lz

README

Module LZ

The LZ module provides sequencers that convert byte streams into blocks of Lempel-Ziv 77 sequences. It is designed to support multiple compression methods that differ in the way they are encoding those LZ77 sequences.

Documentation

Overview

Package lz supports encoding and decoding of LZ77 sequences. A sequence, as described in the Zstandard specification, consists of a literal copy command followed by a match copy command. The literal copy command is described by the length in literal bytes to be copied, and the match command consists of the distance of the match to copy and the length of the match in bytes.

A Parser converts a byte stream into blocks of sequences. The Decoder converts the block of sequences into the original decompressed byte stream.

The module provides multiple parser implementations that offer different combinations of encoding speed and compression ratios. Usually, a slower parser will generate a better compression ratio.

The package has a generic parser that needs a path finder, which selects the sequences for data block and a mapper that finds potential matches in the byte stream.

For optimization we may provide custom implementations that integrate matcher or path finder into the parser and avoids the calling overhead through interfaces.

The library supports the implementation of parsers outside of this package.

Index

• Variables
• type Block
  • func (b *Block) Len() int64
  • func (b *Block) LenCheck() (n int64, err error)
• type Buffer
  • func (b *Buffer) ByteAt(off int64) (c byte, err error)
  • func (b *Buffer) Init(size, retentionSize int, shift ShiftFunc) error
  • func (b *Buffer) Parsable() int
  • func (b *Buffer) ReadAt(p []byte, off int64) (n int, err error)
  • func (b *Buffer) ReadFrom(r io.Reader) (n int64, err error)
  • func (b *Buffer) Reset(data []byte) error
  • func (b *Buffer) Write(p []byte) (n int, err error)
• type Decoder
  • func NewDecoder(opts DecoderOptions) (*Decoder, error)
  • func (d *Decoder) ByteAtEnd(off int) byte
  • func (d *Decoder) Init(opts DecoderOptions) error
  • func (d *Decoder) Read(p []byte) (n int, err error)
  • func (d *Decoder) Reset()
  • func (d *Decoder) Write(p []byte) (n int, err error)
  • func (d *Decoder) WriteBlock(blk *Block) (n int, err error)
  • func (d *Decoder) WriteByte(c byte) error
  • func (d *Decoder) WriteMatch(mu, ou uint32) (n int, err error)
  • func (d *Decoder) WriteTo(w io.Writer) (n int64, err error)
• type DecoderOptions
• type Entry
• type Mapper
  • func NewMapper(name string) (Mapper, error)
• type Matcher
• type Parser
  • func NewParser(opts ParserOptions) (Parser, error)
• type ParserFlags
• type ParserOptions
  • func (o ParserOptions) MarshalJSON() ([]byte, error)
  • func (o *ParserOptions) SetDefaults()
  • func (o *ParserOptions) UnmarshalJSON(data []byte) error
• type PathFinder
  • func NewPathFinder(name string, m Matcher) (PathFinder, error)
• type Seq
  • func (s Seq) Len() int64
• type ShiftFunc

Variables

var ErrEndOfBuffer = errors.New("lz: end of buffer")

ErrEndOfBuffer is returned at the end of the buffer.

var ErrFullBuffer = errors.New("lz: full buffer")

ErrFullBuffer is returned when the buffer is full and no more data can be written to it.

var ErrOutOfBuffer = errors.New("lz: offset outside of buffer")

ErrOutOfBuffer is returned when the offset is outside of the buffer.

var ErrStartOfBuffer = errors.New("lz: start of buffer")

ErrStartOfBuffer is returned at the start of the buffer.

Types

type Block

type Block struct {
	Sequences []Seq
	Literals  []byte
}

Block stores sequences and literals. Note that the sequences stored in the Sequences slice might not consume the entire Literals slice. The remaining literal bytes must be added to the decoded text after all sequences have been decoded.

func (*Block) Len

func (b *Block) Len() int64

Len computes the length of the block in bytes. It assumes that the sum of the literal lengths in the sequences does not exceed the length of the Literals byte slice.

func (*Block) LenCheck

func (b *Block) LenCheck() (n int64, err error)

LenCheck computes the length of the block in bytes and verifies that the sum of the literal lengths in the sequences is less than the bytes in the Literals field. If that is not the case an error is returned.

type Buffer

type Buffer struct {
	Data []byte
	// Window end index
	W int
	// maximum buffer size
	Size int
	// RetentionSize is the number of bytes to keep when pruning the buffer.
	RetentionSize int
	// offset of Data
	Off int64

	// Shift will be called when the buffer is pruned to inform other
	// components about the number of bytes removed from the start of
	// the buffer.
	Shift ShiftFunc
}

Buffer is the Buffer used for LZ parsing.

The Off field describes the offset of Data[0] in the original stream. The W points to the end of sliding window used for copying matches.

