| 1 | /* Low-level bidirectional buffer-scanning functions for GNU Emacs. |
| 2 | Copyright (C) 2000, 2001, 2004, 2005, 2009, 2010 |
| 3 | Free Software Foundation, Inc. |
| 4 | |
| 5 | This file is part of GNU Emacs. |
| 6 | |
| 7 | GNU Emacs is free software: you can redistribute it and/or modify |
| 8 | it under the terms of the GNU General Public License as published by |
| 9 | the Free Software Foundation, either version 3 of the License, or |
| 10 | (at your option) any later version. |
| 11 | |
| 12 | GNU Emacs is distributed in the hope that it will be useful, |
| 13 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 14 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 15 | GNU General Public License for more details. |
| 16 | |
| 17 | You should have received a copy of the GNU General Public License |
| 18 | along with GNU Emacs. If not, see <http://www.gnu.org/licenses/>. */ |
| 19 | |
| 20 | /* Written by Eli Zaretskii <eliz@gnu.org>. |
| 21 | |
| 22 | A sequential implementation of the Unicode Bidirectional algorithm, |
| 23 | as per UAX#9, a part of the Unicode Standard. |
| 24 | |
| 25 | Unlike the reference and most other implementations, this one is |
| 26 | designed to be called once for every character in the buffer or |
| 27 | string. |
| 28 | |
| 29 | The main entry point is bidi_move_to_visually_next. Each time it |
| 30 | is called, it finds the next character in the visual order, and |
| 31 | returns its information in a special structure. The caller is then |
| 32 | expected to process this character for display or any other |
| 33 | purposes, and call bidi_move_to_visually_next for the next |
| 34 | character. See the comments in bidi_move_to_visually_next for more |
| 35 | details about its algorithm that finds the next visual-order |
| 36 | character by resolving their levels on the fly. |
| 37 | |
| 38 | The two other entry points are bidi_paragraph_init and |
| 39 | bidi_mirror_char. The first determines the base direction of a |
| 40 | paragraph, while the second returns the mirrored version of its |
| 41 | argument character. |
| 42 | |
| 43 | If you want to understand the code, you will have to read it |
| 44 | together with the relevant portions of UAX#9. The comments include |
| 45 | references to UAX#9 rules, for that very reason. |
| 46 | |
| 47 | A note about references to UAX#9 rules: if the reference says |
| 48 | something like "X9/Retaining", it means that you need to refer to |
| 49 | rule X9 and to its modifications decribed in the "Implementation |
| 50 | Notes" section of UAX#9, under "Retaining Format Codes". */ |
| 51 | |
| 52 | #ifdef HAVE_CONFIG_H |
| 53 | #include <config.h> |
| 54 | #endif |
| 55 | |
| 56 | #include <stdio.h> |
| 57 | |
| 58 | #ifdef HAVE_STRING_H |
| 59 | #include <string.h> |
| 60 | #endif |
| 61 | |
| 62 | #include <setjmp.h> |
| 63 | |
| 64 | #include "lisp.h" |
| 65 | #include "buffer.h" |
| 66 | #include "character.h" |
| 67 | #include "dispextern.h" |
| 68 | |
| 69 | static int bidi_initialized = 0; |
| 70 | |
| 71 | static Lisp_Object bidi_type_table; |
| 72 | |
| 73 | /* FIXME: Remove these when bidi_explicit_dir_char uses a lookup table. */ |
| 74 | #define LRM_CHAR 0x200E |
| 75 | #define RLM_CHAR 0x200F |
| 76 | #define LRE_CHAR 0x202A |
| 77 | #define RLE_CHAR 0x202B |
| 78 | #define PDF_CHAR 0x202C |
| 79 | #define LRO_CHAR 0x202D |
| 80 | #define RLO_CHAR 0x202E |
| 81 | |
| 82 | #define BIDI_EOB -1 |
| 83 | #define BIDI_BOB -2 /* FIXME: Is this needed? */ |
| 84 | |
| 85 | /* Local data structures. (Look in dispextern.h for the rest.) */ |
| 86 | |
| 87 | /* What we need to know about the current paragraph. */ |
| 88 | struct bidi_paragraph_info { |
| 89 | int start_bytepos; /* byte position where it begins */ |
| 90 | int end_bytepos; /* byte position where it ends */ |
| 91 | int embedding_level; /* its basic embedding level */ |
| 92 | bidi_dir_t base_dir; /* its base direction */ |
| 93 | }; |
| 94 | |
| 95 | /* Data type for describing the bidirectional character categories. */ |
| 96 | typedef enum { |
| 97 | UNKNOWN_BC, |
| 98 | NEUTRAL, |
| 99 | WEAK, |
| 100 | STRONG |
| 101 | } bidi_category_t; |
| 102 | |
| 103 | int bidi_ignore_explicit_marks_for_paragraph_level = 1; |
| 104 | |
| 105 | static Lisp_Object paragraph_start_re, paragraph_separate_re; |
| 106 | static Lisp_Object Qparagraph_start, Qparagraph_separate; |
| 107 | |
| 108 | static void |
| 109 | bidi_initialize () |
| 110 | { |
| 111 | /* FIXME: This should come from the Unicode Database. */ |
| 112 | struct { |
| 113 | int from, to; |
| 114 | bidi_type_t type; |
| 115 | } bidi_type[] = |
| 116 | { { 0x0000, 0x0008, WEAK_BN }, |
| 117 | { 0x0009, 0x0000, NEUTRAL_S }, |
| 118 | { 0x000A, 0x0000, NEUTRAL_B }, |
| 119 | { 0x000B, 0x0000, NEUTRAL_S }, |
| 120 | { 0x000C, 0x0000, NEUTRAL_WS }, |
| 121 | { 0x000D, 0x0000, NEUTRAL_B }, |
| 122 | { 0x000E, 0x001B, WEAK_BN }, |
| 123 | { 0x001C, 0x001E, NEUTRAL_B }, |
| 124 | { 0x001F, 0x0000, NEUTRAL_S }, |
| 125 | { 0x0020, 0x0000, NEUTRAL_WS }, |
| 126 | { 0x0021, 0x0022, NEUTRAL_ON }, |
| 127 | { 0x0023, 0x0025, WEAK_ET }, |
| 128 | { 0x0026, 0x002A, NEUTRAL_ON }, |
| 129 | { 0x002B, 0x0000, WEAK_ES }, |
| 130 | { 0x002C, 0x0000, WEAK_CS }, |
| 131 | { 0x002D, 0x0000, WEAK_ES }, |
| 132 | { 0x002E, 0x002F, WEAK_CS }, |
| 133 | { 0x0030, 0x0039, WEAK_EN }, |
| 134 | { 0x003A, 0x0000, WEAK_CS }, |
| 135 | { 0x003B, 0x0040, NEUTRAL_ON }, |
| 136 | { 0x005B, 0x0060, NEUTRAL_ON }, |
| 137 | { 0x007B, 0x007E, NEUTRAL_ON }, |
| 138 | { 0x007F, 0x0084, WEAK_BN }, |
| 139 | { 0x0085, 0x0000, NEUTRAL_B }, |
| 140 | { 0x0086, 0x009F, WEAK_BN }, |
| 141 | { 0x00A0, 0x0000, WEAK_CS }, |
| 142 | { 0x00A1, 0x0000, NEUTRAL_ON }, |
| 143 | { 0x00A2, 0x00A5, WEAK_ET }, |
| 144 | { 0x00A6, 0x00A9, NEUTRAL_ON }, |
| 145 | { 0x00AB, 0x00AC, NEUTRAL_ON }, |
| 146 | { 0x00AD, 0x0000, WEAK_BN }, |
| 147 | { 0x00AE, 0x00Af, NEUTRAL_ON }, |
| 148 | { 0x00B0, 0x00B1, WEAK_ET }, |
| 149 | { 0x00B2, 0x00B3, WEAK_EN }, |
| 150 | { 0x00B4, 0x0000, NEUTRAL_ON }, |
| 151 | { 0x00B6, 0x00B8, NEUTRAL_ON }, |
| 152 | { 0x00B9, 0x0000, WEAK_EN }, |
| 153 | { 0x00BB, 0x00BF, NEUTRAL_ON }, |
| 154 | { 0x00D7, 0x0000, NEUTRAL_ON }, |
| 155 | { 0x00F7, 0x0000, NEUTRAL_ON }, |
| 156 | { 0x02B9, 0x02BA, NEUTRAL_ON }, |
| 157 | { 0x02C2, 0x02CF, NEUTRAL_ON }, |
| 158 | { 0x02D2, 0x02DF, NEUTRAL_ON }, |
| 159 | { 0x02E5, 0x02ED, NEUTRAL_ON }, |
| 160 | { 0x0300, 0x036F, WEAK_NSM }, |
| 161 | { 0x0374, 0x0375, NEUTRAL_ON }, |
| 162 | { 0x037E, 0x0385, NEUTRAL_ON }, |
| 163 | { 0x0387, 0x0000, NEUTRAL_ON }, |
| 164 | { 0x03F6, 0x0000, NEUTRAL_ON }, |
| 165 | { 0x0483, 0x0489, WEAK_NSM }, |
| 166 | { 0x058A, 0x0000, NEUTRAL_ON }, |
| 167 | { 0x0591, 0x05BD, WEAK_NSM }, |
| 168 | { 0x05BE, 0x0000, STRONG_R }, |
| 169 | { 0x05BF, 0x0000, WEAK_NSM }, |
| 170 | { 0x05C0, 0x0000, STRONG_R }, |
| 171 | { 0x05C1, 0x05C2, WEAK_NSM }, |
| 172 | { 0x05C3, 0x0000, STRONG_R }, |
| 173 | { 0x05C4, 0x05C5, WEAK_NSM }, |
| 174 | { 0x05C6, 0x0000, STRONG_R }, |
| 175 | { 0x05C7, 0x0000, WEAK_NSM }, |
| 176 | { 0x05D0, 0x05F4, STRONG_R }, |
| 177 | { 0x060C, 0x0000, WEAK_CS }, |
| 178 | { 0x061B, 0x064A, STRONG_AL }, |
| 179 | { 0x064B, 0x0655, WEAK_NSM }, |
| 180 | { 0x0660, 0x0669, WEAK_AN }, |
| 181 | { 0x066A, 0x0000, WEAK_ET }, |
| 182 | { 0x066B, 0x066C, WEAK_AN }, |
| 183 | { 0x066D, 0x066F, STRONG_AL }, |
| 184 | { 0x0670, 0x0000, WEAK_NSM }, |
| 185 | { 0x0671, 0x06D5, STRONG_AL }, |
| 186 | { 0x06D6, 0x06DC, WEAK_NSM }, |
| 187 | { 0x06DD, 0x0000, STRONG_AL }, |
| 188 | { 0x06DE, 0x06E4, WEAK_NSM }, |
| 189 | { 0x06E5, 0x06E6, STRONG_AL }, |
| 190 | { 0x06E7, 0x06E8, WEAK_NSM }, |
| 191 | { 0x06E9, 0x0000, NEUTRAL_ON }, |
| 192 | { 0x06EA, 0x06ED, WEAK_NSM }, |
| 193 | { 0x06F0, 0x06F9, WEAK_EN }, |
| 194 | { 0x06FA, 0x070D, STRONG_AL }, |
| 195 | { 0x070F, 0x0000, WEAK_BN }, |
| 196 | { 0x0710, 0x0000, STRONG_AL }, |
| 197 | { 0x0711, 0x0000, WEAK_NSM }, |
| 198 | { 0x0712, 0x072C, STRONG_AL }, |
| 199 | { 0x0730, 0x074A, WEAK_NSM }, |
| 200 | { 0x0780, 0x07A5, STRONG_AL }, |
| 201 | { 0x07A6, 0x07B0, WEAK_NSM }, |
| 202 | { 0x07B1, 0x0000, STRONG_AL }, |
| 203 | { 0x0901, 0x0902, WEAK_NSM }, |
| 204 | { 0x093C, 0x0000, WEAK_NSM }, |
| 205 | { 0x0941, 0x0948, WEAK_NSM }, |
| 206 | { 0x094D, 0x0000, WEAK_NSM }, |
| 207 | { 0x0951, 0x0954, WEAK_NSM }, |
| 208 | { 0x0962, 0x0963, WEAK_NSM }, |
| 209 | { 0x0981, 0x0000, WEAK_NSM }, |
| 210 | { 0x09BC, 0x0000, WEAK_NSM }, |
| 211 | { 0x09C1, 0x09C4, WEAK_NSM }, |
| 212 | { 0x09CD, 0x0000, WEAK_NSM }, |
| 213 | { 0x09E2, 0x09E3, WEAK_NSM }, |
| 214 | { 0x09F2, 0x09F3, WEAK_ET }, |
| 215 | { 0x0A02, 0x0000, WEAK_NSM }, |
| 216 | { 0x0A3C, 0x0000, WEAK_NSM }, |
| 217 | { 0x0A41, 0x0A4D, WEAK_NSM }, |
| 218 | { 0x0A70, 0x0A71, WEAK_NSM }, |
| 219 | { 0x0A81, 0x0A82, WEAK_NSM }, |
| 220 | { 0x0ABC, 0x0000, WEAK_NSM }, |
| 221 | { 0x0AC1, 0x0AC8, WEAK_NSM }, |
| 222 | { 0x0ACD, 0x0000, WEAK_NSM }, |
| 223 | { 0x0B01, 0x0000, WEAK_NSM }, |
| 224 | { 0x0B3C, 0x0000, WEAK_NSM }, |
| 225 | { 0x0B3F, 0x0000, WEAK_NSM }, |
| 226 | { 0x0B41, 0x0B43, WEAK_NSM }, |
| 227 | { 0x0B4D, 0x0B56, WEAK_NSM }, |
| 228 | { 0x0B82, 0x0000, WEAK_NSM }, |
| 229 | { 0x0BC0, 0x0000, WEAK_NSM }, |
| 230 | { 0x0BCD, 0x0000, WEAK_NSM }, |
| 231 | { 0x0C3E, 0x0C40, WEAK_NSM }, |
| 232 | { 0x0C46, 0x0C56, WEAK_NSM }, |
| 233 | { 0x0CBF, 0x0000, WEAK_NSM }, |
| 234 | { 0x0CC6, 0x0000, WEAK_NSM }, |
| 235 | { 0x0CCC, 0x0CCD, WEAK_NSM }, |
| 236 | { 0x0D41, 0x0D43, WEAK_NSM }, |
| 237 | { 0x0D4D, 0x0000, WEAK_NSM }, |
| 238 | { 0x0DCA, 0x0000, WEAK_NSM }, |
| 239 | { 0x0DD2, 0x0DD6, WEAK_NSM }, |
| 240 | { 0x0E31, 0x0000, WEAK_NSM }, |
| 241 | { 0x0E34, 0x0E3A, WEAK_NSM }, |
| 242 | { 0x0E3F, 0x0000, WEAK_ET }, |
| 243 | { 0x0E47, 0x0E4E, WEAK_NSM }, |
| 244 | { 0x0EB1, 0x0000, WEAK_NSM }, |
| 245 | { 0x0EB4, 0x0EBC, WEAK_NSM }, |
| 246 | { 0x0EC8, 0x0ECD, WEAK_NSM }, |
| 247 | { 0x0F18, 0x0F19, WEAK_NSM }, |
| 248 | { 0x0F35, 0x0000, WEAK_NSM }, |
| 249 | { 0x0F37, 0x0000, WEAK_NSM }, |
| 250 | { 0x0F39, 0x0000, WEAK_NSM }, |
