Documentation
¶
Index ¶
- type Reader
- func (r *Reader) Bit() uint8
- func (r *Reader) Bits32(width uint8) uint32
- func (r *Reader) Bits64(width uint8) uint64
- func (r *Reader) Byte() uint8
- func (r *Reader) ByteAt(pos int) uint32
- func (r *Reader) BytePos() int
- func (r *Reader) Init(buffer []byte, position int32, limit int32)
- func (r *Reader) Overrun() bool
- func (r *Reader) Position() int32
- func (r *Reader) ReadBit() (uint8, error)
- func (r *Reader) ReadBits32(width uint8) (uint32, error)
- func (r *Reader) ReadBits64(width uint8) (uint64, error)
- func (r *Reader) ReadByte() (byte, error)
- func (r *Reader) ReadSigned32(width uint8) (int32, error)
- func (r *Reader) ReadSigned64(width uint8) (int64, error)
- func (r *Reader) ReadUnary64() (uint64, error)
- func (r *Reader) Remaining() int32
- func (r *Reader) Rice64(param uint8) uint64
- func (r *Reader) Seek(pos int32)
- func (r *Reader) Signed32(width uint8) int32
- func (r *Reader) Signed64(width uint8) int64
- func (r *Reader) SkipToByte()
- func (r *Reader) Unary64() uint64
- func (r *Reader) Unread() []byte
- type Writer
- func (w *Writer) Bit(value uint8)
- func (w *Writer) Bits64(value uint64, width uint8)
- func (w *Writer) Byte(value byte)
- func (w *Writer) BytePos() int
- func (w *Writer) Bytes() []byte
- func (w *Writer) BytesAppend(values []byte)
- func (w *Writer) DetachBytes() []byte
- func (w *Writer) Grow(n int)
- func (w *Writer) Init()
- func (w *Writer) PadToByte()
- func (w *Writer) Position() int32
- func (w *Writer) Signed64(value int64, width uint8)
- func (w *Writer) Unary64(value uint64)
- func (w *Writer) UnaryBits64(value uint64, width uint8)
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type Reader ¶
type Reader struct {
// contains filtered or unexported fields
}
Reader reads bits from a byte buffer. It exposes two API tiers:
- Fast tier (no "Read" prefix): never returns an error. Reads past the end of the buffer are zero-filled/clamped and recorded via Overrun. Intended for hot decode loops.
- Checked tier ("Read" prefix): thin wrappers around the Fast tier that surface Overrun as an error. Intended for cold paths (header parsing) where a precise per-call error is worth the branch.
func (*Reader) Bit ¶
Bit reads a single bit (0 or 1). Past the limit it sets Overrun and returns 0. It returns uint8 rather than bool so Bits64/Unary64/Signed32 can accumulate it with a branchless shift-or instead of a conditional.
func (*Reader) Bits32 ¶
Bits32 reads a width up to 32 bits using a byte-cache algorithm ported unchanged from MP3's original GetBits (used by codec-mp3's layer1/2 dequantization hot path). It is not decomposed into Bit, since doing so would change the performance characteristics MP3 currently relies on.
func (*Reader) Byte ¶
Byte reads one byte at the current (byte-aligned) position. Alignment is a caller invariant, not a data-driven condition, so it is checked via assertf rather than returned as an error.
func (*Reader) ByteAt ¶
ByteAt returns the byte at index pos, or 0 if pos is out of range. It is the bounds-checked primitive shared by Bit/Byte/Bits32 and by MP3's Huffman decoding hot path.
func (*Reader) Init ¶
Init (re)initializes the reader without allocating, so it can be reused across frames.
func (*Reader) Overrun ¶
the limit. It is a sticky flag reset by Init/New: since the read position only advances, once it goes true every later read is also out of range, so a single check after a batch of Fast-tier reads is equivalent to checking every individual call.
func (*Reader) ReadBits32 ¶
ReadBits32 is the Checked-tier wrapper around Bits32.
func (*Reader) ReadBits64 ¶
ReadBits64 is the Checked-tier wrapper around Bits64.
func (*Reader) ReadByte ¶
ReadByte is the Checked-tier wrapper around Byte, named to match Go's io.ByteReader convention (e.g. bufio.Reader.ReadByte).
func (*Reader) ReadSigned32 ¶
ReadSigned32 is the Checked-tier wrapper around Signed32.
func (*Reader) ReadSigned64 ¶
ReadSigned64 is the Checked-tier wrapper around Signed64.
func (*Reader) ReadUnary64 ¶
ReadUnary64 is the Checked-tier wrapper around Unary64.
func (*Reader) Remaining ¶
Remaining returns the number of unread bits between the current position and the limit.
func (*Reader) Seek ¶
Seek jumps to an absolute bit position. It exists for the one place that needs it (Huffman decoding aligning to a granule's declared bit boundary), not as a general-purpose field setter.
func (*Reader) SkipToByte ¶
func (r *Reader) SkipToByte()
SkipToByte advances the position to the next byte boundary without reading a value. This is a distinct operation from Byte/ReadByte (which read a byte), kept separate so the two are never confused for one another.
func (*Reader) Unary64 ¶
Unary64 reads a unary-coded value: the number of 0 bits before the next 1 bit. It stops at the limit instead of looping forever on truncated data.
The scan runs in three phases using math/bits.LeadingZeros: (1) inspect the remainder of an unaligned first byte, (2) fast-scan whole words then bytes that are fully within both the limit and the physical buffer, (3) finish the remaining (<8-bit) tail with the bit-at-a-time path, which also supplies the sticky-overrun bookkeeping on truncated data.
type Writer ¶
type Writer struct {
// contains filtered or unexported fields
}
Writer writes MSB-first bits to an in-memory byte buffer.
func (*Writer) Bits64 ¶
Bits64 writes width bits (width <= 64) MSB-first from value. It merges into the in-progress last byte once (if the writer isn't byte-aligned), appends any whole bytes directly, and appends one final partial byte — replacing what used to be `width` separate per-bit calls, each with its own division/modulo and append-growth check.
func (*Writer) Bytes ¶
Bytes returns the bytes written so far. If the writer is not byte-aligned, the final byte contains zero padding in the remaining least-significant bits.
func (*Writer) BytesAppend ¶
BytesAppend writes bytes at the current byte-aligned position.
func (*Writer) DetachBytes ¶
DetachBytes transfers the written buffer to the caller and resets the writer. The returned bytes must not be retained by the writer's caller after transfer.
func (*Writer) Init ¶
func (w *Writer) Init()
Init resets the writer to an empty buffer, retaining capacity for reuse.
func (*Writer) PadToByte ¶
func (w *Writer) PadToByte()
PadToByte writes zero bits until the next byte boundary.
func (*Writer) Signed64 ¶
Signed64 writes width bits (width <= 64) as a two's-complement signed value.
func (*Writer) Unary64 ¶
Unary64 writes a unary-coded value: value zero bits followed by a one bit. Rice-coded residuals (its main caller) are dominated by this: each sample writes one unary quotient, so batching the zero run into whole-byte appends instead of `value` separate per-bit calls matters even when the typical run is short.
func (*Writer) UnaryBits64 ¶
UnaryBits64 writes a unary-prefixed bit-field: value>>width zero bits, then a one bit, then the low width bits of value. It is equivalent to Unary64(value>>width) followed by Bits64(value, width) — the pattern each Rice-coded residual sample uses — but folds both into a single Bits64 call when the combined field fits in 64 bits (the common case once the Rice parameter is well chosen), instead of two separate alignment/append passes per sample.
Source Files