mirror of
https://git.uploadfilter24.eu/lerentis/terraform-provider-gitea.git
synced 2024-11-14 14:08:12 +00:00
135 lines
3.7 KiB
Go
135 lines
3.7 KiB
Go
// Copyright 2014-2017 Ulrich Kunitz. All rights reserved.
|
|
// Use of this source code is governed by a BSD-style
|
|
// license that can be found in the LICENSE file.
|
|
|
|
package lzma
|
|
|
|
import (
|
|
"errors"
|
|
"fmt"
|
|
)
|
|
|
|
// decoderDict provides the dictionary for the decoder. The whole
|
|
// dictionary is used as reader buffer.
|
|
type decoderDict struct {
|
|
buf buffer
|
|
head int64
|
|
}
|
|
|
|
// newDecoderDict creates a new decoder dictionary. The whole dictionary
|
|
// will be used as reader buffer.
|
|
func newDecoderDict(dictCap int) (d *decoderDict, err error) {
|
|
// lower limit supports easy test cases
|
|
if !(1 <= dictCap && int64(dictCap) <= MaxDictCap) {
|
|
return nil, errors.New("lzma: dictCap out of range")
|
|
}
|
|
d = &decoderDict{buf: *newBuffer(dictCap)}
|
|
return d, nil
|
|
}
|
|
|
|
// Reset clears the dictionary. The read buffer is not changed, so the
|
|
// buffered data can still be read.
|
|
func (d *decoderDict) Reset() {
|
|
d.head = 0
|
|
}
|
|
|
|
// WriteByte writes a single byte into the dictionary. It is used to
|
|
// write literals into the dictionary.
|
|
func (d *decoderDict) WriteByte(c byte) error {
|
|
if err := d.buf.WriteByte(c); err != nil {
|
|
return err
|
|
}
|
|
d.head++
|
|
return nil
|
|
}
|
|
|
|
// pos returns the position of the dictionary head.
|
|
func (d *decoderDict) pos() int64 { return d.head }
|
|
|
|
// dictLen returns the actual length of the dictionary.
|
|
func (d *decoderDict) dictLen() int {
|
|
capacity := d.buf.Cap()
|
|
if d.head >= int64(capacity) {
|
|
return capacity
|
|
}
|
|
return int(d.head)
|
|
}
|
|
|
|
// byteAt returns a byte stored in the dictionary. If the distance is
|
|
// non-positive or exceeds the current length of the dictionary the zero
|
|
// byte is returned.
|
|
func (d *decoderDict) byteAt(dist int) byte {
|
|
if !(0 < dist && dist <= d.dictLen()) {
|
|
return 0
|
|
}
|
|
i := d.buf.front - dist
|
|
if i < 0 {
|
|
i += len(d.buf.data)
|
|
}
|
|
return d.buf.data[i]
|
|
}
|
|
|
|
// writeMatch writes the match at the top of the dictionary. The given
|
|
// distance must point in the current dictionary and the length must not
|
|
// exceed the maximum length 273 supported in LZMA.
|
|
//
|
|
// The error value ErrNoSpace indicates that no space is available in
|
|
// the dictionary for writing. You need to read from the dictionary
|
|
// first.
|
|
func (d *decoderDict) writeMatch(dist int64, length int) error {
|
|
if !(0 < dist && dist <= int64(d.dictLen())) {
|
|
return errors.New("writeMatch: distance out of range")
|
|
}
|
|
if !(0 < length && length <= maxMatchLen) {
|
|
return errors.New("writeMatch: length out of range")
|
|
}
|
|
if length > d.buf.Available() {
|
|
return ErrNoSpace
|
|
}
|
|
d.head += int64(length)
|
|
|
|
i := d.buf.front - int(dist)
|
|
if i < 0 {
|
|
i += len(d.buf.data)
|
|
}
|
|
for length > 0 {
|
|
var p []byte
|
|
if i >= d.buf.front {
|
|
p = d.buf.data[i:]
|
|
i = 0
|
|
} else {
|
|
p = d.buf.data[i:d.buf.front]
|
|
i = d.buf.front
|
|
}
|
|
if len(p) > length {
|
|
p = p[:length]
|
|
}
|
|
if _, err := d.buf.Write(p); err != nil {
|
|
panic(fmt.Errorf("d.buf.Write returned error %s", err))
|
|
}
|
|
length -= len(p)
|
|
}
|
|
return nil
|
|
}
|
|
|
|
// Write writes the given bytes into the dictionary and advances the
|
|
// head.
|
|
func (d *decoderDict) Write(p []byte) (n int, err error) {
|
|
n, err = d.buf.Write(p)
|
|
d.head += int64(n)
|
|
return n, err
|
|
}
|
|
|
|
// Available returns the number of available bytes for writing into the
|
|
// decoder dictionary.
|
|
func (d *decoderDict) Available() int { return d.buf.Available() }
|
|
|
|
// Read reads data from the buffer contained in the decoder dictionary.
|
|
func (d *decoderDict) Read(p []byte) (n int, err error) { return d.buf.Read(p) }
|
|
|
|
// Buffered returns the number of bytes currently buffered in the
|
|
// decoder dictionary.
|
|
func (d *decoderDict) buffered() int { return d.buf.Buffered() }
|
|
|
|
// Peek gets data from the buffer without advancing the rear index.
|
|
func (d *decoderDict) peek(p []byte) (n int, err error) { return d.buf.Peek(p) }
|