1
0
Fork 0
forked from forgejo/forgejo

Macaron 1.5 (#12596)

* update macaron to v1.5 of fork

* update macaron to v1.5 of fork

* test gzip PR

* add push method impl to context_tests

* use proper gzip commit

Co-authored-by: zeripath <art27@cantab.net>
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
This commit is contained in:
techknowlogick 2020-08-27 22:47:17 -04:00 committed by GitHub
parent 211321fb93
commit c5d5d63c9c
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
53 changed files with 2622 additions and 665 deletions

View file

@ -12,8 +12,6 @@ but it can be used as a secondary step to compressors (like Snappy) that does no
* [Godoc documentation](https://godoc.org/github.com/klauspost/compress/huff0)
THIS PACKAGE IS NOT CONSIDERED STABLE AND API OR ENCODING MAY CHANGE IN THE FUTURE.
## News
* Mar 2018: First implementation released. Consider this beta software for now.
@ -75,6 +73,8 @@ which can be given to the decompressor.
Decompressing is done by calling the [`Decompress1X`](https://godoc.org/github.com/klauspost/compress/huff0#Scratch.Decompress1X)
or [`Decompress4X`](https://godoc.org/github.com/klauspost/compress/huff0#Scratch.Decompress4X) function.
For concurrently decompressing content with a fixed table a stateless [`Decoder`](https://godoc.org/github.com/klauspost/compress/huff0#Decoder) can be requested which will remain correct as long as the scratch is unchanged. The capacity of the provided slice indicates the expected output size.
You must provide the output from the compression stage, at exactly the size you got back. If you receive an error back
your input was likely corrupted.
@ -84,4 +84,4 @@ There are no integrity checks, so relying on errors from the decompressor does n
# Contributing
Contributions are always welcome. Be aware that adding public functions will require good justification and breaking
changes will likely not be accepted. If in doubt open an issue before writing the PR.
changes will likely not be accepted. If in doubt open an issue before writing the PR.

