compress/flate: dictionary support

R=bradfitzwork
CC=golang-dev
https://golang.org/cl/4397043

src/pkg/compress/flate/deflate.go

@@ -477,6 +477,33 @@ func NewWriter(w io.Writer, level int) *Writer {
 	return &Writer{pw, &d}
 }
 
+// NewWriterDict is like NewWriter but initializes the new
+// Writer with a preset dictionary.  The returned Writer behaves
+// as if the dictionary had been written to it without producing
+// any compressed output.  The compressed data written to w
+// can only be decompressed by a Reader initialized with the
+// same dictionary.
+func NewWriterDict(w io.Writer, level int, dict []byte) *Writer {
+	dw := &dictWriter{w, false}
+	zw := NewWriter(dw, level)
+	zw.Write(dict)
+	zw.Flush()
+	dw.enabled = true
+	return zw
+}
+
+type dictWriter struct {
+	w       io.Writer
+	enabled bool
+}
+
+func (w *dictWriter) Write(b []byte) (n int, err os.Error) {
+	if w.enabled {
+		return w.w.Write(b)
+	}
+	return len(b), nil
+}
+
 // A Writer takes data written to it and writes the compressed
 // form of that data to an underlying writer (see NewWriter).
 type Writer struct {

src/pkg/compress/flate/deflate_test.go

@@ -275,3 +275,49 @@ func TestDeflateInflateString(t *testing.T) {
 	}
 	testToFromWithLevel(t, 1, gold, "2.718281828...")
 }
+
+func TestReaderDict(t *testing.T) {
+	const (
+		dict = "hello world"
+		text = "hello again world"
+	)
+	var b bytes.Buffer
+	w := NewWriter(&b, 5)
+	w.Write([]byte(dict))
+	w.Flush()
+	b.Reset()
+	w.Write([]byte(text))
+	w.Close()
+
+	r := NewReaderDict(&b, []byte(dict))
+	data, err := ioutil.ReadAll(r)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if string(data) != "hello again world" {
+		t.Fatalf("read returned %q want %q", string(data), text)
+	}
+}
+
+func TestWriterDict(t *testing.T) {
+	const (
+		dict = "hello world"
+		text = "hello again world"
+	)
+	var b bytes.Buffer
+	w := NewWriter(&b, 5)
+	w.Write([]byte(dict))
+	w.Flush()
+	b.Reset()
+	w.Write([]byte(text))
+	w.Close()
+
+	var b1 bytes.Buffer
+	w = NewWriterDict(&b1, 5, []byte(dict))
+	w.Write([]byte(text))
+	w.Close()
+
+	if !bytes.Equal(b1.Bytes(), b.Bytes()) {
+		t.Fatalf("writer wrote %q want %q", b1.Bytes(), b.Bytes())
+	}
+}

src/pkg/compress/flate/inflate.go

@@ -526,6 +526,20 @@ func (f *decompressor) dataBlock() os.Error {
 	return nil
 }
 
+func (f *decompressor) setDict(dict []byte) {
+	if len(dict) > len(f.hist) {
+		// Will only remember the tail.
+		dict = dict[len(dict)-len(f.hist):]
+	}
+
+	f.hp = copy(f.hist[:], dict)
+	if f.hp == len(f.hist) {
+		f.hp = 0
+		f.hfull = true
+	}
+	f.hw = f.hp
+}
+
 func (f *decompressor) moreBits() os.Error {
 	c, err := f.r.ReadByte()
 	if err != nil {
@@ -618,3 +632,16 @@ func NewReader(r io.Reader) io.ReadCloser {
 	go func() { pw.CloseWithError(f.decompress(r, pw)) }()
 	return pr
 }
+
+// NewReaderDict is like NewReader but initializes the reader
+// with a preset dictionary.  The returned Reader behaves as if
+// the uncompressed data stream started with the given dictionary,
+// which has already been read.  NewReaderDict is typically used
+// to read data compressed by NewWriterDict.
+func NewReaderDict(r io.Reader, dict []byte) io.ReadCloser {
+	var f decompressor
+	f.setDict(dict)
+	pr, pw := io.Pipe()
+	go func() { pw.CloseWithError(f.decompress(r, pw)) }()
+	return pr
+}