1 /* vi:set ts=8 sts=4 sw=4: 2 * 3 * VIM - Vi IMproved by Bram Moolenaar 4 * Multibyte extensions partly by Sung-Hoon Baek 5 * 6 * Do ":help uganda" in Vim to read copying and usage conditions. 7 * Do ":help credits" in Vim to see a list of people who contributed. 8 * See README.txt for an overview of the Vim source code. 9 */ 10 /* 11 * mbyte.c: Code specifically for handling multi-byte characters. 12 * 13 * The encoding used in the core is set with 'encoding'. When 'encoding' is 14 * changed, the following four variables are set (for speed). 15 * Currently these types of character encodings are supported: 16 * 17 * "enc_dbcs" When non-zero it tells the type of double byte character 18 * encoding (Chinese, Korean, Japanese, etc.). 19 * The cell width on the display is equal to the number of 20 * bytes. (exception: DBCS_JPNU with first byte 0x8e) 21 * Recognizing the first or second byte is difficult, it 22 * requires checking a byte sequence from the start. 23 * "enc_utf8" When TRUE use Unicode characters in UTF-8 encoding. 24 * The cell width on the display needs to be determined from 25 * the character value. 26 * Recognizing bytes is easy: 0xxx.xxxx is a single-byte 27 * char, 10xx.xxxx is a trailing byte, 11xx.xxxx is a leading 28 * byte of a multi-byte character. 29 * To make things complicated, up to two composing characters 30 * are allowed. These are drawn on top of the first char. 31 * For most editing the sequence of bytes with composing 32 * characters included is considered to be one character. 33 * "enc_unicode" When 2 use 16-bit Unicode characters (or UTF-16). 34 * When 4 use 32-but Unicode characters. 35 * Internally characters are stored in UTF-8 encoding to 36 * avoid NUL bytes. Conversion happens when doing I/O. 37 * "enc_utf8" will also be TRUE. 38 * 39 * "has_mbyte" is set when "enc_dbcs" or "enc_utf8" is non-zero. 40 * 41 * If none of these is TRUE, 8-bit bytes are used for a character. The 42 * encoding isn't currently specified (TODO). 43 * 44 * 'encoding' specifies the encoding used in the core. This is in registers, 45 * text manipulation, buffers, etc. Conversion has to be done when characters 46 * in another encoding are received or send: 47 * 48 * clipboard 49 * ^ 50 * | (2) 51 * V 52 * +---------------+ 53 * (1) | | (3) 54 * keyboard ----->| core |-----> display 55 * | | 56 * +---------------+ 57 * ^ 58 * | (4) 59 * V 60 * file 61 * 62 * (1) Typed characters arrive in the current locale. Conversion is to be 63 * done when 'encoding' is different from 'termencoding'. 64 * (2) Text will be made available with the encoding specified with 65 * 'encoding'. If this is not sufficient, system-specific conversion 66 * might be required. 67 * (3) For the GUI the correct font must be selected, no conversion done. 68 * Otherwise, conversion is to be done when 'encoding' differs from 69 * 'termencoding'. (Different in the GTK+ 2 port -- 'termencoding' 70 * is always used for both input and output and must always be set to 71 * "utf-8". gui_mch_init() does this automatically.) 72 * (4) The encoding of the file is specified with 'fileencoding'. Conversion 73 * is to be done when it's different from 'encoding'. 74 * 75 * The viminfo file is a special case: Only text is converted, not file names. 76 * Vim scripts may contain an ":encoding" command. This has an effect for 77 * some commands, like ":menutrans" 78 */ 79 80 #include "vim.h" 81 82 #ifdef WIN32UNIX 83 # ifndef WIN32_LEAN_AND_MEAN 84 # define WIN32_LEAN_AND_MEAN 85 # endif 86 # include <windows.h> 87 # ifdef WIN32 88 # undef WIN32 /* Some windows.h define WIN32, we don't want that here. */ 89 # endif 90 #endif 91 92 #if (defined(WIN3264) || defined(WIN32UNIX)) && !defined(__MINGW32__) 93 # include <winnls.h> 94 #endif 95 96 #ifdef FEAT_GUI_X11 97 # include <X11/Intrinsic.h> 98 #endif 99 #ifdef X_LOCALE 100 #include <X11/Xlocale.h> 101 #endif 102 103 #if defined(FEAT_XIM) && defined(HAVE_GTK2) 104 # include <gdk/gdkkeysyms.h> 105 # ifdef WIN3264 106 # include <gdk/gdkwin32.h> 107 # else 108 # include <gdk/gdkx.h> 109 # endif 110 #endif 111 112 #ifdef HAVE_WCHAR_H 113 # include <wchar.h> 114 #endif 115 116 #if 0 117 /* This has been disabled, because several people reported problems with the 118 * wcwidth() and iswprint() library functions, esp. for Hebrew. */ 119 # ifdef __STDC_ISO_10646__ 120 # define USE_WCHAR_FUNCTIONS 121 # endif 122 #endif 123 124 #if defined(FEAT_MBYTE) || defined(PROTO) 125 126 static int enc_canon_search __ARGS((char_u *name)); 127 static int dbcs_char2len __ARGS((int c)); 128 static int dbcs_char2bytes __ARGS((int c, char_u *buf)); 129 static int dbcs_ptr2len __ARGS((char_u *p)); 130 static int dbcs_char2cells __ARGS((int c)); 131 static int dbcs_ptr2char __ARGS((char_u *p)); 132 133 /* Lookup table to quickly get the length in bytes of a UTF-8 character from 134 * the first byte of a UTF-8 string. Bytes which are illegal when used as the 135 * first byte have a one, because these will be used separately. */ 136 static char utf8len_tab[256] = 137 { 138 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 139 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 140 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 141 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 142 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /*bogus*/ 143 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /*bogus*/ 144 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, 145 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,4,4,4,4,4,4,4,4,5,5,5,5,6,6,1,1, 146 }; 147 148 /* 149 * XIM often causes trouble. Define XIM_DEBUG to get a log of XIM callbacks 150 * in the "xim.log" file. 151 */ 152 /* #define XIM_DEBUG */ 153 #ifdef XIM_DEBUG 154 static void 155 xim_log(char *s, ...) 156 { 157 va_list arglist; 158 static FILE *fd = NULL; 159 160 if (fd == (FILE *)-1) 161 return; 162 if (fd == NULL) 163 { 164 fd = mch_fopen("xim.log", "w"); 165 if (fd == NULL) 166 { 167 EMSG("Cannot open xim.log"); 168 fd = (FILE *)-1; 169 return; 170 } 171 } 172 173 va_start(arglist, s); 174 vfprintf(fd, s, arglist); 175 va_end(arglist); 176 } 177 #endif 178 179 #endif 180 181 #if defined(FEAT_MBYTE) || defined(FEAT_POSTSCRIPT) || defined(PROTO) 182 /* 183 * Canonical encoding names and their properties. 184 * "iso-8859-n" is handled by enc_canonize() directly. 185 */ 186 static struct 187 { char *name; int prop; int codepage;} 188 enc_canon_table[] = 189 { 190 #define IDX_LATIN_1 0 191 {"latin1", ENC_8BIT + ENC_LATIN1, 1252}, 192 #define IDX_ISO_2 1 193 {"iso-8859-2", ENC_8BIT, 0}, 194 #define IDX_ISO_3 2 195 {"iso-8859-3", ENC_8BIT, 0}, 196 #define IDX_ISO_4 3 197 {"iso-8859-4", ENC_8BIT, 0}, 198 #define IDX_ISO_5 4 199 {"iso-8859-5", ENC_8BIT, 0}, 200 #define IDX_ISO_6 5 201 {"iso-8859-6", ENC_8BIT, 0}, 202 #define IDX_ISO_7 6 203 {"iso-8859-7", ENC_8BIT, 0}, 204 #define IDX_ISO_8 7 205 {"iso-8859-8", ENC_8BIT, 0}, 206 #define IDX_ISO_9 8 207 {"iso-8859-9", ENC_8BIT, 0}, 208 #define IDX_ISO_10 9 209 {"iso-8859-10", ENC_8BIT, 0}, 210 #define IDX_ISO_11 10 211 {"iso-8859-11", ENC_8BIT, 0}, 212 #define IDX_ISO_13 11 213 {"iso-8859-13", ENC_8BIT, 0}, 214 #define IDX_ISO_14 12 215 {"iso-8859-14", ENC_8BIT, 0}, 216 #define IDX_ISO_15 13 217 {"iso-8859-15", ENC_8BIT + ENC_LATIN9, 0}, 218 #define IDX_KOI8_R 14 219 {"koi8-r", ENC_8BIT, 0}, 220 #define IDX_KOI8_U 15 221 {"koi8-u", ENC_8BIT, 0}, 222 #define IDX_UTF8 16 223 {"utf-8", ENC_UNICODE, 0}, 224 #define IDX_UCS2 17 225 {"ucs-2", ENC_UNICODE + ENC_ENDIAN_B + ENC_2BYTE, 0}, 226 #define IDX_UCS2LE 18 227 {"ucs-2le", ENC_UNICODE + ENC_ENDIAN_L + ENC_2BYTE, 0}, 228 #define IDX_UTF16 19 229 {"utf-16", ENC_UNICODE + ENC_ENDIAN_B + ENC_2WORD, 0}, 230 #define IDX_UTF16LE 20 231 {"utf-16le", ENC_UNICODE + ENC_ENDIAN_L + ENC_2WORD, 0}, 232 #define IDX_UCS4 21 233 {"ucs-4", ENC_UNICODE + ENC_ENDIAN_B + ENC_4BYTE, 0}, 234 #define IDX_UCS4LE 22 235 {"ucs-4le", ENC_UNICODE + ENC_ENDIAN_L + ENC_4BYTE, 0}, 236 237 /* For debugging DBCS encoding on Unix. */ 238 #define IDX_DEBUG 23 239 {"debug", ENC_DBCS, DBCS_DEBUG}, 240 #define IDX_EUC_JP 24 241 {"euc-jp", ENC_DBCS, DBCS_JPNU}, 242 #define IDX_SJIS 25 243 {"sjis", ENC_DBCS, DBCS_JPN}, 244 #define IDX_EUC_KR 26 245 {"euc-kr", ENC_DBCS, DBCS_KORU}, 246 #define IDX_EUC_CN 27 247 {"euc-cn", ENC_DBCS, DBCS_CHSU}, 248 #define IDX_EUC_TW 28 249 {"euc-tw", ENC_DBCS, DBCS_CHTU}, 250 #define IDX_BIG5 29 251 {"big5", ENC_DBCS, DBCS_CHT}, 252 253 /* MS-DOS and MS-Windows codepages are included here, so that they can be 254 * used on Unix too. Most of them are similar to ISO-8859 encodings, but 255 * not exactly the same. */ 256 #define IDX_CP437 30 257 {"cp437", ENC_8BIT, 437}, /* like iso-8859-1 */ 258 #define IDX_CP737 31 259 {"cp737", ENC_8BIT, 737}, /* like iso-8859-7 */ 260 #define IDX_CP775 32 261 {"cp775", ENC_8BIT, 775}, /* Baltic */ 262 #define IDX_CP850 33 263 {"cp850", ENC_8BIT, 850}, /* like iso-8859-4 */ 264 #define IDX_CP852 34 265 {"cp852", ENC_8BIT, 852}, /* like iso-8859-1 */ 266 #define IDX_CP855 35 267 {"cp855", ENC_8BIT, 855}, /* like iso-8859-2 */ 268 #define IDX_CP857 36 269 {"cp857", ENC_8BIT, 857}, /* like iso-8859-5 */ 270 #define IDX_CP860 37 271 {"cp860", ENC_8BIT, 860}, /* like iso-8859-9 */ 272 #define IDX_CP861 38 273 {"cp861", ENC_8BIT, 861}, /* like iso-8859-1 */ 274 #define IDX_CP862 39 275 {"cp862", ENC_8BIT, 862}, /* like iso-8859-1 */ 276 #define IDX_CP863 40 277 {"cp863", ENC_8BIT, 863}, /* like iso-8859-8 */ 278 #define IDX_CP865 41 279 {"cp865", ENC_8BIT, 865}, /* like iso-8859-1 */ 280 #define IDX_CP866 42 281 {"cp866", ENC_8BIT, 866}, /* like iso-8859-5 */ 282 #define IDX_CP869 43 283 {"cp869", ENC_8BIT, 869}, /* like iso-8859-7 */ 284 #define IDX_CP874 44 285 {"cp874", ENC_8BIT, 874}, /* Thai */ 286 #define IDX_CP932 45 287 {"cp932", ENC_DBCS, DBCS_JPN}, 288 #define IDX_CP936 46 289 {"cp936", ENC_DBCS, DBCS_CHS}, 290 #define IDX_CP949 47 291 {"cp949", ENC_DBCS, DBCS_KOR}, 292 #define IDX_CP950 48 293 {"cp950", ENC_DBCS, DBCS_CHT}, 294 #define IDX_CP1250 49 295 {"cp1250", ENC_8BIT, 1250}, /* Czech, Polish, etc. */ 296 #define IDX_CP1251 50 297 {"cp1251", ENC_8BIT, 1251}, /* Cyrillic */ 298 /* cp1252 is considered to be equal to latin1 */ 299 #define IDX_CP1253 51 300 {"cp1253", ENC_8BIT, 1253}, /* Greek */ 301 #define IDX_CP1254 52 302 {"cp1254", ENC_8BIT, 1254}, /* Turkish */ 303 #define IDX_CP1255 53 304 {"cp1255", ENC_8BIT, 1255}, /* Hebrew */ 305 #define IDX_CP1256 54 306 {"cp1256", ENC_8BIT, 1256}, /* Arabic */ 307 #define IDX_CP1257 55 308 {"cp1257", ENC_8BIT, 1257}, /* Baltic */ 309 #define IDX_CP1258 56 310 {"cp1258", ENC_8BIT, 1258}, /* Vietnamese */ 311 312 #define IDX_MACROMAN 57 313 {"macroman", ENC_8BIT + ENC_MACROMAN, 0}, /* Mac OS */ 314 #define IDX_COUNT 58 315 }; 316 317 /* 318 * Aliases for encoding names. 319 */ 320 static struct 321 { char *name; int canon;} 322 enc_alias_table[] = 323 { 324 {"ansi", IDX_LATIN_1}, 325 {"iso-8859-1", IDX_LATIN_1}, 326 {"latin2", IDX_ISO_2}, 327 {"latin3", IDX_ISO_3}, 328 {"latin4", IDX_ISO_4}, 329 {"cyrillic", IDX_ISO_5}, 330 {"arabic", IDX_ISO_6}, 331 {"greek", IDX_ISO_7}, 332 #ifdef WIN3264 333 {"hebrew", IDX_CP1255}, 334 #else 335 {"hebrew", IDX_ISO_8}, 336 #endif 337 {"latin5", IDX_ISO_9}, 338 {"turkish", IDX_ISO_9}, /* ? */ 339 {"latin6", IDX_ISO_10}, 340 {"nordic", IDX_ISO_10}, /* ? */ 341 {"thai", IDX_ISO_11}, /* ? */ 342 {"latin7", IDX_ISO_13}, 343 {"latin8", IDX_ISO_14}, 344 {"latin9", IDX_ISO_15}, 345 {"utf8", IDX_UTF8}, 346 {"unicode", IDX_UCS2}, 347 {"ucs2", IDX_UCS2}, 348 {"ucs2be", IDX_UCS2}, 349 {"ucs-2be", IDX_UCS2}, 350 {"ucs2le", IDX_UCS2LE}, 351 {"utf16", IDX_UTF16}, 352 {"utf16be", IDX_UTF16}, 353 {"utf-16be", IDX_UTF16}, 354 {"utf16le", IDX_UTF16LE}, 355 {"ucs4", IDX_UCS4}, 356 {"ucs4be", IDX_UCS4}, 357 {"ucs-4be", IDX_UCS4}, 358 {"ucs4le", IDX_UCS4LE}, 359 {"932", IDX_CP932}, 360 {"949", IDX_CP949}, 361 {"936", IDX_CP936}, 362 {"950", IDX_CP950}, 363 {"eucjp", IDX_EUC_JP}, 364 {"unix-jis", IDX_EUC_JP}, 365 {"ujis", IDX_EUC_JP}, 366 {"shift-jis", IDX_SJIS}, 367 {"euckr", IDX_EUC_KR}, 368 {"5601", IDX_EUC_KR}, /* Sun: KS C 5601 */ 369 {"euccn", IDX_EUC_CN}, 370 {"gb2312", IDX_EUC_CN}, 371 {"euctw", IDX_EUC_TW}, 372 #if defined(WIN3264) || defined(WIN32UNIX) || defined(MACOS) 373 {"japan", IDX_CP932}, 374 {"korea", IDX_CP949}, 375 {"prc", IDX_CP936}, 376 {"chinese", IDX_CP936}, 377 {"taiwan", IDX_CP950}, 378 {"big5", IDX_CP950}, 379 #else 380 {"japan", IDX_EUC_JP}, 381 {"korea", IDX_EUC_KR}, 382 {"prc", IDX_EUC_CN}, 383 {"chinese", IDX_EUC_CN}, 384 {"taiwan", IDX_EUC_TW}, 385 {"cp950", IDX_BIG5}, 386 {"950", IDX_BIG5}, 387 #endif 388 {"mac", IDX_MACROMAN}, 389 {NULL, 0} 390 }; 391 392 #ifndef CP_UTF8 393 # define CP_UTF8 65001 /* magic number from winnls.h */ 394 #endif 395 396 /* 397 * Find encoding "name" in the list of canonical encoding names. 398 * Returns -1 if not found. 399 */ 400 static int 401 enc_canon_search(name) 402 char_u *name; 403 { 404 int i; 405 406 for (i = 0; i < IDX_COUNT; ++i) 407 if (STRCMP(name, enc_canon_table[i].name) == 0) 408 return i; 409 return -1; 410 } 411 412 #endif 413 414 #if defined(FEAT_MBYTE) || defined(PROTO) 415 416 /* 417 * Find canonical encoding "name" in the list and return its properties. 418 * Returns 0 if not found. 419 */ 420 int 421 enc_canon_props(name) 422 char_u *name; 423 { 424 int i; 425 426 i = enc_canon_search(name); 427 if (i >= 0) 428 return enc_canon_table[i].prop; 429 #ifdef WIN3264 430 if (name[0] == 'c' && name[1] == 'p' && VIM_ISDIGIT(name[2])) 431 { 432 CPINFO cpinfo; 433 434 /* Get info on this codepage to find out what it is. */ 435 if (GetCPInfo(atoi(name + 2), &cpinfo) != 0) 436 { 437 if (cpinfo.MaxCharSize == 1) /* some single-byte encoding */ 438 return ENC_8BIT; 439 if (cpinfo.MaxCharSize == 2 440 && (cpinfo.LeadByte[0] != 0 || cpinfo.LeadByte[1] != 0)) 441 /* must be a DBCS encoding */ 442 return ENC_DBCS; 443 } 444 return 0; 445 } 446 #endif 447 if (STRNCMP(name, "2byte-", 6) == 0) 448 return ENC_DBCS; 449 if (STRNCMP(name, "8bit-", 5) == 0 || STRNCMP(name, "iso-8859-", 9) == 0) 450 return ENC_8BIT; 451 return 0; 452 } 453 454 /* 455 * Set up for using multi-byte characters. 