Mercurial > projects > dil
annotate trunk/src/dil/Unicode.d @ 737:f88b5285b86b
Implemented DDocEmitter.
Fixed quite a few bugs.
| author | Aziz K?ksal <aziz.koeksal@gmail.com> |
|---|---|
| date | Sat, 09 Feb 2008 02:00:20 +0100 |
| parents | ceaac6a24258 |
| children | 49fe21aa387c |
| rev | line source |
|---|---|
| 510 | 1 /++ |
| 2 Author: Aziz Köksal | |
| 3 License: GPL3 | |
| 4 +/ | |
| 5 module dil.Unicode; | |
|
629
d050e211402b
Moved files in src/std/ to src/util/.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
518
diff
changeset
|
6 public import util.uni : isUniAlpha; |
| 510 | 7 |
| 8 /// U+FFFD = �. Used to replace invalid Unicode characters. | |
| 9 const dchar REPLACEMENT_CHAR = '\uFFFD'; | |
| 10 /// Invalid character, returned on errors. | |
| 11 const dchar ERROR_CHAR = 0xD800; | |
| 12 | |
| 13 /++ | |
| 14 Returns true if this character is not a surrogate | |
| 15 code point and not higher than 0x10FFFF. | |
| 16 +/ | |
| 17 bool isValidChar(dchar d) | |
| 18 { | |
| 19 return d < 0xD800 || d > 0xDFFF && d <= 0x10FFFF; | |
| 20 } | |
| 21 | |
| 22 /++ | |
| 23 Returns true if this is one of the | |
| 24 There are a total of 66 noncharacters. | |
| 25 See_also: Chapter 16.7 Noncharacters in Unicode 5.0 | |
| 26 +/ | |
| 27 bool isNoncharacter(dchar d) | |
| 28 { | |
| 29 return 0xFDD0 <= d && d <= 0xFDEF || // 32 | |
| 30 d <= 0x10FFFF && (d & 0xFFFF) >= 0xFFFE; // 34 | |
| 31 } | |
| 32 | |
| 33 /// Returns true if this is a trail byte of a UTF-8 sequence? | |
| 34 bool isTrailByte(ubyte b) | |
| 35 { | |
| 36 return (b & 0xC0) == 0x80; // 10xx_xxxx | |
| 37 } | |
| 38 | |
| 39 /// Returns true if this is a lead byte of a UTF-8 sequence. | |
| 40 bool isLeadByte(ubyte b) | |
| 41 { | |
| 42 return (b & 0xC0) == 0xC0; // 11xx_xxxx | |
| 43 } | |
| 44 | |
|
722
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
45 /// Advances ref_p only if this is a valid Unicode alpha character. |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
46 bool isUnicodeAlpha(ref char* ref_p, char* end) |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
47 in { assert(ref_p && ref_p < end); } |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
48 body |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
49 { |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
50 if (*ref_p < 0x80) |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
51 return false; |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
52 auto p = ref_p; |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
53 auto c = decode(p, end); |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
54 if (!isUniAlpha(c)) |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
55 return false; |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
56 ref_p = p; |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
57 return true; |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
58 } |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
59 |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
60 /// index is set one past the last trail byte of the valid UTF-8 sequence. |
| 510 | 61 dchar decode(char[] str, ref size_t index) |
|
722
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
62 in { assert(str.length && index < str.length); } |
|
737
f88b5285b86b
Implemented DDocEmitter.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
722
diff
changeset
|
63 out(c) { assert(isValidChar(c) || c == ERROR_CHAR); } |
| 510 | 64 body |
| 65 { | |
| 66 char* p = str.ptr + index; | |
| 67 char* end = str.ptr + str.length; | |
|
722
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
68 dchar c = decode(p, end); |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
69 if (c != ERROR_CHAR) |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
70 index = p - str.ptr + 1; |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
71 return c; |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
72 } |
| 510 | 73 |
|
722
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
74 /// ref_p is set to the last trail byte of the valid UTF-8 sequence. |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
75 dchar decode(ref char* ref_p, char* end) |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
76 in { assert(ref_p && ref_p < end); } |
|
737
f88b5285b86b
Implemented DDocEmitter.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
722
diff
changeset
|
77 out(c) { assert(isValidChar(c) || c == ERROR_CHAR); } |
|
722
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
78 body |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
79 { |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
80 char* p = ref_p; |
|
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
81 dchar c = *p; |
| 510 | 82 |
| 83 if (c < 0x80) | |
| 84 { | |
|
722
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
85 ref_p++; |
| 510 | 86 return c; |
| 87 } | |
| 88 | |
|
722
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
89 p++; // Move to second byte. |
| 510 | 90 if (!(p < end)) |
| 91 return ERROR_CHAR; | |
| 92 | |
| 93 // Error if second byte is not a trail byte. | |
| 94 if (!isTrailByte(*p)) | |
| 95 return ERROR_CHAR; | |
| 96 | |
| 97 // Check for overlong sequences. | |
| 98 switch (c) | |
| 99 { | |
| 100 case 0xE0, // 11100000 100xxxxx | |
| 101 0xF0, // 11110000 1000xxxx | |
| 102 0xF8, // 11111000 10000xxx | |
| 103 0xFC: // 11111100 100000xx | |
| 104 if ((*p & c) == 0x80) | |
| 105 return ERROR_CHAR; | |
| 106 default: | |
| 107 if ((c & 0xFE) == 0xC0) // 1100000x | |
| 108 return ERROR_CHAR; | |
| 109 } | |
| 110 | |
|
722
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
