zstd: Add delta encoding support

This adds support for delta encoding, compatible with the --patch-from
option that was introduced in zstd reference v1.4.5:
https://github.com/facebook/zstd/wiki/Zstandard-as-a-patching-engine

zstd/decoder.go

@@ -341,15 +341,8 @@ func (d *Decoder) DecodeAll(input, dst []byte) ([]byte, error) {
 			}
 			return dst, err
 		}
-		if frame.DictionaryID != nil {
-			dict, ok := d.dicts[*frame.DictionaryID]
-			if !ok {
-				return nil, ErrUnknownDictionary
-			}
-			if debugDecoder {
-				println("setting dict", frame.DictionaryID)
-			}
-			frame.history.setDict(&dict)
+		if err = d.setDict(frame); err != nil {
+			return nil, err
 		}
 		if frame.WindowSize > d.o.maxWindowSize {
 			if debugDecoder {
@@ -495,18 +488,12 @@ func (d *Decoder) nextBlockSync() (ok bool) {
 		if !d.syncStream.inFrame {
 			d.frame.history.reset()
 			d.current.err = d.frame.reset(&d.syncStream.br)
+			if d.current.err == nil {
+				d.current.err = d.setDict(d.frame)
+			}
 			if d.current.err != nil {
 				return false
 			}
-			if d.frame.DictionaryID != nil {
-				dict, ok := d.dicts[*d.frame.DictionaryID]
-				if !ok {
-					d.current.err = ErrUnknownDictionary
-					return false
-				} else {
-					d.frame.history.setDict(&dict)
-				}
-			}
 			if d.frame.WindowSize > d.o.maxDecodedSize || d.frame.WindowSize > d.o.maxWindowSize {
 				d.current.err = ErrDecoderSizeExceeded
 				return false
@@ -865,13 +852,8 @@ decodeStream:
 		if debugDecoder && err != nil {
 			println("Frame decoder returned", err)
 		}
-		if err == nil && frame.DictionaryID != nil {
-			dict, ok := d.dicts[*frame.DictionaryID]
-			if !ok {
-				err = ErrUnknownDictionary
-			} else {
-				frame.history.setDict(&dict)
-			}
+		if err == nil {
+			err = d.setDict(frame)
 		}
 		if err == nil && d.frame.WindowSize > d.o.maxWindowSize {
 			if debugDecoder {
@@ -953,3 +935,20 @@ decodeStream:
 	hist.reset()
 	d.frame.history.b = frameHistCache
 }
+
+func (d *Decoder) setDict(frame *frameDec) (err error) {
+	dict, ok := d.dicts[frame.DictionaryID]
+	if ok {
+		if debugDecoder {
+			println("setting dict", frame.DictionaryID)
+		}
+		frame.history.setDict(&dict)
+	} else if frame.DictionaryID != 0 {
+		// A zero (or missing) dictionary id means unspecified: the decoder
+		// must know whether a dictionary is used "by other means".
+		// In particular, zstd --patch-from uses this id.
+		// So only return an error if the dictionary id is not zero.
+		err = ErrUnknownDictionary
+	}
+	return err
+}

zstd/decoder_options.go

@@ -85,7 +85,13 @@ func WithDecoderMaxMemory(n uint64) DOption {
 }
 
 // WithDecoderDicts allows to register one or more dictionaries for the decoder.
-// If several dictionaries with the same ID is provided the last one will be used.
+//
+// Each slice in dict must be in the [dictionary format] produced by
+// "zstd --train" from the Zstandard reference implementation.
+//
+// If several dictionaries with the same ID are provided, the last one will be used.
+//
+// [dictionary format]: https://github.com/facebook/zstd/blob/dev/doc/zstd_compression_format.md#dictionary-format
 func WithDecoderDicts(dicts ...[]byte) DOption {
 	return func(o *decoderOptions) error {
 		for _, b := range dicts {
@@ -99,6 +105,15 @@ func WithDecoderDicts(dicts ...[]byte) DOption {
 	}
 }
 
+// WithEncoderDictRaw registers a dictionary that may be used by the decoder.
+// The slice content can be arbitrary data.
+func WithDecoderDictRaw(id uint32, content []byte) DOption {
+	return func(o *decoderOptions) error {
+		o.dicts = append(o.dicts, dict{id: id, content: content})
+		return nil
+	}
+}
+
 // WithDecoderMaxWindow allows to set a maximum window size for decodes.
 // This allows rejecting packets that will cause big memory usage.
 // The Decoder will likely allocate more memory based on the WithDecoderLowmem setting.