456 * Called in three cases: 457 * - by main() to initialize (p_enc == NULL) 458 * - by set_init_1() after 'encoding' was set to its default. 459 * - by do_set() when 'encoding' has been set. 460 * p_enc must have been passed through enc_canonize() already. 461 * Sets the "enc_unicode", "enc_utf8", "enc_dbcs" and "has_mbyte" flags. 462 * Fills mb_bytelen_tab[] and returns NULL when there are no problems. 463 * When there is something wrong: Returns an error message and doesn't change 464 * anything. 465 */ 466 char_u * 467 mb_init() 468 { 469 int i; 470 int idx; 471 int n; 472 int enc_dbcs_new = 0; 473 #if defined(USE_ICONV) && !defined(WIN3264) && !defined(WIN32UNIX) \ 474 && !defined(MACOS) 475 # define LEN_FROM_CONV 476 vimconv_T vimconv; 477 char_u *p; 478 #endif 479 480 if (p_enc == NULL) 481 { 482 /* Just starting up: set the whole table to one's. */ 483 for (i = 0; i < 256; ++i) 484 mb_bytelen_tab[i] = 1; 485 input_conv.vc_type = CONV_NONE; 486 input_conv.vc_factor = 1; 487 output_conv.vc_type = CONV_NONE; 488 return NULL; 489 } 490 491 #ifdef WIN3264 492 if (p_enc[0] == 'c' && p_enc[1] == 'p' && VIM_ISDIGIT(p_enc[2])) 493 { 494 CPINFO cpinfo; 495 496 /* Get info on this codepage to find out what it is. */ 497 if (GetCPInfo(atoi(p_enc + 2), &cpinfo) != 0) 498 { 499 if (cpinfo.MaxCharSize == 1) 500 { 501 /* some single-byte encoding */ 502 enc_unicode = 0; 503 enc_utf8 = FALSE; 504 } 505 else if (cpinfo.MaxCharSize == 2 506 && (cpinfo.LeadByte[0] != 0 || cpinfo.LeadByte[1] != 0)) 507 { 508 /* must be a DBCS encoding, check below */ 509 enc_dbcs_new = atoi(p_enc + 2); 510 } 511 else 512 goto codepage_invalid; 513 } 514 else if (GetLastError() == ERROR_INVALID_PARAMETER) 515 { 516 codepage_invalid: 517 return (char_u *)N_("E543: Not a valid codepage"); 518 } 519 } 520 #endif 521 else if (STRNCMP(p_enc, "8bit-", 5) == 0 522 || STRNCMP(p_enc, "iso-8859-", 9) == 0) 523 { 524 /* Accept any "8bit-" or "iso-8859-" name. */ 525 enc_unicode = 0; 526 enc_utf8 = FALSE; 527 } 528 else if (STRNCMP(p_enc, "2byte-", 6) == 0) 529 { 530 #ifdef WIN3264 531 /* Windows: accept only valid codepage numbers, check below. */ 532 if (p_enc[6] != 'c' || p_enc[7] != 'p' 533 || (enc_dbcs_new = atoi(p_enc + 8)) == 0) 534 return e_invarg; 535 #else 536 /* Unix: accept any "2byte-" name, assume current locale. */ 537 enc_dbcs_new = DBCS_2BYTE; 538 #endif 539 } 540 else if ((idx = enc_canon_search(p_enc)) >= 0) 541 { 542 i = enc_canon_table[idx].prop; 543 if (i & ENC_UNICODE) 544 { 545 /* Unicode */ 546 enc_utf8 = TRUE; 547 if (i & (ENC_2BYTE | ENC_2WORD)) 548 enc_unicode = 2; 549 else if (i & ENC_4BYTE) 550 enc_unicode = 4; 551 else 552 enc_unicode = 0; 553 } 554 else if (i & ENC_DBCS) 555 { 556 /* 2byte, handle below */ 557 enc_dbcs_new = enc_canon_table[idx].codepage; 558 } 559 else 560 { 561 /* Must be 8-bit. */ 562 enc_unicode = 0; 563 enc_utf8 = FALSE; 564 } 565 } 566 else /* Don't know what encoding this is, reject it. */ 567 return e_invarg; 568 569 if (enc_dbcs_new != 0) 570 { 571 #ifdef WIN3264 572 /* Check if the DBCS code page is OK. */ 573 if (!IsValidCodePage(enc_dbcs_new)) 574 goto codepage_invalid; 575 #endif 576 enc_unicode = 0; 577 enc_utf8 = FALSE; 578 } 579 enc_dbcs = enc_dbcs_new; 580 has_mbyte = (enc_dbcs != 0 || enc_utf8); 581 582 #ifdef WIN3264 583 enc_codepage = encname2codepage(p_enc); 584 enc_latin9 = (STRCMP(p_enc, "iso-8859-15") == 0); 585 #endif 586 587 /* Detect an encoding that uses latin1 characters. */ 588 enc_latin1like = (enc_utf8 || STRCMP(p_enc, "latin1") == 0 589 || STRCMP(p_enc, "iso-8859-15") == 0); 590 591 /* 592 * Set the function pointers. 593 */ 594 if (enc_utf8) 595 { 596 mb_ptr2len = utfc_ptr2len; 597 mb_char2len = utf_char2len; 598 mb_char2bytes = utf_char2bytes; 599 mb_ptr2cells = utf_ptr2cells; 600 mb_char2cells = utf_char2cells; 601 mb_off2cells = utf_off2cells; 602 mb_ptr2char = utf_ptr2char; 603 mb_head_off = utf_head_off; 604 } 605 else if (enc_dbcs != 0) 606 { 607 mb_ptr2len = dbcs_ptr2len; 608 mb_char2len = dbcs_char2len; 609 mb_char2bytes = dbcs_char2bytes; 610 mb_ptr2cells = dbcs_ptr2cells; 611 mb_char2cells = dbcs_char2cells; 612 mb_off2cells = dbcs_off2cells; 613 mb_ptr2char = dbcs_ptr2char; 614 mb_head_off = dbcs_head_off; 615 } 616 else 617 { 618 mb_ptr2len = latin_ptr2len; 619 mb_char2len = latin_char2len; 620 mb_char2bytes = latin_char2bytes; 621 mb_ptr2cells = latin_ptr2cells; 622 mb_char2cells = latin_char2cells; 623 mb_off2cells = latin_off2cells; 624 mb_ptr2char = latin_ptr2char; 625 mb_head_off = latin_head_off; 626 } 627 628 /* 629 * Fill the mb_bytelen_tab[] for MB_BYTE2LEN(). 630 */ 631 #ifdef LEN_FROM_CONV 632 /* When 'encoding' is different from the current locale mblen() won't 633 * work. Use conversion to "utf-8" instead. */ 634 vimconv.vc_type = CONV_NONE; 635 if (enc_dbcs) 636 { 637 p = enc_locale(); 638 if (p == NULL || STRCMP(p, p_enc) != 0) 639 { 640 convert_setup(&vimconv, p_enc, (char_u *)"utf-8"); 641 vimconv.vc_fail = TRUE; 642 } 643 vim_free(p); 644 } 645 #endif 646 647 for (i = 0; i < 256; ++i) 648 { 649 /* Our own function to reliably check the length of UTF-8 characters, 650 * independent of mblen(). */ 651 if (enc_utf8) 652 n = utf8len_tab[i]; 653 else if (enc_dbcs == 0) 654 n = 1; 655 else 656 { 657 #if defined(WIN3264) || defined(WIN32UNIX) 658 /* enc_dbcs is set by setting 'fileencoding'. It becomes a Windows 659 * CodePage identifier, which we can pass directly in to Windows 660 * API */ 661 n = IsDBCSLeadByteEx(enc_dbcs, (BYTE)i) ? 2 : 1; 662 #else 663 # ifdef MACOS 664 /* 665 * if mblen() is not available, character which MSB is turned on 666 * are treated as leading byte character. (note : This assumption 667 * is not always true.) 668 */ 669 n = (i & 0x80) ? 2 : 1; 670 # else 671 char buf[MB_MAXBYTES]; 672 # ifdef X_LOCALE 673 # ifndef mblen 674 # define mblen _Xmblen 675 # endif 676 # endif 677 if (i == NUL) /* just in case mblen() can't handle "" */ 678 n = 1; 679 else 680 { 681 buf[0] = i; 682 buf[1] = 0; 683 #ifdef LEN_FROM_CONV 684 if (vimconv.vc_type != CONV_NONE) 685 { 686 /* 687 * string_convert() should fail when converting the first 688 * byte of a double-byte character. 689 */ 690 p = string_convert(&vimconv, (char_u *)buf, NULL); 691 if (p != NULL) 692 { 693 vim_free(p); 694 n = 1; 695 } 696 else 697 n = 2; 698 } 699 else 700 #endif 701 { 702 /* 703 * mblen() should return -1 for invalid (means the leading 704 * multibyte) character. However there are some platforms 705 * where mblen() returns 0 for invalid character. 706 * Therefore, following condition includes 0. 707 */ 708 (void)mblen(NULL, 0); /* First reset the state. */ 709 if (mblen(buf, (size_t)1) <= 0) 710 n = 2; 711 else 712 n = 1; 713 } 714 } 715 # endif 716 #endif 717 } 718 719 mb_bytelen_tab[i] = n; 720 } 721 722 #ifdef LEN_FROM_CONV 723 convert_setup(&vimconv, NULL, NULL); 724 #endif 725 726 /* The cell width depends on the type of multi-byte characters. */ 727 (void)init_chartab(); 728 729 /* When enc_utf8 is set or reset, (de)allocate ScreenLinesUC[] */ 730 screenalloc(FALSE); 731 732 /* When using Unicode, set default for 'fileencodings'. */ 733 if (enc_utf8 && !option_was_set((char_u *)"fencs")) 734 set_string_option_direct((char_u *)"fencs", -1, 735 (char_u *)"ucs-bom,utf-8,default,latin1", OPT_FREE); 736 #if defined(HAVE_BIND_TEXTDOMAIN_CODESET) && defined(FEAT_GETTEXT) 737 /* GNU gettext 0.10.37 supports this feature: set the codeset used for 738 * translated messages independently from the current locale. */ 739 (void)bind_textdomain_codeset(VIMPACKAGE, 740 enc_utf8 ? "utf-8" : (char *)p_enc); 741 #endif 742 743 #ifdef WIN32 744 /* When changing 'encoding' while starting up, then convert the command 745 * line arguments from the active codepage to 'encoding'. */ 746 if (starting != 0) 747 fix_arg_enc(); 748 #endif 749 750 #ifdef FEAT_AUTOCMD 751 /* Fire an autocommand to let people do custom font setup. This must be 752 * after Vim has been setup for the new encoding. */ 753 apply_autocmds(EVENT_ENCODINGCHANGED, NULL, (char_u *)"", FALSE, curbuf); 754 #endif 755 756 #ifdef FEAT_SYN_HL 757 /* Need to reload spell dictionaries */ 758 spell_reload(); 759 #endif 760 761 return NULL; 762 } 763 764 /* 765 * Return the size of the BOM for the current buffer: 766 * 0 - no BOM 767 * 2 - UCS-2 or UTF-16 BOM 768 * 4 - UCS-4 BOM 769 * 3 - UTF-8 BOM 770 */ 771 int 772 bomb_size() 773 { 774 int n = 0; 775 776 if (curbuf->b_p_bomb && !curbuf->b_p_bin) 777 { 778 if (*curbuf->b_p_fenc == NUL) 779 { 780 if (enc_utf8) 781 { 782 if (enc_unicode != 0) 783 n = enc_unicode; 784 else 785 n = 3; 786 } 787 } 788 else if (STRCMP(curbuf->b_p_fenc, "utf-8") == 0) 789 n = 3; 790 else if (STRNCMP(curbuf->b_p_fenc, "ucs-2", 5) == 0 791 || STRNCMP(curbuf->b_p_fenc, "utf-16", 6) == 0) 792 n = 2; 793 else if (STRNCMP(curbuf->b_p_fenc, "ucs-4", 5) == 0) 794 n = 4; 795 } 796 return n; 797 } 798 799 /* 800 * Get class of pointer: 801 * 0 for blank or NUL 802 * 1 for punctuation 803 * 2 for an (ASCII) word character 804 * >2 for other word characters 805 */ 806 int 807 mb_get_class(p) 808 char_u *p; 809 { 810 if (MB_BYTE2LEN(p[0]) == 1) 811 { 812 if (p[0] == NUL || vim_iswhite(p[0])) 813 return 0; 814 if (vim_iswordc(p[0])) 815 return 2; 816 return 1; 817 } 818 if (enc_dbcs != 0 && p[0] != NUL && p[1] != NUL) 819 return dbcs_class(p[0], p[1]); 820 if (enc_utf8) 821 return utf_class(utf_ptr2char(p)); 822 return 0; 823 } 824 825 /* 826 * Get class of a double-byte character. This always returns 3 or bigger. 827 * TODO: Should return 1 for punctuation. 828 */ 829 int 830 dbcs_class(lead, trail) 831 unsigned lead; 832 unsigned trail; 833 { 834 switch (enc_dbcs) 835 { 836 /* please add classfy routine for your language in here */ 837 838 case DBCS_JPNU: /* ? */ 839 case DBCS_JPN: 840 { 841 /* JIS code classification */ 842 unsigned char lb = lead; 843 unsigned char tb = trail; 844 845 /* convert process code to JIS */ 846 # if defined(WIN3264) || defined(WIN32UNIX) || defined(MACOS) 847 /* process code is SJIS */ 848 if (lb <= 0x9f) 849 lb = (lb - 0x81) * 2 + 0x21; 850 else 851 lb = (lb - 0xc1) * 2 + 0x21; 852 if (tb <= 0x7e) 853 tb -= 0x1f; 854 else if (tb <= 0x9e) 855 tb -= 0x20; 856 else 857 { 858 tb -= 0x7e; 859 lb += 1; 860 } 861 # else 862 /* 863 * XXX: Code page identification can not use with all 864 * system! So, some other encoding information 865 * will be needed. 866 * In japanese: SJIS,EUC,UNICODE,(JIS) 867 * Note that JIS-code system don't use as 868 * process code in most system because it uses 869 * escape sequences(JIS is context depend encoding). 870 */ 871 /* assume process code is JAPANESE-EUC */ 872 lb &= 0x7f; 873 tb &= 0x7f; 874 # endif 875 /* exceptions */ 876 switch (lb << 8 | tb) 877 { 878 case 0x2121: /* ZENKAKU space */ 879 return 0; 880 case 0x2122: /* KU-TEN (Japanese comma) */ 881 case 0x2123: /* TOU-TEN (Japanese period) */ 882 case 0x2124: /* ZENKAKU comma */ 883 case 0x2125: /* ZENKAKU period */ 884 return 1; 885 case 0x213c: /* prolongedsound handled as KATAKANA */ 886 return 13; 887 } 888 /* sieved by KU code */ 889 switch (lb) 890 { 891 case 0x21: 892 case 0x22: 893 /* special symbols */ 894 return 10; 895 case 0x23: 896 /* alpha-numeric */ 897 return 11; 898 case 0x24: 899 /* hiragana */ 900 return 12; 901 case 0x25: 902 /* katakana */ 903 return 13; 904 case 0x26: 905 /* greek */ 906 return 14; 907 case 0x27: 908 /* russian */ 909 return 15; 910 case 0x28: 911 /* lines */ 912 return 16; 913 default: 914 /* kanji */ 915 return 17; 916 } 917 } 918 919 case DBCS_KORU: /* ? */ 920 case DBCS_KOR: 921 { 922 /* KS code classification */ 923 unsigned char c1 = lead; 924 unsigned char c2 = trail; 925 926 /* 927 * 20 : Hangul 928 * 21 : Hanja 929 * 22 : Symbols 930 * 23 : Alpha-numeric/Roman Letter (Full width) 931 * 24 : Hangul Letter(Alphabet) 932 * 25 : Roman Numeral/Greek Letter 933 * 26 : Box Drawings 934 * 27 : Unit Symbols 935 * 28 : Circled/Parenthesized Letter 936 * 29 : Hirigana/Katakana 937 * 30 : Cyrillic Letter 938 */ 939 940 if (c1 >= 0xB0 && c1 <= 0xC8) 941 /* Hangul */ 942 return 20; 943 #if defined(WIN3264) || defined(WIN32UNIX) 944 else if (c1 <= 0xA0 || c2 <= 0xA0) 945 /* Extended Hangul Region : MS UHC(Unified Hangul Code) */ 946 /* c1: 0x81-0xA0 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xFE 947 * c1: 0xA1-0xC6 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xA0 948 */ 949 return 20; 950 #endif 951 952 else if (c1 >= 0xCA && c1 <= 0xFD) 953 /* Hanja */ 954 return 21; 955 else switch (c1) 956 { 957 case 0xA1: 958 case 0xA2: 959 /* Symbols */ 960 return 22; 961 case 0xA3: 962 /* Alpha-numeric */ 963 return 23; 964 case 0xA4: 965 /* Hangul Letter(Alphabet) */ 966 return 24; 967 case 0xA5: 968 /* Roman Numeral/Greek Letter */ 969 return 25; 970 case 0xA6: 971 /* Box Drawings */ 972 return 26; 973 case 0xA7: 974 /* Unit Symbols */ 975 return 27; 976 case 0xA8: 977 case 0xA9: 978 if (c2 <= 0xAF) 979 return 25; /* Roman Letter */ 980 else if (c2 >= 0xF6) 981 return 22; /* Symbols */ 982 else 983 /* Circled/Parenthesized Letter */ 984 return 28; 985 case 0xAA: 986 case 0xAB: 987 /* Hirigana/Katakana */ 988 return 29; 989 case 0xAC: 990 /* Cyrillic Letter */ 991 return 30; 992 } 993 } 994 default: 995 break; 996 } 997 return 3; 998 } 999 1000 /* 1001 * mb_char2len() function pointer. 1002 * Return length in bytes of character "c". 1003 * Returns 1 for a single-byte character. 1004 */ 1005 /* ARGSUSED */ 1006 int 1007 latin_char2len(c) 1008 int c; 1009 { 1010 return 1; 1011 } 1012 1013 static int 1014 dbcs_char2len(c) 1015 int c; 1016 { 1017 if (c >= 0x100) 1018 return 2; 1019 return 1; 1020 } 1021 1022 /* 1023 * mb_char2bytes() function pointer. 1024 * Convert a character to its bytes. 1025 * Returns the length in bytes. 1026 */ 1027 int 1028 latin_char2bytes(c, buf) 1029 int c; 1030 char_u *buf; 1031 { 1032 buf[0] = c; 1033 return 1; 1034 } 1035 1036 static int 1037 dbcs_char2bytes(c, buf) 1038 int c; 1039 char_u *buf; 1040 { 1041 if (c >= 0x100) 1042 { 1043 buf[0] = (unsigned)c >> 8; 1044 buf[1] = c; 1045 /* Never use a NUL byte, it causes lots of trouble. It's an invalid 1046 * character anyway. */ 1047 if (buf[1] == NUL) 1048 buf[1] = '\n'; 1049 return 2; 1050 } 1051 buf[0] = c; 1052 return 1; 1053 } 1054 1055 /* 1056 * mb_ptr2len() function pointer. 