| 251 | { 0x0F3A, 0x0F3D, NEUTRAL_ON }, |
| 252 | { 0x0F71, 0x0F7E, WEAK_NSM }, |
| 253 | { 0x0F80, 0x0F84, WEAK_NSM }, |
| 254 | { 0x0F86, 0x0F87, WEAK_NSM }, |
| 255 | { 0x0F90, 0x0FBC, WEAK_NSM }, |
| 256 | { 0x0FC6, 0x0000, WEAK_NSM }, |
| 257 | { 0x102D, 0x1030, WEAK_NSM }, |
| 258 | { 0x1032, 0x1037, WEAK_NSM }, |
| 259 | { 0x1039, 0x0000, WEAK_NSM }, |
| 260 | { 0x1058, 0x1059, WEAK_NSM }, |
| 261 | { 0x1680, 0x0000, NEUTRAL_WS }, |
| 262 | { 0x169B, 0x169C, NEUTRAL_ON }, |
| 263 | { 0x1712, 0x1714, WEAK_NSM }, |
| 264 | { 0x1732, 0x1734, WEAK_NSM }, |
| 265 | { 0x1752, 0x1753, WEAK_NSM }, |
| 266 | { 0x1772, 0x1773, WEAK_NSM }, |
| 267 | { 0x17B7, 0x17BD, WEAK_NSM }, |
| 268 | { 0x17C6, 0x0000, WEAK_NSM }, |
| 269 | { 0x17C9, 0x17D3, WEAK_NSM }, |
| 270 | { 0x17DB, 0x0000, WEAK_ET }, |
| 271 | { 0x1800, 0x180A, NEUTRAL_ON }, |
| 272 | { 0x180B, 0x180D, WEAK_NSM }, |
| 273 | { 0x180E, 0x0000, WEAK_BN }, |
| 274 | { 0x18A9, 0x0000, WEAK_NSM }, |
| 275 | { 0x1FBD, 0x0000, NEUTRAL_ON }, |
| 276 | { 0x1FBF, 0x1FC1, NEUTRAL_ON }, |
| 277 | { 0x1FCD, 0x1FCF, NEUTRAL_ON }, |
| 278 | { 0x1FDD, 0x1FDF, NEUTRAL_ON }, |
| 279 | { 0x1FED, 0x1FEF, NEUTRAL_ON }, |
| 280 | { 0x1FFD, 0x1FFE, NEUTRAL_ON }, |
| 281 | { 0x2000, 0x200A, NEUTRAL_WS }, |
| 282 | { 0x200B, 0x200D, WEAK_BN }, |
| 283 | { 0x200F, 0x0000, STRONG_R }, |
| 284 | { 0x2010, 0x2027, NEUTRAL_ON }, |
| 285 | { 0x2028, 0x0000, NEUTRAL_WS }, |
| 286 | { 0x2029, 0x0000, NEUTRAL_B }, |
| 287 | { 0x202A, 0x0000, LRE }, |
| 288 | { 0x202B, 0x0000, RLE }, |
| 289 | { 0x202C, 0x0000, PDF }, |
| 290 | { 0x202D, 0x0000, LRO }, |
| 291 | { 0x202E, 0x0000, RLO }, |
| 292 | { 0x202F, 0x0000, NEUTRAL_WS }, |
| 293 | { 0x2030, 0x2034, WEAK_ET }, |
| 294 | { 0x2035, 0x2057, NEUTRAL_ON }, |
| 295 | { 0x205F, 0x0000, NEUTRAL_WS }, |
| 296 | { 0x2060, 0x206F, WEAK_BN }, |
| 297 | { 0x2070, 0x0000, WEAK_EN }, |
| 298 | { 0x2074, 0x2079, WEAK_EN }, |
| 299 | { 0x207A, 0x207B, WEAK_ET }, |
| 300 | { 0x207C, 0x207E, NEUTRAL_ON }, |
| 301 | { 0x2080, 0x2089, WEAK_EN }, |
| 302 | { 0x208A, 0x208B, WEAK_ET }, |
| 303 | { 0x208C, 0x208E, NEUTRAL_ON }, |
| 304 | { 0x20A0, 0x20B1, WEAK_ET }, |
| 305 | { 0x20D0, 0x20EA, WEAK_NSM }, |
| 306 | { 0x2100, 0x2101, NEUTRAL_ON }, |
| 307 | { 0x2103, 0x2106, NEUTRAL_ON }, |
| 308 | { 0x2108, 0x2109, NEUTRAL_ON }, |
| 309 | { 0x2114, 0x0000, NEUTRAL_ON }, |
| 310 | { 0x2116, 0x2118, NEUTRAL_ON }, |
| 311 | { 0x211E, 0x2123, NEUTRAL_ON }, |
| 312 | { 0x2125, 0x0000, NEUTRAL_ON }, |
| 313 | { 0x2127, 0x0000, NEUTRAL_ON }, |
| 314 | { 0x2129, 0x0000, NEUTRAL_ON }, |
| 315 | { 0x212E, 0x0000, WEAK_ET }, |
| 316 | { 0x2132, 0x0000, NEUTRAL_ON }, |
| 317 | { 0x213A, 0x0000, NEUTRAL_ON }, |
| 318 | { 0x2140, 0x2144, NEUTRAL_ON }, |
| 319 | { 0x214A, 0x215F, NEUTRAL_ON }, |
| 320 | { 0x2190, 0x2211, NEUTRAL_ON }, |
| 321 | { 0x2212, 0x2213, WEAK_ET }, |
| 322 | { 0x2214, 0x2335, NEUTRAL_ON }, |
| 323 | { 0x237B, 0x2394, NEUTRAL_ON }, |
| 324 | { 0x2396, 0x244A, NEUTRAL_ON }, |
| 325 | { 0x2460, 0x249B, WEAK_EN }, |
| 326 | { 0x24EA, 0x0000, WEAK_EN }, |
| 327 | { 0x24EB, 0x2FFB, NEUTRAL_ON }, |
| 328 | { 0x3000, 0x0000, NEUTRAL_WS }, |
| 329 | { 0x3001, 0x3004, NEUTRAL_ON }, |
| 330 | { 0x3008, 0x3020, NEUTRAL_ON }, |
| 331 | { 0x302A, 0x302F, WEAK_NSM }, |
| 332 | { 0x3030, 0x0000, NEUTRAL_ON }, |
| 333 | { 0x3036, 0x3037, NEUTRAL_ON }, |
| 334 | { 0x303D, 0x303F, NEUTRAL_ON }, |
| 335 | { 0x3099, 0x309A, WEAK_NSM }, |
| 336 | { 0x309B, 0x309C, NEUTRAL_ON }, |
| 337 | { 0x30A0, 0x0000, NEUTRAL_ON }, |
| 338 | { 0x30FB, 0x0000, NEUTRAL_ON }, |
| 339 | { 0x3251, 0x325F, NEUTRAL_ON }, |
| 340 | { 0x32B1, 0x32BF, NEUTRAL_ON }, |
| 341 | { 0xA490, 0xA4C6, NEUTRAL_ON }, |
| 342 | { 0xFB1D, 0x0000, STRONG_R }, |
| 343 | { 0xFB1E, 0x0000, WEAK_NSM }, |
| 344 | { 0xFB1F, 0xFB28, STRONG_R }, |
| 345 | { 0xFB29, 0x0000, WEAK_ET }, |
| 346 | { 0xFB2A, 0xFB4F, STRONG_R }, |
| 347 | { 0xFB50, 0xFD3D, STRONG_AL }, |
| 348 | { 0xFD3E, 0xFD3F, NEUTRAL_ON }, |
| 349 | { 0xFD50, 0xFDFC, STRONG_AL }, |
| 350 | { 0xFE00, 0xFE23, WEAK_NSM }, |
| 351 | { 0xFE30, 0xFE4F, NEUTRAL_ON }, |
| 352 | { 0xFE50, 0x0000, WEAK_CS }, |
| 353 | { 0xFE51, 0x0000, NEUTRAL_ON }, |
| 354 | { 0xFE52, 0x0000, WEAK_CS }, |
| 355 | { 0xFE54, 0x0000, NEUTRAL_ON }, |
| 356 | { 0xFE55, 0x0000, WEAK_CS }, |
| 357 | { 0xFE56, 0xFE5E, NEUTRAL_ON }, |
| 358 | { 0xFE5F, 0x0000, WEAK_ET }, |
| 359 | { 0xFE60, 0xFE61, NEUTRAL_ON }, |
| 360 | { 0xFE62, 0xFE63, WEAK_ET }, |
| 361 | { 0xFE64, 0xFE68, NEUTRAL_ON }, |
| 362 | { 0xFE69, 0xFE6A, WEAK_ET }, |
| 363 | { 0xFE6B, 0x0000, NEUTRAL_ON }, |
| 364 | { 0xFE70, 0xFEFC, STRONG_AL }, |
| 365 | { 0xFEFF, 0x0000, WEAK_BN }, |
| 366 | { 0xFF01, 0xFF02, NEUTRAL_ON }, |
| 367 | { 0xFF03, 0xFF05, WEAK_ET }, |
| 368 | { 0xFF06, 0xFF0A, NEUTRAL_ON }, |
| 369 | { 0xFF0B, 0x0000, WEAK_ET }, |
| 370 | { 0xFF0C, 0x0000, WEAK_CS }, |
| 371 | { 0xFF0D, 0x0000, WEAK_ET }, |
| 372 | { 0xFF0E, 0x0000, WEAK_CS }, |
| 373 | { 0xFF0F, 0x0000, WEAK_ES }, |
| 374 | { 0xFF10, 0xFF19, WEAK_EN }, |
| 375 | { 0xFF1A, 0x0000, WEAK_CS }, |
| 376 | { 0xFF1B, 0xFF20, NEUTRAL_ON }, |
| 377 | { 0xFF3B, 0xFF40, NEUTRAL_ON }, |
| 378 | { 0xFF5B, 0xFF65, NEUTRAL_ON }, |
| 379 | { 0xFFE0, 0xFFE1, WEAK_ET }, |
| 380 | { 0xFFE2, 0xFFE4, NEUTRAL_ON }, |
| 381 | { 0xFFE5, 0xFFE6, WEAK_ET }, |
| 382 | { 0xFFE8, 0xFFEE, NEUTRAL_ON }, |
| 383 | { 0xFFF9, 0xFFFB, WEAK_BN }, |
| 384 | { 0xFFFC, 0xFFFD, NEUTRAL_ON }, |
| 385 | { 0x1D167, 0x1D169, WEAK_NSM }, |
| 386 | { 0x1D173, 0x1D17A, WEAK_BN }, |
| 387 | { 0x1D17B, 0x1D182, WEAK_NSM }, |
| 388 | { 0x1D185, 0x1D18B, WEAK_NSM }, |
| 389 | { 0x1D1AA, 0x1D1AD, WEAK_NSM }, |
| 390 | { 0x1D7CE, 0x1D7FF, WEAK_EN }, |
| 391 | { 0xE0001, 0xE007F, WEAK_BN } }; |
| 392 | int i; |
| 393 | |
| 394 | bidi_type_table = Fmake_char_table (Qnil, make_number (STRONG_L)); |
| 395 | staticpro (&bidi_type_table); |
| 396 | |
| 397 | for (i = 0; i < sizeof bidi_type / sizeof bidi_type[0]; i++) |
| 398 | char_table_set_range (bidi_type_table, bidi_type[i].from, |
| 399 | bidi_type[i].to ? bidi_type[i].to : bidi_type[i].from, |
| 400 | make_number (bidi_type[i].type)); |
| 401 | |
| 402 | Qparagraph_start = intern ("paragraph-start"); |
| 403 | staticpro (&Qparagraph_start); |
| 404 | paragraph_start_re = Fsymbol_value (Qparagraph_start); |
| 405 | if (!STRINGP (paragraph_start_re)) |
| 406 | paragraph_start_re = build_string ("\f\\|[ \t]*$"); |
| 407 | staticpro (¶graph_start_re); |
| 408 | Qparagraph_separate = intern ("paragraph-separate"); |
| 409 | staticpro (&Qparagraph_separate); |
| 410 | paragraph_separate_re = Fsymbol_value (Qparagraph_separate); |
| 411 | if (!STRINGP (paragraph_separate_re)) |
| 412 | paragraph_separate_re = build_string ("[ \t\f]*$"); |
| 413 | staticpro (¶graph_separate_re); |
| 414 | bidi_initialized = 1; |
| 415 | } |
| 416 | |
| 417 | /* Return the bidi type of a character CH, subject to the current |
| 418 | directional OVERRIDE. */ |
| 419 | static INLINE bidi_type_t |
| 420 | bidi_get_type (int ch, bidi_dir_t override) |
| 421 | { |
| 422 | bidi_type_t default_type; |
| 423 | |
| 424 | if (ch == BIDI_EOB) |
| 425 | return NEUTRAL_B; |
| 426 | if (ch < 0 || ch > MAX_CHAR) |
| 427 | abort (); |
| 428 | |
| 429 | default_type = (bidi_type_t) XINT (CHAR_TABLE_REF (bidi_type_table, ch)); |
| 430 | |
| 431 | if (override == NEUTRAL_DIR) |
| 432 | return default_type; |
| 433 | |
| 434 | switch (default_type) |
| 435 | { |
| 436 | /* Although UAX#9 does not tell, it doesn't make sense to |
| 437 | override NEUTRAL_B and LRM/RLM characters. */ |
| 438 | case NEUTRAL_B: |
| 439 | case LRE: |
| 440 | case LRO: |
| 441 | case RLE: |
| 442 | case RLO: |
| 443 | case PDF: |
| 444 | return default_type; |
| 445 | default: |
| 446 | switch (ch) |
| 447 | { |
| 448 | case LRM_CHAR: |
| 449 | case RLM_CHAR: |
| 450 | return default_type; |
| 451 | default: |
| 452 | if (override == L2R) /* X6 */ |
| 453 | return STRONG_L; |
| 454 | else if (override == R2L) |
| 455 | return STRONG_R; |
| 456 | else |
| 457 | abort (); /* can't happen: handled above */ |
| 458 | } |
| 459 | } |
| 460 | } |
| 461 | |
| 462 | void |
| 463 | bidi_check_type (bidi_type_t type) |
| 464 | { |
| 465 | if (type < UNKNOWN_BT || type > NEUTRAL_ON) |
| 466 | abort (); |
| 467 | } |
| 468 | |
| 469 | /* Given a bidi TYPE of a character, return its category. */ |
| 470 | static INLINE bidi_category_t |
| 471 | bidi_get_category (bidi_type_t type) |
| 472 | { |
| 473 | switch (type) |
| 474 | { |
| 475 | case UNKNOWN_BT: |
| 476 | return UNKNOWN_BC; |
| 477 | case STRONG_L: |
| 478 | case STRONG_R: |
| 479 | case STRONG_AL: |
| 480 | case LRE: |
| 481 | case LRO: |
| 482 | case RLE: |
| 483 | case RLO: |
| 484 | return STRONG; |
| 485 | case PDF: /* ??? really?? */ |
| 486 | case WEAK_EN: |
| 487 | case WEAK_ES: |
| 488 | case WEAK_ET: |
| 489 | case WEAK_AN: |
| 490 | case WEAK_CS: |
| 491 | case WEAK_NSM: |
| 492 | case WEAK_BN: |
| 493 | return WEAK; |
| 494 | case NEUTRAL_B: |
| 495 | case NEUTRAL_S: |
| 496 | case NEUTRAL_WS: |
| 497 | case NEUTRAL_ON: |
| 498 | return NEUTRAL; |
| 499 | default: |
| 500 | abort (); |
| 501 | } |
| 502 | } |
| 503 | |
| 504 | /* Return the mirrored character of C, if any. |
| 505 | |
| 506 | Note: The conditions in UAX#9 clause L4 must be tested by the |
| 507 | caller. */ |
| 508 | /* FIXME: exceedingly temporary! Should consult the Unicode database |
| 509 | of character properties. */ |
| 510 | int |
| 511 | bidi_mirror_char (int c) |
| 512 | { |
| 513 | static const char mirrored_pairs[] = "()<>[]{}"; |
| 514 | const char *p = c > 0 && c < 128 ? strchr (mirrored_pairs, c) : NULL; |
| 515 | |
| 516 | if (p) |
| 517 | { |
| 518 | size_t i = p - mirrored_pairs; |
| 519 | |
| 520 | return mirrored_pairs [(i ^ 1)]; |
| 521 | } |
| 522 | return c; |
| 523 | } |
| 524 | |
| 525 | /* Copy the bidi iterator from FROM to TO. To save cycles, this only |
| 526 | copies the part of the level stack that is actually in use. */ |
| 527 | static INLINE void |
| 528 | bidi_copy_it (struct bidi_it *to, struct bidi_it *from) |