View file

@ -6,6 +6,7 @@
package huff0
import (
"encoding/binary"
"errors"
"io"
)
@ -34,29 +35,16 @@ func (b *bitReader) init(in []byte) error {
}
b.bitsRead = 64
b.value = 0
b.fill()
b.fill()
if len(in) >= 8 {
b.fillFastStart()
} else {
b.fill()
b.fill()
}
b.bitsRead += 8 - uint8(highBit32(uint32(v)))
return nil
}
// getBits will return n bits. n can be 0.
func (b *bitReader) getBits(n uint8) uint16 {
if n == 0 || b.bitsRead >= 64 {
return 0
}
return b.getBitsFast(n)
}
// getBitsFast requires that at least one bit is requested every time.
// There are no checks if the buffer is filled.
func (b *bitReader) getBitsFast(n uint8) uint16 {
const regMask = 64 - 1
v := uint16((b.value << (b.bitsRead & regMask)) >> ((regMask + 1 - n) & regMask))
b.bitsRead += n
return v
}
// peekBitsFast requires that at least one bit is requested every time.
// There are no checks if the buffer is filled.
func (b *bitReader) peekBitsFast(n uint8) uint16 {
@ -71,21 +59,36 @@ func (b *bitReader) fillFast() {
if b.bitsRead < 32 {
return
}
// Do single re-slice to avoid bounds checks.
// 2 bounds checks.
v := b.in[b.off-4 : b.off]
v = v[:4]
low := (uint32(v[0])) | (uint32(v[1]) << 8) | (uint32(v[2]) << 16) | (uint32(v[3]) << 24)
b.value = (b.value << 32) | uint64(low)
b.bitsRead -= 32
b.off -= 4
}
func (b *bitReader) advance(n uint8) {
b.bitsRead += n
}
// fillFastStart() assumes the bitreader is empty and there is at least 8 bytes to read.
func (b *bitReader) fillFastStart() {
// Do single re-slice to avoid bounds checks.
b.value = binary.LittleEndian.Uint64(b.in[b.off-8:])
b.bitsRead = 0
b.off -= 8
}
// fill() will make sure at least 32 bits are available.
func (b *bitReader) fill() {
if b.bitsRead < 32 {
return
}
if b.off > 4 {
v := b.in[b.off-4 : b.off]
v := b.in[b.off-4:]
v = v[:4]
low := (uint32(v[0])) | (uint32(v[1]) << 8) | (uint32(v[2]) << 16) | (uint32(v[3]) << 24)
b.value = (b.value << 32) | uint64(low)
b.bitsRead -= 32
@ -113,3 +116,214 @@ func (b *bitReader) close() error {
}
return nil
}
// bitReader reads a bitstream in reverse.
// The last set bit indicates the start of the stream and is used
// for aligning the input.
type bitReaderBytes struct {
in []byte
off uint // next byte to read is at in[off - 1]
value uint64
bitsRead uint8
}
// init initializes and resets the bit reader.
func (b *bitReaderBytes) init(in []byte) error {
if len(in) < 1 {
return errors.New("corrupt stream: too short")
}
b.in = in
b.off = uint(len(in))
// The highest bit of the last byte indicates where to start
v := in[len(in)-1]
if v == 0 {
return errors.New("corrupt stream, did not find end of stream")
}
b.bitsRead = 64
b.value = 0
if len(in) >= 8 {
b.fillFastStart()
} else {
b.fill()
b.fill()
}
b.advance(8 - uint8(highBit32(uint32(v))))
return nil
}
// peekBitsFast requires that at least one bit is requested every time.
// There are no checks if the buffer is filled.
func (b *bitReaderBytes) peekByteFast() uint8 {
got := uint8(b.value >> 56)
return got
}
func (b *bitReaderBytes) advance(n uint8) {
b.bitsRead += n
b.value <<= n & 63
}
// fillFast() will make sure at least 32 bits are available.
// There must be at least 4 bytes available.
func (b *bitReaderBytes) fillFast() {
if b.bitsRead < 32 {
return
}
// 2 bounds checks.
v := b.in[b.off-4 : b.off]
v = v[:4]
low := (uint32(v[0])) | (uint32(v[1]) << 8) | (uint32(v[2]) << 16) | (uint32(v[3]) << 24)
b.value |= uint64(low) << (b.bitsRead - 32)
b.bitsRead -= 32
b.off -= 4
}
// fillFastStart() assumes the bitReaderBytes is empty and there is at least 8 bytes to read.
func (b *bitReaderBytes) fillFastStart() {
// Do single re-slice to avoid bounds checks.
b.value = binary.LittleEndian.Uint64(b.in[b.off-8:])
b.bitsRead = 0
b.off -= 8
}
// fill() will make sure at least 32 bits are available.
func (b *bitReaderBytes) fill() {
if b.bitsRead < 32 {
return
}
if b.off > 4 {
v := b.in[b.off-4:]
v = v[:4]
low := (uint32(v[0])) | (uint32(v[1]) << 8) | (uint32(v[2]) << 16) | (uint32(v[3]) << 24)
b.value |= uint64(low) << (b.bitsRead - 32)
b.bitsRead -= 32
b.off -= 4
return
}
for b.off > 0 {
b.value |= uint64(b.in[b.off-1]) << (b.bitsRead - 8)
b.bitsRead -= 8
b.off--
}
}
// finished returns true if all bits have been read from the bit stream.
func (b *bitReaderBytes) finished() bool {
return b.off == 0 && b.bitsRead >= 64
}
// close the bitstream and returns an error if out-of-buffer reads occurred.
func (b *bitReaderBytes) close() error {
// Release reference.
b.in = nil
if b.bitsRead > 64 {
return io.ErrUnexpectedEOF
}
return nil
}
// bitReaderShifted reads a bitstream in reverse.
// The last set bit indicates the start of the stream and is used
// for aligning the input.
type bitReaderShifted struct {
in []byte
off uint // next byte to read is at in[off - 1]
value uint64
bitsRead uint8
}
// init initializes and resets the bit reader.
func (b *bitReaderShifted) init(in []byte) error {
if len(in) < 1 {
return errors.New("corrupt stream: too short")
}
b.in = in
b.off = uint(len(in))
// The highest bit of the last byte indicates where to start
v := in[len(in)-1]
if v == 0 {
return errors.New("corrupt stream, did not find end of stream")
}
b.bitsRead = 64
b.value = 0
if len(in) >= 8 {
b.fillFastStart()
} else {
b.fill()
b.fill()
}
b.advance(8 - uint8(highBit32(uint32(v))))
return nil
}
// peekBitsFast requires that at least one bit is requested every time.
// There are no checks if the buffer is filled.
func (b *bitReaderShifted) peekBitsFast(n uint8) uint16 {
return uint16(b.value >> ((64 - n) & 63))
}
func (b *bitReaderShifted) advance(n uint8) {
b.bitsRead += n
b.value <<= n & 63
}
// fillFast() will make sure at least 32 bits are available.
// There must be at least 4 bytes available.
func (b *bitReaderShifted) fillFast() {
if b.bitsRead < 32 {
return
}
// 2 bounds checks.
v := b.in[b.off-4 : b.off]
v = v[:4]
low := (uint32(v[0])) | (uint32(v[1]) << 8) | (uint32(v[2]) << 16) | (uint32(v[3]) << 24)
b.value |= uint64(low) << ((b.bitsRead - 32) & 63)
b.bitsRead -= 32
b.off -= 4
}
// fillFastStart() assumes the bitReaderShifted is empty and there is at least 8 bytes to read.
func (b *bitReaderShifted) fillFastStart() {
// Do single re-slice to avoid bounds checks.
b.value = binary.LittleEndian.Uint64(b.in[b.off-8:])
b.bitsRead = 0
b.off -= 8
}
// fill() will make sure at least 32 bits are available.
func (b *bitReaderShifted) fill() {
if b.bitsRead < 32 {
return
}
if b.off > 4 {
v := b.in[b.off-4:]
v = v[:4]
low := (uint32(v[0])) | (uint32(v[1]) << 8) | (uint32(v[2]) << 16) | (uint32(v[3]) << 24)
b.value |= uint64(low) << ((b.bitsRead - 32) & 63)
b.bitsRead -= 32
b.off -= 4
return
}
for b.off > 0 {
b.value |= uint64(b.in[b.off-1]) << ((b.bitsRead - 8) & 63)
b.bitsRead -= 8
b.off--
}
}
// finished returns true if all bits have been read from the bit stream.
func (b *bitReaderShifted) finished() bool {
return b.off == 0 && b.bitsRead >= 64
}
// close the bitstream and returns an error if out-of-buffer reads occurred.
func (b *bitReaderShifted) close() error {
// Release reference.
b.in = nil
if b.bitsRead > 64 {
return io.ErrUnexpectedEOF
}
return nil
}