111 const char[] checkNextByte = "if (!(++p < end && isTrailByte(*p)))" |
| 510 | 112 " return ERROR_CHAR;"; |
| 113 const char[] appendSixBits = "c = (c << 6) | *p & 0b0011_1111;"; | |
| 114 | |
| 115 // Decode | |
| 116 if ((c & 0b1110_0000) == 0b1100_0000) | |
| 117 { | |
| 118 // 110xxxxx 10xxxxxx | |
| 119 c &= 0b0001_1111; | |
| 120 mixin(appendSixBits); | |
| 121 } | |
| 122 else if ((c & 0b1111_0000) == 0b1110_0000) | |
| 123 { | |
| 124 // 1110xxxx 10xxxxxx 10xxxxxx | |
| 125 c &= 0b0000_1111; | |
| 126 mixin(appendSixBits ~ | |
| 127 checkNextByte ~ appendSixBits); | |
| 128 } | |
| 129 else if ((c & 0b1111_1000) == 0b1111_0000) | |
| 130 { | |
| 131 // 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx | |
| 132 c &= 0b0000_0111; | |
| 133 mixin(appendSixBits ~ | |
| 134 checkNextByte ~ appendSixBits ~ | |
| 135 checkNextByte ~ appendSixBits); | |
| 136 } | |
| 137 else | |
| 138 // 5 and 6 byte UTF-8 sequences are not allowed yet. | |
| 139 // 111110xx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx | |
| 140 // 1111110x 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx | |
| 141 return ERROR_CHAR; | |
| 142 | |
| 143 assert(isTrailByte(*p)); | |
| 144 | |
| 145 if (!isValidChar(c)) | |
| 146 return ERROR_CHAR; | |
|
722
ceaac6a24258
Added isUnicodeAlpha() for DDocParser and MacroParser.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
629
diff
changeset
|
147 ref_p = p; |
| 510 | 148 return c; |
| 149 } | |
| 150 | |
| 151 /// Encodes a character and appends it to str. | |
|
518
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
152 void encode(ref char[] str, dchar c) |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
153 { |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
154 assert(isValidChar(c), "check if character is valid before calling encode()."); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
155 |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
156 char[6] b = void; |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
157 if (c < 0x80) |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
158 str ~= c; |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
159 if (c < 0x800) |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
160 { |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
161 b[0] = 0xC0 | (c >> 6); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
162 b[1] = 0x80 | (c & 0x3F); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
163 str ~= b[0..2]; |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
164 } |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
165 else if (c < 0x10000) |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
166 { |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
167 b[0] = 0xE0 | (c >> 12); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
168 b[1] = 0x80 | ((c >> 6) & 0x3F); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
169 b[2] = 0x80 | (c & 0x3F); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
170 str ~= b[0..3]; |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
171 } |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
172 else if (c < 0x200000) |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
173 { |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
174 b[0] = 0xF0 | (c >> 18); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
175 b[1] = 0x80 | ((c >> 12) & 0x3F); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
176 b[2] = 0x80 | ((c >> 6) & 0x3F); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
177 b[3] = 0x80 | (c & 0x3F); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
178 str ~= b[0..4]; |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
179 } |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
180 /+ // There are no 5 and 6 byte UTF-8 sequences yet. |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
181 else if (c < 0x4000000) |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
182 { |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
183 b[0] = 0xF8 | (c >> 24); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
184 b[1] = 0x80 | ((c >> 18) & 0x3F); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
185 b[2] = 0x80 | ((c >> 12) & 0x3F); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
186 b[3] = 0x80 | ((c >> 6) & 0x3F); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
187 b[4] = 0x80 | (c & 0x3F); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
188 str ~= b[0..5]; |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
189 } |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
190 else if (c < 0x80000000) |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
191 { |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
192 b[0] = 0xFC | (c >> 30); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
193 b[1] = 0x80 | ((c >> 24) & 0x3F); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
194 b[2] = 0x80 | ((c >> 18) & 0x3F); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
195 b[3] = 0x80 | ((c >> 12) & 0x3F); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
196 b[4] = 0x80 | ((c >> 6) & 0x3F); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
197 b[5] = 0x80 | (c & 0x3F); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
198 str ~= b[0..6]; |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
199 } |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
200 +/ |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
201 else |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
202 assert(0); |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
203 } |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
204 |
|
8f86bb9ef715
Added module dil.Converter and dil.FileBOM.
