mirror of
https://github.com/superseriousbusiness/gotosocial.git
synced 2024-11-01 15:00:00 +00:00
105 lines
3 KiB
Go
105 lines
3 KiB
Go
|
// Package chunkreader provides an io.Reader wrapper that minimizes IO reads and memory allocations.
|
||
|
package chunkreader
|
||
|
|
||
|
import (
|
||
|
"io"
|
||
|
)
|
||
|
|
||
|
// ChunkReader is a io.Reader wrapper that minimizes IO reads and memory allocations. It allocates memory in chunks and
|
||
|
// will read as much as will fit in the current buffer in a single call regardless of how large a read is actually
|
||
|
// requested. The memory returned via Next is owned by the caller. This avoids the need for an additional copy.
|
||
|
//
|
||
|
// The downside of this approach is that a large buffer can be pinned in memory even if only a small slice is
|
||
|
// referenced. For example, an entire 4096 byte block could be pinned in memory by even a 1 byte slice. In these rare
|
||
|
// cases it would be advantageous to copy the bytes to another slice.
|
||
|
type ChunkReader struct {
|
||
|
r io.Reader
|
||
|
|
||
|
buf []byte
|
||
|
rp, wp int // buf read position and write position
|
||
|
|
||
|
config Config
|
||
|
}
|
||
|
|
||
|
// Config contains configuration parameters for ChunkReader.
|
||
|
type Config struct {
|
||
|
MinBufLen int // Minimum buffer length
|
||
|
}
|
||
|
|
||
|
// New creates and returns a new ChunkReader for r with default configuration.
|
||
|
func New(r io.Reader) *ChunkReader {
|
||
|
cr, err := NewConfig(r, Config{})
|
||
|
if err != nil {
|
||
|
panic("default config can't be bad")
|
||
|
}
|
||
|
|
||
|
return cr
|
||
|
}
|
||
|
|
||
|
// NewConfig creates and a new ChunkReader for r configured by config.
|
||
|
func NewConfig(r io.Reader, config Config) (*ChunkReader, error) {
|
||
|
if config.MinBufLen == 0 {
|
||
|
// By historical reasons Postgres currently has 8KB send buffer inside,
|
||
|
// so here we want to have at least the same size buffer.
|
||
|
// @see https://github.com/postgres/postgres/blob/249d64999615802752940e017ee5166e726bc7cd/src/backend/libpq/pqcomm.c#L134
|
||
|
// @see https://www.postgresql.org/message-id/0cdc5485-cb3c-5e16-4a46-e3b2f7a41322%40ya.ru
|
||
|
config.MinBufLen = 8192
|
||
|
}
|
||
|
|
||
|
return &ChunkReader{
|
||
|
r: r,
|
||
|
buf: make([]byte, config.MinBufLen),
|
||
|
config: config,
|
||
|
}, nil
|
||
|
}
|
||
|
|
||
|
// Next returns buf filled with the next n bytes. The caller gains ownership of buf. It is not necessary to make a copy
|
||
|
// of buf. If an error occurs, buf will be nil.
|
||
|
func (r *ChunkReader) Next(n int) (buf []byte, err error) {
|
||
|
// n bytes already in buf
|
||
|
if (r.wp - r.rp) >= n {
|
||
|
buf = r.buf[r.rp : r.rp+n]
|
||
|
r.rp += n
|
||
|
return buf, err
|
||
|
}
|
||
|
|
||
|
// available space in buf is less than n
|
||
|
if len(r.buf) < n {
|
||
|
r.copyBufContents(r.newBuf(n))
|
||
|
}
|
||
|
|
||
|
// buf is large enough, but need to shift filled area to start to make enough contiguous space
|
||
|
minReadCount := n - (r.wp - r.rp)
|
||
|
if (len(r.buf) - r.wp) < minReadCount {
|
||
|
newBuf := r.newBuf(n)
|
||
|
r.copyBufContents(newBuf)
|
||
|
}
|
||
|
|
||
|
if err := r.appendAtLeast(minReadCount); err != nil {
|
||
|
return nil, err
|
||
|
}
|
||
|
|
||
|
buf = r.buf[r.rp : r.rp+n]
|
||
|
r.rp += n
|
||
|
return buf, nil
|
||
|
}
|
||
|
|
||
|
func (r *ChunkReader) appendAtLeast(fillLen int) error {
|
||
|
n, err := io.ReadAtLeast(r.r, r.buf[r.wp:], fillLen)
|
||
|
r.wp += n
|
||
|
return err
|
||
|
}
|
||
|
|
||
|
func (r *ChunkReader) newBuf(size int) []byte {
|
||
|
if size < r.config.MinBufLen {
|
||
|
size = r.config.MinBufLen
|
||
|
}
|
||
|
return make([]byte, size)
|
||
|
}
|
||
|
|
||
|
func (r *ChunkReader) copyBufContents(dest []byte) {
|
||
|
r.wp = copy(dest, r.buf[r.rp:r.wp])
|
||
|
r.rp = 0
|
||
|
r.buf = dest
|
||
|
}
|