1057 * Get byte length of character at "*p" but stop at a NUL. 1058 * For UTF-8 this includes following composing characters. 1059 * Returns 0 when *p is NUL. 1060 * 1061 */ 1062 int 1063 latin_ptr2len(p) 1064 char_u *p; 1065 { 1066 return MB_BYTE2LEN(*p); 1067 } 1068 1069 static int 1070 dbcs_ptr2len(p) 1071 char_u *p; 1072 { 1073 int len; 1074 1075 /* Check if second byte is not missing. */ 1076 len = MB_BYTE2LEN(*p); 1077 if (len == 2 && p[1] == NUL) 1078 len = 1; 1079 return len; 1080 } 1081 1082 struct interval 1083 { 1084 unsigned short first; 1085 unsigned short last; 1086 }; 1087 static int intable __ARGS((struct interval *table, size_t size, int c)); 1088 1089 /* 1090 * Return TRUE if "c" is in "table[size / sizeof(struct interval)]". 1091 */ 1092 static int 1093 intable(table, size, c) 1094 struct interval *table; 1095 size_t size; 1096 int c; 1097 { 1098 int mid, bot, top; 1099 1100 /* first quick check for Latin1 etc. characters */ 1101 if (c < table[0].first) 1102 return FALSE; 1103 1104 /* binary search in table */ 1105 bot = 0; 1106 top = size / sizeof(struct interval) - 1; 1107 while (top >= bot) 1108 { 1109 mid = (bot + top) / 2; 1110 if (table[mid].last < c) 1111 bot = mid + 1; 1112 else if (table[mid].first > c) 1113 top = mid - 1; 1114 else 1115 return TRUE; 1116 } 1117 return FALSE; 1118 } 1119 1120 /* 1121 * For UTF-8 character "c" return 2 for a double-width character, 1 for others. 1122 * Returns 4 or 6 for an unprintable character. 1123 * Is only correct for characters >= 0x80. 1124 * When p_ambw is "double", return 2 for a character with East Asian Width 1125 * class 'A'(mbiguous). 1126 */ 1127 int 1128 utf_char2cells(c) 1129 int c; 1130 { 1131 /* sorted list of non-overlapping intervals of East Asian Ambiguous 1132 * characters, generated with: 1133 * "uniset +WIDTH-A -cat=Me -cat=Mn -cat=Cf c" */ 1134 static struct interval ambiguous[] = { 1135 {0x00A1, 0x00A1}, {0x00A4, 0x00A4}, {0x00A7, 0x00A8}, 1136 {0x00AA, 0x00AA}, {0x00AE, 0x00AE}, {0x00B0, 0x00B4}, 1137 {0x00B6, 0x00BA}, {0x00BC, 0x00BF}, {0x00C6, 0x00C6}, 1138 {0x00D0, 0x00D0}, {0x00D7, 0x00D8}, {0x00DE, 0x00E1}, 1139 {0x00E6, 0x00E6}, {0x00E8, 0x00EA}, {0x00EC, 0x00ED}, 1140 {0x00F0, 0x00F0}, {0x00F2, 0x00F3}, {0x00F7, 0x00FA}, 1141 {0x00FC, 0x00FC}, {0x00FE, 0x00FE}, {0x0101, 0x0101}, 1142 {0x0111, 0x0111}, {0x0113, 0x0113}, {0x011B, 0x011B}, 1143 {0x0126, 0x0127}, {0x012B, 0x012B}, {0x0131, 0x0133}, 1144 {0x0138, 0x0138}, {0x013F, 0x0142}, {0x0144, 0x0144}, 1145 {0x0148, 0x014B}, {0x014D, 0x014D}, {0x0152, 0x0153}, 1146 {0x0166, 0x0167}, {0x016B, 0x016B}, {0x01CE, 0x01CE}, 1147 {0x01D0, 0x01D0}, {0x01D2, 0x01D2}, {0x01D4, 0x01D4}, 1148 {0x01D6, 0x01D6}, {0x01D8, 0x01D8}, {0x01DA, 0x01DA}, 1149 {0x01DC, 0x01DC}, {0x0251, 0x0251}, {0x0261, 0x0261}, 1150 {0x02C4, 0x02C4}, {0x02C7, 0x02C7}, {0x02C9, 0x02CB}, 1151 {0x02CD, 0x02CD}, {0x02D0, 0x02D0}, {0x02D8, 0x02DB}, 1152 {0x02DD, 0x02DD}, {0x02DF, 0x02DF}, {0x0391, 0x03A1}, 1153 {0x03A3, 0x03A9}, {0x03B1, 0x03C1}, {0x03C3, 0x03C9}, 1154 {0x0401, 0x0401}, {0x0410, 0x044F}, {0x0451, 0x0451}, 1155 {0x2010, 0x2010}, {0x2013, 0x2016}, {0x2018, 0x2019}, 1156 {0x201C, 0x201D}, {0x2020, 0x2022}, {0x2024, 0x2027}, 1157 {0x2030, 0x2030}, {0x2032, 0x2033}, {0x2035, 0x2035}, 1158 {0x203B, 0x203B}, {0x203E, 0x203E}, {0x2074, 0x2074}, 1159 {0x207F, 0x207F}, {0x2081, 0x2084}, {0x20AC, 0x20AC}, 1160 {0x2103, 0x2103}, {0x2105, 0x2105}, {0x2109, 0x2109}, 1161 {0x2113, 0x2113}, {0x2116, 0x2116}, {0x2121, 0x2122}, 1162 {0x2126, 0x2126}, {0x212B, 0x212B}, {0x2153, 0x2154}, 1163 {0x215B, 0x215E}, {0x2160, 0x216B}, {0x2170, 0x2179}, 1164 {0x2190, 0x2199}, {0x21B8, 0x21B9}, {0x21D2, 0x21D2}, 1165 {0x21D4, 0x21D4}, {0x21E7, 0x21E7}, {0x2200, 0x2200}, 1166 {0x2202, 0x2203}, {0x2207, 0x2208}, {0x220B, 0x220B}, 1167 {0x220F, 0x220F}, {0x2211, 0x2211}, {0x2215, 0x2215}, 1168 {0x221A, 0x221A}, {0x221D, 0x2220}, {0x2223, 0x2223}, 1169 {0x2225, 0x2225}, {0x2227, 0x222C}, {0x222E, 0x222E}, 1170 {0x2234, 0x2237}, {0x223C, 0x223D}, {0x2248, 0x2248}, 1171 {0x224C, 0x224C}, {0x2252, 0x2252}, {0x2260, 0x2261}, 1172 {0x2264, 0x2267}, {0x226A, 0x226B}, {0x226E, 0x226F}, 1173 {0x2282, 0x2283}, {0x2286, 0x2287}, {0x2295, 0x2295}, 1174 {0x2299, 0x2299}, {0x22A5, 0x22A5}, {0x22BF, 0x22BF}, 1175 {0x2312, 0x2312}, {0x2460, 0x24E9}, {0x24EB, 0x254B}, 1176 {0x2550, 0x2573}, {0x2580, 0x258F}, {0x2592, 0x2595}, 1177 {0x25A0, 0x25A1}, {0x25A3, 0x25A9}, {0x25B2, 0x25B3}, 1178 {0x25B6, 0x25B7}, {0x25BC, 0x25BD}, {0x25C0, 0x25C1}, 1179 {0x25C6, 0x25C8}, {0x25CB, 0x25CB}, {0x25CE, 0x25D1}, 1180 {0x25E2, 0x25E5}, {0x25EF, 0x25EF}, {0x2605, 0x2606}, 1181 {0x2609, 0x2609}, {0x260E, 0x260F}, {0x2614, 0x2615}, 1182 {0x261C, 0x261C}, {0x261E, 0x261E}, {0x2640, 0x2640}, 1183 {0x2642, 0x2642}, {0x2660, 0x2661}, {0x2663, 0x2665}, 1184 {0x2667, 0x266A}, {0x266C, 0x266D}, {0x266F, 0x266F}, 1185 {0x273D, 0x273D}, {0x2776, 0x277F}, {0xE000, 0xF8FF}, 1186 {0xFFFD, 0xFFFD}, /* {0xF0000, 0xFFFFD}, {0x100000, 0x10FFFD} */ 1187 }; 1188 1189 if (c >= 0x100) 1190 { 1191 #ifdef USE_WCHAR_FUNCTIONS 1192 /* 1193 * Assume the library function wcwidth() works better than our own 1194 * stuff. It should return 1 for ambiguous width chars! 1195 */ 1196 int n = wcwidth(c); 1197 1198 if (n < 0) 1199 return 6; /* unprintable, displays <xxxx> */ 1200 if (n > 1) 1201 return n; 1202 #else 1203 if (!utf_printable(c)) 1204 return 6; /* unprintable, displays <xxxx> */ 1205 if (c >= 0x1100 1206 && (c <= 0x115f /* Hangul Jamo */ 1207 || c == 0x2329 1208 || c == 0x232a 1209 || (c >= 0x2e80 && c <= 0xa4cf 1210 && c != 0x303f) /* CJK ... Yi */ 1211 || (c >= 0xac00 && c <= 0xd7a3) /* Hangul Syllables */ 1212 || (c >= 0xf900 && c <= 0xfaff) /* CJK Compatibility 1213 Ideographs */ 1214 || (c >= 0xfe30 && c <= 0xfe6f) /* CJK Compatibility Forms */ 1215 || (c >= 0xff00 && c <= 0xff60) /* Fullwidth Forms */ 1216 || (c >= 0xffe0 && c <= 0xffe6) 1217 || (c >= 0x20000 && c <= 0x2fffd) 1218 || (c >= 0x30000 && c <= 0x3fffd))) 1219 return 2; 1220 #endif 1221 } 1222 1223 /* Characters below 0x100 are influenced by 'isprint' option */ 1224 else if (c >= 0x80 && !vim_isprintc(c)) 1225 return 4; /* unprintable, displays <xx> */ 1226 1227 if (c >= 0x80 && *p_ambw == 'd' && intable(ambiguous, sizeof(ambiguous), c)) 1228 return 2; 1229 1230 return 1; 1231 } 1232 1233 /* 1234 * mb_ptr2cells() function pointer. 1235 * Return the number of display cells character at "*p" occupies. 1236 * This doesn't take care of unprintable characters, use ptr2cells() for that. 1237 */ 1238 /*ARGSUSED*/ 1239 int 1240 latin_ptr2cells(p) 1241 char_u *p; 1242 { 1243 return 1; 1244 } 1245 1246 int 1247 utf_ptr2cells(p) 1248 char_u *p; 1249 { 1250 int c; 1251 1252 /* Need to convert to a wide character. */ 1253 if (*p >= 0x80) 1254 { 1255 c = utf_ptr2char(p); 1256 /* An illegal byte is displayed as <xx>. */ 1257 if (utf_ptr2len(p) == 1 || c == NUL) 1258 return 4; 1259 /* If the char is ASCII it must be an overlong sequence. */ 1260 if (c < 0x80) 1261 return char2cells(c); 1262 return utf_char2cells(c); 1263 } 1264 return 1; 1265 } 1266 1267 int 1268 dbcs_ptr2cells(p) 1269 char_u *p; 1270 { 1271 /* Number of cells is equal to number of bytes, except for euc-jp when 1272 * the first byte is 0x8e. */ 1273 if (enc_dbcs == DBCS_JPNU && *p == 0x8e) 1274 return 1; 1275 return MB_BYTE2LEN(*p); 1276 } 1277 1278 /* 1279 * mb_char2cells() function pointer. 1280 * Return the number of display cells character "c" occupies. 1281 * Only takes care of multi-byte chars, not "^C" and such. 1282 */ 1283 /*ARGSUSED*/ 1284 int 1285 latin_char2cells(c) 1286 int c; 1287 { 1288 return 1; 1289 } 1290 1291 static int 1292 dbcs_char2cells(c) 1293 int c; 1294 { 1295 /* Number of cells is equal to number of bytes, except for euc-jp when 1296 * the first byte is 0x8e. */ 1297 if (enc_dbcs == DBCS_JPNU && ((unsigned)c >> 8) == 0x8e) 1298 return 1; 1299 /* use the first byte */ 1300 return MB_BYTE2LEN((unsigned)c >> 8); 1301 } 1302 1303 /* 1304 * mb_off2cells() function pointer. 1305 * Return number of display cells for char at ScreenLines[off]. 1306 * Caller must make sure "off" and "off + 1" are valid! 1307 */ 1308 /*ARGSUSED*/ 1309 int 1310 latin_off2cells(off) 1311 unsigned off; 1312 { 1313 return 1; 1314 } 1315 1316 int 1317 dbcs_off2cells(off) 1318 unsigned off; 1319 { 1320 /* Number of cells is equal to number of bytes, except for euc-jp when 1321 * the first byte is 0x8e. */ 1322 if (enc_dbcs == DBCS_JPNU && ScreenLines[off] == 0x8e) 1323 return 1; 1324 return MB_BYTE2LEN(ScreenLines[off]); 1325 } 1326 1327 int 1328 utf_off2cells(off) 1329 unsigned off; 1330 { 1331 return ScreenLines[off + 1] == 0 ? 2 : 1; 1332 } 1333 1334 /* 1335 * mb_ptr2char() function pointer. 1336 * Convert a byte sequence into a character. 1337 */ 1338 int 1339 latin_ptr2char(p) 1340 char_u *p; 1341 { 1342 return *p; 1343 } 1344 1345 static int 1346 dbcs_ptr2char(p) 1347 char_u *p; 1348 { 1349 if (MB_BYTE2LEN(*p) > 1 && p[1] != NUL) 1350 return (p[0] << 8) + p[1]; 1351 return *p; 1352 } 1353 1354 /* 1355 * Convert a UTF-8 byte sequence to a wide character. 1356 * If the sequence is illegal or truncated by a NUL the first byte is 1357 * returned. 1358 * Does not include composing characters, of course. 1359 */ 1360 int 1361 utf_ptr2char(p) 1362 char_u *p; 1363 { 1364 int len; 1365 1366 if (p[0] < 0x80) /* be quick for ASCII */ 1367 return p[0]; 1368 1369 len = utf8len_tab[p[0]]; 1370 if ((p[1] & 0xc0) == 0x80) 1371 { 1372 if (len == 2) 1373 return ((p[0] & 0x1f) << 6) + (p[1] & 0x3f); 1374 if ((p[2] & 0xc0) == 0x80) 1375 { 1376 if (len == 3) 1377 return ((p[0] & 0x0f) << 12) + ((p[1] & 0x3f) << 6) 1378 + (p[2] & 0x3f); 1379 if ((p[3] & 0xc0) == 0x80) 1380 { 1381 if (len == 4) 1382 return ((p[0] & 0x07) << 18) + ((p[1] & 0x3f) << 12) 1383 + ((p[2] & 0x3f) << 6) + (p[3] & 0x3f); 1384 if ((p[4] & 0xc0) == 0x80) 1385 { 1386 if (len == 5) 1387 return ((p[0] & 0x03) << 24) + ((p[1] & 0x3f) << 18) 1388 + ((p[2] & 0x3f) << 12) + ((p[3] & 0x3f) << 6) 1389 + (p[4] & 0x3f); 1390 if ((p[5] & 0xc0) == 0x80 && len == 6) 1391 return ((p[0] & 0x01) << 30) + ((p[1] & 0x3f) << 24) 1392 + ((p[2] & 0x3f) << 18) + ((p[3] & 0x3f) << 12) 1393 + ((p[4] & 0x3f) << 6) + (p[5] & 0x3f); 1394 } 1395 } 1396 } 1397 } 1398 /* Illegal value, just return the first byte */ 1399 return p[0]; 1400 } 1401 1402 /* 1403 * Get character at **pp and advance *pp to the next character. 1404 * Note: composing characters are skipped! 1405 */ 1406 int 1407 mb_ptr2char_adv(pp) 1408 char_u **pp; 1409 { 1410 int c; 1411 1412 c = (*mb_ptr2char)(*pp); 1413 *pp += (*mb_ptr2len)(*pp); 1414 return c; 1415 } 1416 1417 /* 1418 * Get character at **pp and advance *pp to the next character. 1419 * Note: composing characters are returned as separate characters. 1420 */ 1421 int 1422 mb_cptr2char_adv(pp) 1423 char_u **pp; 1424 { 1425 int c; 1426 1427 c = (*mb_ptr2char)(*pp); 1428 if (enc_utf8) 1429 *pp += utf_ptr2len(*pp); 1430 else 1431 *pp += (*mb_ptr2len)(*pp); 1432 return c; 1433 } 1434 1435 #if defined(FEAT_ARABIC) || defined(PROTO) 1436 /* 1437 * Check whether we are dealing with Arabic combining characters. 1438 * Note: these are NOT really composing characters! 1439 */ 1440 int 1441 arabic_combine(one, two) 1442 int one; /* first character */ 1443 int two; /* character just after "one" */ 1444 { 1445 if (one == a_LAM) 1446 return arabic_maycombine(two); 1447 return FALSE; 1448 } 1449 1450 /* 1451 * Check whether we are dealing with a character that could be regarded as an 1452 * Arabic combining character, need to check the character before this. 1453 */ 1454 int 1455 arabic_maycombine(two) 1456 int two; 1457 { 1458 if (p_arshape && !p_tbidi) 1459 return (two == a_ALEF_MADDA 1460 || two == a_ALEF_HAMZA_ABOVE 1461 || two == a_ALEF_HAMZA_BELOW 1462 || two == a_ALEF); 1463 return FALSE; 1464 } 1465 1466 /* 1467 * Check if the character pointed to by "p2" is a composing character when it 1468 * comes after "p1". For Arabic sometimes "ab" is replaced with "c", which 1469 * behaves like a composing character. 1470 */ 1471 int 1472 utf_composinglike(p1, p2) 1473 char_u *p1; 1474 char_u *p2; 1475 { 1476 int c2; 1477 1478 c2 = utf_ptr2char(p2); 1479 if (utf_iscomposing(c2)) 1480 return TRUE; 1481 if (!arabic_maycombine(c2)) 1482 return FALSE; 1483 return arabic_combine(utf_ptr2char(p1), c2); 1484 } 1485 #endif 1486 1487 /* 1488 * Convert a UTF-8 byte string to a wide chararacter. Also get up to two 1489 * composing characters. 1490 */ 1491 int 1492 utfc_ptr2char(p, p1, p2) 1493 char_u *p; 1494 int *p1; /* return: first composing char or 0 */ 1495 int *p2; /* return: second composing char or 0 */ 1496 { 1497 int len; 1498 int c; 1499 int cc; 1500 1501 c = utf_ptr2char(p); 1502 len = utf_ptr2len(p); 1503 /* Only accept a composing char when the first char isn't illegal. */ 1504 if ((len > 1 || *p < 0x80) 1505 && p[len] >= 0x80 1506 && UTF_COMPOSINGLIKE(p, p + len)) 1507 { 1508 *p1 = utf_ptr2char(p + len); 1509 len += utf_ptr2len(p + len); 1510 if (p[len] >= 0x80 && utf_iscomposing(cc = utf_ptr2char(p + len))) 1511 *p2 = cc; 1512 else 1513 *p2 = 0; 1514 } 1515 else 1516 { 1517 *p1 = 0; 1518 *p2 = 0; 1519 } 1520 return c; 1521 } 1522 1523 /* 1524 * Convert a UTF-8 byte string to a wide chararacter. Also get up to two 1525 * composing characters. Use no more than p[maxlen]. 1526 */ 1527 int 1528 utfc_ptr2char_len(p, p1, p2, maxlen) 1529 char_u *p; 1530 int *p1; /* return: first composing char or 0 */ 1531 int *p2; /* return: second composing char or 0 */ 1532 int maxlen; 1533 { 1534 int len; 1535 int c; 1536 int cc; 1537 1538 c = utf_ptr2char(p); 1539 len = utf_ptr2len_len(p, maxlen); 1540 /* Only accept a composing char when the first char isn't illegal. */ 1541 if ((len > 1 || *p < 0x80) 1542 && len < maxlen 1543 && p[len] >= 0x80 1544 && UTF_COMPOSINGLIKE(p, p + len)) 1545 { 1546 *p1 = utf_ptr2char(p + len); 1547 len += utf_ptr2len_len(p + len, maxlen - len); 1548 if (len < maxlen 1549 && p[len] >= 0x80 1550 && utf_iscomposing(cc = utf_ptr2char(p + len))) 1551 *p2 = cc; 1552 else 1553 *p2 = 0; 1554 } 1555 else 1556 { 1557 *p1 = 0; 1558 *p2 = 0; 1559 } 1560 return c; 1561 } 1562 1563 /* 1564 * Convert the character at screen position "off" to a sequence of bytes. 