View file

@ -43,6 +43,11 @@ func (b *bitWriter) addBits16Clean(value uint16, bits uint8) {
func (b *bitWriter) encSymbol(ct cTable, symbol byte) {
enc := ct[symbol]
b.bitContainer |= uint64(enc.val) << (b.nBits & 63)
if false {
if enc.nBits == 0 {
panic("nbits 0")
}
}
b.nBits += enc.nBits
}
@ -54,6 +59,14 @@ func (b *bitWriter) encTwoSymbols(ct cTable, av, bv byte) {
sh := b.nBits & 63
combined := uint64(encA.val) | (uint64(encB.val) << (encA.nBits & 63))
b.bitContainer |= combined << sh
if false {
if encA.nBits == 0 {
panic("nbitsA 0")
}
if encB.nBits == 0 {
panic("nbitsB 0")
}
}
b.nBits += encA.nBits + encB.nBits
}

View file

@ -77,8 +77,11 @@ func compress(in []byte, s *Scratch, compressor func(src []byte) ([]byte, error)
// Each symbol present maximum once or too well distributed.
return nil, false, ErrIncompressible
}
if s.Reuse == ReusePolicyPrefer && canReuse {
if s.Reuse == ReusePolicyMust && !canReuse {
// We must reuse, but we can't.
return nil, false, ErrIncompressible
}
if (s.Reuse == ReusePolicyPrefer || s.Reuse == ReusePolicyMust) && canReuse {
keepTable := s.cTable
keepTL := s.actualTableLog
s.cTable = s.prevTable
@ -90,6 +93,9 @@ func compress(in []byte, s *Scratch, compressor func(src []byte) ([]byte, error)
s.OutData = s.Out
return s.Out, true, nil
}
if s.Reuse == ReusePolicyMust {
return nil, false, ErrIncompressible
}
// Do not attempt to re-use later.
s.prevTable = s.prevTable[:0]
}

File diff suppressed because it is too large Load diff

View file

@ -55,6 +55,9 @@ const (
// ReusePolicyNone will disable re-use of tables.
// This is slightly faster than ReusePolicyAllow but may produce larger output.
ReusePolicyNone
// ReusePolicyMust must allow reuse and produce smaller output.
ReusePolicyMust
)
type Scratch struct {
@ -79,6 +82,13 @@ type Scratch struct {
// Slice of the returned data.
OutData []byte
// MaxDecodedSize will set the maximum allowed output size.
// This value will automatically be set to BlockSizeMax if not set.
// Decoders will return ErrMaxDecodedSizeExceeded is this limit is exceeded.
MaxDecodedSize int
br byteReader
// MaxSymbolValue will override the maximum symbol value of the next block.
MaxSymbolValue uint8
@ -95,12 +105,6 @@ type Scratch struct {
// If WantLogLess == 0 any improvement will do.
WantLogLess uint8
// MaxDecodedSize will set the maximum allowed output size.
// This value will automatically be set to BlockSizeMax if not set.
// Decoders will return ErrMaxDecodedSizeExceeded is this limit is exceeded.
MaxDecodedSize int
br byteReader
symbolLen uint16 // Length of active part of the symbol table.
maxCount int // count of the most probable symbol
clearCount bool // clear count