zstd/dict_test.go

@@ -459,3 +459,38 @@ func readDicts(tb testing.TB, zr *zip.Reader) [][]byte {
 	}
 	return dicts
 }
+
+// Test decoding of zstd --patch-from output.
+func TestDecoderRawDict(t *testing.T) {
+	t.Parallel()
+
+	dict, err := os.ReadFile("testdata/delta/source.txt")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	delta, err := os.Open("testdata/delta/target.txt.zst")
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer delta.Close()
+
+	dec, err := NewReader(delta, WithDecoderDictRaw(0, dict))
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	out, err := io.ReadAll(dec)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	ref, err := os.ReadFile("testdata/delta/target.txt")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if !bytes.Equal(out, ref) {
+		t.Errorf("mismatch: got %q, wanted %q", out, ref)
+	}
+}

zstd/encoder_options.go

@@ -305,7 +305,13 @@ func WithLowerEncoderMem(b bool) EOption {
 }
 
 // WithEncoderDict allows to register a dictionary that will be used for the encode.
+//
+// The slice dict must be in the [dictionary format] produced by
+// "zstd --train" from the Zstandard reference implementation.
+//
 // The encoder *may* choose to use no dictionary instead for certain payloads.
+//
+// [dictionary format]: https://github.com/facebook/zstd/blob/dev/doc/zstd_compression_format.md#dictionary-format
 func WithEncoderDict(dict []byte) EOption {
 	return func(o *encoderOptions) error {
 		d, err := loadDict(dict)
@@ -316,3 +322,14 @@ func WithEncoderDict(dict []byte) EOption {
 		return nil
 	}
 }
+
+// WithEncoderDictRaw registers a dictionary that may be used by the encoder.
+//
+// The slice content may contain arbitrary data. It will be used as an initial
+// history.
+func WithEncoderDictRaw(id uint32, content []byte) EOption {
+	return func(o *encoderOptions) error {
+		o.dict = &dict{id: id, content: content}
+		return nil
+	}
+}

zstd/example_test.go

@@ -0,0 +1,47 @@
+package zstd_test
+
+import (
+	"bytes"
+	"fmt"
+
+	"github.com/klauspost/compress/zstd"
+)
+
+func ExampleWithEncoderDictRaw() {
+	// "Raw" dictionaries can be used for compressed delta encoding.
+
+	source := []byte(`
+		This is the source file. Compression of the target file with
+		the source file as the dictionary will produce a compressed
+		delta encoding of the target file.`)
+	target := []byte(`
+		This is the target file. Decompression of the delta encoding with
+		the source file as the dictionary will produce this file.`)
+
+	// The dictionary id is arbitrary. We use zero for compatibility
+	// with zstd --patch-from, but applications can use any id
+	// not in the range [32768, 1<<31).
+	const id = 0
+
+	bestLevel := zstd.WithEncoderLevel(zstd.SpeedBestCompression)
+
+	w, _ := zstd.NewWriter(nil, bestLevel,
+		zstd.WithEncoderDictRaw(id, source))
+	delta := w.EncodeAll(target, nil)
+	fmt.Printf("delta encoding:   %4d bytes\n", len(delta))
+
+	r, _ := zstd.NewReader(nil, zstd.WithDecoderDictRaw(id, source))
+	out, err := r.DecodeAll(delta, nil)
+	if err != nil || !bytes.Equal(out, target) {
+		panic("decoding error")
+	}
+
+	// Ordinary compression, for reference.
+	w, _ = zstd.NewWriter(nil, bestLevel)
+	z := w.EncodeAll(target, nil)
+	fmt.Printf("compression only: %4d bytes\n", len(z))
+
+	// Output:
+	// delta encoding:     44 bytes
+	// compression only:  111 bytes
+}

zstd/framedec.go

@@ -29,7 +29,7 @@ type frameDec struct {
 
 	FrameContentSize uint64
 
-	DictionaryID  *uint32
+	DictionaryID  uint32
 	HasCheckSum   bool
 	SingleSegment bool
 }
@@ -155,7 +155,7 @@ func (d *frameDec) reset(br byteBuffer) error {
 
 	// Read Dictionary_ID
 	// https://github.com/facebook/zstd/blob/dev/doc/zstd_compression_format.md#dictionary_id
-	d.DictionaryID = nil
+	d.DictionaryID = 0
 	if size := fhd & 3; size != 0 {
 		if size == 3 {
 			size = 4
@@ -178,11 +178,7 @@ func (d *frameDec) reset(br byteBuffer) error {
 		if debugDecoder {
 			println("Dict size", size, "ID:", id)
 		}
-		if id > 0 {
-			// ID 0 means "sorry, no dictionary anyway".
-			// https://github.com/facebook/zstd/blob/dev/doc/zstd_compression_format.md#dictionary-format
-			d.DictionaryID = &id
-		}
+		d.DictionaryID = id
 	}
 
 	// Read Frame_Content_Size

zstd/testdata/delta/source.txt

@@ -0,0 +1,5 @@
+0000000000000000
+
+This file is to be used as the dictionary for compressing target.txt:
+
+	zstd -19 --patch-from=source.txt target.txt

zstd/testdata/delta/target.txt

@@ -0,0 +1,5 @@
+0000000000000000
+
+This file is to be compressed with source.txt as the dictionary:
+
+	zstd -19 --patch-from=source.txt target.txt

zstd/zstd.go

@@ -72,7 +72,6 @@ var (
 	ErrDecoderSizeExceeded = errors.New("decompressed size exceeds configured limit")
 
 	// ErrUnknownDictionary is returned if the dictionary ID is unknown.
-	// For the time being dictionaries are not supported.
 	ErrUnknownDictionary = errors.New("unknown dictionary")
 
 	// ErrFrameSizeExceeded is returned if the stated frame size is exceeded.