| // Copyright 2013 The Go Authors. All rights reserved. |
| // Use of this source code is governed by a BSD-style |
| // license that can be found in the LICENSE file. |
| |
| package simplifiedchinese |
| |
| import ( |
| "errors" |
| "unicode/utf8" |
| |
| "golang.org/x/text/encoding" |
| "golang.org/x/text/encoding/internal" |
| "golang.org/x/text/encoding/internal/identifier" |
| "golang.org/x/text/transform" |
| ) |
| |
| // HZGB2312 is the HZ-GB2312 encoding. |
| var HZGB2312 encoding.Encoding = &hzGB2312 |
| |
| var hzGB2312 = internal.Encoding{ |
| internal.FuncEncoding{hzGB2312NewDecoder, hzGB2312NewEncoder}, |
| "HZ-GB2312", |
| identifier.HZGB2312, |
| } |
| |
| func hzGB2312NewDecoder() transform.Transformer { |
| return new(hzGB2312Decoder) |
| } |
| |
| func hzGB2312NewEncoder() transform.Transformer { |
| return new(hzGB2312Encoder) |
| } |
| |
| var errInvalidHZGB2312 = errors.New("simplifiedchinese: invalid HZ-GB2312 encoding") |
| |
| const ( |
| asciiState = iota |
| gbState |
| ) |
| |
| type hzGB2312Decoder int |
| |
| func (d *hzGB2312Decoder) Reset() { |
| *d = asciiState |
| } |
| |
| func (d *hzGB2312Decoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) { |
| r, size := rune(0), 0 |
| loop: |
| for ; nSrc < len(src); nSrc += size { |
| c0 := src[nSrc] |
| if c0 >= utf8.RuneSelf { |
| err = errInvalidHZGB2312 |
| break loop |
| } |
| |
| if c0 == '~' { |
| if nSrc+1 >= len(src) { |
| err = transform.ErrShortSrc |
| break loop |
| } |
| size = 2 |
| switch src[nSrc+1] { |
| case '{': |
| *d = gbState |
| continue |
| case '}': |
| *d = asciiState |
| continue |
| case '~': |
| if nDst >= len(dst) { |
| err = transform.ErrShortDst |
| break loop |
| } |
| dst[nDst] = '~' |
| nDst++ |
| continue |
| case '\n': |
| continue |
| default: |
| err = errInvalidHZGB2312 |
| break loop |
| } |
| } |
| |
| if *d == asciiState { |
| r, size = rune(c0), 1 |
| } else { |
| if nSrc+1 >= len(src) { |
| err = transform.ErrShortSrc |
| break loop |
| } |
| c1 := src[nSrc+1] |
| if c0 < 0x21 || 0x7e <= c0 || c1 < 0x21 || 0x7f <= c1 { |
| err = errInvalidHZGB2312 |
| break loop |
| } |
| |
| r, size = '\ufffd', 2 |
| if i := int(c0-0x01)*190 + int(c1+0x3f); i < len(decode) { |
| r = rune(decode[i]) |
| if r == 0 { |
| r = '\ufffd' |
| } |
| } |
| } |
| |
| if nDst+utf8.RuneLen(r) > len(dst) { |
| err = transform.ErrShortDst |
| break loop |
| } |
| nDst += utf8.EncodeRune(dst[nDst:], r) |
| } |
| if atEOF && err == transform.ErrShortSrc { |
| err = errInvalidHZGB2312 |
| } |
| return nDst, nSrc, err |
| } |
| |
| type hzGB2312Encoder int |
| |
| func (d *hzGB2312Encoder) Reset() { |
| *d = asciiState |
| } |
| |
| func (e *hzGB2312Encoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) { |
| r, size := rune(0), 0 |
| for ; nSrc < len(src); nSrc += size { |
| r = rune(src[nSrc]) |
| |
| // Decode a 1-byte rune. |
| if r < utf8.RuneSelf { |
| size = 1 |
| if r == '~' { |
| if nDst+2 > len(dst) { |
| err = transform.ErrShortDst |
| break |
| } |
| dst[nDst+0] = '~' |
| dst[nDst+1] = '~' |
| nDst += 2 |
| continue |
| } |
| |
| } else { |
| // Decode a multi-byte rune. |
| r, size = utf8.DecodeRune(src[nSrc:]) |
| if size == 1 { |
| // All valid runes of size 1 (those below utf8.RuneSelf) were |
| // handled above. We have invalid UTF-8 or we haven't seen the |
| // full character yet. |
| if !atEOF && !utf8.FullRune(src[nSrc:]) { |
| err = transform.ErrShortSrc |
| break |
| } |
| } |
| |
| // func init checks that the switch covers all tables. |
| switch { |
| case encode0Low <= r && r < encode0High: |
| if r = rune(encode0[r-encode0Low]); r != 0 { |
| goto writeGB |
| } |
| case encode1Low <= r && r < encode1High: |
| if r = rune(encode1[r-encode1Low]); r != 0 { |
| goto writeGB |
| } |
| case encode2Low <= r && r < encode2High: |
| if r = rune(encode2[r-encode2Low]); r != 0 { |
| goto writeGB |
| } |
| case encode3Low <= r && r < encode3High: |
| if r = rune(encode3[r-encode3Low]); r != 0 { |
| goto writeGB |
| } |
| case encode4Low <= r && r < encode4High: |
| if r = rune(encode4[r-encode4Low]); r != 0 { |
| goto writeGB |
| } |
| } |
| r = encoding.ASCIISub |
| } |
| |
| writeASCII: |
| if *e != asciiState { |
| if nDst+3 > len(dst) { |
| err = transform.ErrShortDst |
| break |
| } |
| *e = asciiState |
| dst[nDst+0] = '~' |
| dst[nDst+1] = '}' |
| nDst += 2 |
| } else if nDst >= len(dst) { |
| err = transform.ErrShortDst |
| break |
| } |
| dst[nDst] = uint8(r) |
| nDst++ |
| continue |
| |
| writeGB: |
| c0 := uint8(r>>8) - 0x80 |
| c1 := uint8(r) - 0x80 |
| if c0 < 0x21 || 0x7e <= c0 || c1 < 0x21 || 0x7f <= c1 { |
| r = encoding.ASCIISub |
| goto writeASCII |
| } |
| if *e == asciiState { |
| if nDst+4 > len(dst) { |
| err = transform.ErrShortDst |
| break |
| } |
| *e = gbState |
| dst[nDst+0] = '~' |
| dst[nDst+1] = '{' |
| nDst += 2 |
| } else if nDst+2 > len(dst) { |
| err = transform.ErrShortDst |
| break |
| } |
| dst[nDst+0] = c0 |
| dst[nDst+1] = c1 |
| nDst += 2 |
| continue |
| } |
| return nDst, nSrc, err |
| } |