1565 * Includes the composing characters. 1566 * "buf" must at least have the length MB_MAXBYTES. 1567 * Returns the produced number of bytes. 1568 */ 1569 int 1570 utfc_char2bytes(off, buf) 1571 int off; 1572 char_u *buf; 1573 { 1574 int len; 1575 1576 len = utf_char2bytes(ScreenLinesUC[off], buf); 1577 if (ScreenLinesC1[off] != 0) 1578 { 1579 len += utf_char2bytes(ScreenLinesC1[off], buf + len); 1580 if (ScreenLinesC2[off] != 0) 1581 len += utf_char2bytes(ScreenLinesC2[off], buf + len); 1582 } 1583 return len; 1584 } 1585 1586 /* 1587 * Get the length of a UTF-8 byte sequence, not including any following 1588 * composing characters. 1589 * Returns 0 for "". 1590 * Returns 1 for an illegal byte sequence. 1591 */ 1592 int 1593 utf_ptr2len(p) 1594 char_u *p; 1595 { 1596 int len; 1597 int i; 1598 1599 if (*p == NUL) 1600 return 0; 1601 len = utf8len_tab[*p]; 1602 for (i = 1; i < len; ++i) 1603 if ((p[i] & 0xc0) != 0x80) 1604 return 1; 1605 return len; 1606 } 1607 1608 /* 1609 * Return length of UTF-8 character, obtained from the first byte. 1610 * "b" must be between 0 and 255! 1611 */ 1612 int 1613 utf_byte2len(b) 1614 int b; 1615 { 1616 return utf8len_tab[b]; 1617 } 1618 1619 /* 1620 * Get the length of UTF-8 byte sequence "p[size]". Does not include any 1621 * following composing characters. 1622 * Returns 1 for "". 1623 * Returns 1 for an illegal byte sequence. 1624 * Returns number > "size" for an incomplete byte sequence. 1625 */ 1626 int 1627 utf_ptr2len_len(p, size) 1628 char_u *p; 1629 int size; 1630 { 1631 int len; 1632 int i; 1633 1634 if (*p == NUL) 1635 return 1; 1636 len = utf8len_tab[*p]; 1637 if (len > size) 1638 return len; /* incomplete byte sequence. */ 1639 for (i = 1; i < len; ++i) 1640 if ((p[i] & 0xc0) != 0x80) 1641 return 1; 1642 return len; 1643 } 1644 1645 /* 1646 * Return the number of bytes the UTF-8 encoding of the character at "p" takes. 1647 * This includes following composing characters. 1648 */ 1649 int 1650 utfc_ptr2len(p) 1651 char_u *p; 1652 { 1653 int len; 1654 int b0 = *p; 1655 #ifdef FEAT_ARABIC 1656 int prevlen; 1657 #endif 1658 1659 if (b0 == NUL) 1660 return 0; 1661 if (b0 < 0x80 && p[1] < 0x80) /* be quick for ASCII */ 1662 return 1; 1663 1664 /* Skip over first UTF-8 char, stopping at a NUL byte. */ 1665 len = utf_ptr2len(p); 1666 1667 /* Check for illegal byte. */ 1668 if (len == 1 && b0 >= 0x80) 1669 return 1; 1670 1671 /* 1672 * Check for composing characters. We can handle only the first two, but 1673 * skip all of them (otherwise the cursor would get stuck). 1674 */ 1675 #ifdef FEAT_ARABIC 1676 prevlen = 0; 1677 #endif 1678 for (;;) 1679 { 1680 if (p[len] < 0x80 || !UTF_COMPOSINGLIKE(p + prevlen, p + len)) 1681 return len; 1682 1683 /* Skip over composing char */ 1684 #ifdef FEAT_ARABIC 1685 prevlen = len; 1686 #endif 1687 len += utf_ptr2len(p + len); 1688 } 1689 } 1690 1691 /* 1692 * Return the number of bytes the UTF-8 encoding of the character at "p[size]" 1693 * takes. This includes following composing characters. 1694 * Returns 1 for an illegal char or an incomplete byte sequence. 1695 */ 1696 int 1697 utfc_ptr2len_len(p, size) 1698 char_u *p; 1699 int size; 1700 { 1701 int len; 1702 #ifdef FEAT_ARABIC 1703 int prevlen; 1704 #endif 1705 1706 if (*p == NUL) 1707 return 0; 1708 if (p[0] < 0x80 && (size == 1 || p[1] < 0x80)) /* be quick for ASCII */ 1709 return 1; 1710 1711 /* Skip over first UTF-8 char, stopping at a NUL byte. */ 1712 len = utf_ptr2len_len(p, size); 1713 1714 /* Check for illegal byte and incomplete byte sequence. */ 1715 if ((len == 1 && p[0] >= 0x80) || len > size) 1716 return 1; 1717 1718 /* 1719 * Check for composing characters. We can handle only the first two, but 1720 * skip all of them (otherwise the cursor would get stuck). 1721 */ 1722 #ifdef FEAT_ARABIC 1723 prevlen = 0; 1724 #endif 1725 while (len < size) 1726 { 1727 if (p[len] < 0x80 || !UTF_COMPOSINGLIKE(p + prevlen, p + len)) 1728 break; 1729 1730 /* Skip over composing char */ 1731 #ifdef FEAT_ARABIC 1732 prevlen = len; 1733 #endif 1734 len += utf_ptr2len_len(p + len, size - len); 1735 } 1736 return len; 1737 } 1738 1739 /* 1740 * Return the number of bytes the UTF-8 encoding of character "c" takes. 1741 * This does not include composing characters. 1742 */ 1743 int 1744 utf_char2len(c) 1745 int c; 1746 { 1747 if (c < 0x80) 1748 return 1; 1749 if (c < 0x800) 1750 return 2; 1751 if (c < 0x10000) 1752 return 3; 1753 if (c < 0x200000) 1754 return 4; 1755 if (c < 0x4000000) 1756 return 5; 1757 return 6; 1758 } 1759 1760 /* 1761 * Convert Unicode character "c" to UTF-8 string in "buf[]". 1762 * Returns the number of bytes. 1763 * This does not include composing characters. 1764 */ 1765 int 1766 utf_char2bytes(c, buf) 1767 int c; 1768 char_u *buf; 1769 { 1770 if (c < 0x80) /* 7 bits */ 1771 { 1772 buf[0] = c; 1773 return 1; 1774 } 1775 if (c < 0x800) /* 11 bits */ 1776 { 1777 buf[0] = 0xc0 + ((unsigned)c >> 6); 1778 buf[1] = 0x80 + (c & 0x3f); 1779 return 2; 1780 } 1781 if (c < 0x10000) /* 16 bits */ 1782 { 1783 buf[0] = 0xe0 + ((unsigned)c >> 12); 1784 buf[1] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1785 buf[2] = 0x80 + (c & 0x3f); 1786 return 3; 1787 } 1788 if (c < 0x200000) /* 21 bits */ 1789 { 1790 buf[0] = 0xf0 + ((unsigned)c >> 18); 1791 buf[1] = 0x80 + (((unsigned)c >> 12) & 0x3f); 1792 buf[2] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1793 buf[3] = 0x80 + (c & 0x3f); 1794 return 4; 1795 } 1796 if (c < 0x4000000) /* 26 bits */ 1797 { 1798 buf[0] = 0xf8 + ((unsigned)c >> 24); 1799 buf[1] = 0x80 + (((unsigned)c >> 18) & 0x3f); 1800 buf[2] = 0x80 + (((unsigned)c >> 12) & 0x3f); 1801 buf[3] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1802 buf[4] = 0x80 + (c & 0x3f); 1803 return 5; 1804 } 1805 /* 31 bits */ 1806 buf[0] = 0xfc + ((unsigned)c >> 30); 1807 buf[1] = 0x80 + (((unsigned)c >> 24) & 0x3f); 1808 buf[2] = 0x80 + (((unsigned)c >> 18) & 0x3f); 1809 buf[3] = 0x80 + (((unsigned)c >> 12) & 0x3f); 1810 buf[4] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1811 buf[5] = 0x80 + (c & 0x3f); 1812 return 6; 1813 } 1814 1815 /* 1816 * Return TRUE if "c" is a composing UTF-8 character. This means it will be 1817 * drawn on top of the preceding character. 1818 * Based on code from Markus Kuhn. 1819 */ 1820 int 1821 utf_iscomposing(c) 1822 int c; 1823 { 1824 /* sorted list of non-overlapping intervals */ 1825 static struct interval combining[] = 1826 { 1827 {0x0300, 0x034f}, {0x0360, 0x036f}, {0x0483, 0x0486}, {0x0488, 0x0489}, 1828 {0x0591, 0x05a1}, {0x05a3, 0x05b9}, {0x05bb, 0x05bd}, {0x05bf, 0x05bf}, 1829 {0x05c1, 0x05c2}, {0x05c4, 0x05c4}, {0x0610, 0x0615}, {0x064b, 0x0658}, 1830 {0x0670, 0x0670}, {0x06d6, 0x06dc}, {0x06de, 0x06e4}, {0x06e7, 0x06e8}, 1831 {0x06ea, 0x06ed}, {0x0711, 0x0711}, {0x0730, 0x074a}, {0x07a6, 0x07b0}, 1832 {0x0901, 0x0903}, {0x093c, 0x093c}, {0x093e, 0x094d}, {0x0951, 0x0954}, 1833 {0x0962, 0x0963}, {0x0981, 0x0983}, {0x09bc, 0x09bc}, {0x09be, 0x09c4}, 1834 {0x09c7, 0x09c8}, {0x09cb, 0x09cd}, {0x09d7, 0x09d7}, {0x09e2, 0x09e3}, 1835 {0x0a01, 0x0a03}, {0x0a3c, 0x0a3c}, {0x0a3e, 0x0a42}, {0x0a47, 0x0a48}, 1836 {0x0a4b, 0x0a4d}, {0x0a70, 0x0a71}, {0x0a81, 0x0a83}, {0x0abc, 0x0abc}, 1837 {0x0abe, 0x0ac5}, {0x0ac7, 0x0ac9}, {0x0acb, 0x0acd}, {0x0ae2, 0x0ae3}, 1838 {0x0b01, 0x0b03}, {0x0b3c, 0x0b3c}, {0x0b3e, 0x0b43}, {0x0b47, 0x0b48}, 1839 {0x0b4b, 0x0b4d}, {0x0b56, 0x0b57}, {0x0b82, 0x0b82}, {0x0bbe, 0x0bc2}, 1840 {0x0bc6, 0x0bc8}, {0x0bca, 0x0bcd}, {0x0bd7, 0x0bd7}, {0x0c01, 0x0c03}, 1841 {0x0c3e, 0x0c44}, {0x0c46, 0x0c48}, {0x0c4a, 0x0c4d}, {0x0c55, 0x0c56}, 1842 {0x0c82, 0x0c83}, {0x0cbc, 0x0cbc}, {0x0cbe, 0x0cc4}, {0x0cc6, 0x0cc8}, 1843 {0x0cca, 0x0ccd}, {0x0cd5, 0x0cd6}, {0x0d02, 0x0d03}, {0x0d3e, 0x0d43}, 1844 {0x0d46, 0x0d48}, {0x0d4a, 0x0d4d}, {0x0d57, 0x0d57}, {0x0d82, 0x0d83}, 1845 {0x0dca, 0x0dca}, {0x0dcf, 0x0dd4}, {0x0dd6, 0x0dd6}, {0x0dd8, 0x0ddf}, 1846 {0x0df2, 0x0df3}, {0x0e31, 0x0e31}, {0x0e34, 0x0e3a}, {0x0e47, 0x0e4e}, 1847 {0x0eb1, 0x0eb1}, {0x0eb4, 0x0eb9}, {0x0ebb, 0x0ebc}, {0x0ec8, 0x0ecd}, 1848 {0x0f18, 0x0f19}, {0x0f35, 0x0f35}, {0x0f37, 0x0f37}, {0x0f39, 0x0f39}, 1849 {0x0f3e, 0x0f3f}, {0x0f71, 0x0f84}, {0x0f86, 0x0f87}, {0x0f90, 0x0f97}, 1850 {0x0f99, 0x0fbc}, {0x0fc6, 0x0fc6}, {0x102c, 0x1032}, {0x1036, 0x1039}, 1851 {0x1056, 0x1059}, {0x1712, 0x1714}, {0x1732, 0x1734}, {0x1752, 0x1753}, 1852 {0x1772, 0x1773}, {0x17b6, 0x17d3}, {0x17dd, 0x17dd}, {0x180b, 0x180d}, 1853 {0x18a9, 0x18a9}, {0x1920, 0x192b}, {0x1930, 0x193b}, {0x20d0, 0x20ea}, 1854 {0x302a, 0x302f}, {0x3099, 0x309a}, {0xfb1e, 0xfb1e}, {0xfe00, 0xfe0f}, 1855 {0xfe20, 0xfe23}, 1856 }; 1857 1858 return intable(combining, sizeof(combining), c); 1859 } 1860 1861 /* 1862 * Return TRUE for characters that can be displayed in a normal way. 1863 * Only for characters of 0x100 and above! 1864 */ 1865 int 1866 utf_printable(c) 1867 int c; 1868 { 1869 #ifdef USE_WCHAR_FUNCTIONS 1870 /* 1871 * Assume the iswprint() library function works better than our own stuff. 1872 */ 1873 return iswprint(c); 1874 #else 1875 /* Sorted list of non-overlapping intervals. 1876 * 0xd800-0xdfff is reserved for UTF-16, actually illegal. */ 1877 static struct interval nonprint[] = 1878 { 1879 {0x070f, 0x070f}, {0x180b, 0x180e}, {0x200b, 0x200f}, {0x202a, 0x202e}, 1880 {0x206a, 0x206f}, {0xd800, 0xdfff}, {0xfeff, 0xfeff}, {0xfff9, 0xfffb}, 1881 {0xfffe, 0xffff} 1882 }; 1883 1884 return !intable(nonprint, sizeof(nonprint), c); 1885 #endif 1886 } 1887 1888 /* 1889 * Get class of a Unicode character. 1890 * 0: white space 1891 * 1: punctuation 1892 * 2 or bigger: some class of word character. 1893 */ 1894 int 1895 utf_class(c) 1896 int c; 1897 { 1898 /* sorted list of non-overlapping intervals */ 1899 static struct clinterval 1900 { 1901 unsigned short first; 1902 unsigned short last; 1903 unsigned short class; 1904 } classes[] = 1905 { 1906 {0x037e, 0x037e, 1}, /* Greek question mark */ 1907 {0x0387, 0x0387, 1}, /* Greek ano teleia */ 1908 {0x055a, 0x055f, 1}, /* Armenian punctuation */ 1909 {0x0589, 0x0589, 1}, /* Armenian full stop */ 1910 {0x05be, 0x05be, 1}, 1911 {0x05c0, 0x05c0, 1}, 1912 {0x05c3, 0x05c3, 1}, 1913 {0x05f3, 0x05f4, 1}, 1914 {0x060c, 0x060c, 1}, 1915 {0x061b, 0x061b, 1}, 1916 {0x061f, 0x061f, 1}, 1917 {0x066a, 0x066d, 1}, 1918 {0x06d4, 0x06d4, 1}, 1919 {0x0700, 0x070d, 1}, /* Syriac punctuation */ 1920 {0x0964, 0x0965, 1}, 1921 {0x0970, 0x0970, 1}, 1922 {0x0df4, 0x0df4, 1}, 1923 {0x0e4f, 0x0e4f, 1}, 1924 {0x0e5a, 0x0e5b, 1}, 1925 {0x0f04, 0x0f12, 1}, 1926 {0x0f3a, 0x0f3d, 1}, 1927 {0x0f85, 0x0f85, 1}, 1928 {0x104a, 0x104f, 1}, /* Myanmar punctuation */ 1929 {0x10fb, 0x10fb, 1}, /* Georgian punctuation */ 1930 {0x1361, 0x1368, 1}, /* Ethiopic punctuation */ 1931 {0x166d, 0x166e, 1}, /* Canadian Syl. punctuation */ 1932 {0x1680, 0x1680, 0}, 1933 {0x169b, 0x169c, 1}, 1934 {0x16eb, 0x16ed, 1}, 1935 {0x1735, 0x1736, 1}, 1936 {0x17d4, 0x17dc, 1}, /* Khmer punctuation */ 1937 {0x1800, 0x180a, 1}, /* Mongolian punctuation */ 1938 {0x2000, 0x200b, 0}, /* spaces */ 1939 {0x200c, 0x2027, 1}, /* punctuation and symbols */ 1940 {0x2028, 0x2029, 0}, 1941 {0x202a, 0x202e, 1}, /* punctuation and symbols */ 1942 {0x202f, 0x202f, 0}, 1943 {0x2030, 0x205e, 1}, /* punctuation and symbols */ 1944 {0x205f, 0x205f, 0}, 1945 {0x2060, 0x27ff, 1}, /* punctuation and symbols */ 1946 {0x2070, 0x207f, 0x2070}, /* superscript */ 1947 {0x2080, 0x208f, 0x2080}, /* subscript */ 1948 {0x2983, 0x2998, 1}, 1949 {0x29d8, 0x29db, 1}, 1950 {0x29fc, 0x29fd, 1}, 1951 {0x3000, 0x3000, 0}, /* ideographic space */ 1952 {0x3001, 0x3020, 1}, /* ideographic punctuation */ 1953 {0x3030, 0x3030, 1}, 1954 {0x303d, 0x303d, 1}, 1955 {0x3040, 0x309f, 0x3040}, /* Hiragana */ 1956 {0x30a0, 0x30ff, 0x30a0}, /* Katakana */ 1957 {0x3300, 0x9fff, 0x4e00}, /* CJK Ideographs */ 1958 {0xac00, 0xd7a3, 0xac00}, /* Hangul Syllables */ 1959 {0xf900, 0xfaff, 0x4e00}, /* CJK Ideographs */ 1960 {0xfd3e, 0xfd3f, 1}, 1961 {0xfe30, 0xfe6b, 1}, /* punctuation forms */ 1962 {0xff00, 0xff0f, 1}, /* half/fullwidth ASCII */ 1963 {0xff1a, 0xff20, 1}, /* half/fullwidth ASCII */ 1964 {0xff3b, 0xff40, 1}, /* half/fullwidth ASCII */ 1965 {0xff5b, 0xff65, 1}, /* half/fullwidth ASCII */ 1966 }; 1967 int bot = 0; 1968 int top = sizeof(classes) / sizeof(struct clinterval) - 1; 1969 int mid; 1970 1971 /* First quick check for Latin1 characters, use 'iskeyword'. */ 1972 if (c < 0x100) 1973 { 1974 if (c == ' ' || c == '\t' || c == NUL || c == 0xa0) 1975 return 0; /* blank */ 1976 if (vim_iswordc(c)) 1977 return 2; /* word character */ 1978 return 1; /* punctuation */ 1979 } 1980 1981 /* binary search in table */ 1982 while (top >= bot) 1983 { 1984 mid = (bot + top) / 2; 1985 if (classes[mid].last < c) 1986 bot = mid + 1; 1987 else if (classes[mid].first > c) 1988 top = mid - 1; 1989 else 1990 return (int)classes[mid].class; 1991 } 1992 1993 /* most other characters are "word" characters */ 1994 return 2; 1995 } 1996 1997 /* 1998 * Code for Unicode case-dependent operations. Based on notes in 1999 * http://www.unicode.org/Public/UNIDATA/CaseFolding.txt 2000 * This code uses simple case folding, not full case folding. 2001 */ 2002 2003 /* 2004 * The following table is built by foldExtract.pl < CaseFolding.txt . 2005 * It must be in numeric order, because we use binary search on it. 2006 * An entry such as {0x41,0x5a,1,32} means that UCS-4 characters in the range 2007 * from 0x41 to 0x5a inclusive, stepping by 1, are folded by adding 32. 