Aziz K?ksal <aziz.koeksal@gmail.com>
parents:
510
diff
changeset
|
205 /// Encodes a character and appends it to str. |
| 510 | 206 void encode(ref wchar[] str, dchar c) |
| 207 in { assert(isValidChar(c)); } | |
| 208 body | |
| 209 { | |
| 210 if (c < 0x10000) | |
| 211 str ~= cast(wchar)c; | |
| 212 else | |
| 213 { | |
| 214 // Encode with surrogate pair. | |
| 215 wchar[2] pair = void; | |
| 216 c -= 0x10000; // c' | |
| 217 // higher10bits(c') | 0b1101_10xx_xxxx_xxxx | |
| 218 pair[0] = (c >> 10) | 0xD800; | |
| 219 // lower10bits(c') | 0b1101_11yy_yyyy_yyyy | |
| 220 pair[1] = (c & 0x3FF) | 0xDC00; | |
| 221 str ~= pair; | |
| 222 } | |
| 223 } | |
| 224 | |
| 225 /++ | |
| 226 Returns a decoded character from a UTF-16 sequence. | |
| 227 In case of an error in the sequence 0xD800 is returned. | |
| 228 Params: | |
| 229 str = the UTF-16 sequence. | |
| 230 index = where to start from. | |
| 231 +/ | |
| 232 dchar decode(wchar[] str, ref size_t index) | |
| 233 { | |
| 234 assert(str.length && index < str.length); | |
| 235 dchar c = str[index]; | |
| 236 if (0xD800 > c || c > 0xDFFF) | |
| 237 { | |
| 238 ++index; | |
| 239 return c; | |
| 240 } | |
| 241 if (c <= 0xDBFF && index+1 != str.length) | |
| 242 { | |
| 243 wchar c2 = str[index+1]; | |
| 244 if (0xDC00 <= c2 && c2 <= 0xDFFF) | |
| 245 { | |
| 246 // (c - 0xD800) << 10 + 0x10000 -> | |
| 247 // (c - 0xD800 + 0x40) << 10 -> | |
| 248 c = (c - 0xD7C0) << 10; | |
| 249 c |= (c2 & 0x3FF); | |
| 250 index += 2; | |
| 251 return c; | |
| 252 } | |
| 253 } | |
| 254 return ERROR_CHAR; | |
| 255 } | |
| 256 | |
| 257 /++ | |
| 258 Returns a decoded character from a UTF-16 sequence. | |
| 259 In case of an error in the sequence 0xD800 is returned. | |
| 260 Params: | |
| 261 p = start of the UTF-16 sequence. | |
| 262 end = one past the end of the sequence. | |
| 263 +/ | |
| 264 dchar decode(ref wchar* p, wchar* end) | |
| 265 { | |
| 266 assert(p && p < end); | |
| 267 dchar c = *p; | |
| 268 if (0xD800 > c || c > 0xDFFF) | |
| 269 { | |
| 270 ++p; | |
| 271 return c; | |
| 272 } | |
| 273 if (c <= 0xDBFF && p+1 != end) | |
| 274 { | |
| 275 wchar c2 = p[1]; | |
| 276 if (0xDC00 <= c2 && c2 <= 0xDFFF) | |
| 277 { | |
| 278 c = (c - 0xD7C0) << 10; | |
| 279 c |= (c2 & 0x3FF); | |
| 280 p += 2; | |
| 281 return c; | |
| 282 } | |
| 283 } | |
| 284 return ERROR_CHAR; | |
| 285 } | |
| 286 | |
| 287 /// Decode a character from a zero-terminated string. | |
| 288 dchar decode(ref wchar* p) | |
| 289 { | |
| 290 assert(p); | |
| 291 dchar c = *p; | |
| 292 if (0xD800 > c || c > 0xDFFF) | |
| 293 { | |
| 294 ++p; | |
| 295 return c; | |
| 296 } | |
| 297 if (c <= 0xDBFF) | |
| 298 { | |
| 299 wchar c2 = p[1]; | |
| 300 if (0xDC00 <= c2 && c2 <= 0xDFFF) | |
| 301 { | |
| 302 c = (c - 0xD7C0) << 10; | |
| 303 c |= (c2 & 0x3FF); | |
| 304 p += 2; | |
| 305 return c; | |
| 306 } | |
| 307 } | |
| 308 return ERROR_CHAR; | |
| 309 } |