| 529 | { |
| 530 | int i; |
| 531 | |
| 532 | /* Copy everything except the level stack and beyond. */ |
| 533 | memcpy (to, from, ((size_t)&((struct bidi_it *)0)->level_stack[0])); |
| 534 | |
| 535 | /* Copy the active part of the level stack. */ |
| 536 | to->level_stack[0] = from->level_stack[0]; /* level zero is always in use */ |
| 537 | for (i = 1; i <= from->stack_idx; i++) |
| 538 | to->level_stack[i] = from->level_stack[i]; |
| 539 | } |
| 540 | |
| 541 | /* Caching the bidi iterator states. */ |
| 542 | |
| 543 | static struct bidi_it bidi_cache[1000]; /* FIXME: make this dynamically allocated! */ |
| 544 | static int bidi_cache_idx; |
| 545 | static int bidi_cache_last_idx; |
| 546 | |
| 547 | static INLINE void |
| 548 | bidi_cache_reset (void) |
| 549 | { |
| 550 | bidi_cache_idx = 0; |
| 551 | bidi_cache_last_idx = -1; |
| 552 | } |
| 553 | |
| 554 | static INLINE void |
| 555 | bidi_cache_fetch_state (int idx, struct bidi_it *bidi_it) |
| 556 | { |
| 557 | int current_scan_dir = bidi_it->scan_dir; |
| 558 | |
| 559 | if (idx < 0 || idx >= bidi_cache_idx) |
| 560 | abort (); |
| 561 | |
| 562 | bidi_copy_it (bidi_it, &bidi_cache[idx]); |
| 563 | bidi_it->scan_dir = current_scan_dir; |
| 564 | bidi_cache_last_idx = idx; |
| 565 | } |
| 566 | |
| 567 | /* Find a cached state with a given CHARPOS and resolved embedding |
| 568 | level less or equal to LEVEL. if LEVEL is -1, disregard the |
| 569 | resolved levels in cached states. DIR, if non-zero, means search |
| 570 | in that direction from the last cache hit. */ |
| 571 | static INLINE int |
| 572 | bidi_cache_search (int charpos, int level, int dir) |
| 573 | { |
| 574 | int i, i_start; |
| 575 | |
| 576 | if (bidi_cache_idx) |
| 577 | { |
| 578 | if (charpos < bidi_cache[bidi_cache_last_idx].charpos) |
| 579 | dir = -1; |
| 580 | else if (charpos > bidi_cache[bidi_cache_last_idx].charpos) |
| 581 | dir = 1; |
| 582 | if (dir) |
| 583 | i_start = bidi_cache_last_idx; |
| 584 | else |
| 585 | { |
| 586 | dir = -1; |
| 587 | i_start = bidi_cache_idx - 1; |
| 588 | } |
| 589 | |
| 590 | if (dir < 0) |
| 591 | { |
| 592 | /* Linear search for now; FIXME! */ |
| 593 | for (i = i_start; i >= 0; i--) |
| 594 | if (bidi_cache[i].charpos == charpos |
| 595 | && (level == -1 || bidi_cache[i].resolved_level <= level)) |
| 596 | return i; |
| 597 | } |
| 598 | else |
| 599 | { |
| 600 | for (i = i_start; i < bidi_cache_idx; i++) |
| 601 | if (bidi_cache[i].charpos == charpos |
| 602 | && (level == -1 || bidi_cache[i].resolved_level <= level)) |
| 603 | return i; |
| 604 | } |
| 605 | } |
| 606 | |
| 607 | return -1; |
| 608 | } |
| 609 | |
| 610 | /* Find a cached state where the resolved level changes to a value |
| 611 | that is lower than LEVEL, and return its cache slot index. DIR is |
| 612 | the direction to search, starting with the last used cache slot. |
| 613 | BEFORE, if non-zero, means return the index of the slot that is |
| 614 | ``before'' the level change in the search direction. That is, |
| 615 | given the cached levels like this: |
| 616 | |
| 617 | 1122333442211 |
| 618 | AB C |
| 619 | |
| 620 | and assuming we are at the position cached at the slot marked with |
| 621 | C, searching backwards (DIR = -1) for LEVEL = 2 will return the |
| 622 | index of slot B or A, depending whether BEFORE is, respectively, |
| 623 | non-zero or zero. */ |
| 624 | static int |
| 625 | bidi_cache_find_level_change (int level, int dir, int before) |
| 626 | { |
| 627 | if (bidi_cache_idx) |
| 628 | { |
| 629 | int i = dir ? bidi_cache_last_idx : bidi_cache_idx - 1; |
| 630 | int incr = before ? 1 : 0; |
| 631 | |
| 632 | if (!dir) |
| 633 | dir = -1; |
| 634 | else if (!incr) |
| 635 | i += dir; |
| 636 | |
| 637 | if (dir < 0) |
| 638 | { |
| 639 | while (i >= incr) |
| 640 | { |
| 641 | if (bidi_cache[i - incr].resolved_level >= 0 |
| 642 | && bidi_cache[i - incr].resolved_level < level) |
| 643 | return i; |
| 644 | i--; |
| 645 | } |
| 646 | } |
| 647 | else |
| 648 | { |
| 649 | while (i < bidi_cache_idx - incr) |
| 650 | { |
| 651 | if (bidi_cache[i + incr].resolved_level >= 0 |
| 652 | && bidi_cache[i + incr].resolved_level < level) |
| 653 | return i; |
| 654 | i++; |
| 655 | } |
| 656 | } |
| 657 | } |
| 658 | |
| 659 | return -1; |
| 660 | } |
| 661 | |
| 662 | static INLINE void |
| 663 | bidi_cache_iterator_state (struct bidi_it *bidi_it, int resolved) |
| 664 | { |
| 665 | int idx; |
| 666 | |
| 667 | /* We should never cache on backward scans. */ |
| 668 | if (bidi_it->scan_dir == -1) |
| 669 | abort (); |
| 670 | idx = bidi_cache_search (bidi_it->charpos, -1, 1); |
| 671 | |
| 672 | if (idx < 0) |
| 673 | { |
| 674 | idx = bidi_cache_idx; |
| 675 | /* Don't overrun the cache limit. */ |
| 676 | if (idx > sizeof (bidi_cache) / sizeof (bidi_cache[0]) - 1) |
| 677 | abort (); |
| 678 | /* Character positions should correspond to cache positions 1:1. |
| 679 | If we are outside the range of cached positions, the cache is |
| 680 | useless and must be reset. */ |
| 681 | if (idx > 0 && |
| 682 | (bidi_it->charpos > bidi_cache[idx - 1].charpos + 1 |
| 683 | || bidi_it->charpos < bidi_cache[0].charpos)) |
| 684 | { |
| 685 | bidi_cache_reset (); |
| 686 | idx = 0; |
| 687 | } |
| 688 | bidi_copy_it (&bidi_cache[idx], bidi_it); |
| 689 | if (!resolved) |
| 690 | bidi_cache[idx].resolved_level = -1; |
| 691 | bidi_cache[idx].new_paragraph = 0; |
| 692 | } |
| 693 | else |
| 694 | { |
| 695 | /* Copy only the members which could have changed, to avoid |
| 696 | costly copying of the entire struct. */ |
| 697 | bidi_cache[idx].type = bidi_it->type; |
| 698 | bidi_check_type (bidi_it->type); |
| 699 | bidi_cache[idx].type_after_w1 = bidi_it->type_after_w1; |
| 700 | bidi_check_type (bidi_it->type_after_w1); |
| 701 | if (resolved) |
| 702 | bidi_cache[idx].resolved_level = bidi_it->resolved_level; |
| 703 | else |
| 704 | bidi_cache[idx].resolved_level = -1; |
| 705 | bidi_cache[idx].invalid_levels = bidi_it->invalid_levels; |
| 706 | bidi_cache[idx].invalid_rl_levels = bidi_it->invalid_rl_levels; |
| 707 | bidi_cache[idx].next_for_neutral = bidi_it->next_for_neutral; |
| 708 | bidi_cache[idx].next_for_ws = bidi_it->next_for_ws; |
| 709 | bidi_cache[idx].ignore_bn_limit = bidi_it->ignore_bn_limit; |
| 710 | } |
| 711 | |
| 712 | bidi_cache_last_idx = idx; |
| 713 | if (idx >= bidi_cache_idx) |
| 714 | bidi_cache_idx = idx + 1; |
| 715 | } |
| 716 | |
| 717 | static INLINE bidi_type_t |
| 718 | bidi_cache_find (int charpos, int level, struct bidi_it *bidi_it) |
| 719 | { |
| 720 | int i = bidi_cache_search (charpos, level, bidi_it->scan_dir); |
| 721 | |
| 722 | if (i >= 0) |
| 723 | { |
| 724 | bidi_dir_t current_scan_dir = bidi_it->scan_dir; |
| 725 | |
| 726 | bidi_copy_it (bidi_it, &bidi_cache[i]); |
| 727 | bidi_cache_last_idx = i; |
| 728 | /* Don't let scan direction from from the cached state override |
| 729 | the current scan direction. */ |
| 730 | bidi_it->scan_dir = current_scan_dir; |
| 731 | return bidi_it->type; |
| 732 | } |
| 733 | |
| 734 | return UNKNOWN_BT; |
| 735 | } |
| 736 | |
| 737 | static INLINE int |
| 738 | bidi_peek_at_next_level (struct bidi_it *bidi_it) |
| 739 | { |
| 740 | if (bidi_cache_idx == 0 || bidi_cache_last_idx == -1) |
| 741 | abort (); |
| 742 | return bidi_cache[bidi_cache_last_idx + bidi_it->scan_dir].resolved_level; |
| 743 | } |
| 744 | |
| 745 | /* Check if buffer position CHARPOS/BYTEPOS is the end of a paragraph. |
| 746 | Value is the non-negative length of the paragraph separator |
| 747 | following the buffer position, -1 if position is at the beginning |
| 748 | of a new paragraph, or -2 if position is neither at beginning nor |
| 749 | at end of a paragraph. */ |
| 750 | static EMACS_INT |
| 751 | bidi_at_paragraph_end (EMACS_INT charpos, EMACS_INT bytepos) |
| 752 | { |
| 753 | /* FIXME: Why Fbuffer_local_value rather than just Fsymbol_value? */ |
| 754 | Lisp_Object sep_re; |
| 755 | Lisp_Object start_re; |
| 756 | EMACS_INT val; |
| 757 | |
| 758 | sep_re = paragraph_separate_re; |
| 759 | start_re = paragraph_start_re; |
| 760 | |
| 761 | val = fast_looking_at (sep_re, charpos, bytepos, ZV, ZV_BYTE, Qnil); |
| 762 | if (val < 0) |
| 763 | { |
| 764 | if (fast_looking_at (start_re, charpos, bytepos, ZV, ZV_BYTE, Qnil) >= 0) |
| 765 | val = -1; |
| 766 | else |
| 767 | val = -2; |
| 768 | } |
| 769 | |
| 770 | return val; |
| 771 | } |
| 772 | |
| 773 | /* Determine the start-of-run (sor) directional type given the two |
| 774 | embedding levels on either side of the run boundary. Also, update |
| 775 | the saved info about previously seen characters, since that info is |
| 776 | generally valid for a single level run. */ |
| 777 | static INLINE void |
| 778 | bidi_set_sor_type (struct bidi_it *bidi_it, int level_before, int level_after) |
| 779 | { |
| 780 | int higher_level = level_before > level_after ? level_before : level_after; |
| 781 | |
| 782 | /* The prev_was_pdf gork is required for when we have several PDFs |
| 783 | in a row. In that case, we want to compute the sor type for the |
| 784 | next level run only once: when we see the first PDF. That's |
| 785 | because the sor type depends only on the higher of the two levels |
| 786 | that we find on the two sides of the level boundary (see UAX#9, |
| 787 | clause X10), and so we don't need to know the final embedding |
| 788 | level to which we descend after processing all the PDFs. */ |
| 789 | if (!bidi_it->prev_was_pdf || level_before < level_after) |
| 790 | /* FIXME: should the default sor direction be user selectable? */ |
| 791 | bidi_it->sor = (higher_level & 1) != 0 ? R2L : L2R; |
| 792 | if (level_before > level_after) |
| 793 | bidi_it->prev_was_pdf = 1; |
| 794 | |
| 795 | bidi_it->prev.type = UNKNOWN_BT; |
| 796 | bidi_it->last_strong.type = bidi_it->last_strong.type_after_w1 = |
| 797 | bidi_it->last_strong.orig_type = UNKNOWN_BT; |
| 798 | bidi_it->prev_for_neutral.type = bidi_it->sor == R2L ? STRONG_R : STRONG_L; |
| 799 | bidi_it->prev_for_neutral.charpos = bidi_it->charpos; |
