implement decoding charsets (other than ascii and utf-8) while reading textual message parts, and improve search

message.Part now has a ReaderUTF8OrBinary() along with the existing Reader().
the new function returns a reader of decoded content. we now use it in a few
places, including search. we only support the charsets in
golang.org/x/text/encoding/ianaindex.

search has also been changed to not read the entire message in memory. instead,
we make one 8k buffer for reading and search in that, and we keep the buffer
around for all messages. saves quite some allocations when searching large
mailboxes.

vendor/golang.org/x/text/encoding/simplifiedchinese/all.go

@@ -0,0 +1,12 @@
+// Copyright 2015 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 (
+	"golang.org/x/text/encoding"
+)
+
+// All is a list of all defined encodings in this package.
+var All = []encoding.Encoding{GB18030, GBK, HZGB2312}

vendor/golang.org/x/text/encoding/simplifiedchinese/gbk.go

@@ -0,0 +1,273 @@
+// 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 (
+	"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"
+)
+
+var (
+	// GB18030 is the GB18030 encoding.
+	GB18030 encoding.Encoding = &gbk18030
+	// GBK is the GBK encoding. It encodes an extension of the GB2312 character set
+	// and is also known as Code Page 936.
+	GBK encoding.Encoding = &gbk
+)
+
+var gbk = internal.Encoding{
+	&internal.SimpleEncoding{
+		gbkDecoder{gb18030: false},
+		gbkEncoder{gb18030: false},
+	},
+	"GBK",
+	identifier.GBK,
+}
+
+var gbk18030 = internal.Encoding{
+	&internal.SimpleEncoding{
+		gbkDecoder{gb18030: true},
+		gbkEncoder{gb18030: true},
+	},
+	"GB18030",
+	identifier.GB18030,
+}
+
+type gbkDecoder struct {
+	transform.NopResetter
+	gb18030 bool
+}
+
+func (d gbkDecoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, size := rune(0), 0
+loop:
+	for ; nSrc < len(src); nSrc += size {
+		switch c0 := src[nSrc]; {
+		case c0 < utf8.RuneSelf:
+			r, size = rune(c0), 1
+
+		// Microsoft's Code Page 936 extends GBK 1.0 to encode the euro sign U+20AC
+		// as 0x80. The HTML5 specification at http://encoding.spec.whatwg.org/#gbk
+		// says to treat "gbk" as Code Page 936.
+		// GBK’s decoder is gb18030’s decoder. https://encoding.spec.whatwg.org/#gbk-decoder
+		// If byte is 0x80, return code point U+20AC. https://encoding.spec.whatwg.org/#gb18030-decoder
+		case c0 == 0x80:
+			r, size = '€', 1
+
+		case c0 < 0xff:
+			if nSrc+1 >= len(src) {
+				if !atEOF {
+					err = transform.ErrShortSrc
+					break loop
+				}
+				r, size = utf8.RuneError, 1
+				goto write
+			}
+			c1 := src[nSrc+1]
+			switch {
+			case 0x40 <= c1 && c1 < 0x7f:
+				c1 -= 0x40
+			case 0x80 <= c1 && c1 < 0xff:
+				c1 -= 0x41
+			case d.gb18030 && 0x30 <= c1 && c1 < 0x40:
+				if nSrc+3 >= len(src) {
+					if !atEOF {
+						err = transform.ErrShortSrc
+						break loop
+					}
+					// The second byte here is always ASCII, so we can set size
+					// to 1 in all cases.
+					r, size = utf8.RuneError, 1
+					goto write
+				}
+				c2 := src[nSrc+2]
+				if c2 < 0x81 || 0xff <= c2 {
+					r, size = utf8.RuneError, 1
+					goto write
+				}
+				c3 := src[nSrc+3]
+				if c3 < 0x30 || 0x3a <= c3 {
+					r, size = utf8.RuneError, 1
+					goto write
+				}
+				size = 4
+				r = ((rune(c0-0x81)*10+rune(c1-0x30))*126+rune(c2-0x81))*10 + rune(c3-0x30)
+				if r < 39420 {
+					i, j := 0, len(gb18030)
+					for i < j {
+						h := i + (j-i)/2
+						if r >= rune(gb18030[h][0]) {
+							i = h + 1
+						} else {
+							j = h
+						}
+					}
+					dec := &gb18030[i-1]
+					r += rune(dec[1]) - rune(dec[0])
+					goto write
+				}
+				r -= 189000
+				if 0 <= r && r < 0x100000 {
+					r += 0x10000
+				} else {
+					r, size = utf8.RuneError, 1
+				}
+				goto write
+			default:
+				r, size = utf8.RuneError, 1
+				goto write
+			}
+			r, size = '\ufffd', 2
+			if i := int(c0-0x81)*190 + int(c1); i < len(decode) {
+				r = rune(decode[i])
+				if r == 0 {
+					r = '\ufffd'
+				}
+			}
+
+		default:
+			r, size = utf8.RuneError, 1
+		}
+
+	write:
+		if nDst+utf8.RuneLen(r) > len(dst) {
+			err = transform.ErrShortDst
+			break loop
+		}
+		nDst += utf8.EncodeRune(dst[nDst:], r)
+	}
+	return nDst, nSrc, err
+}
+
+type gbkEncoder struct {
+	transform.NopResetter
+	gb18030 bool
+}
+
+func (e gbkEncoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, r2, size := rune(0), rune(0), 0
+	for ; nSrc < len(src); nSrc += size {
+		r = rune(src[nSrc])
+
+		// Decode a 1-byte rune.
+		if r < utf8.RuneSelf {
+			size = 1
+
+		} 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 r2 = rune(encode0[r-encode0Low]); r2 != 0 {
+					goto write2
+				}
+			case encode1Low <= r && r < encode1High:
+				// Microsoft's Code Page 936 extends GBK 1.0 to encode the euro sign U+20AC
+				// as 0x80. The HTML5 specification at http://encoding.spec.whatwg.org/#gbk
+				// says to treat "gbk" as Code Page 936.
+				// GBK’s encoder is gb18030’s encoder with its _is GBK_ set to true. https://encoding.spec.whatwg.org/#gbk-encoder
+				// If _is GBK_ is true and code point is U+20AC, return byte 0x80. https://encoding.spec.whatwg.org/#gb18030-encoder
+				if !e.gb18030 && r == '€' {
+					r = 0x80
+					goto write1
+				}
+				if r2 = rune(encode1[r-encode1Low]); r2 != 0 {
+					goto write2
+				}
+			case encode2Low <= r && r < encode2High:
+				if r2 = rune(encode2[r-encode2Low]); r2 != 0 {
+					goto write2
+				}
+			case encode3Low <= r && r < encode3High:
+				if r2 = rune(encode3[r-encode3Low]); r2 != 0 {
+					goto write2
+				}
+			case encode4Low <= r && r < encode4High:
+				if r2 = rune(encode4[r-encode4Low]); r2 != 0 {
+					goto write2
+				}
+			}
+
+			if e.gb18030 {
+				if r < 0x10000 {
+					i, j := 0, len(gb18030)
+					for i < j {
+						h := i + (j-i)/2
+						if r >= rune(gb18030[h][1]) {
+							i = h + 1
+						} else {
+							j = h
+						}
+					}
+					dec := &gb18030[i-1]
+					r += rune(dec[0]) - rune(dec[1])
+					goto write4
+				} else if r < 0x110000 {
+					r += 189000 - 0x10000
+					goto write4
+				}
+			}
+			err = internal.ErrASCIIReplacement
+			break
+		}
+
+	write1:
+		if nDst >= len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		dst[nDst] = uint8(r)
+		nDst++
+		continue
+
+	write2:
+		if nDst+2 > len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		dst[nDst+0] = uint8(r2 >> 8)
+		dst[nDst+1] = uint8(r2)
+		nDst += 2
+		continue
+
+	write4:
+		if nDst+4 > len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		dst[nDst+3] = uint8(r%10 + 0x30)
+		r /= 10
+		dst[nDst+2] = uint8(r%126 + 0x81)
+		r /= 126
+		dst[nDst+1] = uint8(r%10 + 0x30)
+		r /= 10
+		dst[nDst+0] = uint8(r + 0x81)
+		nDst += 4
+		continue
+	}
+	return nDst, nSrc, err
+}
+
+func init() {
+	// Check that the hard-coded encode switch covers all tables.
+	if numEncodeTables != 5 {
+		panic("bad numEncodeTables")
+	}
+}