2008 */ 2009 2010 typedef struct 2011 { 2012 int rangeStart; 2013 int rangeEnd; 2014 int step; 2015 int offset; 2016 } convertStruct; 2017 2018 static convertStruct foldCase[] = 2019 { 2020 {0x41,0x5a,1,32}, {0xc0,0xd6,1,32}, {0xd8,0xde,1,32}, 2021 {0x100,0x12e,2,1}, {0x130,0x130,-1,-199}, {0x132,0x136,2,1}, 2022 {0x139,0x147,2,1}, {0x14a,0x176,2,1}, {0x178,0x178,-1,-121}, 2023 {0x179,0x17d,2,1}, {0x181,0x181,-1,210}, {0x182,0x184,2,1}, 2024 {0x186,0x186,-1,206}, {0x187,0x187,-1,1}, {0x189,0x18a,1,205}, 2025 {0x18b,0x18b,-1,1}, {0x18e,0x18e,-1,79}, {0x18f,0x18f,-1,202}, 2026 {0x190,0x190,-1,203}, {0x191,0x191,-1,1}, {0x193,0x193,-1,205}, 2027 {0x194,0x194,-1,207}, {0x196,0x196,-1,211}, {0x197,0x197,-1,209}, 2028 {0x198,0x198,-1,1}, {0x19c,0x19c,-1,211}, {0x19d,0x19d,-1,213}, 2029 {0x19f,0x19f,-1,214}, {0x1a0,0x1a4,2,1}, {0x1a6,0x1a6,-1,218}, 2030 {0x1a7,0x1a7,-1,1}, {0x1a9,0x1a9,-1,218}, {0x1ac,0x1ac,-1,1}, 2031 {0x1ae,0x1ae,-1,218}, {0x1af,0x1af,-1,1}, {0x1b1,0x1b2,1,217}, 2032 {0x1b3,0x1b5,2,1}, {0x1b7,0x1b7,-1,219}, {0x1b8,0x1bc,4,1}, 2033 {0x1c4,0x1c4,-1,2}, {0x1c5,0x1c5,-1,1}, {0x1c7,0x1c7,-1,2}, 2034 {0x1c8,0x1c8,-1,1}, {0x1ca,0x1ca,-1,2}, {0x1cb,0x1db,2,1}, 2035 {0x1de,0x1ee,2,1}, {0x1f1,0x1f1,-1,2}, {0x1f2,0x1f4,2,1}, 2036 {0x1f6,0x1f6,-1,-97}, {0x1f7,0x1f7,-1,-56}, {0x1f8,0x21e,2,1}, 2037 {0x220,0x220,-1,-130}, {0x222,0x232,2,1}, {0x386,0x386,-1,38}, 2038 {0x388,0x38a,1,37}, {0x38c,0x38c,-1,64}, {0x38e,0x38f,1,63}, 2039 {0x391,0x3a1,1,32}, {0x3a3,0x3ab,1,32}, {0x3d8,0x3ee,2,1}, 2040 {0x3f4,0x3f4,-1,-60}, {0x3f7,0x3f7,-1,1}, {0x3f9,0x3f9,-1,-7}, 2041 {0x3fa,0x3fa,-1,1}, {0x400,0x40f,1,80}, {0x410,0x42f,1,32}, 2042 {0x460,0x480,2,1}, {0x48a,0x4be,2,1}, {0x4c1,0x4cd,2,1}, 2043 {0x4d0,0x4f4,2,1}, {0x4f8,0x500,8,1}, {0x502,0x50e,2,1}, 2044 {0x531,0x556,1,48}, {0x1e00,0x1e94,2,1}, {0x1ea0,0x1ef8,2,1}, 2045 {0x1f08,0x1f0f,1,-8}, {0x1f18,0x1f1d,1,-8}, {0x1f28,0x1f2f,1,-8}, 2046 {0x1f38,0x1f3f,1,-8}, {0x1f48,0x1f4d,1,-8}, {0x1f59,0x1f5f,2,-8}, 2047 {0x1f68,0x1f6f,1,-8}, {0x1f88,0x1f8f,1,-8}, {0x1f98,0x1f9f,1,-8}, 2048 {0x1fa8,0x1faf,1,-8}, {0x1fb8,0x1fb9,1,-8}, {0x1fba,0x1fbb,1,-74}, 2049 {0x1fbc,0x1fbc,-1,-9}, {0x1fc8,0x1fcb,1,-86}, {0x1fcc,0x1fcc,-1,-9}, 2050 {0x1fd8,0x1fd9,1,-8}, {0x1fda,0x1fdb,1,-100}, {0x1fe8,0x1fe9,1,-8}, 2051 {0x1fea,0x1feb,1,-112}, {0x1fec,0x1fec,-1,-7}, {0x1ff8,0x1ff9,1,-128}, 2052 {0x1ffa,0x1ffb,1,-126}, {0x1ffc,0x1ffc,-1,-9}, {0x2126,0x2126,-1,-7517}, 2053 {0x212a,0x212a,-1,-8383}, {0x212b,0x212b,-1,-8262}, 2054 {0x2160,0x216f,1,16}, {0x24b6,0x24cf,1,26}, {0xff21,0xff3a,1,32}, 2055 {0x10400,0x10427,1,40} 2056 }; 2057 2058 static int utf_convert(int a, convertStruct table[], int tableSize); 2059 2060 /* 2061 * Generic conversion function for case operations. 2062 * Return the converted equivalent of "a", which is a UCS-4 character. Use 2063 * the given conversion "table". Uses binary search on "table". 2064 */ 2065 static int 2066 utf_convert(a, table, tableSize) 2067 int a; 2068 convertStruct table[]; 2069 int tableSize; 2070 { 2071 int start, mid, end; /* indices into table */ 2072 2073 start = 0; 2074 end = tableSize / sizeof(convertStruct); 2075 while (start < end) 2076 { 2077 /* need to search further */ 2078 mid = (end + start) /2; 2079 if (table[mid].rangeEnd < a) 2080 start = mid + 1; 2081 else 2082 end = mid; 2083 } 2084 if (table[start].rangeStart <= a && a <= table[start].rangeEnd 2085 && (a - table[start].rangeStart) % table[start].step == 0) 2086 return (a + table[start].offset); 2087 else 2088 return a; 2089 } 2090 2091 /* 2092 * Return the folded-case equivalent of "a", which is a UCS-4 character. Uses 2093 * simple case folding. 2094 */ 2095 int 2096 utf_fold(a) 2097 int a; 2098 { 2099 return utf_convert(a, foldCase, sizeof(foldCase)); 2100 } 2101 2102 /* 2103 * The following tables are built by upperLowerExtract.pl < UnicodeData.txt . 2104 * They must be in numeric order, because we use binary search on them. 2105 * An entry such as {0x41,0x5a,1,32} means that UCS-4 characters in the range 2106 * from 0x41 to 0x5a inclusive, stepping by 1, are switched to lower (for 2107 * example) by adding 32. 2108 */ 2109 static convertStruct toLower[] = 2110 { 2111 {0x41,0x5a,1,32}, {0xc0,0xd6,1,32}, {0xd8,0xde,1,32}, 2112 {0x100,0x12e,2,1}, {0x130,0x130,-1,-199}, {0x132,0x136,2,1}, 2113 {0x139,0x147,2,1}, {0x14a,0x176,2,1}, {0x178,0x178,-1,-121}, 2114 {0x179,0x17d,2,1}, {0x181,0x181,-1,210}, {0x182,0x184,2,1}, 2115 {0x186,0x186,-1,206}, {0x187,0x187,-1,1}, {0x189,0x18a,1,205}, 2116 {0x18b,0x18b,-1,1}, {0x18e,0x18e,-1,79}, {0x18f,0x18f,-1,202}, 2117 {0x190,0x190,-1,203}, {0x191,0x191,-1,1}, {0x193,0x193,-1,205}, 2118 {0x194,0x194,-1,207}, {0x196,0x196,-1,211}, {0x197,0x197,-1,209}, 2119 {0x198,0x198,-1,1}, {0x19c,0x19c,-1,211}, {0x19d,0x19d,-1,213}, 2120 {0x19f,0x19f,-1,214}, {0x1a0,0x1a4,2,1}, {0x1a6,0x1a6,-1,218}, 2121 {0x1a7,0x1a7,-1,1}, {0x1a9,0x1a9,-1,218}, {0x1ac,0x1ac,-1,1}, 2122 {0x1ae,0x1ae,-1,218}, {0x1af,0x1af,-1,1}, {0x1b1,0x1b2,1,217}, 2123 {0x1b3,0x1b5,2,1}, {0x1b7,0x1b7,-1,219}, {0x1b8,0x1bc,4,1}, 2124 {0x1c4,0x1ca,3,2}, {0x1cd,0x1db,2,1}, {0x1de,0x1ee,2,1}, 2125 {0x1f1,0x1f1,-1,2}, {0x1f4,0x1f4,-1,1}, {0x1f6,0x1f6,-1,-97}, 2126 {0x1f7,0x1f7,-1,-56}, {0x1f8,0x21e,2,1}, {0x220,0x220,-1,-130}, 2127 {0x222,0x232,2,1}, {0x386,0x386,-1,38}, {0x388,0x38a,1,37}, 2128 {0x38c,0x38c,-1,64}, {0x38e,0x38f,1,63}, {0x391,0x3a1,1,32}, 2129 {0x3a3,0x3ab,1,32}, {0x3d8,0x3ee,2,1}, {0x3f4,0x3f4,-1,-60}, 2130 {0x3f7,0x3f7,-1,1}, {0x3f9,0x3f9,-1,-7}, {0x3fa,0x3fa,-1,1}, 2131 {0x400,0x40f,1,80}, {0x410,0x42f,1,32}, {0x460,0x480,2,1}, 2132 {0x48a,0x4be,2,1}, {0x4c1,0x4cd,2,1}, {0x4d0,0x4f4,2,1}, 2133 {0x4f8,0x500,8,1}, {0x502,0x50e,2,1}, {0x531,0x556,1,48}, 2134 {0x1e00,0x1e94,2,1}, {0x1ea0,0x1ef8,2,1}, {0x1f08,0x1f0f,1,-8}, 2135 {0x1f18,0x1f1d,1,-8}, {0x1f28,0x1f2f,1,-8}, {0x1f38,0x1f3f,1,-8}, 2136 {0x1f48,0x1f4d,1,-8}, {0x1f59,0x1f5f,2,-8}, {0x1f68,0x1f6f,1,-8}, 2137 {0x1fb8,0x1fb9,1,-8}, {0x1fba,0x1fbb,1,-74}, {0x1fc8,0x1fcb,1,-86}, 2138 {0x1fd8,0x1fd9,1,-8}, {0x1fda,0x1fdb,1,-100}, {0x1fe8,0x1fe9,1,-8}, 2139 {0x1fea,0x1feb,1,-112}, {0x1fec,0x1fec,-1,-7}, {0x1ff8,0x1ff9,1,-128}, 2140 {0x1ffa,0x1ffb,1,-126}, {0x2126,0x2126,-1,-7517}, {0x212a,0x212a,-1,-8383}, 2141 {0x212b,0x212b,-1,-8262}, {0xff21,0xff3a,1,32}, {0x10400,0x10427,1,40} 2142 }; 2143 2144 static convertStruct toUpper[] = 2145 { 2146 {0x61,0x7a,1,-32}, {0xb5,0xb5,-1,743}, {0xe0,0xf6,1,-32}, 2147 {0xf8,0xfe,1,-32}, {0xff,0xff,-1,121}, {0x101,0x12f,2,-1}, 2148 {0x131,0x131,-1,-232}, {0x133,0x137,2,-1}, {0x13a,0x148,2,-1}, 2149 {0x14b,0x177,2,-1}, {0x17a,0x17e,2,-1}, {0x17f,0x17f,-1,-300}, 2150 {0x183,0x185,2,-1}, {0x188,0x18c,4,-1}, {0x192,0x192,-1,-1}, 2151 {0x195,0x195,-1,97}, {0x199,0x199,-1,-1}, {0x19e,0x19e,-1,130}, 2152 {0x1a1,0x1a5,2,-1}, {0x1a8,0x1ad,5,-1}, {0x1b0,0x1b4,4,-1}, 2153 {0x1b6,0x1b9,3,-1}, {0x1bd,0x1bd,-1,-1}, {0x1bf,0x1bf,-1,56}, 2154 {0x1c5,0x1c6,1,-1}, {0x1c8,0x1c9,1,-1}, {0x1cb,0x1cc,1,-1}, 2155 {0x1ce,0x1dc,2,-1}, {0x1dd,0x1dd,-1,-79}, {0x1df,0x1ef,2,-1}, 2156 {0x1f2,0x1f3,1,-1}, {0x1f5,0x1f9,4,-1}, {0x1fb,0x21f,2,-1}, 2157 {0x223,0x233,2,-1}, {0x253,0x253,-1,-210}, {0x254,0x254,-1,-206}, 2158 {0x256,0x257,1,-205}, {0x259,0x259,-1,-202}, {0x25b,0x25b,-1,-203}, 2159 {0x260,0x260,-1,-205}, {0x263,0x263,-1,-207}, {0x268,0x268,-1,-209}, 2160 {0x269,0x26f,6,-211}, {0x272,0x272,-1,-213}, {0x275,0x275,-1,-214}, 2161 {0x280,0x283,3,-218}, {0x288,0x288,-1,-218}, {0x28a,0x28b,1,-217}, 2162 {0x292,0x292,-1,-219}, {0x3ac,0x3ac,-1,-38}, {0x3ad,0x3af,1,-37}, 2163 {0x3b1,0x3c1,1,-32}, {0x3c2,0x3c2,-1,-31}, {0x3c3,0x3cb,1,-32}, 2164 {0x3cc,0x3cc,-1,-64}, {0x3cd,0x3ce,1,-63}, {0x3d0,0x3d0,-1,-62}, 2165 {0x3d1,0x3d1,-1,-57}, {0x3d5,0x3d5,-1,-47}, {0x3d6,0x3d6,-1,-54}, 2166 {0x3d9,0x3ef,2,-1}, {0x3f0,0x3f0,-1,-86}, {0x3f1,0x3f1,-1,-80}, 2167 {0x3f2,0x3f2,-1,7}, {0x3f5,0x3f5,-1,-96}, {0x3f8,0x3fb,3,-1}, 2168 {0x430,0x44f,1,-32}, {0x450,0x45f,1,-80}, {0x461,0x481,2,-1}, 2169 {0x48b,0x4bf,2,-1}, {0x4c2,0x4ce,2,-1}, {0x4d1,0x4f5,2,-1}, 2170 {0x4f9,0x501,8,-1}, {0x503,0x50f,2,-1}, {0x561,0x586,1,-48}, 2171 {0x1e01,0x1e95,2,-1}, {0x1e9b,0x1e9b,-1,-59}, {0x1ea1,0x1ef9,2,-1}, 2172 {0x1f00,0x1f07,1,8}, {0x1f10,0x1f15,1,8}, {0x1f20,0x1f27,1,8}, 2173 {0x1f30,0x1f37,1,8}, {0x1f40,0x1f45,1,8}, {0x1f51,0x1f57,2,8}, 2174 {0x1f60,0x1f67,1,8}, {0x1f70,0x1f71,1,74}, {0x1f72,0x1f75,1,86}, 2175 {0x1f76,0x1f77,1,100}, {0x1f78,0x1f79,1,128}, {0x1f7a,0x1f7b,1,112}, 2176 {0x1f7c,0x1f7d,1,126}, {0x1f80,0x1f87,1,8}, {0x1f90,0x1f97,1,8}, 2177 {0x1fa0,0x1fa7,1,8}, {0x1fb0,0x1fb1,1,8}, {0x1fb3,0x1fb3,-1,9}, 2178 {0x1fbe,0x1fbe,-1,-7205}, {0x1fc3,0x1fc3,-1,9}, {0x1fd0,0x1fd1,1,8}, 2179 {0x1fe0,0x1fe1,1,8}, {0x1fe5,0x1fe5,-1,7}, {0x1ff3,0x1ff3,-1,9}, 2180 {0xff41,0xff5a,1,-32}, {0x10428,0x1044f,1,-40} 2181 }; 2182 2183 /* 2184 * Return the upper-case equivalent of "a", which is a UCS-4 character. Use 2185 * simple case folding. 2186 */ 2187 int 2188 utf_toupper(a) 2189 int a; 2190 { 2191 /* If 'casemap' contains "keepascii" use ASCII style toupper(). */ 2192 if (a < 128 && (cmp_flags & CMP_KEEPASCII)) 2193 return TOUPPER_ASC(a); 2194 2195 #if defined(HAVE_TOWUPPER) && defined(__STDC__ISO_10646__) 2196 /* If towupper() is availble and handles Unicode, use it. */ 2197 if (!(cmp_flags & CMP_INTERNAL)) 2198 return towupper(a); 2199 #endif 2200 2201 /* For characters below 128 use locale sensitive toupper(). */ 2202 if (a < 128) 2203 return TOUPPER_LOC(a); 2204 2205 /* For any other characters use the above mapping table. */ 2206 return utf_convert(a, toUpper, sizeof(toUpper)); 2207 } 2208 2209 int 2210 utf_islower(a) 2211 int a; 2212 { 2213 return (utf_toupper(a) != a); 2214 } 2215 2216 /* 2217 * Return the lower-case equivalent of "a", which is a UCS-4 character. Use 2218 * simple case folding. 2219 */ 2220 int 2221 utf_tolower(a) 2222 int a; 2223 { 2224 /* If 'casemap' contains "keepascii" use ASCII style tolower(). */ 2225 if (a < 128 && (cmp_flags & CMP_KEEPASCII)) 2226 return TOLOWER_ASC(a); 2227 2228 #if defined(HAVE_TOWLOWER) && defined(__STDC__ISO_10646__) 2229 /* If towlower() is available and handles Unicode, use it. */ 2230 if (!(cmp_flags & CMP_INTERNAL)) 2231 return towlower(a); 2232 #endif 2233 2234 /* For characters below 128 use locale sensitive tolower(). */ 2235 if (a < 128) 2236 return TOLOWER_LOC(a); 2237 2238 /* For any other characters use the above mapping table. */ 2239 return utf_convert(a, toLower, sizeof(toLower)); 2240 } 2241 2242 int 2243 utf_isupper(a) 2244 int a; 2245 { 2246 return (utf_tolower(a) != a); 2247 } 2248 2249 /* 2250 * Version of strnicmp() that handles multi-byte characters. 2251 * Needed for Big5, Sjift-JIS and UTF-8 encoding. Other DBCS encodings can 2252 * probably use strnicmp(), because there are no ASCII characters in the 2253 * second byte. 2254 * Returns zero if s1 and s2 are equal (ignoring case), the difference between 2255 * two characters otherwise. 2256 */ 2257 int 2258 mb_strnicmp(s1, s2, nn) 2259 char_u *s1, *s2; 2260 size_t nn; 2261 { 2262 int i, j, l; 2263 int cdiff; 2264 int incomplete = FALSE; 2265 int n = nn; 2266 2267 for (i = 0; i < n; i += l) 2268 { 2269 if (s1[i] == NUL && s2[i] == NUL) /* both strings end */ 2270 return 0; 2271 if (enc_utf8) 2272 { 2273 l = utf_byte2len(s1[i]); 2274 if (l > n - i) 2275 { 2276 l = n - i; /* incomplete character */ 2277 incomplete = TRUE; 2278 } 2279 /* Check directly first, it's faster. */ 2280 for (j = 0; j < l; ++j) 2281 if (s1[i + j] != s2[i + j]) 2282 break; 2283 if (j < l) 2284 { 2285 /* If one of the two characters is incomplete return -1. */ 2286 if (incomplete || i + utf_byte2len(s2[i]) > n) 2287 return -1; 2288 cdiff = utf_fold(utf_ptr2char(s1 + i)) 2289 - utf_fold(utf_ptr2char(s2 + i)); 2290 if (cdiff != 0) 2291 return cdiff; 2292 } 2293 } 2294 else 2295 { 2296 l = (*mb_ptr2len)(s1 + i); 2297 if (l <= 1) 2298 { 2299 /* Single byte: first check normally, then with ignore case. */ 2300 if (s1[i] != s2[i]) 2301 { 2302 cdiff = TOLOWER_LOC(s1[i]) - TOLOWER_LOC(s2[i]); 2303 if (cdiff != 0) 2304 return cdiff; 2305 } 2306 } 2307 else 2308 { 2309 /* For non-Unicode multi-byte don't ignore case. */ 2310 if (l > n - i) 2311 l = n - i; 2312 cdiff = STRNCMP(s1 + i, s2 + i, l); 2313 if (cdiff != 0) 2314 return cdiff; 2315 } 2316 } 2317 } 2318 return 0; 2319 } 2320 2321 /* 2322 * "g8": show bytes of the UTF-8 char under the cursor. Doesn't matter what 2323 * 'encoding' has been set to. 2324 */ 2325 void 2326 show_utf8() 2327 { 2328 int len; 2329 int rlen = 0; 2330 char_u *line; 2331 int clen; 2332 int i; 2333 2334 /* Get the byte length of the char under the cursor, including composing 2335 * characters. */ 2336 line = ml_get_cursor(); 2337 len = utfc_ptr2len(line); 2338 if (len == 0) 2339 { 2340 MSG("NUL"); 2341 return; 2342 } 2343 2344 clen = 0; 2345 for (i = 0; i < len; ++i) 2346 { 2347 if (clen == 0) 2348 { 2349 /* start of (composing) character, get its length */ 2350 if (i > 0) 2351 { 2352 STRCPY(IObuff + rlen, "+ "); 2353 rlen += 2; 2354 } 2355 clen = utf_ptr2len(line + i); 2356 } 2357 sprintf((char *)IObuff + rlen, "%02x ", line[i]); 2358 --clen; 2359 rlen += STRLEN(IObuff + rlen); 2360 if (rlen > IOSIZE - 20) 2361 break; 2362 } 2363 2364 msg(IObuff); 2365 } 2366 2367 /* 2368 * mb_head_off() function pointer. 2369 * Return offset from "p" to the first byte of the character it points into. 2370 * Returns 0 when already at the first byte of a character. 2371 */ 2372 /*ARGSUSED*/ 2373 int 2374 latin_head_off(base, p) 2375 char_u *base; 2376 char_u *p; 2377 { 2378 return 0; 2379 } 2380 2381 int 2382 dbcs_head_off(base, p) 2383 char_u *base; 2384 char_u *p; 2385 { 2386 char_u *q; 2387 2388 /* It can't be a trailing byte when not using DBCS, at the start of the 2389 * string or the previous byte can't start a double-byte. */ 2390 if (p <= base || MB_BYTE2LEN(p[-1]) == 1) 2391 return 0; 2392 2393 /* This is slow: need to start at the base and go forward until the 2394 * byte we are looking for. Return 1 when we went past it, 0 otherwise. */ 2395 q = base; 2396 while (q < p) 2397 q += dbcs_ptr2len(q); 2398 return (q == p) ? 0 : 1; 2399 } 2400 2401 #if defined(FEAT_CLIPBOARD) || defined(FEAT_GUI) || defined(FEAT_RIGHTLEFT) \ 2402 || defined(PROTO) 2403 /* 2404 * Special version of dbcs_head_off() that works for ScreenLines[], where 2405 * single-width DBCS_JPNU characters are stored separately. 