| 800 | bidi_it->prev_for_neutral.bytepos = bidi_it->bytepos; |
| 801 | bidi_it->next_for_neutral.type = bidi_it->next_for_neutral.type_after_w1 = |
| 802 | bidi_it->next_for_neutral.orig_type = UNKNOWN_BT; |
| 803 | bidi_it->ignore_bn_limit = 0; /* meaning it's unknown */ |
| 804 | } |
| 805 | |
| 806 | static void |
| 807 | bidi_line_init (struct bidi_it *bidi_it) |
| 808 | { |
| 809 | bidi_it->scan_dir = 1; /* FIXME: do we need to have control on this? */ |
| 810 | bidi_it->resolved_level = bidi_it->level_stack[0].level; |
| 811 | bidi_it->level_stack[0].override = NEUTRAL_DIR; /* X1 */ |
| 812 | bidi_it->invalid_levels = 0; |
| 813 | bidi_it->invalid_rl_levels = -1; |
| 814 | bidi_it->next_en_pos = -1; |
| 815 | bidi_it->next_for_ws.type = UNKNOWN_BT; |
| 816 | bidi_set_sor_type (bidi_it, |
| 817 | bidi_it->paragraph_dir == R2L ? 1 : 0, |
| 818 | bidi_it->level_stack[0].level); /* X10 */ |
| 819 | |
| 820 | bidi_cache_reset (); |
| 821 | } |
| 822 | |
| 823 | /* Find the beginning of this paragraph by looking back in the buffer. |
| 824 | Value is the byte position of the paragraph's beginning. */ |
| 825 | static EMACS_INT |
| 826 | bidi_find_paragraph_start (EMACS_INT pos, EMACS_INT pos_byte) |
| 827 | { |
| 828 | Lisp_Object re = paragraph_start_re; |
| 829 | EMACS_INT limit = ZV, limit_byte = ZV_BYTE; |
| 830 | |
| 831 | while (pos_byte > BEGV_BYTE |
| 832 | && fast_looking_at (re, pos, pos_byte, limit, limit_byte, Qnil) < 0) |
| 833 | { |
| 834 | pos = find_next_newline_no_quit (pos - 1, -1); |
| 835 | pos_byte = CHAR_TO_BYTE (pos); |
| 836 | } |
| 837 | return pos_byte; |
| 838 | } |
| 839 | |
| 840 | /* Determine the direction, a.k.a. base embedding level, of the |
| 841 | paragraph we are about to iterate through. If DIR is either L2R or |
| 842 | R2L, just use that. Otherwise, determine the paragraph direction |
| 843 | from the first strong character of the paragraph. |
| 844 | |
| 845 | Note that this gives the paragraph separator the same direction as |
| 846 | the preceding paragraph, even though Emacs generally views the |
| 847 | separartor as not belonging to any paragraph. */ |
| 848 | void |
| 849 | bidi_paragraph_init (bidi_dir_t dir, struct bidi_it *bidi_it) |
| 850 | { |
| 851 | EMACS_INT bytepos = bidi_it->bytepos; |
| 852 | |
| 853 | /* Special case for an empty buffer. */ |
| 854 | if (bytepos == BEGV_BYTE && bytepos == ZV_BYTE) |
| 855 | dir = L2R; |
| 856 | /* We should never be called at EOB or before BEGV. */ |
| 857 | else if (bytepos >= ZV_BYTE || bytepos < BEGV_BYTE) |
| 858 | abort (); |
| 859 | |
| 860 | if (dir == L2R) |
| 861 | { |
| 862 | bidi_it->paragraph_dir = L2R; |
| 863 | bidi_it->new_paragraph = 0; |
| 864 | } |
| 865 | else if (dir == R2L) |
| 866 | { |
| 867 | bidi_it->paragraph_dir = R2L; |
| 868 | bidi_it->new_paragraph = 0; |
| 869 | } |
| 870 | else if (dir == NEUTRAL_DIR) /* P2 */ |
| 871 | { |
| 872 | int ch, ch_len; |
| 873 | EMACS_INT pos; |
| 874 | bidi_type_t type; |
| 875 | |
| 876 | /* If we are inside a paragraph separator, we are just waiting |
| 877 | for the separator to be exhausted; use the previous paragraph |
| 878 | direction. But don't do that if we have been just reseated, |
| 879 | because we need to reinitialize below in that case. */ |
| 880 | if (!bidi_it->first_elt |
| 881 | && bidi_it->charpos < bidi_it->separator_limit) |
| 882 | return; |
| 883 | |
| 884 | /* If we are on a newline, get past it to where the next |
| 885 | paragraph might start. But don't do that at BEGV since then |
| 886 | we are potentially in a new paragraph that doesn't yet |
| 887 | exist. */ |
| 888 | pos = bidi_it->charpos; |
| 889 | if (bytepos > BEGV_BYTE && FETCH_CHAR (bytepos) == '\n') |
| 890 | { |
| 891 | bytepos++; |
| 892 | pos++; |
| 893 | } |
| 894 | |
| 895 | /* We are either at the beginning of a paragraph or in the |
| 896 | middle of it. Find where this paragraph starts. */ |
| 897 | bytepos = bidi_find_paragraph_start (pos, bytepos); |
| 898 | |
| 899 | /* We should always be at the beginning of a new line at this |
| 900 | point. */ |
| 901 | if (!(bytepos == BEGV_BYTE || FETCH_CHAR (bytepos - 1) == '\n')) |
| 902 | abort (); |
| 903 | |
| 904 | bidi_it->separator_limit = -1; |
| 905 | bidi_it->new_paragraph = 0; |
| 906 | ch = FETCH_CHAR (bytepos); |
| 907 | ch_len = CHAR_BYTES (ch); |
| 908 | pos = BYTE_TO_CHAR (bytepos); |
| 909 | type = bidi_get_type (ch, NEUTRAL_DIR); |
| 910 | |
| 911 | for (pos++, bytepos += ch_len; |
| 912 | /* NOTE: UAX#9 says to search only for L, AL, or R types of |
| 913 | characters, and ignore RLE, RLO, LRE, and LRO. However, |
| 914 | I'm not sure it makes sense to omit those 4; should try |
| 915 | with and without that to see the effect. */ |
| 916 | (bidi_get_category (type) != STRONG) |
| 917 | || (bidi_ignore_explicit_marks_for_paragraph_level |
| 918 | && (type == RLE || type == RLO |
| 919 | || type == LRE || type == LRO)); |
| 920 | type = bidi_get_type (ch, NEUTRAL_DIR)) |
| 921 | { |
| 922 | if (type == NEUTRAL_B && bidi_at_paragraph_end (pos, bytepos) >= -1) |
| 923 | break; |
| 924 | if (bytepos >= ZV_BYTE) |
| 925 | { |
| 926 | /* Pretend there's a paragraph separator at end of buffer. */ |
| 927 | type = NEUTRAL_B; |
| 928 | break; |
| 929 | } |
| 930 | FETCH_CHAR_ADVANCE (ch, pos, bytepos); |
| 931 | } |
| 932 | if (type == STRONG_R || type == STRONG_AL) /* P3 */ |
| 933 | bidi_it->paragraph_dir = R2L; |
| 934 | else if (type == STRONG_L) |
| 935 | bidi_it->paragraph_dir = L2R; |
| 936 | } |
| 937 | else |
| 938 | abort (); |
| 939 | |
| 940 | /* Contrary to UAX#9 clause P3, we only default the paragraph |
| 941 | direction to L2R if we have no previous usable paragraph |
| 942 | direction. */ |
| 943 | if (bidi_it->paragraph_dir == NEUTRAL_DIR) |
| 944 | bidi_it->paragraph_dir = L2R; /* P3 and ``higher protocols'' */ |
| 945 | if (bidi_it->paragraph_dir == R2L) |
| 946 | bidi_it->level_stack[0].level = 1; |
| 947 | else |
| 948 | bidi_it->level_stack[0].level = 0; |
| 949 | |
| 950 | bidi_line_init (bidi_it); |
| 951 | } |
| 952 | |
| 953 | /* Do whatever UAX#9 clause X8 says should be done at paragraph's |
| 954 | end. */ |
| 955 | static INLINE void |
| 956 | bidi_set_paragraph_end (struct bidi_it *bidi_it) |
| 957 | { |
| 958 | bidi_it->invalid_levels = 0; |
| 959 | bidi_it->invalid_rl_levels = -1; |
| 960 | bidi_it->stack_idx = 0; |
| 961 | bidi_it->resolved_level = bidi_it->level_stack[0].level; |
| 962 | } |
| 963 | |
| 964 | /* Initialize the bidi iterator from buffer position CHARPOS. */ |
| 965 | void |
| 966 | bidi_init_it (EMACS_INT charpos, EMACS_INT bytepos, struct bidi_it *bidi_it) |
| 967 | { |
| 968 | if (! bidi_initialized) |
| 969 | bidi_initialize (); |
| 970 | bidi_it->charpos = charpos; |
| 971 | bidi_it->bytepos = bytepos; |
| 972 | bidi_it->first_elt = 1; |
| 973 | bidi_set_paragraph_end (bidi_it); |
| 974 | bidi_it->new_paragraph = 1; |
| 975 | bidi_it->separator_limit = -1; |
| 976 | bidi_it->type = NEUTRAL_B; |
| 977 | bidi_it->type_after_w1 = NEUTRAL_B; |
| 978 | bidi_it->orig_type = NEUTRAL_B; |
| 979 | bidi_it->prev_was_pdf = 0; |
| 980 | bidi_it->prev.type = bidi_it->prev.type_after_w1 = |
| 981 | bidi_it->prev.orig_type = UNKNOWN_BT; |
| 982 | bidi_it->last_strong.type = bidi_it->last_strong.type_after_w1 = |
| 983 | bidi_it->last_strong.orig_type = UNKNOWN_BT; |
| 984 | bidi_it->next_for_neutral.charpos = -1; |
| 985 | bidi_it->next_for_neutral.type = |
| 986 | bidi_it->next_for_neutral.type_after_w1 = |
| 987 | bidi_it->next_for_neutral.orig_type = UNKNOWN_BT; |
| 988 | bidi_it->prev_for_neutral.charpos = -1; |
| 989 | bidi_it->prev_for_neutral.type = |
| 990 | bidi_it->prev_for_neutral.type_after_w1 = |
| 991 | bidi_it->prev_for_neutral.orig_type = UNKNOWN_BT; |
| 992 | bidi_it->sor = L2R; /* FIXME: should it be user-selectable? */ |
| 993 | } |
| 994 | |
| 995 | /* Push the current embedding level and override status; reset the |
| 996 | current level to LEVEL and the current override status to OVERRIDE. */ |
| 997 | static INLINE void |
| 998 | bidi_push_embedding_level (struct bidi_it *bidi_it, |
| 999 | int level, bidi_dir_t override) |
| 1000 | { |
| 1001 | bidi_it->stack_idx++; |
| 1002 | if (bidi_it->stack_idx >= BIDI_MAXLEVEL) |
| 1003 | abort (); |
| 1004 | bidi_it->level_stack[bidi_it->stack_idx].level = level; |
| 1005 | bidi_it->level_stack[bidi_it->stack_idx].override = override; |
| 1006 | } |
| 1007 | |
| 1008 | /* Pop the embedding level and directional override status from the |
| 1009 | stack, and return the new level. */ |
| 1010 | static INLINE int |
| 1011 | bidi_pop_embedding_level (struct bidi_it *bidi_it) |
| 1012 | { |
| 1013 | /* UAX#9 says to ignore invalid PDFs. */ |
| 1014 | if (bidi_it->stack_idx > 0) |
| 1015 | bidi_it->stack_idx--; |
| 1016 | return bidi_it->level_stack[bidi_it->stack_idx].level; |
| 1017 | } |
| 1018 | |
| 1019 | /* Record in SAVED_INFO the information about the current character. */ |
| 1020 | static INLINE void |
| 1021 | bidi_remember_char (struct bidi_saved_info *saved_info, |
| 1022 | struct bidi_it *bidi_it) |
| 1023 | { |
| 1024 | saved_info->charpos = bidi_it->charpos; |
| 1025 | saved_info->bytepos = bidi_it->bytepos; |
| 1026 | saved_info->type = bidi_it->type; |
| 1027 | bidi_check_type (bidi_it->type); |
| 1028 | saved_info->type_after_w1 = bidi_it->type_after_w1; |
| 1029 | bidi_check_type (bidi_it->type_after_w1); |
| 1030 | saved_info->orig_type = bidi_it->orig_type; |
| 1031 | bidi_check_type (bidi_it->orig_type); |
| 1032 | } |
| 1033 | |
| 1034 | /* Resolve the type of a neutral character according to the type of |
| 1035 | surrounding strong text and the current embedding level. */ |
| 1036 | static INLINE bidi_type_t |
| 1037 | bidi_resolve_neutral_1 (bidi_type_t prev_type, bidi_type_t next_type, int lev) |
| 1038 | { |
| 1039 | /* N1: European and Arabic numbers are treated as though they were R. */ |
| 1040 | if (next_type == WEAK_EN || next_type == WEAK_AN) |
| 1041 | next_type = STRONG_R; |
| 1042 | if (prev_type == WEAK_EN || prev_type == WEAK_AN) |