vendor/golang.org/x/text/encoding/simplifiedchinese/hzgb2312.go

@@ -0,0 +1,245 @@
+// 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 (
+	"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)
+}
+
+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 {
+			r, size = utf8.RuneError, 1
+			goto write
+		}
+
+		if c0 == '~' {
+			if nSrc+1 >= len(src) {
+				if !atEOF {
+					err = transform.ErrShortSrc
+					break loop
+				}
+				r, size = utf8.RuneError, 1
+				goto write
+			}
+			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:
+				r = utf8.RuneError
+				goto write
+			}
+		}
+
+		if *d == asciiState {
+			r, size = rune(c0), 1
+		} else {
+			if nSrc+1 >= len(src) {
+				if !atEOF {
+					err = transform.ErrShortSrc
+					break loop
+				}
+				r, size = utf8.RuneError, 1
+				goto write
+			}
+			size = 2
+			c1 := src[nSrc+1]
+			if c0 < 0x21 || 0x7e <= c0 || c1 < 0x21 || 0x7f <= c1 {
+				// error
+			} else if i := int(c0-0x01)*190 + int(c1+0x3f); i < len(decode) {
+				r = rune(decode[i])
+				if r != 0 {
+					goto write
+				}
+			}
+			if c1 > utf8.RuneSelf {
+				// Be consistent and always treat non-ASCII as a single error.
+				size = 1
+			}
+			r = utf8.RuneError
+		}
+
+	write:
+		if nDst+utf8.RuneLen(r) > len(dst) {
+			err = transform.ErrShortDst
+			break loop
+		}
+		nDst += utf8.EncodeRune(dst[nDst:], r)
+	}
+	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 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 += 1
+			continue
+
+		}
+
+		// 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
+			}
+		}
+
+	terminateInASCIIState:
+		// Switch back to ASCII state in case of error so that an ASCII
+		// replacement character can be written in the correct state.
+		if *e != asciiState {
+			if nDst+2 > len(dst) {
+				err = transform.ErrShortDst
+				break
+			}
+			dst[nDst+0] = '~'
+			dst[nDst+1] = '}'
+			nDst += 2
+		}
+		err = internal.ErrASCIIReplacement
+		break
+
+	writeGB:
+		c0 := uint8(r>>8) - 0x80
+		c1 := uint8(r) - 0x80
+		if c0 < 0x21 || 0x7e <= c0 || c1 < 0x21 || 0x7f <= c1 {
+			goto terminateInASCIIState
+		}
+		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
+	}
+	// TODO: should one always terminate in ASCII state to make it safe to
+	// concatenate two HZ-GB2312-encoded strings?
+	return nDst, nSrc, err
+}