2406 */ 2407 int 2408 dbcs_screen_head_off(base, p) 2409 char_u *base; 2410 char_u *p; 2411 { 2412 char_u *q; 2413 2414 /* It can't be a trailing byte when not using DBCS, at the start of the 2415 * string or the previous byte can't start a double-byte. 2416 * For euc-jp an 0x8e byte in the previous cell always means we have a 2417 * lead byte in the current cell. */ 2418 if (p <= base 2419 || (enc_dbcs == DBCS_JPNU && p[-1] == 0x8e) 2420 || MB_BYTE2LEN(p[-1]) == 1) 2421 return 0; 2422 2423 /* This is slow: need to start at the base and go forward until the 2424 * byte we are looking for. Return 1 when we went past it, 0 otherwise. 2425 * For DBCS_JPNU look out for 0x8e, which means the second byte is not 2426 * stored as the next byte. */ 2427 q = base; 2428 while (q < p) 2429 { 2430 if (enc_dbcs == DBCS_JPNU && *q == 0x8e) 2431 ++q; 2432 else 2433 q += dbcs_ptr2len(q); 2434 } 2435 return (q == p) ? 0 : 1; 2436 } 2437 #endif 2438 2439 int 2440 utf_head_off(base, p) 2441 char_u *base; 2442 char_u *p; 2443 { 2444 char_u *q; 2445 char_u *s; 2446 int c; 2447 #ifdef FEAT_ARABIC 2448 char_u *j; 2449 #endif 2450 2451 if (*p < 0x80) /* be quick for ASCII */ 2452 return 0; 2453 2454 /* Skip backwards over trailing bytes: 10xx.xxxx 2455 * Skip backwards again if on a composing char. */ 2456 for (q = p; ; --q) 2457 { 2458 /* Move s to the last byte of this char. */ 2459 for (s = q; (s[1] & 0xc0) == 0x80; ++s) 2460 ; 2461 /* Move q to the first byte of this char. */ 2462 while (q > base && (*q & 0xc0) == 0x80) 2463 --q; 2464 /* Check for illegal sequence. Do allow an illegal byte after where we 2465 * started. */ 2466 if (utf8len_tab[*q] != (int)(s - q + 1) 2467 && utf8len_tab[*q] != (int)(p - q + 1)) 2468 return 0; 2469 2470 if (q <= base) 2471 break; 2472 2473 c = utf_ptr2char(q); 2474 if (utf_iscomposing(c)) 2475 continue; 2476 2477 #ifdef FEAT_ARABIC 2478 if (arabic_maycombine(c)) 2479 { 2480 /* Advance to get a sneak-peak at the next char */ 2481 j = q; 2482 --j; 2483 /* Move j to the first byte of this char. */ 2484 while (j > base && (*j & 0xc0) == 0x80) 2485 --j; 2486 if (arabic_combine(utf_ptr2char(j), c)) 2487 continue; 2488 } 2489 #endif 2490 break; 2491 } 2492 2493 return (int)(p - q); 2494 } 2495 2496 #if defined(FEAT_EVAL) || defined(PROTO) 2497 /* 2498 * Copy a character from "*fp" to "*tp" and advance the pointers. 2499 */ 2500 void 2501 mb_copy_char(fp, tp) 2502 char_u **fp; 2503 char_u **tp; 2504 { 2505 int l = (*mb_ptr2len)(*fp); 2506 2507 mch_memmove(*tp, *fp, (size_t)l); 2508 *tp += l; 2509 *fp += l; 2510 } 2511 #endif 2512 2513 /* 2514 * Return the offset from "p" to the first byte of a character. When "p" is 2515 * at the start of a character 0 is returned, otherwise the offset to the next 2516 * character. Can start anywhere in a stream of bytes. 2517 */ 2518 int 2519 mb_off_next(base, p) 2520 char_u *base; 2521 char_u *p; 2522 { 2523 int i; 2524 int j; 2525 2526 if (enc_utf8) 2527 { 2528 if (*p < 0x80) /* be quick for ASCII */ 2529 return 0; 2530 2531 /* Find the next character that isn't 10xx.xxxx */ 2532 for (i = 0; (p[i] & 0xc0) == 0x80; ++i) 2533 ; 2534 if (i > 0) 2535 { 2536 /* Check for illegal sequence. */ 2537 for (j = 0; p - j > base; ++j) 2538 if ((p[-j] & 0xc0) != 0x80) 2539 break; 2540 if (utf8len_tab[p[-j]] != i + j) 2541 return 0; 2542 } 2543 return i; 2544 } 2545 2546 /* Only need to check if we're on a trail byte, it doesn't matter if we 2547 * want the offset to the next or current character. */ 2548 return (*mb_head_off)(base, p); 2549 } 2550 2551 /* 2552 * Return the offset from "p" to the last byte of the character it points 2553 * into. Can start anywhere in a stream of bytes. 2554 */ 2555 int 2556 mb_tail_off(base, p) 2557 char_u *base; 2558 char_u *p; 2559 { 2560 int i; 2561 int j; 2562 2563 if (*p == NUL) 2564 return 0; 2565 2566 if (enc_utf8) 2567 { 2568 /* Find the last character that is 10xx.xxxx */ 2569 for (i = 0; (p[i + 1] & 0xc0) == 0x80; ++i) 2570 ; 2571 /* Check for illegal sequence. */ 2572 for (j = 0; p - j > base; ++j) 2573 if ((p[-j] & 0xc0) != 0x80) 2574 break; 2575 if (utf8len_tab[p[-j]] != i + j + 1) 2576 return 0; 2577 return i; 2578 } 2579 2580 /* It can't be the first byte if a double-byte when not using DBCS, at the 2581 * end of the string or the byte can't start a double-byte. */ 2582 if (enc_dbcs == 0 || p[1] == NUL || MB_BYTE2LEN(*p) == 1) 2583 return 0; 2584 2585 /* Return 1 when on the lead byte, 0 when on the tail byte. */ 2586 return 1 - dbcs_head_off(base, p); 2587 } 2588 2589 #if defined(HAVE_GTK2) || defined(PROTO) 2590 /* 2591 * Return TRUE if string "s" is a valid utf-8 string. 2592 * When "end" is NULL stop at the first NUL. 2593 * When "end" is positive stop there. 2594 */ 2595 int 2596 utf_valid_string(s, end) 2597 char_u *s; 2598 char_u *end; 2599 { 2600 int l; 2601 char_u *p = s; 2602 2603 while (end == NULL ? *p != NUL : p < end) 2604 { 2605 if ((*p & 0xc0) == 0x80) 2606 return FALSE; /* invalid lead byte */ 2607 l = utf8len_tab[*p]; 2608 if (end != NULL && p + l > end) 2609 return FALSE; /* incomplete byte sequence */ 2610 ++p; 2611 while (--l > 0) 2612 if ((*p++ & 0xc0) != 0x80) 2613 return FALSE; /* invalid trail byte */ 2614 } 2615 return TRUE; 2616 } 2617 #endif 2618 2619 #if defined(FEAT_GUI) || defined(PROTO) 2620 /* 2621 * Special version of mb_tail_off() for use in ScreenLines[]. 2622 */ 2623 int 2624 dbcs_screen_tail_off(base, p) 2625 char_u *base; 2626 char_u *p; 2627 { 2628 /* It can't be the first byte if a double-byte when not using DBCS, at the 2629 * end of the string or the byte can't start a double-byte. 2630 * For euc-jp an 0x8e byte always means we have a lead byte in the current 2631 * cell. */ 2632 if (*p == NUL || p[1] == NUL 2633 || (enc_dbcs == DBCS_JPNU && *p == 0x8e) 2634 || MB_BYTE2LEN(*p) == 1) 2635 return 0; 2636 2637 /* Return 1 when on the lead byte, 0 when on the tail byte. */ 2638 return 1 - dbcs_screen_head_off(base, p); 2639 } 2640 #endif 2641 2642 /* 2643 * If the cursor moves on an trail byte, set the cursor on the lead byte. 2644 * Thus it moves left if necessary. 2645 * Return TRUE when the cursor was adjusted. 2646 */ 2647 void 2648 mb_adjust_cursor() 2649 { 2650 mb_adjustpos(&curwin->w_cursor); 2651 } 2652 2653 /* 2654 * Adjust position "*lp" to point to the first byte of a multi-byte character. 2655 * If it points to a tail byte it's moved backwards to the head byte. 2656 */ 2657 void 2658 mb_adjustpos(lp) 2659 pos_T *lp; 2660 { 2661 char_u *p; 2662 2663 if (lp->col > 0 2664 #ifdef FEAT_VIRTUALEDIT 2665 || lp->coladd > 1 2666 #endif 2667 ) 2668 { 2669 p = ml_get(lp->lnum); 2670 lp->col -= (*mb_head_off)(p, p + lp->col); 2671 #ifdef FEAT_VIRTUALEDIT 2672 /* Reset "coladd" when the cursor would be on the right half of a 2673 * double-wide character. */ 2674 if (lp->coladd == 1 2675 && p[lp->col] != TAB 2676 && vim_isprintc((*mb_ptr2char)(p + lp->col)) 2677 && ptr2cells(p + lp->col) > 1) 2678 lp->coladd = 0; 2679 #endif 2680 } 2681 } 2682 2683 /* 2684 * Return a pointer to the character before "*p", if there is one. 2685 */ 2686 char_u * 2687 mb_prevptr(line, p) 2688 char_u *line; /* start of the string */ 2689 char_u *p; 2690 { 2691 if (p > line) 2692 mb_ptr_back(line, p); 2693 return p; 2694 } 2695 2696 /* 2697 * Return the character length of "str". Each multi-byte character (with 2698 * following composing characters) counts as one. 2699 */ 2700 int 2701 mb_charlen(str) 2702 char_u *str; 2703 { 2704 char_u *p = str; 2705 int count; 2706 2707 if (p == NULL) 2708 return 0; 2709 2710 for (count = 0; *p != NUL; count++) 2711 p += (*mb_ptr2len)(p); 2712 2713 return count; 2714 } 2715 2716 #if defined(FEAT_SYN_HL) || defined(PROTO) 2717 /* 2718 * Like mb_charlen() but for a string with specified length. 2719 */ 2720 int 2721 mb_charlen_len(str, len) 2722 char_u *str; 2723 int len; 2724 { 2725 char_u *p = str; 2726 int count; 2727 2728 for (count = 0; *p != NUL && p < str + len; count++) 2729 p += (*mb_ptr2len)(p); 2730 2731 return count; 2732 } 2733 #endif 2734 2735 /* 2736 * Try to un-escape a multi-byte character. 2737 * Used for the "to" and "from" part of a mapping. 2738 * Return the un-escaped string if it is a multi-byte character, and advance 2739 * "pp" to just after the bytes that formed it. 2740 * Return NULL if no multi-byte char was found. 2741 */ 2742 char_u * 2743 mb_unescape(pp) 2744 char_u **pp; 2745 { 2746 static char_u buf[MB_MAXBYTES + 1]; 2747 int n, m = 0; 2748 char_u *str = *pp; 2749 2750 /* Must translate K_SPECIAL KS_SPECIAL KE_FILLER to K_SPECIAL and CSI 2751 * KS_EXTRA KE_CSI to CSI. */ 2752 for (n = 0; str[n] != NUL && m <= MB_MAXBYTES; ++n) 2753 { 2754 if (str[n] == K_SPECIAL 2755 && str[n + 1] == KS_SPECIAL 2756 && str[n + 2] == KE_FILLER) 2757 { 2758 buf[m++] = K_SPECIAL; 2759 n += 2; 2760 } 2761 # ifdef FEAT_GUI 2762 else if (str[n] == CSI 2763 && str[n + 1] == KS_EXTRA 2764 && str[n + 2] == (int)KE_CSI) 2765 { 2766 buf[m++] = CSI; 2767 n += 2; 2768 } 2769 # endif 2770 else if (str[n] == K_SPECIAL 2771 # ifdef FEAT_GUI 2772 || str[n] == CSI 2773 # endif 2774 ) 2775 break; /* a special key can't be a multibyte char */ 2776 else 2777 buf[m++] = str[n]; 2778 buf[m] = NUL; 2779 2780 /* Return a multi-byte character if it's found. An illegal sequence 2781 * will result in a 1 here. */ 2782 if ((*mb_ptr2len)(buf) > 1) 2783 { 2784 *pp = str + n + 1; 2785 return buf; 2786 } 2787 } 2788 return NULL; 2789 } 2790 2791 /* 2792 * Return TRUE if the character at "row"/"col" on the screen is the left side 2793 * of a double-width character. 2794 * Caller must make sure "row" and "col" are not invalid! 2795 */ 2796 int 2797 mb_lefthalve(row, col) 2798 int row; 2799 int col; 2800 { 2801 #ifdef FEAT_HANGULIN 2802 if (composing_hangul) 2803 return TRUE; 2804 #endif 2805 if (enc_dbcs != 0) 2806 return dbcs_off2cells(LineOffset[row] + col) > 1; 2807 if (enc_utf8) 2808 return (col + 1 < Columns 2809 && ScreenLines[LineOffset[row] + col + 1] == 0); 2810 return FALSE; 2811 } 2812 2813 # if defined(FEAT_CLIPBOARD) || defined(FEAT_GUI) || defined(FEAT_RIGHTLEFT) \ 2814 || defined(PROTO) 2815 /* 2816 * Correct a position on the screen, if it's the right halve of a double-wide 2817 * char move it to the left halve. Returns the corrected column. 2818 */ 2819 int 2820 mb_fix_col(col, row) 2821 int col; 2822 int row; 2823 { 2824 col = check_col(col); 2825 row = check_row(row); 2826 if (has_mbyte && ScreenLines != NULL && col > 0 2827 && ((enc_dbcs 2828 && ScreenLines[LineOffset[row] + col] != NUL 2829 && dbcs_screen_head_off(ScreenLines + LineOffset[row], 2830 ScreenLines + LineOffset[row] + col)) 2831 || (enc_utf8 && ScreenLines[LineOffset[row] + col] == 0))) 2832 --col; 2833 return col; 2834 } 2835 # endif 2836 #endif 2837 2838 #if defined(FEAT_MBYTE) || defined(FEAT_POSTSCRIPT) || defined(PROTO) 2839 static int enc_alias_search __ARGS((char_u *name)); 2840 2841 /* 2842 * Skip the Vim specific head of a 'encoding' name. 2843 */ 2844 char_u * 2845 enc_skip(p) 2846 char_u *p; 2847 { 2848 if (STRNCMP(p, "2byte-", 6) == 0) 2849 return p + 6; 2850 if (STRNCMP(p, "8bit-", 5) == 0) 2851 return p + 5; 2852 return p; 2853 } 2854 2855 /* 2856 * Find the canonical name for encoding "enc". 2857 * When the name isn't recognized, returns "enc" itself, but with all lower 2858 * case characters and '_' replaced with '-'. 2859 * Returns an allocated string. NULL for out-of-memory. 2860 */ 2861 char_u * 2862 enc_canonize(enc) 2863 char_u *enc; 2864 { 2865 char_u *r; 2866 char_u *p, *s; 2867 int i; 2868 2869 # ifdef FEAT_MBYTE 2870 if (STRCMP(enc, "default") == 0) 2871 { 2872 /* Use the default encoding as it's found by set_init_1(). */ 2873 r = get_encoding_default(); 2874 if (r == NULL) 2875 r = (char_u *)"latin1"; 2876 return vim_strsave(r); 2877 } 2878 # endif 2879 2880 /* copy "enc" to allocted memory, with room for two '-' */ 2881 r = alloc((unsigned)(STRLEN(enc) + 3)); 2882 if (r != NULL) 2883 { 2884 /* Make it all lower case and replace '_' with '-'. */ 2885 p = r; 2886 for (s = enc; *s != NUL; ++s) 2887 { 2888 if (*s == '_') 2889 *p++ = '-'; 2890 else 2891 *p++ = TOLOWER_ASC(*s); 2892 } 2893 *p = NUL; 2894 2895 /* Skip "2byte-" and "8bit-". */ 2896 p = enc_skip(r); 2897 2898 /* Change "microsoft-cp" to "cp". Used in some spell files. */ 2899 if (STRNCMP(p, "microsoft-cp", 12) == 0) 2900 mch_memmove(p, p + 10, STRLEN(p + 10) + 1); 2901 2902 /* "iso8859" -> "iso-8859" */ 2903 if (STRNCMP(p, "iso8859", 7) == 0) 2904 { 2905 mch_memmove(p + 4, p + 3, STRLEN(p + 2)); 2906 p[3] = '-'; 2907 } 2908 2909 /* "iso-8859n" -> "iso-8859-n" */ 2910 if (STRNCMP(p, "iso-8859", 8) == 0 && p[8] != '-') 2911 { 2912 mch_memmove(p + 9, p + 8, STRLEN(p + 7)); 2913 p[8] = '-'; 2914 } 2915 2916 /* "latin-N" -> "latinN" */ 2917 if (STRNCMP(p, "latin-", 6) == 0) 2918 mch_memmove(p + 5, p + 6, STRLEN(p + 5)); 2919 2920 if (enc_canon_search(p) >= 0) 2921 { 2922 /* canonical name can be used unmodified */ 2923 if (p != r) 2924 mch_memmove(r, p, STRLEN(p) + 1); 2925 } 2926 else if ((i = enc_alias_search(p)) >= 0) 2927 { 2928 /* alias recognized, get canonical name */ 2929 vim_free(r); 2930 r = vim_strsave((char_u *)enc_canon_table[i].name); 2931 } 2932 } 2933 return r; 2934 } 2935 2936 /* 2937 * Search for an encoding alias of "name". 2938 * Returns -1 when not found. 2939 */ 2940 static int 2941 enc_alias_search(name) 2942 char_u *name; 2943 { 2944 int i; 2945 2946 for (i = 0; enc_alias_table[i].name != NULL; ++i) 2947 if (STRCMP(name, enc_alias_table[i].name) == 0) 2948 return enc_alias_table[i].canon; 2949 return -1; 2950 } 2951 #endif 2952 2953 #if defined(FEAT_MBYTE) || defined(PROTO) 2954 2955 #ifdef HAVE_LANGINFO_H 2956 # include <langinfo.h> 2957 #endif 2958 2959 /* 2960 * Get the canonicalized encoding of the current locale. 