| 1043 | prev_type = STRONG_R; |
| 1044 | |
| 1045 | if (next_type == prev_type) /* N1 */ |
| 1046 | return next_type; |
| 1047 | else if ((lev & 1) == 0) /* N2 */ |
| 1048 | return STRONG_L; |
| 1049 | else |
| 1050 | return STRONG_R; |
| 1051 | } |
| 1052 | |
| 1053 | static INLINE int |
| 1054 | bidi_explicit_dir_char (int c) |
| 1055 | { |
| 1056 | /* FIXME: this should be replaced with a lookup table with suitable |
| 1057 | bits set, like standard C ctype macros do. */ |
| 1058 | return (c == LRE_CHAR || c == LRO_CHAR |
| 1059 | || c == RLE_CHAR || c == RLO_CHAR || c == PDF_CHAR); |
| 1060 | } |
| 1061 | |
| 1062 | /* A helper function for bidi_resolve_explicit. It advances to the |
| 1063 | next character in logical order and determines the new embedding |
| 1064 | level and directional override, but does not take into account |
| 1065 | empty embeddings. */ |
| 1066 | static int |
| 1067 | bidi_resolve_explicit_1 (struct bidi_it *bidi_it) |
| 1068 | { |
| 1069 | int curchar; |
| 1070 | bidi_type_t type; |
| 1071 | int current_level; |
| 1072 | int new_level; |
| 1073 | bidi_dir_t override; |
| 1074 | |
| 1075 | if (bidi_it->bytepos < BEGV_BYTE /* after reseat to BEGV? */ |
| 1076 | || bidi_it->first_elt) |
| 1077 | { |
| 1078 | bidi_it->first_elt = 0; |
| 1079 | if (bidi_it->charpos < BEGV) |
| 1080 | bidi_it->charpos = BEGV; |
| 1081 | bidi_it->bytepos = CHAR_TO_BYTE (bidi_it->charpos); |
| 1082 | } |
| 1083 | else if (bidi_it->bytepos < ZV_BYTE) /* don't move at ZV */ |
| 1084 | { |
| 1085 | bidi_it->charpos++; |
| 1086 | if (bidi_it->ch_len == 0) |
| 1087 | abort (); |
| 1088 | bidi_it->bytepos += bidi_it->ch_len; |
| 1089 | } |
| 1090 | |
| 1091 | current_level = bidi_it->level_stack[bidi_it->stack_idx].level; /* X1 */ |
| 1092 | override = bidi_it->level_stack[bidi_it->stack_idx].override; |
| 1093 | new_level = current_level; |
| 1094 | |
| 1095 | /* in case it is a unibyte character (not yet implemented) */ |
| 1096 | /* _fetch_multibyte_char_len = 1; */ |
| 1097 | if (bidi_it->bytepos >= ZV_BYTE) |
| 1098 | { |
| 1099 | curchar = BIDI_EOB; |
| 1100 | bidi_it->ch_len = 1; |
| 1101 | } |
| 1102 | else |
| 1103 | { |
| 1104 | curchar = FETCH_CHAR (bidi_it->bytepos); |
| 1105 | bidi_it->ch_len = CHAR_BYTES (curchar); |
| 1106 | } |
| 1107 | bidi_it->ch = curchar; |
| 1108 | |
| 1109 | /* Don't apply directional override here, as all the types we handle |
| 1110 | below will not be affected by the override anyway, and we need |
| 1111 | the original type unaltered. The override will be applied in |
| 1112 | bidi_resolve_weak. */ |
| 1113 | type = bidi_get_type (curchar, NEUTRAL_DIR); |
| 1114 | bidi_it->orig_type = type; |
| 1115 | bidi_check_type (bidi_it->orig_type); |
| 1116 | |
| 1117 | if (type != PDF) |
| 1118 | bidi_it->prev_was_pdf = 0; |
| 1119 | |
| 1120 | bidi_it->type_after_w1 = UNKNOWN_BT; |
| 1121 | |
| 1122 | switch (type) |
| 1123 | { |
| 1124 | case RLE: /* X2 */ |
| 1125 | case RLO: /* X4 */ |
| 1126 | bidi_it->type_after_w1 = type; |
| 1127 | bidi_check_type (bidi_it->type_after_w1); |
| 1128 | type = WEAK_BN; /* X9/Retaining */ |
| 1129 | if (bidi_it->ignore_bn_limit <= 0) |
| 1130 | { |
| 1131 | if (current_level <= BIDI_MAXLEVEL - 4) |
| 1132 | { |
| 1133 | /* Compute the least odd embedding level greater than |
| 1134 | the current level. */ |
| 1135 | new_level = ((current_level + 1) & ~1) + 1; |
| 1136 | if (bidi_it->type_after_w1 == RLE) |
| 1137 | override = NEUTRAL_DIR; |
| 1138 | else |
| 1139 | override = R2L; |
| 1140 | if (current_level == BIDI_MAXLEVEL - 4) |
| 1141 | bidi_it->invalid_rl_levels = 0; |
| 1142 | bidi_push_embedding_level (bidi_it, new_level, override); |
| 1143 | } |
| 1144 | else |
| 1145 | { |
| 1146 | bidi_it->invalid_levels++; |
| 1147 | /* See the commentary about invalid_rl_levels below. */ |
| 1148 | if (bidi_it->invalid_rl_levels < 0) |
| 1149 | bidi_it->invalid_rl_levels = 0; |
| 1150 | bidi_it->invalid_rl_levels++; |
| 1151 | } |
| 1152 | } |
| 1153 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
| 1154 | || bidi_it->next_en_pos > bidi_it->charpos) |
| 1155 | type = WEAK_EN; |
| 1156 | break; |
| 1157 | case LRE: /* X3 */ |
| 1158 | case LRO: /* X5 */ |
| 1159 | bidi_it->type_after_w1 = type; |
| 1160 | bidi_check_type (bidi_it->type_after_w1); |
| 1161 | type = WEAK_BN; /* X9/Retaining */ |
| 1162 | if (bidi_it->ignore_bn_limit <= 0) |
| 1163 | { |
| 1164 | if (current_level <= BIDI_MAXLEVEL - 5) |
| 1165 | { |
| 1166 | /* Compute the least even embedding level greater than |
| 1167 | the current level. */ |
| 1168 | new_level = ((current_level + 2) & ~1); |
| 1169 | if (bidi_it->type_after_w1 == LRE) |
| 1170 | override = NEUTRAL_DIR; |
| 1171 | else |
| 1172 | override = L2R; |
| 1173 | bidi_push_embedding_level (bidi_it, new_level, override); |
| 1174 | } |
| 1175 | else |
| 1176 | { |
| 1177 | bidi_it->invalid_levels++; |
| 1178 | /* invalid_rl_levels counts invalid levels encountered |
| 1179 | while the embedding level was already too high for |
| 1180 | LRE/LRO, but not for RLE/RLO. That is because |
| 1181 | there may be exactly one PDF which we should not |
| 1182 | ignore even though invalid_levels is non-zero. |
| 1183 | invalid_rl_levels helps to know what PDF is |
| 1184 | that. */ |
| 1185 | if (bidi_it->invalid_rl_levels >= 0) |
| 1186 | bidi_it->invalid_rl_levels++; |
| 1187 | } |
| 1188 | } |
| 1189 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
| 1190 | || bidi_it->next_en_pos > bidi_it->charpos) |
| 1191 | type = WEAK_EN; |
| 1192 | break; |
| 1193 | case PDF: /* X7 */ |
| 1194 | bidi_it->type_after_w1 = type; |
| 1195 | bidi_check_type (bidi_it->type_after_w1); |
| 1196 | type = WEAK_BN; /* X9/Retaining */ |
| 1197 | if (bidi_it->ignore_bn_limit <= 0) |
| 1198 | { |
| 1199 | if (!bidi_it->invalid_rl_levels) |
| 1200 | { |
| 1201 | new_level = bidi_pop_embedding_level (bidi_it); |
| 1202 | bidi_it->invalid_rl_levels = -1; |
| 1203 | if (bidi_it->invalid_levels) |
| 1204 | bidi_it->invalid_levels--; |
| 1205 | /* else nothing: UAX#9 says to ignore invalid PDFs */ |
| 1206 | } |
| 1207 | if (!bidi_it->invalid_levels) |
| 1208 | new_level = bidi_pop_embedding_level (bidi_it); |
| 1209 | else |
| 1210 | { |
| 1211 | bidi_it->invalid_levels--; |
| 1212 | bidi_it->invalid_rl_levels--; |
| 1213 | } |
| 1214 | } |
| 1215 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
| 1216 | || bidi_it->next_en_pos > bidi_it->charpos) |
| 1217 | type = WEAK_EN; |
| 1218 | break; |
| 1219 | default: |
| 1220 | /* Nothing. */ |
| 1221 | break; |
| 1222 | } |
| 1223 | |
| 1224 | bidi_it->type = type; |
| 1225 | bidi_check_type (bidi_it->type); |
| 1226 | |
| 1227 | return new_level; |
| 1228 | } |
| 1229 | |
| 1230 | /* Given an iterator state in BIDI_IT, advance one character position |
| 1231 | in the buffer to the next character (in the logical order), resolve |
| 1232 | any explicit embeddings and directional overrides, and return the |
| 1233 | embedding level of the character after resolving explicit |
| 1234 | directives and ignoring empty embeddings. */ |
| 1235 | static int |
| 1236 | bidi_resolve_explicit (struct bidi_it *bidi_it) |
| 1237 | { |
| 1238 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; |
| 1239 | int new_level = bidi_resolve_explicit_1 (bidi_it); |
| 1240 | |
| 1241 | if (prev_level < new_level |
| 1242 | && bidi_it->type == WEAK_BN |
| 1243 | && bidi_it->ignore_bn_limit == 0 /* only if not already known */ |
| 1244 | && bidi_it->bytepos < ZV_BYTE /* not already at EOB */ |
| 1245 | && bidi_explicit_dir_char (FETCH_CHAR (bidi_it->bytepos |
| 1246 | + bidi_it->ch_len))) |
| 1247 | { |
| 1248 | /* Avoid pushing and popping embedding levels if the level run |
| 1249 | is empty, as this breaks level runs where it shouldn't. |
| 1250 | UAX#9 removes all the explicit embedding and override codes, |
| 1251 | so empty embeddings disappear without a trace. We need to |
| 1252 | behave as if we did the same. */ |
| 1253 | struct bidi_it saved_it; |
| 1254 | int level = prev_level; |
| 1255 | |
| 1256 | bidi_copy_it (&saved_it, bidi_it); |
| 1257 | |
| 1258 | while (bidi_explicit_dir_char (FETCH_CHAR (bidi_it->bytepos |
| 1259 | + bidi_it->ch_len))) |
| 1260 | { |
| 1261 | level = bidi_resolve_explicit_1 (bidi_it); |
| 1262 | } |
| 1263 | |
| 1264 | if (level == prev_level) /* empty embedding */ |
| 1265 | saved_it.ignore_bn_limit = bidi_it->charpos + 1; |
| 1266 | else /* this embedding is non-empty */ |
| 1267 | saved_it.ignore_bn_limit = -1; |
| 1268 | |
| 1269 | bidi_copy_it (bidi_it, &saved_it); |
| 1270 | if (bidi_it->ignore_bn_limit > 0) |
| 1271 | { |
| 1272 | /* We pushed a level, but we shouldn't have. Undo that. */ |
| 1273 | if (!bidi_it->invalid_rl_levels) |
| 1274 | { |
| 1275 | new_level = bidi_pop_embedding_level (bidi_it); |
| 1276 | bidi_it->invalid_rl_levels = -1; |
| 1277 | if (bidi_it->invalid_levels) |
| 1278 | bidi_it->invalid_levels--; |
| 1279 | } |
| 1280 | if (!bidi_it->invalid_levels) |
| 1281 | new_level = bidi_pop_embedding_level (bidi_it); |
| 1282 | else |
| 1283 | { |
| 1284 | bidi_it->invalid_levels--; |
| 1285 | bidi_it->invalid_rl_levels--; |
| 1286 | } |
| 1287 | } |
| 1288 | } |
| 1289 | |
| 1290 | if (bidi_it->type == NEUTRAL_B) /* X8 */ |
| 1291 | { |
| 1292 | bidi_set_paragraph_end (bidi_it); |
| 1293 | /* This is needed by bidi_resolve_weak below, and in L1. */ |
| 1294 | bidi_it->type_after_w1 = bidi_it->type; |
| 1295 | bidi_check_type (bidi_it->type_after_w1); |
| 1296 | } |
| 1297 | |
| 1298 | return new_level; |
| 1299 | } |
| 1300 | |
| 1301 | /* Advance in the buffer, resolve weak types and return the type of |
| 1302 | the next character after weak type resolution. */ |
| 1303 | static bidi_type_t |
| 1304 | bidi_resolve_weak (struct bidi_it *bidi_it) |
| 1305 | { |
| 1306 | bidi_type_t type; |
| 1307 | bidi_dir_t override; |
| 1308 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; |
| 1309 | int new_level = bidi_resolve_explicit (bidi_it); |
| 1310 | int next_char; |
| 1311 | bidi_type_t type_of_next; |