Data is not fully allocated at the beginning. It grows with the usage. There must be always 7 extra bytes allocated at the end of Data to allow easy reads of data from the Buffer.

func (*Buffer) ByteAt

func (b *Buffer) ByteAt(off int64) (c byte, err error)

ByteAt returns the byte at offset off. If off is outside of the buffer, ErrOutOfBuffer is returned.

func (*Buffer) Init

func (b *Buffer) Init(size, retentionSize int, shift ShiftFunc) error

Init initializes the buffer. The old data slice is reused and the capacity might be larger than the new buffer size.

func (*Buffer) Parsable

func (b *Buffer) Parsable() int

Parsable returns the number of bytes that can be parsed from the buffer.

func (*Buffer) ReadAt

func (b *Buffer) ReadAt(p []byte, off int64) (n int, err error)

ReadAt reads len(p) bytes from the buffer starting at byte offset off. It returns the number of bytes read and any error encountered. If off is outside of the buffer, ErrOutOfBuffer is returned.

func (*Buffer) ReadFrom

func (b *Buffer) ReadFrom(r io.Reader) (n int64, err error)

ReadFrom reads data from r until EOF or error. It returns the number of bytes read and any error encountered.

func (*Buffer) Reset

func (b *Buffer) Reset(data []byte) error

Reset resets the buffer with the provided data slice. If the data slice is larger than the buffer size, the buffer size will be updated. Note that the data slice should have 7 extra bytes, len(data)+7 <= cap(data). Otherwise the old slice will be used or a new one need to be allocated.

func (*Buffer) Write

func (b *Buffer) Write(p []byte) (n int, err error)

Write writes data to the buffer. If not all data can be written, ErrFullBuffer is returned.

type Decoder

type Decoder struct {
	// Data is the actual buffer. The end of the slice is also the head of
	// the dictionary window.
	Data []byte
	// R tracks the position of the reads from the buffer and must be less
	// or equal to the length of the Data slice.
	R int
	// Off records the total offset and marks the end of the Data slice,
	// which is also the end of the dictionary window.
	Off int64

	// DecoderOptions provides the configuration parameters WindowSize and
	// BufferSize.
	DecoderOptions
}

Decoder provides a simple buffer for decoding LZ77 sequences. Data is the actual buffer. The end of the slice is also the head of the dictionary window. R tracks the read position in the buffer and must be less than or equal to the length of the Data slice. Off records the total offset and marks the end of the Data slice, which is also the end of the dictionary window. DecoderConfig provides the configuration parameters WindowSize and BufferSize.

func NewDecoder

func NewDecoder(opts DecoderOptions) (*Decoder, error)

func (*Decoder) ByteAtEnd

func (d *Decoder) ByteAtEnd(off int) byte

ByteAtEnd returns the byte at the end of the buffer.

func (*Decoder) Init

func (d *Decoder) Init(opts DecoderOptions) error

Init initializes the DecoderBuffer.

func (*Decoder) Read

func (d *Decoder) Read(p []byte) (n int, err error)

Read reads decoded data from the buffer.

func (*Decoder) Reset

func (d *Decoder) Reset()

Reset returns the DecoderBuffer to its initialized state.

func (*Decoder) Write

func (d *Decoder) Write(p []byte) (n int, err error)

Write inserts the slice into the buffer. The method will write the entire slice or return 0 and ErrFullBuffer.

func (*Decoder) WriteBlock

func (d *Decoder) WriteBlock(blk *Block) (n int, err error)

WriteBlock writes sequences from the block into the buffer. Each sequence is written atomically, as the block value is not modified. If there is not enough space in the buffer, ErrFullBuffer will be returned. All written sequences and literals will be removed from the block.

The capacity of the block slices will not be maintained. You have to keep a copy of the block to achieve that.

The growth of the array is limited to BufferSize.

The function returns the number of bytes written.

func (*Decoder) WriteByte

func (d *Decoder) WriteByte(c byte) error

WriteByte writes a single byte into the buffer.

func (*Decoder) WriteMatch

func (d *Decoder) WriteMatch(mu, ou uint32) (n int, err error)

WriteMatch appends the match to the end of the buffer. The match will be written completely, or n=0 and ErrFullBuffer will be returned.

func (*Decoder) WriteTo

func (d *Decoder) WriteTo(w io.Writer) (n int64, err error)

WriteTo writes the decoded data to the writer.

type DecoderOptions

type DecoderOptions struct {
	// Size of the sliding dictionary window in bytes.
	WindowSize int
	// Maximum size of the buffer in bytes.
	BufferSize int
}

DecoderOptions contains the parameters for the DecoderBuffer and decoder types. WindowSize must be smaller than BufferSize. It is recommended to set BufferSize to twice the WindowSize.

type Entry

type Entry struct {
	// contains filtered or unexported fields
}

Entry is returned by a Mapper for a found match. It provides the position i of the match in the Data slice of the buffer and v contains the leading 4 bytes of the match to avoid a lookup in the buffer.

type Mapper

type Mapper interface {
	// InputLen returns the length of the input data into the table. We are
	// supporting length from 2 to 8 bytes.
	InputLen() int

	// Reset resets the internal state of the mapper.
	Reset()