2961 * Returns an allocated string when successful, NULL when not. 2962 */ 2963 char_u * 2964 enc_locale() 2965 { 2966 #ifndef WIN3264 2967 char *s; 2968 char *p; 2969 int i; 2970 #endif 2971 char buf[50]; 2972 #ifdef WIN3264 2973 long acp = GetACP(); 2974 2975 if (acp == 1200) 2976 STRCPY(buf, "ucs-2le"); 2977 else if (acp == 1252) /* cp1252 is used as latin1 */ 2978 STRCPY(buf, "latin1"); 2979 else 2980 sprintf(buf, "cp%ld", acp); 2981 #else 2982 # ifdef HAVE_NL_LANGINFO_CODESET 2983 if ((s = nl_langinfo(CODESET)) == NULL || *s == NUL) 2984 # endif 2985 # if defined(HAVE_LOCALE_H) || defined(X_LOCALE) 2986 if ((s = setlocale(LC_CTYPE, NULL)) == NULL || *s == NUL) 2987 # endif 2988 if ((s = getenv("LC_ALL")) == NULL || *s == NUL) 2989 if ((s = getenv("LC_CTYPE")) == NULL || *s == NUL) 2990 s = getenv("LANG"); 2991 2992 if (s == NULL || *s == NUL) 2993 return FAIL; 2994 2995 /* The most generic locale format is: 2996 * language[_territory][.codeset][@modifier][+special][,[sponsor][_revision]] 2997 * If there is a '.' remove the part before it. 2998 * if there is something after the codeset, remove it. 2999 * Make the name lowercase and replace '_' with '-'. 3000 * Exception: "ja_JP.EUC" == "euc-jp", "zh_CN.EUC" = "euc-cn", 3001 * "ko_KR.EUC" == "euc-kr" 3002 */ 3003 if ((p = (char *)vim_strchr((char_u *)s, '.')) != NULL) 3004 { 3005 if (p > s + 2 && STRNICMP(p + 1, "EUC", 3) == 0 3006 && !isalnum((int)p[4]) && p[4] != '-' && p[-3] == '_') 3007 { 3008 /* copy "XY.EUC" to "euc-XY" to buf[10] */ 3009 STRCPY(buf + 10, "euc-"); 3010 buf[14] = p[-2]; 3011 buf[15] = p[-1]; 3012 buf[16] = 0; 3013 s = buf + 10; 3014 } 3015 else 3016 s = p + 1; 3017 } 3018 for (i = 0; s[i] != NUL && i < sizeof(buf) - 1; ++i) 3019 { 3020 if (s[i] == '_' || s[i] == '-') 3021 buf[i] = '-'; 3022 else if (isalnum((int)s[i])) 3023 buf[i] = TOLOWER_ASC(s[i]); 3024 else 3025 break; 3026 } 3027 buf[i] = NUL; 3028 #endif 3029 3030 return enc_canonize((char_u *)buf); 3031 } 3032 3033 #if defined(WIN3264) || defined(PROTO) 3034 /* 3035 * Convert an encoding name to an MS-Windows codepage. 3036 * Returns zero if no codepage can be figured out. 3037 */ 3038 int 3039 encname2codepage(name) 3040 char_u *name; 3041 { 3042 int cp; 3043 char_u *p = name; 3044 int idx; 3045 3046 if (STRNCMP(p, "8bit-", 5) == 0) 3047 p += 5; 3048 else if (STRNCMP(p_enc, "2byte-", 6) == 0) 3049 p += 6; 3050 3051 if (p[0] == 'c' && p[1] == 'p') 3052 cp = atoi(p + 2); 3053 else if ((idx = enc_canon_search(p)) >= 0) 3054 cp = enc_canon_table[idx].codepage; 3055 else 3056 return 0; 3057 if (IsValidCodePage(cp)) 3058 return cp; 3059 return 0; 3060 } 3061 #endif 3062 3063 # if defined(USE_ICONV) || defined(PROTO) 3064 3065 static char_u *iconv_string __ARGS((vimconv_T *vcp, char_u *str, int slen, int *unconvlenp)); 3066 3067 /* 3068 * Call iconv_open() with a check if iconv() works properly (there are broken 3069 * versions). 3070 * Returns (void *)-1 if failed. 3071 * (should return iconv_t, but that causes problems with prototypes). 3072 */ 3073 void * 3074 my_iconv_open(to, from) 3075 char_u *to; 3076 char_u *from; 3077 { 3078 iconv_t fd; 3079 #define ICONV_TESTLEN 400 3080 char_u tobuf[ICONV_TESTLEN]; 3081 char *p; 3082 size_t tolen; 3083 static int iconv_ok = -1; 3084 3085 if (iconv_ok == FALSE) 3086 return (void *)-1; /* detected a broken iconv() previously */ 3087 3088 #ifdef DYNAMIC_ICONV 3089 /* Check if the iconv.dll can be found. */ 3090 if (!iconv_enabled(TRUE)) 3091 return (void *)-1; 3092 #endif 3093 3094 fd = iconv_open((char *)enc_skip(to), (char *)enc_skip(from)); 3095 3096 if (fd != (iconv_t)-1 && iconv_ok == -1) 3097 { 3098 /* 3099 * Do a dummy iconv() call to check if it actually works. There is a 3100 * version of iconv() on Linux that is broken. We can't ignore it, 3101 * because it's wide-spread. The symptoms are that after outputting 3102 * the initial shift state the "to" pointer is NULL and conversion 3103 * stops for no apparent reason after about 8160 characters. 3104 */ 3105 p = (char *)tobuf; 3106 tolen = ICONV_TESTLEN; 3107 (void)iconv(fd, NULL, NULL, &p, &tolen); 3108 if (p == NULL) 3109 { 3110 iconv_ok = FALSE; 3111 iconv_close(fd); 3112 fd = (iconv_t)-1; 3113 } 3114 else 3115 iconv_ok = TRUE; 3116 } 3117 3118 return (void *)fd; 3119 } 3120 3121 /* 3122 * Convert the string "str[slen]" with iconv(). 3123 * If "unconvlenp" is not NULL handle the string ending in an incomplete 3124 * sequence and set "*unconvlenp" to the length of it. 3125 * Returns the converted string in allocated memory. NULL for an error. 3126 */ 3127 static char_u * 3128 iconv_string(vcp, str, slen, unconvlenp) 3129 vimconv_T *vcp; 3130 char_u *str; 3131 int slen; 3132 int *unconvlenp; 3133 { 3134 const char *from; 3135 size_t fromlen; 3136 char *to; 3137 size_t tolen; 3138 size_t len = 0; 3139 size_t done = 0; 3140 char_u *result = NULL; 3141 char_u *p; 3142 int l; 3143 3144 from = (char *)str; 3145 fromlen = slen; 3146 for (;;) 3147 { 3148 if (len == 0 || ICONV_ERRNO == ICONV_E2BIG) 3149 { 3150 /* Allocate enough room for most conversions. When re-allocating 3151 * increase the buffer size. */ 3152 len = len + fromlen * 2 + 40; 3153 p = alloc((unsigned)len); 3154 if (p != NULL && done > 0) 3155 mch_memmove(p, result, done); 3156 vim_free(result); 3157 result = p; 3158 if (result == NULL) /* out of memory */ 3159 break; 3160 } 3161 3162 to = (char *)result + done; 3163 tolen = len - done - 2; 3164 /* Avoid a warning for systems with a wrong iconv() prototype by 3165 * casting the second argument to void *. */ 3166 if (iconv(vcp->vc_fd, (void *)&from, &fromlen, &to, &tolen) 3167 != (size_t)-1) 3168 { 3169 /* Finished, append a NUL. */ 3170 *to = NUL; 3171 break; 3172 } 3173 3174 /* Check both ICONV_EINVAL and EINVAL, because the dynamically loaded 3175 * iconv library may use one of them. */ 3176 if (!vcp->vc_fail && unconvlenp != NULL 3177 && (ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL)) 3178 { 3179 /* Handle an incomplete sequence at the end. */ 3180 *to = NUL; 3181 *unconvlenp = fromlen; 3182 break; 3183 } 3184 3185 /* Check both ICONV_EILSEQ and EILSEQ, because the dynamically loaded 3186 * iconv library may use one of them. */ 3187 else if (!vcp->vc_fail 3188 && (ICONV_ERRNO == ICONV_EILSEQ || ICONV_ERRNO == EILSEQ 3189 || ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL)) 3190 { 3191 /* Can't convert: insert a '?' and skip a character. This assumes 3192 * conversion from 'encoding' to something else. In other 3193 * situations we don't know what to skip anyway. */ 3194 *to++ = '?'; 3195 if ((*mb_ptr2cells)((char_u *)from) > 1) 3196 *to++ = '?'; 3197 if (enc_utf8) 3198 l = utfc_ptr2len_len((char_u *)from, fromlen); 3199 else 3200 { 3201 l = (*mb_ptr2len)((char_u *)from); 3202 if (l > (int)fromlen) 3203 l = fromlen; 3204 } 3205 from += l; 3206 fromlen -= l; 3207 } 3208 else if (ICONV_ERRNO != ICONV_E2BIG) 3209 { 3210 /* conversion failed */ 3211 vim_free(result); 3212 result = NULL; 3213 break; 3214 } 3215 /* Not enough room or skipping illegal sequence. */ 3216 done = to - (char *)result; 3217 } 3218 return result; 3219 } 3220 3221 # if defined(DYNAMIC_ICONV) || defined(PROTO) 3222 /* 3223 * Dynamically load the "iconv.dll" on Win32. 3224 */ 3225 3226 #ifndef DYNAMIC_ICONV /* just generating prototypes */ 3227 # define HINSTANCE int 3228 #endif 3229 static HINSTANCE hIconvDLL = 0; 3230 static HINSTANCE hMsvcrtDLL = 0; 3231 3232 # ifndef DYNAMIC_ICONV_DLL 3233 # define DYNAMIC_ICONV_DLL "iconv.dll" 3234 # define DYNAMIC_ICONV_DLL_ALT "libiconv.dll" 3235 # endif 3236 # ifndef DYNAMIC_MSVCRT_DLL 3237 # define DYNAMIC_MSVCRT_DLL "msvcrt.dll" 3238 # endif 3239 3240 /* 3241 * Try opening the iconv.dll and return TRUE if iconv() can be used. 3242 */ 3243 int 3244 iconv_enabled(verbose) 3245 int verbose; 3246 { 3247 if (hIconvDLL != 0 && hMsvcrtDLL != 0) 3248 return TRUE; 3249 hIconvDLL = LoadLibrary(DYNAMIC_ICONV_DLL); 3250 if (hIconvDLL == 0) /* sometimes it's called libiconv.dll */ 3251 hIconvDLL = LoadLibrary(DYNAMIC_ICONV_DLL_ALT); 3252 if (hIconvDLL != 0) 3253 hMsvcrtDLL = LoadLibrary(DYNAMIC_MSVCRT_DLL); 3254 if (hIconvDLL == 0 || hMsvcrtDLL == 0) 3255 { 3256 /* Only give the message when 'verbose' is set, otherwise it might be 3257 * done whenever a conversion is attempted. */ 3258 if (verbose && p_verbose > 0) 3259 { 3260 verbose_enter(); 3261 EMSG2(_(e_loadlib), 3262 hIconvDLL == 0 ? DYNAMIC_ICONV_DLL : DYNAMIC_MSVCRT_DLL); 3263 verbose_leave(); 3264 } 3265 iconv_end(); 3266 return FALSE; 3267 } 3268 3269 iconv = (void *)GetProcAddress(hIconvDLL, "libiconv"); 3270 iconv_open = (void *)GetProcAddress(hIconvDLL, "libiconv_open"); 3271 iconv_close = (void *)GetProcAddress(hIconvDLL, "libiconv_close"); 3272 iconvctl = (void *)GetProcAddress(hIconvDLL, "libiconvctl"); 3273 iconv_errno = (void *)GetProcAddress(hMsvcrtDLL, "_errno"); 3274 if (iconv == NULL || iconv_open == NULL || iconv_close == NULL 3275 || iconvctl == NULL || iconv_errno == NULL) 3276 { 3277 iconv_end(); 3278 if (verbose && p_verbose > 0) 3279 { 3280 verbose_enter(); 3281 EMSG2(_(e_loadfunc), "for libiconv"); 3282 verbose_leave(); 3283 } 3284 return FALSE; 3285 } 3286 return TRUE; 3287 } 3288 3289 void 3290 iconv_end() 3291 { 3292 /* Don't use iconv() when inputting or outputting characters. */ 3293 if (input_conv.vc_type == CONV_ICONV) 3294 convert_setup(&input_conv, NULL, NULL); 3295 if (output_conv.vc_type == CONV_ICONV) 3296 convert_setup(&output_conv, NULL, NULL); 3297 3298 if (hIconvDLL != 0) 3299 FreeLibrary(hIconvDLL); 3300 if (hMsvcrtDLL != 0) 3301 FreeLibrary(hMsvcrtDLL); 3302 hIconvDLL = 0; 3303 hMsvcrtDLL = 0; 3304 } 3305 # endif /* DYNAMIC_ICONV */ 3306 # endif /* USE_ICONV */ 3307 3308 #endif /* FEAT_MBYTE */ 3309 3310 #if defined(FEAT_XIM) || defined(PROTO) 3311 3312 # ifdef FEAT_GUI_GTK 3313 static int xim_has_preediting INIT(= FALSE); /* IM current status */ 3314 3315 /* 3316 * Set preedit_start_col to the current cursor position. 3317 */ 3318 static void 3319 init_preedit_start_col(void) 3320 { 3321 if (State & CMDLINE) 3322 preedit_start_col = cmdline_getvcol_cursor(); 3323 else if (curwin != NULL) 3324 getvcol(curwin, &curwin->w_cursor, &preedit_start_col, NULL, NULL); 3325 /* Prevent that preediting marks the buffer as changed. */ 3326 xim_changed_while_preediting = curbuf->b_changed; 3327 } 3328 # endif 3329 3330 # if defined(HAVE_GTK2) && !defined(PROTO) 3331 3332 static int im_is_active = FALSE; /* IM is enabled for current mode */ 3333 static int im_preedit_cursor = 0; /* cursor offset in characters */ 3334 static int im_preedit_trailing = 0; /* number of characters after cursor */ 3335 3336 static unsigned long im_commit_handler_id = 0; 3337 static unsigned int im_activatekey_keyval = GDK_VoidSymbol; 3338 static unsigned int im_activatekey_state = 0; 3339 3340 void 3341 im_set_active(int active) 3342 { 3343 int was_active; 3344 3345 was_active = !!im_is_active; 3346 im_is_active = (active && !p_imdisable); 3347 3348 if (im_is_active != was_active) 3349 xim_reset(); 3350 } 3351 3352 void 3353 xim_set_focus(int focus) 3354 { 3355 if (xic != NULL) 3356 { 3357 if (focus) 3358 gtk_im_context_focus_in(xic); 3359 else 3360 gtk_im_context_focus_out(xic); 3361 } 3362 } 3363 3364 void 3365 im_set_position(int row, int col) 3366 { 3367 if (xic != NULL) 3368 { 3369 GdkRectangle area; 3370 3371 area.x = FILL_X(col); 3372 area.y = FILL_Y(row); 3373 area.width = gui.char_width * (mb_lefthalve(row, col) ? 2 : 1); 3374 area.height = gui.char_height; 3375 3376 gtk_im_context_set_cursor_location(xic, &area); 3377 } 3378 } 3379 3380 # if 0 || defined(PROTO) /* apparently only used in gui_x11.c */ 3381 void 3382 xim_set_preedit(void) 3383 { 3384 im_set_position(gui.row, gui.col); 3385 } 3386 # endif 3387 3388 static void 3389 im_add_to_input(char_u *str, int len) 3390 { 3391 /* Convert from 'termencoding' (always "utf-8") to 'encoding' */ 3392 if (input_conv.vc_type != CONV_NONE) 3393 { 3394 str = string_convert(&input_conv, str, &len); 3395 g_return_if_fail(str != NULL); 3396 } 3397 3398 add_to_input_buf_csi(str, len); 3399 3400 if (input_conv.vc_type != CONV_NONE) 3401 vim_free(str); 3402 3403 if (p_mh) /* blank out the pointer if necessary */ 3404 gui_mch_mousehide(TRUE); 3405 } 3406 3407 static void 3408 im_delete_preedit(void) 3409 { 3410 char_u bskey[] = {CSI, 'k', 'b'}; 3411 char_u delkey[] = {CSI, 'k', 'D'}; 3412 3413 if (State & NORMAL) 3414 { 3415 im_preedit_cursor = 0; 3416 return; 3417 } 3418 for (; im_preedit_cursor > 0; --im_preedit_cursor) 3419 add_to_input_buf(bskey, (int)sizeof(bskey)); 3420 3421 for (; im_preedit_trailing > 0; --im_preedit_trailing) 3422 add_to_input_buf(delkey, (int)sizeof(delkey)); 3423 } 3424 3425 static void 3426 im_correct_cursor(int num_move_back) 3427 { 3428 char_u backkey[] = {CSI, 'k', 'l'}; 3429 3430 if (State & NORMAL) 3431 return; 3432 # ifdef FEAT_RIGHTLEFT 3433 if ((State & CMDLINE) == 0 && curwin != NULL && curwin->w_p_rl) 3434 backkey[2] = 'r'; 3435 # endif 3436 for (; num_move_back > 0; --num_move_back) 3437 add_to_input_buf(backkey, (int)sizeof(backkey)); 3438 } 3439 3440 static int xim_expected_char = NUL; 3441 static int xim_ignored_char = FALSE; 3442 3443 /* 3444 * Update the mode and cursor while in an IM callback. 3445 */ 3446 static void 3447 im_show_info(void) 3448 { 3449 int old_vgetc_busy; 3450 old_vgetc_busy = vgetc_busy; 3451 vgetc_busy = TRUE; 3452 showmode(); 3453 vgetc_busy = old_vgetc_busy; 3454 setcursor(); 3455 out_flush(); 3456 } 3457 3458 /* 3459 * Callback invoked when the user finished preediting. 3460 * Put the final string into the input buffer. 3461 */ 3462 /*ARGSUSED0*/ 3463 static void 3464 im_commit_cb(GtkIMContext *context, const gchar *str, gpointer data) 3465 { 3466 int slen = (int)STRLEN(str); 3467 int add_to_input = TRUE; 3468 int clen; 3469 int len = slen; 3470 int commit_with_preedit = TRUE; 3471 char_u *im_str, *p; 3472 3473 #ifdef XIM_DEBUG 3474 xim_log("im_commit_cb(): %s\n", str); 3475 #endif 3476 3477 /* The imhangul module doesn't reset the preedit string before 3478 * committing. Call im_delete_preedit() to work around that. */ 3479 im_delete_preedit(); 3480 3481 /* Indicate that preediting has finished. */ 3482 if (preedit_start_col == MAXCOL) 3483 { 3484 init_preedit_start_col(); 3485 commit_with_preedit = FALSE; 3486 } 3487 3488 /* The thing which setting "preedit_start_col" to MAXCOL means that 3489 * "preedit_start_col" will be set forcely when calling 3490 * preedit_changed_cb() next time. 3491 * "preedit_start_col" should not reset with MAXCOL on this part. Vim 3492 * is simulating the preediting by using add_to_input_str(). when 3493 * preedit begin immediately before committed, the typebuf is not 3494 * flushed to screen, then it can't get correct "preedit_start_col". 