| 1312 | struct bidi_it saved_it; |
| 1313 | |
| 1314 | type = bidi_it->type; |
| 1315 | override = bidi_it->level_stack[bidi_it->stack_idx].override; |
| 1316 | |
| 1317 | if (type == UNKNOWN_BT |
| 1318 | || type == LRE |
| 1319 | || type == LRO |
| 1320 | || type == RLE |
| 1321 | || type == RLO |
| 1322 | || type == PDF) |
| 1323 | abort (); |
| 1324 | |
| 1325 | if (new_level != prev_level |
| 1326 | || bidi_it->type == NEUTRAL_B) |
| 1327 | { |
| 1328 | /* We've got a new embedding level run, compute the directional |
| 1329 | type of sor and initialize per-run variables (UAX#9, clause |
| 1330 | X10). */ |
| 1331 | bidi_set_sor_type (bidi_it, prev_level, new_level); |
| 1332 | } |
| 1333 | else if (type == NEUTRAL_S || type == NEUTRAL_WS |
| 1334 | || type == WEAK_BN || type == STRONG_AL) |
| 1335 | bidi_it->type_after_w1 = type; /* needed in L1 */ |
| 1336 | bidi_check_type (bidi_it->type_after_w1); |
| 1337 | |
| 1338 | /* Level and directional override status are already recorded in |
| 1339 | bidi_it, and do not need any change; see X6. */ |
| 1340 | if (override == R2L) /* X6 */ |
| 1341 | type = STRONG_R; |
| 1342 | else if (override == L2R) |
| 1343 | type = STRONG_L; |
| 1344 | else |
| 1345 | { |
| 1346 | if (type == WEAK_NSM) /* W1 */ |
| 1347 | { |
| 1348 | /* Note that we don't need to consider the case where the |
| 1349 | prev character has its type overridden by an RLO or LRO, |
| 1350 | because then either the type of this NSM would have been |
| 1351 | also overridden, or the previous character is outside the |
| 1352 | current level run, and thus not relevant to this NSM. |
| 1353 | This is why NSM gets the type_after_w1 of the previous |
| 1354 | character. */ |
| 1355 | if (bidi_it->prev.type_after_w1 != UNKNOWN_BT |
| 1356 | /* if type_after_w1 is NEUTRAL_B, this NSM is at sor */ |
| 1357 | && bidi_it->prev.type_after_w1 != NEUTRAL_B) |
| 1358 | type = bidi_it->prev.type_after_w1; |
| 1359 | else if (bidi_it->sor == R2L) |
| 1360 | type = STRONG_R; |
| 1361 | else if (bidi_it->sor == L2R) |
| 1362 | type = STRONG_L; |
| 1363 | else /* shouldn't happen! */ |
| 1364 | abort (); |
| 1365 | } |
| 1366 | if (type == WEAK_EN /* W2 */ |
| 1367 | && bidi_it->last_strong.type_after_w1 == STRONG_AL) |
| 1368 | type = WEAK_AN; |
| 1369 | else if (type == STRONG_AL) /* W3 */ |
| 1370 | type = STRONG_R; |
| 1371 | else if ((type == WEAK_ES /* W4 */ |
| 1372 | && bidi_it->prev.type_after_w1 == WEAK_EN |
| 1373 | && bidi_it->prev.orig_type == WEAK_EN) |
| 1374 | || (type == WEAK_CS |
| 1375 | && ((bidi_it->prev.type_after_w1 == WEAK_EN |
| 1376 | && bidi_it->prev.orig_type == WEAK_EN) |
| 1377 | || bidi_it->prev.type_after_w1 == WEAK_AN))) |
| 1378 | { |
| 1379 | next_char = |
| 1380 | bidi_it->bytepos + bidi_it->ch_len >= ZV_BYTE |
| 1381 | ? BIDI_EOB : FETCH_CHAR (bidi_it->bytepos + bidi_it->ch_len); |
| 1382 | type_of_next = bidi_get_type (next_char, override); |
| 1383 | |
| 1384 | if (type_of_next == WEAK_BN |
| 1385 | || bidi_explicit_dir_char (next_char)) |
| 1386 | { |
| 1387 | bidi_copy_it (&saved_it, bidi_it); |
| 1388 | while (bidi_resolve_explicit (bidi_it) == new_level |
| 1389 | && bidi_it->type == WEAK_BN) |
| 1390 | ; |
| 1391 | type_of_next = bidi_it->type; |
| 1392 | bidi_copy_it (bidi_it, &saved_it); |
| 1393 | } |
| 1394 | |
| 1395 | /* If the next character is EN, but the last strong-type |
| 1396 | character is AL, that next EN will be changed to AN when |
| 1397 | we process it in W2 above. So in that case, this ES |
| 1398 | should not be changed into EN. */ |
| 1399 | if (type == WEAK_ES |
| 1400 | && type_of_next == WEAK_EN |
| 1401 | && bidi_it->last_strong.type_after_w1 != STRONG_AL) |
| 1402 | type = WEAK_EN; |
| 1403 | else if (type == WEAK_CS) |
| 1404 | { |
| 1405 | if (bidi_it->prev.type_after_w1 == WEAK_AN |
| 1406 | && (type_of_next == WEAK_AN |
| 1407 | /* If the next character is EN, but the last |
| 1408 | strong-type character is AL, EN will be later |
| 1409 | changed to AN when we process it in W2 above. |
| 1410 | So in that case, this ES should not be |
| 1411 | changed into EN. */ |
| 1412 | || (type_of_next == WEAK_EN |
| 1413 | && bidi_it->last_strong.type_after_w1 == STRONG_AL))) |
| 1414 | type = WEAK_AN; |
| 1415 | else if (bidi_it->prev.type_after_w1 == WEAK_EN |
| 1416 | && type_of_next == WEAK_EN |
| 1417 | && bidi_it->last_strong.type_after_w1 != STRONG_AL) |
| 1418 | type = WEAK_EN; |
| 1419 | } |
| 1420 | } |
| 1421 | else if (type == WEAK_ET /* W5: ET with EN before or after it */ |
| 1422 | || type == WEAK_BN) /* W5/Retaining */ |
| 1423 | { |
| 1424 | if (bidi_it->prev.type_after_w1 == WEAK_EN /* ET/BN w/EN before it */ |
| 1425 | || bidi_it->next_en_pos > bidi_it->charpos) |
| 1426 | type = WEAK_EN; |
| 1427 | else /* W5: ET/BN with EN after it. */ |
| 1428 | { |
| 1429 | EMACS_INT en_pos = bidi_it->charpos + 1; |
| 1430 | |
| 1431 | next_char = |
| 1432 | bidi_it->bytepos + bidi_it->ch_len >= ZV_BYTE |
| 1433 | ? BIDI_EOB : FETCH_CHAR (bidi_it->bytepos + bidi_it->ch_len); |
| 1434 | type_of_next = bidi_get_type (next_char, override); |
| 1435 | |
| 1436 | if (type_of_next == WEAK_ET |
| 1437 | || type_of_next == WEAK_BN |
| 1438 | || bidi_explicit_dir_char (next_char)) |
| 1439 | { |
| 1440 | bidi_copy_it (&saved_it, bidi_it); |
| 1441 | while (bidi_resolve_explicit (bidi_it) == new_level |
| 1442 | && (bidi_it->type == WEAK_BN |
| 1443 | || bidi_it->type == WEAK_ET)) |
| 1444 | ; |
| 1445 | type_of_next = bidi_it->type; |
| 1446 | en_pos = bidi_it->charpos; |
| 1447 | bidi_copy_it (bidi_it, &saved_it); |
| 1448 | } |
| 1449 | if (type_of_next == WEAK_EN) |
| 1450 | { |
| 1451 | /* If the last strong character is AL, the EN we've |
| 1452 | found will become AN when we get to it (W2). */ |
| 1453 | if (bidi_it->last_strong.type_after_w1 != STRONG_AL) |
| 1454 | { |
| 1455 | type = WEAK_EN; |
| 1456 | /* Remember this EN position, to speed up processing |
| 1457 | of the next ETs. */ |
| 1458 | bidi_it->next_en_pos = en_pos; |
| 1459 | } |
| 1460 | else if (type == WEAK_BN) |
| 1461 | type = NEUTRAL_ON; /* W6/Retaining */ |
| 1462 | } |
| 1463 | } |
| 1464 | } |
| 1465 | } |
| 1466 | |
| 1467 | if (type == WEAK_ES || type == WEAK_ET || type == WEAK_CS /* W6 */ |
| 1468 | || (type == WEAK_BN |
| 1469 | && (bidi_it->prev.type_after_w1 == WEAK_CS /* W6/Retaining */ |
| 1470 | || bidi_it->prev.type_after_w1 == WEAK_ES |
| 1471 | || bidi_it->prev.type_after_w1 == WEAK_ET))) |
| 1472 | type = NEUTRAL_ON; |
| 1473 | |
| 1474 | /* Store the type we've got so far, before we clobber it with strong |
| 1475 | types in W7 and while resolving neutral types. But leave alone |
| 1476 | the original types that were recorded above, because we will need |
| 1477 | them for the L1 clause. */ |
| 1478 | if (bidi_it->type_after_w1 == UNKNOWN_BT) |
| 1479 | bidi_it->type_after_w1 = type; |
| 1480 | bidi_check_type (bidi_it->type_after_w1); |
| 1481 | |
| 1482 | if (type == WEAK_EN) /* W7 */ |
| 1483 | { |
| 1484 | if ((bidi_it->last_strong.type_after_w1 == STRONG_L) |
| 1485 | || (bidi_it->last_strong.type == UNKNOWN_BT && bidi_it->sor == L2R)) |
| 1486 | type = STRONG_L; |
| 1487 | } |
| 1488 | |
| 1489 | bidi_it->type = type; |
| 1490 | bidi_check_type (bidi_it->type); |
| 1491 | return type; |
| 1492 | } |
| 1493 | |
| 1494 | static bidi_type_t |
| 1495 | bidi_resolve_neutral (struct bidi_it *bidi_it) |
| 1496 | { |
| 1497 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; |
| 1498 | bidi_type_t type = bidi_resolve_weak (bidi_it); |
| 1499 | int current_level = bidi_it->level_stack[bidi_it->stack_idx].level; |
| 1500 | |
| 1501 | if (!(type == STRONG_R |
| 1502 | || type == STRONG_L |
| 1503 | || type == WEAK_BN |
| 1504 | || type == WEAK_EN |
| 1505 | || type == WEAK_AN |
| 1506 | || type == NEUTRAL_B |
| 1507 | || type == NEUTRAL_S |
| 1508 | || type == NEUTRAL_WS |
| 1509 | || type == NEUTRAL_ON)) |
| 1510 | abort (); |
| 1511 | |
| 1512 | if (bidi_get_category (type) == NEUTRAL |
| 1513 | || (type == WEAK_BN && prev_level == current_level)) |
| 1514 | { |
| 1515 | if (bidi_it->next_for_neutral.type != UNKNOWN_BT) |
| 1516 | type = bidi_resolve_neutral_1 (bidi_it->prev_for_neutral.type, |
| 1517 | bidi_it->next_for_neutral.type, |
| 1518 | current_level); |
| 1519 | else |
| 1520 | { |
| 1521 | /* Arrrgh!! The UAX#9 algorithm is too deeply entrenched in |
| 1522 | the assumption of batch-style processing; see clauses W4, |
| 1523 | W5, and especially N1, which require to look far forward |
| 1524 | (as well as back) in the buffer. May the fleas of a |
| 1525 | thousand camels infest the armpits of those who design |
| 1526 | supposedly general-purpose algorithms by looking at their |
| 1527 | own implementations, and fail to consider other possible |
| 1528 | implementations! */ |
| 1529 | struct bidi_it saved_it; |
| 1530 | bidi_type_t next_type; |
| 1531 | |
| 1532 | if (bidi_it->scan_dir == -1) |
| 1533 | abort (); |
| 1534 | |
| 1535 | bidi_copy_it (&saved_it, bidi_it); |
| 1536 | /* Scan the text forward until we find the first non-neutral |
| 1537 | character, and then use that to resolve the neutral we |
| 1538 | are dealing with now. We also cache the scanned iterator |
| 1539 | states, to salvage some of the effort later. */ |
| 1540 | bidi_cache_iterator_state (bidi_it, 0); |
| 1541 | do { |
| 1542 | /* Record the info about the previous character, so that |
| 1543 | it will be cached below with this state. */ |
| 1544 | if (bidi_it->type_after_w1 != WEAK_BN /* W1/Retaining */ |
| 1545 | && bidi_it->type != WEAK_BN) |
| 1546 | bidi_remember_char (&bidi_it->prev, bidi_it); |
| 1547 | type = bidi_resolve_weak (bidi_it); |
| 1548 | /* Paragraph separators have their levels fully resolved |
| 1549 | at this point, so cache them as resolved. */ |
| 1550 | bidi_cache_iterator_state (bidi_it, type == NEUTRAL_B); |
| 1551 | /* FIXME: implement L1 here, by testing for a newline and |
| 1552 | resetting the level for any sequence of whitespace |
| 1553 | characters adjacent to it. */ |
| 1554 | } while (!(type == NEUTRAL_B |
