadedChunk as it is not used anymore.

6 files changed, 12 insertions, 41 deletions

diff --git a/.golangci.yaml b/.golangci.yaml
index 7c47848..c618020 100644
--- a/.golangci.yaml
+++ b/.golangci.yaml
@@ -6,5 +6,5 @@ severity:
   rules:
     - severity: warning
       linters:
-      - errcheck # for now, make errcheck only a warning, remove it later
+      - errcheck # for now, make errcheck only a warning, TODO: remove it later
       - deadcode
diff --git a/TODO.md b/TODO.md
index 9df11d2..adff59b 100644
--- a/TODO.md
+++ b/TODO.md
@@ -26,7 +26,7 @@ priority 1
         - [x] compress before storing
         - [x] decompress before loading
     - [ ] store compressed chunks into tracks of `trackSize` (1024o)
-- [x] add chunk cache... what was it for again ??
+- [x] add chunk cache to uniquely store chunks in RAM
 - [x] better tests for `(*Repo).Commit`
 
 priority 2
@@ -41,6 +41,11 @@ priority 2
 - [ ] custom binary marshal and unmarshal for chunks
 - [ ] use `loadChunkContent` in `loadChunks`
 - [ ] store hashes for faster maps rebuild
+- [ ] try [Fdelta](https://github.com/amlwwalker/fdelta) and
+    [Xdelta](https://github.com/nine-lives-later/go-xdelta) instead of Bsdiff
+- [ ] maybe use an LRU cache instead of the current FIFO one.
+- [x] remove `LoadedChunk` and only use `StoredChunk` instead now that the cache
+    is implemented
 
 reunion 7/09
 ------------
diff --git a/chunk.go b/chunk.go
index 0d84040..3efaf4a 100644
--- a/chunk.go
+++ b/chunk.go
@@ -36,32 +36,6 @@ func (i *ChunkId) Path(repo string) string {
 	return filepath.Join(repo, fmt.Sprintf(versionFmt, i.Ver), chunksName, fmt.Sprintf(chunkIdFmt, i.Idx))
 }
 
-func NewLoadedChunk(id *ChunkId, value []byte) *LoadedChunk {
-	return &LoadedChunk{Id: id, value: value}
-}
-
-type LoadedChunk struct {
-	Id    *ChunkId
-	value []byte
-}
-
-func (c *LoadedChunk) GetId() *ChunkId {
-	return c.Id
-}
-
-func (c *LoadedChunk) Reader() io.ReadSeeker {
-	// log.Printf("Chunk %d: Reading from in-memory value\n", c.id)
-	return bytes.NewReader(c.value)
-}
-
-func (c *LoadedChunk) Len() int {
-	return len(c.value)
-}
-
-func (c *LoadedChunk) Bytes() []byte {
-	return c.value
-}
-
 func NewStoredChunk(repo *Repo, id *ChunkId) *StoredChunk {
 	return &StoredChunk{repo: repo, Id: id}
 }
diff --git a/delta.go b/delta.go
index 423fb07..ff42e86 100644
--- a/delta.go
+++ b/delta.go
@@ -20,7 +20,7 @@ type Patcher interface {
 	Patch(source io.Reader, target io.Writer, patch io.Reader) error
 }
 
-// TODO: maybe move this in it own file ?
+// TODO: maybe move this in its own file ?
 type Bsdiff struct{}
 
 func (*Bsdiff) Diff(source io.Reader, target io.Reader, patch io.Writer) error {
diff --git a/repo.go b/repo.go
index 50eeb25..f7aa5e9 100644
--- a/repo.go
+++ b/repo.go
@@ -299,18 +299,8 @@ func (r *Repo) loadChunks(versions []string, chunks chan<- IdentifiedChunk) {
 			if e.IsDir() {
 				continue
 			}
-			f := filepath.Join(p, e.Name())
-			buff, err := os.ReadFile(f)
-			if err != nil {
-				log.Printf("Error reading chunk '%s': %s", f, err.Error())
-			}
-			c := NewLoadedChunk(
-				&ChunkId{
-					Ver: i,
-					Idx: uint64(j),
-				},
-				buff,
-			)
+			id := &ChunkId{Ver: i, Idx: uint64(j)}
+			c := NewStoredChunk(r, id)
 			chunks <- c
 		}
 	}
diff --git a/repo_test.go b/repo_test.go
index 39556c6..8748d88 100644
--- a/repo_test.go
+++ b/repo_test.go
@@ -131,6 +131,8 @@ func TestLoadChunks(t *testing.T) {
 	resultDir := t.TempDir()
 	dataDir := filepath.Join("testdata", "logs")
 	repo := NewRepo(resultDir)
+	repo.chunkReadWrapper = dummyReader
+	repo.chunkWriteWrapper = dummyWriter
 	resultVersion := filepath.Join(resultDir, "00000")
 	resultChunks := filepath.Join(resultVersion, chunksName)
 	os.MkdirAll(resultChunks, 0775)