func TestRL2EncodeFullRange(t *testing.T) {
symbolSetData := make([]byte, 256)
for i := range symbolSetData {
symbolSetData[i] = byte(i)
}
// src is the results of MTF where each byte is used in reverse order.
src := make([]byte, 256)
expected := make([]uint16, 257)
for i := range src {
src[i] = '\xff'
expected[i] = '\u0100'
}
expected[len(expected)-1] = uint16(len(expected))
_, reduced := symbols.Get(symbolSetData)
dst := Encode(reduced, src)
if len(dst) != len(expected) {
t.Error("RLE2 length doesn't match expected length")
}
for i, b := range dst {
if b != expected[i] {
t.Error("Value", int(b), "isn't the expected value", int(expected[i]))
}
}
}
func TestMTFTransformAfterBWT(t *testing.T) {
data := []byte("nnbaaa")
_, reduced := symbols.Get(data)
Transform(reduced, data, data)
if string(data) != "\x02\x00\x02\x02\x00\x00" {
t.Error("Output is incorrect")
}
}
func TestMTFTransformOdd(t *testing.T) {
data := []byte("baanana")
_, reduced := symbols.Get(data)
Transform(reduced, data, data)
if string(data) != "\x01\x01\x00\x02\x01\x01\x01" {
t.Error("Output is incorrect")
}
}
func TestFrequencies(t *testing.T) {
data := []uint16{'\x03', '\x00', '\x03', '\x03', '\x00', '\x01', '\x04'}
expectedFreqs := Frequencies{'\x00': 2, '\x01': 1, '\x03': 3, '\x04': 1}
_, reduced := symbols.Get([]byte("banana"))
freqs := GetFrequencies(reduced, data)
for i, f := range freqs {
if f != expectedFreqs[i] {
t.Error("Frequency", i, "isn't the expected value", f, "should be", expectedFreqs[i])
}
}
}
func TestMTFTransformFullRange(t *testing.T) {
data := make([]byte, 256)
for i := range data {
data[i] = byte(255 - i)
}
_, reduced := symbols.Get(data)
Transform(reduced, data, data)
if data[0] != '\xff' {
t.Error("Output is incorrect")
}
}
func BenchmarkRL2Encode(b *testing.B) {
rand.Seed(time.Now().UnixNano())
src := make([]byte, 1000000)
for i := range src {
src[i] = byte(rand.Intn(256))
}
_, reduced := symbols.Get(src)
b.ResetTimer()
for i := 0; i < b.N; i++ {
Encode(reduced, src)
}
}
func BenchmarkMTFTransformLarge(b *testing.B) {
rand.Seed(time.Now().UnixNano())
src := make([]byte, 1000000*6)
dst := make([]byte, len(src))
for i := range src {
src[i] = byte(rand.Intn(256))
}
_, reduced := symbols.Get(src)
b.ResetTimer()
for i := 0; i < b.N; i++ {
Transform(reduced, dst, src)
}
}
func TestRL2EncodeShortRun(t *testing.T) {
src := []byte("\x02\x00\x02\x02\x00\x00")
expected := []byte("\x03\x00\x03\x03\x01\x04")
_, reduced := symbols.Get([]byte("banana"))
dst := Encode(reduced, src)
if len(dst) != len(expected) {
t.Error("RLE2 length doesn't match expected length")
}
for i, b := range dst {
if b != uint16(expected[i]) {
t.Error("Value", int(b), "isn't the expected value", int(expected[i]))
}
}
}
func TestGenerateTreesLowest(t *testing.T) {
_, reduced := symbols.Get([]byte("banana"))
data := []uint16{'\x03', '\x00', '\x03', '\x00', '\x01', '\x04'}
freqs := rle2.GetFrequencies(reduced, data)
trees, selections := GenerateTrees(freqs, data)
if len(trees) < 2 {
t.Error("Not enough huffman trees generated")
}
if len(trees) > 6 {
t.Error("Too many huffman trees generated")
}
if len(selections) != 1 {
t.Error("The wrong number of huffman tree selections was returned")
}
}
func TestTreeCodeLength(t *testing.T) {
_, reduced := symbols.Get([]byte("banana"))
data := []uint16{'\x03', '\x00', '\x03', '\x00', '\x01', '\x04'}
freqs := rle2.GetFrequencies(reduced, data)
lowestlen := 0
tree := NewTree(freqs)
for i, code := range tree.Codes {
if i == 0 || code.Len < lowestlen {
lowestlen = code.Len
}
}
if lowestlen != tree.Codes['\x03'].Len {
t.Error("The lowest code-length isn't the most used symbol")
}
}
func TestGenerateTreesMultipleSelections(t *testing.T) {
_, reduced := symbols.Get([]byte("banana"))
data := []uint16{'\x03', '\x00', '\x03', '\x00', '\x01', '\x04'}
base := data
for len(data) <= 200 {
data = append(data, base...)
}
freqs := rle2.GetFrequencies(reduced, data)
trees, selections := GenerateTrees(freqs, data)
if len(trees) < 2 {
t.Error("Not enough huffman trees generated")
}
if len(trees) > 6 {
t.Error("Too many huffman trees generated")
}
if len(selections) != 5 {
t.Error("The wrong number of huffman tree selections was returned")
}
}
// 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()
}