| 1555 | || (type != WEAK_BN |
| 1556 | && bidi_get_category (type) != NEUTRAL) |
| 1557 | /* This is all per level run, so stop when we |
| 1558 | reach the end of this level run. */ |
| 1559 | || bidi_it->level_stack[bidi_it->stack_idx].level != |
| 1560 | current_level)); |
| 1561 | |
| 1562 | bidi_remember_char (&saved_it.next_for_neutral, bidi_it); |
| 1563 | |
| 1564 | switch (type) |
| 1565 | { |
| 1566 | case STRONG_L: |
| 1567 | case STRONG_R: |
| 1568 | case STRONG_AL: |
| 1569 | next_type = type; |
| 1570 | break; |
| 1571 | case WEAK_EN: |
| 1572 | case WEAK_AN: |
| 1573 | /* N1: ``European and Arabic numbers are treated as |
| 1574 | though they were R.'' */ |
| 1575 | next_type = STRONG_R; |
| 1576 | saved_it.next_for_neutral.type = STRONG_R; |
| 1577 | break; |
| 1578 | case WEAK_BN: |
| 1579 | if (!bidi_explicit_dir_char (bidi_it->ch)) |
| 1580 | abort (); /* can't happen: BNs are skipped */ |
| 1581 | /* FALLTHROUGH */ |
| 1582 | case NEUTRAL_B: |
| 1583 | /* Marched all the way to the end of this level run. |
| 1584 | We need to use the eor type, whose information is |
| 1585 | stored by bidi_set_sor_type in the prev_for_neutral |
| 1586 | member. */ |
| 1587 | if (saved_it.type != WEAK_BN |
| 1588 | || bidi_get_category (bidi_it->prev.type_after_w1) == NEUTRAL) |
| 1589 | { |
| 1590 | next_type = bidi_it->prev_for_neutral.type; |
| 1591 | saved_it.next_for_neutral.type = next_type; |
| 1592 | bidi_check_type (next_type); |
| 1593 | } |
| 1594 | else |
| 1595 | { |
| 1596 | /* This is a BN which does not adjoin neutrals. |
| 1597 | Leave its type alone. */ |
| 1598 | bidi_copy_it (bidi_it, &saved_it); |
| 1599 | return bidi_it->type; |
| 1600 | } |
| 1601 | break; |
| 1602 | default: |
| 1603 | abort (); |
| 1604 | } |
| 1605 | type = bidi_resolve_neutral_1 (saved_it.prev_for_neutral.type, |
| 1606 | next_type, current_level); |
| 1607 | saved_it.type = type; |
| 1608 | bidi_check_type (type); |
| 1609 | bidi_copy_it (bidi_it, &saved_it); |
| 1610 | } |
| 1611 | } |
| 1612 | return type; |
| 1613 | } |
| 1614 | |
| 1615 | /* Given an iterator state in BIDI_IT, advance one character position |
| 1616 | in the buffer to the next character (in the logical order), resolve |
| 1617 | the bidi type of that next character, and return that type. */ |
| 1618 | static bidi_type_t |
| 1619 | bidi_type_of_next_char (struct bidi_it *bidi_it) |
| 1620 | { |
| 1621 | bidi_type_t type; |
| 1622 | |
| 1623 | /* This should always be called during a forward scan. */ |
| 1624 | if (bidi_it->scan_dir != 1) |
| 1625 | abort (); |
| 1626 | |
| 1627 | /* Reset the limit until which to ignore BNs if we step out of the |
| 1628 | area where we found only empty levels. */ |
| 1629 | if ((bidi_it->ignore_bn_limit > 0 |
| 1630 | && bidi_it->ignore_bn_limit <= bidi_it->charpos) |
| 1631 | || (bidi_it->ignore_bn_limit == -1 |
| 1632 | && !bidi_explicit_dir_char (bidi_it->ch))) |
| 1633 | bidi_it->ignore_bn_limit = 0; |
| 1634 | |
| 1635 | type = bidi_resolve_neutral (bidi_it); |
| 1636 | |
| 1637 | return type; |
| 1638 | } |
| 1639 | |
| 1640 | /* Given an iterator state BIDI_IT, advance one character position in |
| 1641 | the buffer to the next character (in the logical order), resolve |
| 1642 | the embedding and implicit levels of that next character, and |
| 1643 | return the resulting level. */ |
| 1644 | static int |
| 1645 | bidi_level_of_next_char (struct bidi_it *bidi_it) |
| 1646 | { |
| 1647 | bidi_type_t type; |
| 1648 | int level, prev_level = -1; |
| 1649 | struct bidi_saved_info next_for_neutral; |
| 1650 | |
| 1651 | if (bidi_it->scan_dir == 1) |
| 1652 | { |
| 1653 | /* There's no sense in trying to advance if we hit end of text. */ |
| 1654 | if (bidi_it->bytepos >= ZV_BYTE) |
| 1655 | return bidi_it->resolved_level; |
| 1656 | |
| 1657 | /* Record the info about the previous character. */ |
| 1658 | if (bidi_it->type_after_w1 != WEAK_BN /* W1/Retaining */ |
| 1659 | && bidi_it->type != WEAK_BN) |
| 1660 | bidi_remember_char (&bidi_it->prev, bidi_it); |
| 1661 | if (bidi_it->type_after_w1 == STRONG_R |
| 1662 | || bidi_it->type_after_w1 == STRONG_L |
| 1663 | || bidi_it->type_after_w1 == STRONG_AL) |
| 1664 | bidi_remember_char (&bidi_it->last_strong, bidi_it); |
| 1665 | /* FIXME: it sounds like we don't need both prev and |
| 1666 | prev_for_neutral members, but I'm leaving them both for now. */ |
| 1667 | if (bidi_it->type == STRONG_R || bidi_it->type == STRONG_L |
| 1668 | || bidi_it->type == WEAK_EN || bidi_it->type == WEAK_AN) |
| 1669 | bidi_remember_char (&bidi_it->prev_for_neutral, bidi_it); |
| 1670 | |
| 1671 | /* If we overstepped the characters used for resolving neutrals |
| 1672 | and whitespace, invalidate their info in the iterator. */ |
| 1673 | if (bidi_it->charpos >= bidi_it->next_for_neutral.charpos) |
| 1674 | bidi_it->next_for_neutral.type = UNKNOWN_BT; |
| 1675 | if (bidi_it->next_en_pos >= 0 |
| 1676 | && bidi_it->charpos >= bidi_it->next_en_pos) |
| 1677 | bidi_it->next_en_pos = -1; |
| 1678 | if (bidi_it->next_for_ws.type != UNKNOWN_BT |
| 1679 | && bidi_it->charpos >= bidi_it->next_for_ws.charpos) |
| 1680 | bidi_it->next_for_ws.type = UNKNOWN_BT; |
| 1681 | |
| 1682 | /* This must be taken before we fill the iterator with the info |
| 1683 | about the next char. If we scan backwards, the iterator |
| 1684 | state must be already cached, so there's no need to know the |
| 1685 | embedding level of the previous character, since we will be |
| 1686 | returning to our caller shortly. */ |
| 1687 | prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; |
| 1688 | } |
| 1689 | next_for_neutral = bidi_it->next_for_neutral; |
| 1690 | |
| 1691 | /* Perhaps it is already cached. */ |
| 1692 | type = bidi_cache_find (bidi_it->charpos + bidi_it->scan_dir, -1, bidi_it); |
| 1693 | if (type != UNKNOWN_BT) |
| 1694 | { |
| 1695 | /* Don't lose the information for resolving neutrals! The |
| 1696 | cached states could have been cached before their |
| 1697 | next_for_neutral member was computed. If we are on our way |
| 1698 | forward, we can simply take the info from the previous |
| 1699 | state. */ |
| 1700 | if (bidi_it->scan_dir == 1 |
| 1701 | && bidi_it->next_for_neutral.type == UNKNOWN_BT) |
| 1702 | bidi_it->next_for_neutral = next_for_neutral; |
| 1703 | |
| 1704 | /* If resolved_level is -1, it means this state was cached |
| 1705 | before it was completely resolved, so we cannot return |
| 1706 | it. */ |
| 1707 | if (bidi_it->resolved_level != -1) |
| 1708 | return bidi_it->resolved_level; |
| 1709 | } |
| 1710 | if (bidi_it->scan_dir == -1) |
| 1711 | /* If we are going backwards, the iterator state is already cached |
| 1712 | from previous scans, and should be fully resolved. */ |
| 1713 | abort (); |
| 1714 | |
| 1715 | if (type == UNKNOWN_BT) |
| 1716 | type = bidi_type_of_next_char (bidi_it); |
| 1717 | |
| 1718 | if (type == NEUTRAL_B) |
| 1719 | return bidi_it->resolved_level; |
| 1720 | |
| 1721 | level = bidi_it->level_stack[bidi_it->stack_idx].level; |
| 1722 | if ((bidi_get_category (type) == NEUTRAL /* && type != NEUTRAL_B */) |
| 1723 | || (type == WEAK_BN && prev_level == level)) |
| 1724 | { |
| 1725 | if (bidi_it->next_for_neutral.type == UNKNOWN_BT) |
| 1726 | abort (); |
| 1727 | |
| 1728 | /* If the cached state shows a neutral character, it was not |
| 1729 | resolved by bidi_resolve_neutral, so do it now. */ |
| 1730 | type = bidi_resolve_neutral_1 (bidi_it->prev_for_neutral.type, |
| 1731 | bidi_it->next_for_neutral.type, |
| 1732 | level); |
| 1733 | } |
| 1734 | |
| 1735 | if (!(type == STRONG_R |
| 1736 | || type == STRONG_L |
| 1737 | || type == WEAK_BN |
| 1738 | || type == WEAK_EN |
| 1739 | || type == WEAK_AN)) |
| 1740 | abort (); |
| 1741 | bidi_it->type = type; |
| 1742 | bidi_check_type (bidi_it->type); |
| 1743 | |
| 1744 | /* For L1 below, we need to know, for each WS character, whether |
| 1745 | it belongs to a sequence of WS characters preceeding a newline |
| 1746 | or a TAB or a paragraph separator. */ |
| 1747 | if (bidi_it->orig_type == NEUTRAL_WS |
| 1748 | && bidi_it->next_for_ws.type == UNKNOWN_BT) |
| 1749 | { |
| 1750 | int ch; |
| 1751 | int clen = bidi_it->ch_len; |
| 1752 | EMACS_INT bpos = bidi_it->bytepos; |
| 1753 | EMACS_INT cpos = bidi_it->charpos; |
| 1754 | bidi_type_t chtype; |
| 1755 | |
| 1756 | do { |
| 1757 | /*_fetch_multibyte_char_len = 1;*/ |
| 1758 | ch = bpos + clen >= ZV_BYTE ? BIDI_EOB : FETCH_CHAR (bpos + clen); |
| 1759 | bpos += clen; |
| 1760 | cpos++; |
| 1761 | clen = (ch == BIDI_EOB ? 1 : CHAR_BYTES (ch)); |
| 1762 | if (ch == '\n' || ch == BIDI_EOB /* || ch == LINESEP_CHAR */) |
| 1763 | chtype = NEUTRAL_B; |
| 1764 | else |
| 1765 | chtype = bidi_get_type (ch, NEUTRAL_DIR); |
| 1766 | } while (chtype == NEUTRAL_WS || chtype == WEAK_BN |
| 1767 | || bidi_explicit_dir_char (ch)); /* L1/Retaining */ |
| 1768 | bidi_it->next_for_ws.type = chtype; |
| 1769 | bidi_check_type (bidi_it->next_for_ws.type); |
| 1770 | bidi_it->next_for_ws.charpos = cpos; |
| 1771 | bidi_it->next_for_ws.bytepos = bpos; |
| 1772 | } |
| 1773 | |
| 1774 | /* Resolve implicit levels, with a twist: PDFs get the embedding |
| 1775 | level of the enbedding they terminate. See below for the |
| 1776 | reason. */ |
| 1777 | if (bidi_it->orig_type == PDF |
| 1778 | /* Don't do this if this formatting code didn't change the |
| 1779 | embedding level due to invalid or empty embeddings. */ |
| 1780 | && prev_level != level) |
| 1781 | { |
| 1782 | /* Don't look in UAX#9 for the reason for this: it's our own |
| 1783 | private quirk. The reason is that we want the formatting |
| 1784 | codes to be delivered so that they bracket the text of their |
| 1785 | embedding. For example, given the text |
| 1786 | |
| 1787 | {RLO}teST{PDF} |
| 1788 | |
| 1789 | we want it to be displayed as |
| 1790 | |
| 1791 | {RLO}STet{PDF} |
| 1792 | |
| 1793 | not as |
| 1794 | |
| 1795 | STet{RLO}{PDF} |
| 1796 | |
| 1797 | which will result because we bump up the embedding level as |
| 1798 | soon as we see the RLO and pop it as soon as we see the PDF, |