	// Shift is called by the number of bytes pruned from the buffer and
	// provide the new extended buffer to the mapper.
	Shift(delta int)

	// Put adds all values between a and w to the mapper. We assume that
	// cap(p) >= len(p) + 7.
	Put(p []byte, a, w int) int

	// Get returns all candidate entries for the provided hash value. The
	// entry value v contains the all 4 bytes stored a position i.
	Get(v uint64) []Entry
}

Mapper provides potential matches for a given position in the byte stream. Is it usually implemented by hash tables.

func NewMapper

func NewMapper(name string) (Mapper, error)

NewMapper creates a new Mapper for the provided name of the algorithm. The mappers supported are described below.

hash_<inputLen>:<hashBits> A hash table with the provided input length and hash bits. The input length is between 2 and 8 bytes, and the hash bits can be 24 bits at maximum.

type Matcher

type Matcher interface {
	// Edges returns the potential sequence at the current position.
	Edges(n int) []Seq

	// Skip is called to advance the current position by n bytes. An error
	// is only returned if the there are not enough bytes in the buffer.
	// Note that n can be negative, to allow to set the current position
	// backwards.
	Skip(n int) (skipped int, err error)

	// Parsable returns the number of bytes that are available in the buffer
	// for parsing.
	Parsable() int
}

Matcher is responsible to find matches or Literal bytes in the byte stream. It is only relevant for the PathFinder.

type Parser

type Parser interface {
	// Parse up to block size bytes from the internal buffer and provides
	// the sequences in the block structure. While slices will be reused,
	// not old information will be maintained.
	Parse(blk *Block, n int, flags ParserFlags) (parsed int, err error)

	// Write writes data into the internal buffer.
	Write(p []byte) (n int, err error)

	// ReadFrom reads data from the provided reader into the internal
	// buffer.
	ReadFrom(r io.Reader) (n int64, err error)

	// ReadAt reads len(p) bytes from the internal buffer at offset off.
	ReadAt(p []byte, off int64) (n int, err error)

	// ByteAt returns the byte at offset off in the internal buffer.
	ByteAt(off int64) (c byte, err error)

	// Reset resets the internal buffer to the provided data.
	Reset(data []byte) error

	// Options returns the options used to create the parser.
	Options() ParserOptions
}

Parser provides the possibility to parse a byte stream into LZ77 sequences.

func NewParser

func NewParser(opts ParserOptions) (Parser, error)

NewParser creates a new parser for the provided options.

type ParserFlags

type ParserFlags int

ParserFlags define optional parser behavior.

const (
	// NoTrailingLiterals indicates that the parser should not generate
	// trailing literal bytes in the output.
	NoTrailingLiterals ParserFlags = 1 << iota
)

type ParserOptions

type ParserOptions struct {
	PathFinder string
	Mapper     string

	WindowSize    int
	RetentionSize int
	BufferSize    int

	MinMatchLen int
	MaxMatchLen int
}

ParserOptions provides the configuration for a parser. PathFinder describes the algorithm to find a path through the different matches. The Mapper is used to find potential matches.

For the list of available path finders see NewPathFinder. For the list of available mappers see NewMapper.

func (ParserOptions) MarshalJSON

func (o ParserOptions) MarshalJSON() ([]byte, error)

MarshalJSON returns the JSON encoding of the parser options. The sizes are encoded in a human readable format, for instance 8 MiB are represented as "8M", 16 KiB as "16K", 2 GiB as "2G".

func (*ParserOptions) SetDefaults

func (o *ParserOptions) SetDefaults()

SetDefaults sets the default values for the parser options if the field is zero or empty.

func (*ParserOptions) UnmarshalJSON

func (o *ParserOptions) UnmarshalJSON(data []byte) error

UnmarshalJSON parses the JSON encoding of the parser options. The sizes are expected to be in a human readable format, for instance 8 MiB are represented as "8M", 16 KiB as "16K", 2 GiB as "2G".

type PathFinder

type PathFinder interface {
	Parse(block *Block, n int, flags ParserFlags) (parsed int, err error)
	Reset()
}

PathFinder implements the central Parse Function of the Parser.

func NewPathFinder

func NewPathFinder(name string, m Matcher) (PathFinder, error)

NewPathFinder creates a new PathFinder for the provided name of the algorithm and the Matcher. The path finders supported are described below.

greedy The greedy path finder selects the longest match at each position.

type Seq

type Seq struct {
	LitLen   uint32
	MatchLen uint32
	Offset   uint32
	Aux      uint32
}

Seq represents a single Lempel-Ziv 77 sequence describing a match, consisting of the offset, the length of the match, and the number of literals preceding the match. The Aux field can be used in upper layers to store additional information. One use case is to store up to 4 bytes of literals.

func (Seq) Len

func (s Seq) Len() int64

Len returns the complete length of the sequence in bytes.

type ShiftFunc

type ShiftFunc func(delta int)

ShiftFunc defines the type of the shift function, called when the buffer is pruned to provide more available space.