コード例 #1
0
ファイル: block.go プロジェクト: larzconwell/bzip2
// writeSymbolBitmaps writes the bitmaps for the used symbols.
func (b *block) writeSymbolBitmaps(bw *bits.Writer, syms symbols.Set) {
	rangesUsed := 0
	ranges := make([]int, 16)

	for i, r := range ranges {
		// Toggle the bits for the 16 symbols in the range.
		for j := 0; j < 16; j++ {
			r = (r << 1) | syms[16*i+j]
		}
		ranges[i] = r

		// Toggle the bit for the range in the bitmap.
		present := 0
		if r > 0 {
			present = 1
		}
		rangesUsed = (rangesUsed << 1) | present
	}

	bw.WriteBits(16, uint64(rangesUsed))
	for _, r := range ranges {
		if r > 0 {
			bw.WriteBits(16, uint64(r))
		}
	}
}
コード例 #2
0
ファイル: block.go プロジェクト: larzconwell/bzip2
// writeTreeSelections writes the huffman tree selections in unary encoding.
func (b *block) writeTreeSelections(bw *bits.Writer, selections []byte) {
	for _, selection := range selections {
		for i := byte(0); i < selection; i++ {
			bw.WriteBits(1, 1)
		}

		bw.WriteBits(1, 0)
	}
}
コード例 #3
0
ファイル: block.go プロジェクト: larzconwell/bzip2
// writeTreeCodes writes the delta encoded code-lengths for
// the huffman trees codes.
func (b *block) writeTreeCodes(bw *bits.Writer, trees []*huffman.Tree) {
	for _, tree := range trees {
		// Get the smallest code-length in the huffman tree.
		codelen := 0
		for i, code := range tree.Codes {
			if i == 0 || code.Len < codelen {
				codelen = code.Len
			}
		}
		bw.WriteBits(5, uint64(codelen))

		// Write the code-lengths as modifications to the current length.
		for _, code := range tree.Codes {
			delta := int(math.Abs(float64(codelen - code.Len)))

			// 2 is increment, 3 is decrement.
			op := uint64(2)
			if codelen > code.Len {
				op = 3
			}
			codelen = code.Len

			for i := 0; i < delta; i++ {
				bw.WriteBits(2, op)
			}

			bw.WriteBits(1, 0)
		}
	}
}
コード例 #4
0
ファイル: block.go プロジェクト: larzconwell/bzip2
// WriteBlock compresses the content buffered and writes
// a block to the bit writer given.
func (b *block) WriteBlock(bw *bits.Writer) error {
	rleData := b.runs.Encode()
	syms, reducedSyms := symbols.Get(rleData)

	// BWT step.
	bwtData := make([]byte, len(rleData))
	bwtidx := bwt.Transform(bwtData, rleData)

	// MTF step.
	mtfData := bwtData
	mtf.Transform(reducedSyms, mtfData, bwtData)

	// RLE2 step.
	rle2Data := rle2.Encode(reducedSyms, mtfData)
	freqs := rle2.GetFrequencies(reducedSyms, rle2Data)

	// Setup the huffman trees required to encode rle2Data.
	trees, selections := huffman.GenerateTrees(freqs, rle2Data)

	// Get the MTF encoded huffman tree selections.
	treeSelectionSymbols := make(symbols.ReducedSet, len(trees))
	for i := range trees {
		treeSelectionSymbols[i] = byte(i)
	}
	treeSelectionBytes := make([]byte, len(selections))
	for i, selection := range selections {
		treeSelectionBytes[i] = byte(selection)
	}
	mtf.Transform(treeSelectionSymbols, treeSelectionBytes, treeSelectionBytes)

	// Write the block header.
	bw.WriteBits(48, blockMagic)
	bw.WriteBits(32, uint64(b.crc))
	bw.WriteBits(1, 0)

	// Write the contents that build the decoding steps.
	bw.WriteBits(24, uint64(bwtidx))
	b.writeSymbolBitmaps(bw, syms)
	bw.WriteBits(3, uint64(len(trees)))
	bw.WriteBits(15, uint64(len(selections)))
	b.writeTreeSelections(bw, treeSelectionBytes)
	b.writeTreeCodes(bw, trees)

	// Write the encoded contents, using the huffman trees generated
	// switching them out every 50 symbols.
	encoded := 0
	idx := 0
	tree := trees[selections[idx]]
	for _, b := range rle2Data {
		if encoded == huffman.TreeSelectionLimit {
			encoded = 0
			idx++
			tree = trees[selections[idx]]
		}
		code := tree.Codes[b]

		bw.WriteBits(uint(code.Len), code.Bits)
		encoded++
	}

	return bw.Err()
}