| 1799 | so RLO itself has the same embedding level as "teST", and |
| 1800 | thus would be normally delivered last, just before the PDF. |
| 1801 | The switch below fiddles with the level of PDF so that this |
| 1802 | ugly side effect does not happen. |
| 1803 | |
| 1804 | (This is, of course, only important if the formatting codes |
| 1805 | are actually displayed, but Emacs does need to display them |
| 1806 | if the user wants to.) */ |
| 1807 | level = prev_level; |
| 1808 | } |
| 1809 | else if (bidi_it->orig_type == NEUTRAL_B /* L1 */ |
| 1810 | || bidi_it->orig_type == NEUTRAL_S |
| 1811 | || bidi_it->ch == '\n' || bidi_it->ch == BIDI_EOB |
| 1812 | /* || bidi_it->ch == LINESEP_CHAR */ |
| 1813 | || (bidi_it->orig_type == NEUTRAL_WS |
| 1814 | && (bidi_it->next_for_ws.type == NEUTRAL_B |
| 1815 | || bidi_it->next_for_ws.type == NEUTRAL_S))) |
| 1816 | level = bidi_it->level_stack[0].level; |
| 1817 | else if ((level & 1) == 0) /* I1 */ |
| 1818 | { |
| 1819 | if (type == STRONG_R) |
| 1820 | level++; |
| 1821 | else if (type == WEAK_EN || type == WEAK_AN) |
| 1822 | level += 2; |
| 1823 | } |
| 1824 | else /* I2 */ |
| 1825 | { |
| 1826 | if (type == STRONG_L || type == WEAK_EN || type == WEAK_AN) |
| 1827 | level++; |
| 1828 | } |
| 1829 | |
| 1830 | bidi_it->resolved_level = level; |
| 1831 | return level; |
| 1832 | } |
| 1833 | |
| 1834 | /* Move to the other edge of a level given by LEVEL. If END_FLAG is |
| 1835 | non-zero, we are at the end of a level, and we need to prepare to |
| 1836 | resume the scan of the lower level. |
| 1837 | |
| 1838 | If this level's other edge is cached, we simply jump to it, filling |
| 1839 | the iterator structure with the iterator state on the other edge. |
| 1840 | Otherwise, we walk the buffer until we come back to the same level |
| 1841 | as LEVEL. |
| 1842 | |
| 1843 | Note: we are not talking here about a ``level run'' in the UAX#9 |
| 1844 | sense of the term, but rather about a ``level'' which includes |
| 1845 | all the levels higher than it. In other words, given the levels |
| 1846 | like this: |
| 1847 | |
| 1848 | 11111112222222333333334443343222222111111112223322111 |
| 1849 | A B C |
| 1850 | |
| 1851 | and assuming we are at point A scanning left to right, this |
| 1852 | function moves to point C, whereas the UAX#9 ``level 2 run'' ends |
| 1853 | at point B. */ |
| 1854 | static void |
| 1855 | bidi_find_other_level_edge (struct bidi_it *bidi_it, int level, int end_flag) |
| 1856 | { |
| 1857 | int dir = end_flag ? -bidi_it->scan_dir : bidi_it->scan_dir; |
| 1858 | int idx; |
| 1859 | |
| 1860 | /* Try the cache first. */ |
| 1861 | if ((idx = bidi_cache_find_level_change (level, dir, end_flag)) >= 0) |
| 1862 | bidi_cache_fetch_state (idx, bidi_it); |
| 1863 | else |
| 1864 | { |
| 1865 | int new_level; |
| 1866 | |
| 1867 | if (end_flag) |
| 1868 | abort (); /* if we are at end of level, its edges must be cached */ |
| 1869 | |
| 1870 | bidi_cache_iterator_state (bidi_it, 1); |
| 1871 | do { |
| 1872 | new_level = bidi_level_of_next_char (bidi_it); |
| 1873 | bidi_cache_iterator_state (bidi_it, 1); |
| 1874 | } while (new_level >= level); |
| 1875 | } |
| 1876 | } |
| 1877 | |
| 1878 | void |
| 1879 | bidi_move_to_visually_next (struct bidi_it *bidi_it) |
| 1880 | { |
| 1881 | int old_level, new_level, next_level; |
| 1882 | struct bidi_it sentinel; |
| 1883 | |
| 1884 | if (bidi_it->scan_dir == 0) |
| 1885 | { |
| 1886 | bidi_it->scan_dir = 1; /* default to logical order */ |
| 1887 | } |
| 1888 | |
| 1889 | /* If we just passed a newline, initialize for the next line. */ |
| 1890 | if (!bidi_it->first_elt && bidi_it->orig_type == NEUTRAL_B) |
| 1891 | bidi_line_init (bidi_it); |
| 1892 | |
| 1893 | /* Prepare the sentinel iterator state. */ |
| 1894 | if (bidi_cache_idx == 0) |
| 1895 | { |
| 1896 | bidi_copy_it (&sentinel, bidi_it); |
| 1897 | if (bidi_it->first_elt) |
| 1898 | { |
| 1899 | sentinel.charpos--; /* cached charpos needs to be monotonic */ |
| 1900 | sentinel.bytepos--; |
| 1901 | sentinel.ch = '\n'; /* doesn't matter, but why not? */ |
| 1902 | sentinel.ch_len = 1; |
| 1903 | } |
| 1904 | } |
| 1905 | |
| 1906 | old_level = bidi_it->resolved_level; |
| 1907 | new_level = bidi_level_of_next_char (bidi_it); |
| 1908 | |
| 1909 | /* Reordering of resolved levels (clause L2) is implemented by |
| 1910 | jumping to the other edge of the level and flipping direction of |
| 1911 | scanning the text whenever we find a level change. */ |
| 1912 | if (new_level != old_level) |
| 1913 | { |
| 1914 | int ascending = new_level > old_level; |
| 1915 | int level_to_search = ascending ? old_level + 1 : old_level; |
| 1916 | int incr = ascending ? 1 : -1; |
| 1917 | int expected_next_level = old_level + incr; |
| 1918 | |
| 1919 | /* If we don't have anything cached yet, we need to cache the |
| 1920 | sentinel state, since we'll need it to record where to jump |
| 1921 | when the last non-base level is exhausted. */ |
| 1922 | if (bidi_cache_idx == 0) |
| 1923 | bidi_cache_iterator_state (&sentinel, 1); |
| 1924 | /* Jump (or walk) to the other edge of this level. */ |
| 1925 | bidi_find_other_level_edge (bidi_it, level_to_search, !ascending); |
| 1926 | /* Switch scan direction and peek at the next character in the |
| 1927 | new direction. */ |
| 1928 | bidi_it->scan_dir = -bidi_it->scan_dir; |
| 1929 | |
| 1930 | /* The following loop handles the case where the resolved level |
| 1931 | jumps by more than one. This is typical for numbers inside a |
| 1932 | run of text with left-to-right embedding direction, but can |
| 1933 | also happen in other situations. In those cases the decision |
| 1934 | where to continue after a level change, and in what direction, |
| 1935 | is tricky. For example, given a text like below: |
| 1936 | |
| 1937 | abcdefgh |
| 1938 | 11336622 |
| 1939 | |
| 1940 | (where the numbers below the text show the resolved levels), |
| 1941 | the result of reordering according to UAX#9 should be this: |
| 1942 | |
| 1943 | efdcghba |
| 1944 | |
| 1945 | This is implemented by the loop below which flips direction |
| 1946 | and jumps to the other edge of the level each time it finds |
| 1947 | the new level not to be the expected one. The expected level |
| 1948 | is always one more or one less than the previous one. */ |
| 1949 | next_level = bidi_peek_at_next_level (bidi_it); |
| 1950 | while (next_level != expected_next_level) |
| 1951 | { |
| 1952 | expected_next_level += incr; |
| 1953 | level_to_search += incr; |
| 1954 | bidi_find_other_level_edge (bidi_it, level_to_search, !ascending); |
| 1955 | bidi_it->scan_dir = -bidi_it->scan_dir; |
| 1956 | next_level = bidi_peek_at_next_level (bidi_it); |
| 1957 | } |
| 1958 | |
| 1959 | /* Finally, deliver the next character in the new direction. */ |
| 1960 | next_level = bidi_level_of_next_char (bidi_it); |
| 1961 | } |
| 1962 | |
| 1963 | /* Take note when we have just processed the newline that precedes |
| 1964 | the end of the paragraph. The next time we are about to be |
| 1965 | called, set_iterator_to_next will automatically reinit the |
| 1966 | paragraph direction, if needed. We do this at the newline before |
| 1967 | the paragraph separator, because the next character might not be |
| 1968 | the first character of the next paragraph, due to the bidi |
| 1969 | reordering, whereas we _must_ know the paragraph base direction |
| 1970 | _before_ we process the paragraph's text, since the base |
| 1971 | direction affects the reordering. */ |
| 1972 | if (bidi_it->scan_dir == 1 |
| 1973 | && bidi_it->orig_type == NEUTRAL_B |
| 1974 | && bidi_it->bytepos < ZV_BYTE) |
| 1975 | { |
| 1976 | EMACS_INT sep_len = |
| 1977 | bidi_at_paragraph_end (bidi_it->charpos + 1, |
| 1978 | bidi_it->bytepos + bidi_it->ch_len); |
| 1979 | if (sep_len >= 0) |
| 1980 | { |
| 1981 | bidi_it->new_paragraph = 1; |
| 1982 | /* Record the buffer position of the last character of the |
| 1983 | paragraph separator. */ |
| 1984 | bidi_it->separator_limit = bidi_it->charpos + 1 + sep_len; |
| 1985 | } |
| 1986 | } |
| 1987 | |
| 1988 | if (bidi_it->scan_dir == 1 && bidi_cache_idx) |
| 1989 | { |
| 1990 | /* If we are at paragraph's base embedding level and beyond the |
| 1991 | last cached position, the cache's job is done and we can |
| 1992 | discard it. */ |
| 1993 | if (bidi_it->resolved_level == bidi_it->level_stack[0].level |
| 1994 | && bidi_it->charpos > bidi_cache[bidi_cache_idx - 1].charpos) |
| 1995 | bidi_cache_reset (); |
| 1996 | /* But as long as we are caching during forward scan, we must |
| 1997 | cache each state, or else the cache integrity will be |
| 1998 | compromised: it assumes cached states correspond to buffer |
| 1999 | positions 1:1. */ |
| 2000 | else |
| 2001 | bidi_cache_iterator_state (bidi_it, 1); |
| 2002 | } |
| 2003 | } |
| 2004 | |
| 2005 | /* This is meant to be called from within the debugger, whenever you |
| 2006 | wish to examine the cache contents. */ |
| 2007 | void |
| 2008 | bidi_dump_cached_states (void) |
| 2009 | { |
| 2010 | int i; |
| 2011 | int ndigits = 1; |
| 2012 | |
| 2013 | if (bidi_cache_idx == 0) |
| 2014 | { |
| 2015 | fprintf (stderr, "The cache is empty.\n"); |
| 2016 | return; |
| 2017 | } |
| 2018 | fprintf (stderr, "Total of %d state%s in cache:\n", |
| 2019 | bidi_cache_idx, bidi_cache_idx == 1 ? "" : "s"); |
| 2020 | |
| 2021 | for (i = bidi_cache[bidi_cache_idx - 1].charpos; i > 0; i /= 10) |
| 2022 | ndigits++; |
| 2023 | fputs ("ch ", stderr); |
| 2024 | for (i = 0; i < bidi_cache_idx; i++) |
| 2025 | fprintf (stderr, "%*c", ndigits, bidi_cache[i].ch); |
| 2026 | fputs ("\n", stderr); |
| 2027 | fputs ("lvl ", stderr); |
| 2028 | for (i = 0; i < bidi_cache_idx; i++) |
| 2029 | fprintf (stderr, "%*d", ndigits, bidi_cache[i].resolved_level); |
| 2030 | fputs ("\n", stderr); |
| 2031 | fputs ("pos ", stderr); |
| 2032 | for (i = 0; i < bidi_cache_idx; i++) |
| 2033 | fprintf (stderr, "%*d", ndigits, bidi_cache[i].charpos); |
| 2034 | fputs ("\n", stderr); |
| 2035 | } |