3495 * Thus, it should calculate the cells by adding cells of the committed 3496 * string. */ 3497 if (input_conv.vc_type != CONV_NONE) 3498 { 3499 im_str = string_convert(&input_conv, (char_u *)str, &len); 3500 g_return_if_fail(im_str != NULL); 3501 } 3502 else 3503 im_str = (char_u *)str; 3504 clen = 0; 3505 for (p = im_str; p < im_str + len; p += (*mb_ptr2len)(p)) 3506 clen += (*mb_ptr2cells)(p); 3507 if (input_conv.vc_type != CONV_NONE) 3508 vim_free(im_str); 3509 preedit_start_col += clen; 3510 3511 /* Is this a single character that matches a keypad key that's just 3512 * been pressed? If so, we don't want it to be entered as such - let 3513 * us carry on processing the raw keycode so that it may be used in 3514 * mappings as <kSomething>. */ 3515 if (xim_expected_char != NUL) 3516 { 3517 /* We're currently processing a keypad or other special key */ 3518 if (slen == 1 && str[0] == xim_expected_char) 3519 { 3520 /* It's a match - don't do it here */ 3521 xim_ignored_char = TRUE; 3522 add_to_input = FALSE; 3523 } 3524 else 3525 { 3526 /* Not a match */ 3527 xim_ignored_char = FALSE; 3528 } 3529 } 3530 3531 if (add_to_input) 3532 im_add_to_input((char_u *)str, slen); 3533 3534 /* Inserting chars while "im_is_active" is set does not cause a change of 3535 * buffer. When the chars are committed the buffer must be marked as 3536 * changed. */ 3537 if (!commit_with_preedit) 3538 preedit_start_col = MAXCOL; 3539 3540 /* This flag is used in changed() at next call. */ 3541 xim_changed_while_preediting = TRUE; 3542 3543 if (gtk_main_level() > 0) 3544 gtk_main_quit(); 3545 } 3546 3547 /* 3548 * Callback invoked after start to the preedit. 3549 */ 3550 /*ARGSUSED*/ 3551 static void 3552 im_preedit_start_cb(GtkIMContext *context, gpointer data) 3553 { 3554 #ifdef XIM_DEBUG 3555 xim_log("im_preedit_start_cb()\n"); 3556 #endif 3557 3558 im_is_active = TRUE; 3559 gui_update_cursor(TRUE, FALSE); 3560 } 3561 3562 /* 3563 * Callback invoked after end to the preedit. 3564 */ 3565 /*ARGSUSED*/ 3566 static void 3567 im_preedit_end_cb(GtkIMContext *context, gpointer data) 3568 { 3569 #ifdef XIM_DEBUG 3570 xim_log("im_preedit_end_cb()\n"); 3571 #endif 3572 im_delete_preedit(); 3573 3574 /* Indicate that preediting has finished */ 3575 preedit_start_col = MAXCOL; 3576 xim_has_preediting = FALSE; 3577 3578 im_is_active = FALSE; 3579 gui_update_cursor(TRUE, FALSE); 3580 im_show_info(); 3581 } 3582 3583 /* 3584 * Callback invoked after changes to the preedit string. If the preedit 3585 * string was empty before, remember the preedit start column so we know 3586 * where to apply feedback attributes. Delete the previous preedit string 3587 * if there was one, save the new preedit cursor offset, and put the new 3588 * string into the input buffer. 3589 * 3590 * TODO: The pragmatic "put into input buffer" approach used here has 3591 * several fundamental problems: 3592 * 3593 * - The characters in the preedit string are subject to remapping. 3594 * That's broken, only the finally committed string should be remapped. 3595 * 3596 * - There is a race condition involved: The retrieved value for the 3597 * current cursor position will be wrong if any unprocessed characters 3598 * are still queued in the input buffer. 3599 * 3600 * - Due to the lack of synchronization between the file buffer in memory 3601 * and any typed characters, it's practically impossible to implement the 3602 * "retrieve_surrounding" and "delete_surrounding" signals reliably. IM 3603 * modules for languages such as Thai are likely to rely on this feature 3604 * for proper operation. 3605 * 3606 * Conclusions: I think support for preediting needs to be moved to the 3607 * core parts of Vim. Ideally, until it has been committed, the preediting 3608 * string should only be displayed and not affect the buffer content at all. 3609 * The question how to deal with the synchronization issue still remains. 3610 * Circumventing the input buffer is probably not desirable. Anyway, I think 3611 * implementing "retrieve_surrounding" is the only hard problem. 3612 * 3613 * One way to solve all of this in a clean manner would be to queue all key 3614 * press/release events "as is" in the input buffer, and apply the IM filtering 3615 * at the receiving end of the queue. This, however, would have a rather large 3616 * impact on the code base. If there is an easy way to force processing of all 3617 * remaining input from within the "retrieve_surrounding" signal handler, this 3618 * might not be necessary. Gotta ask on vim-dev for opinions. 3619 */ 3620 /*ARGSUSED1*/ 3621 static void 3622 im_preedit_changed_cb(GtkIMContext *context, gpointer data) 3623 { 3624 char *preedit_string = NULL; 3625 int cursor_index = 0; 3626 int num_move_back = 0; 3627 char_u *str; 3628 char_u *p; 3629 int i; 3630 3631 gtk_im_context_get_preedit_string(context, 3632 &preedit_string, NULL, 3633 &cursor_index); 3634 3635 #ifdef XIM_DEBUG 3636 xim_log("im_preedit_changed_cb(): %s\n", preedit_string); 3637 #endif 3638 3639 g_return_if_fail(preedit_string != NULL); /* just in case */ 3640 3641 /* If preedit_start_col is MAXCOL set it to the current cursor position. */ 3642 if (preedit_start_col == MAXCOL && preedit_string[0] != '\0') 3643 { 3644 xim_has_preediting = TRUE; 3645 3646 /* Urgh, this breaks if the input buffer isn't empty now */ 3647 init_preedit_start_col(); 3648 } 3649 else if (cursor_index == 0 && preedit_string[0] == '\0') 3650 { 3651 if (preedit_start_col == MAXCOL) 3652 xim_has_preediting = FALSE; 3653 3654 /* If at the start position (after typing backspace) 3655 * preedit_start_col must be reset. */ 3656 preedit_start_col = MAXCOL; 3657 } 3658 3659 im_delete_preedit(); 3660 3661 /* 3662 * Compute the end of the preediting area: "preedit_end_col". 3663 * According to the documentation of gtk_im_context_get_preedit_string(), 3664 * the cursor_pos output argument returns the offset in bytes. This is 3665 * unfortunately not true -- real life shows the offset is in characters, 3666 * and the GTK+ source code agrees with me. Will file a bug later. 3667 */ 3668 if (preedit_start_col != MAXCOL) 3669 preedit_end_col = preedit_start_col; 3670 str = (char_u *)preedit_string; 3671 for (p = str, i = 0; *p != NUL; p += utf_byte2len(*p), ++i) 3672 { 3673 int is_composing; 3674 3675 is_composing = ((*p & 0x80) != 0 && utf_iscomposing(utf_ptr2char(p))); 3676 /* 3677 * These offsets are used as counters when generating <BS> and <Del> 3678 * to delete the preedit string. So don't count composing characters 3679 * unless 'delcombine' is enabled. 3680 */ 3681 if (!is_composing || p_deco) 3682 { 3683 if (i < cursor_index) 3684 ++im_preedit_cursor; 3685 else 3686 ++im_preedit_trailing; 3687 } 3688 if (!is_composing && i >= cursor_index) 3689 { 3690 /* This is essentially the same as im_preedit_trailing, except 3691 * composing characters are not counted even if p_deco is set. */ 3692 ++num_move_back; 3693 } 3694 if (preedit_start_col != MAXCOL) 3695 preedit_end_col += utf_ptr2cells(p); 3696 } 3697 3698 if (p > str) 3699 { 3700 im_add_to_input(str, (int)(p - str)); 3701 im_correct_cursor(num_move_back); 3702 } 3703 3704 g_free(preedit_string); 3705 3706 if (gtk_main_level() > 0) 3707 gtk_main_quit(); 3708 } 3709 3710 /* 3711 * Translate the Pango attributes at iter to Vim highlighting attributes. 3712 * Ignore attributes not supported by Vim highlighting. This shouldn't have 3713 * too much impact -- right now we handle even more attributes than necessary 3714 * for the IM modules I tested with. 3715 */ 3716 static int 3717 translate_pango_attributes(PangoAttrIterator *iter) 3718 { 3719 PangoAttribute *attr; 3720 int char_attr = HL_NORMAL; 3721 3722 attr = pango_attr_iterator_get(iter, PANGO_ATTR_UNDERLINE); 3723 if (attr != NULL && ((PangoAttrInt *)attr)->value 3724 != (int)PANGO_UNDERLINE_NONE) 3725 char_attr |= HL_UNDERLINE; 3726 3727 attr = pango_attr_iterator_get(iter, PANGO_ATTR_WEIGHT); 3728 if (attr != NULL && ((PangoAttrInt *)attr)->value >= (int)PANGO_WEIGHT_BOLD) 3729 char_attr |= HL_BOLD; 3730 3731 attr = pango_attr_iterator_get(iter, PANGO_ATTR_STYLE); 3732 if (attr != NULL && ((PangoAttrInt *)attr)->value 3733 != (int)PANGO_STYLE_NORMAL) 3734 char_attr |= HL_ITALIC; 3735 3736 attr = pango_attr_iterator_get(iter, PANGO_ATTR_BACKGROUND); 3737 if (attr != NULL) 3738 { 3739 const PangoColor *color = &((PangoAttrColor *)attr)->color; 3740 3741 /* Assume inverse if black background is requested */ 3742 if ((color->red | color->green | color->blue) == 0) 3743 char_attr |= HL_INVERSE; 3744 } 3745 3746 return char_attr; 3747 } 3748 3749 /* 3750 * Retrieve the highlighting attributes at column col in the preedit string. 3751 * Return -1 if not in preediting mode or if col is out of range. 3752 */ 3753 int 3754 im_get_feedback_attr(int col) 3755 { 3756 char *preedit_string = NULL; 3757 PangoAttrList *attr_list = NULL; 3758 int char_attr = -1; 3759 3760 if (xic == NULL) 3761 return char_attr; 3762 3763 gtk_im_context_get_preedit_string(xic, &preedit_string, &attr_list, NULL); 3764 3765 if (preedit_string != NULL && attr_list != NULL) 3766 { 3767 int index; 3768 3769 /* Get the byte index as used by PangoAttrIterator */ 3770 for (index = 0; col > 0 && preedit_string[index] != '\0'; --col) 3771 index += utfc_ptr2len((char_u *)preedit_string + index); 3772 3773 if (preedit_string[index] != '\0') 3774 { 3775 PangoAttrIterator *iter; 3776 int start, end; 3777 3778 char_attr = HL_NORMAL; 3779 iter = pango_attr_list_get_iterator(attr_list); 3780 3781 /* Extract all relevant attributes from the list. */ 3782 do 3783 { 3784 pango_attr_iterator_range(iter, &start, &end); 3785 3786 if (index >= start && index < end) 3787 char_attr |= translate_pango_attributes(iter); 3788 } 3789 while (pango_attr_iterator_next(iter)); 3790 3791 pango_attr_iterator_destroy(iter); 3792 } 3793 } 3794 3795 if (attr_list != NULL) 3796 pango_attr_list_unref(attr_list); 3797 g_free(preedit_string); 3798 3799 return char_attr; 3800 } 3801 3802 void 3803 xim_init(void) 3804 { 3805 #ifdef XIM_DEBUG 3806 xim_log("xim_init()\n"); 3807 #endif 3808 3809 g_return_if_fail(gui.drawarea != NULL); 3810 g_return_if_fail(gui.drawarea->window != NULL); 3811 3812 xic = gtk_im_multicontext_new(); 3813 g_object_ref(xic); 3814 3815 im_commit_handler_id = g_signal_connect(G_OBJECT(xic), "commit", 3816 G_CALLBACK(&im_commit_cb), NULL); 3817 g_signal_connect(G_OBJECT(xic), "preedit_changed", 3818 G_CALLBACK(&im_preedit_changed_cb), NULL); 3819 g_signal_connect(G_OBJECT(xic), "preedit_start", 3820 G_CALLBACK(&im_preedit_start_cb), NULL); 3821 g_signal_connect(G_OBJECT(xic), "preedit_end", 3822 G_CALLBACK(&im_preedit_end_cb), NULL); 3823 3824 gtk_im_context_set_client_window(xic, gui.drawarea->window); 3825 } 3826 3827 void 3828 im_shutdown(void) 3829 { 3830 #ifdef XIM_DEBUG 3831 xim_log("im_shutdown()\n"); 3832 #endif 3833 3834 if (xic != NULL) 3835 { 3836 gtk_im_context_focus_out(xic); 3837 g_object_unref(xic); 3838 xic = NULL; 3839 } 3840 im_is_active = FALSE; 3841 im_commit_handler_id = 0; 3842 preedit_start_col = MAXCOL; 3843 xim_has_preediting = FALSE; 3844 } 3845 3846 /* 3847 * Convert the string argument to keyval and state for GdkEventKey. 3848 * If str is valid return TRUE, otherwise FALSE. 3849 * 3850 * See 'imactivatekey' for documentation of the format. 3851 */ 3852 static int 3853 im_string_to_keyval(const char *str, unsigned int *keyval, unsigned int *state) 3854 { 3855 const char *mods_end; 3856 unsigned tmp_keyval; 3857 unsigned tmp_state = 0; 3858 3859 mods_end = strrchr(str, '-'); 3860 mods_end = (mods_end != NULL) ? mods_end + 1 : str; 3861 3862 /* Parse modifier keys */ 3863 while (str < mods_end) 3864 switch (*str++) 3865 { 3866 case '-': break; 3867 case 'S': case 's': tmp_state |= (unsigned)GDK_SHIFT_MASK; break; 3868 case 'L': case 'l': tmp_state |= (unsigned)GDK_LOCK_MASK; break; 3869 case 'C': case 'c': tmp_state |= (unsigned)GDK_CONTROL_MASK;break; 3870 case '1': tmp_state |= (unsigned)GDK_MOD1_MASK; break; 3871 case '2': tmp_state |= (unsigned)GDK_MOD2_MASK; break; 3872 case '3': tmp_state |= (unsigned)GDK_MOD3_MASK; break; 3873 case '4': tmp_state |= (unsigned)GDK_MOD4_MASK; break; 3874 case '5': tmp_state |= (unsigned)GDK_MOD5_MASK; break; 3875 default: 3876 return FALSE; 3877 } 3878 3879 tmp_keyval = gdk_keyval_from_name(str); 3880 3881 if (tmp_keyval == 0 || tmp_keyval == GDK_VoidSymbol) 3882 return FALSE; 3883 3884 if (keyval != NULL) 3885 *keyval = tmp_keyval; 3886 if (state != NULL) 3887 *state = tmp_state; 3888 3889 return TRUE; 3890 } 3891 3892 /* 3893 * Return TRUE if p_imak is valid, otherwise FALSE. As a special case, an 3894 * empty string is also regarded as valid. 3895 * 3896 * Note: The numerical key value of p_imak is cached if it was valid; thus 3897 * boldly assuming im_xim_isvalid_imactivate() will always be called whenever 3898 * 'imak' changes. This is currently the case but not obvious -- should 3899 * probably rename the function for clarity. 3900 */ 3901 int 3902 im_xim_isvalid_imactivate(void) 3903 { 3904 if (p_imak[0] == NUL) 3905 { 3906 im_activatekey_keyval = GDK_VoidSymbol; 3907 im_activatekey_state = 0; 3908 return TRUE; 3909 } 3910 3911 return im_string_to_keyval((const char *)p_imak, 3912 &im_activatekey_keyval, 3913 &im_activatekey_state); 3914 } 3915 3916 static void 3917 im_synthesize_keypress(unsigned int keyval, unsigned int state) 3918 { 3919 GdkEventKey *event; 3920 3921 # ifdef HAVE_GTK_MULTIHEAD 3922 event = (GdkEventKey *)gdk_event_new(GDK_KEY_PRESS); 3923 g_object_ref(gui.drawarea->window); /* unreffed by gdk_event_free() */ 3924 # else 3925 event = (GdkEventKey *)g_malloc0((gulong)sizeof(GdkEvent)); 3926 event->type = GDK_KEY_PRESS; 3927 # endif 3928 event->window = gui.drawarea->window; 3929 event->send_event = TRUE; 3930 event->time = GDK_CURRENT_TIME; 3931 event->state = state; 3932 event->keyval = keyval; 3933 event->hardware_keycode = /* needed for XIM */ 3934 XKeysymToKeycode(GDK_WINDOW_XDISPLAY(event->window), (KeySym)keyval); 3935 event->length = 0; 3936 event->string = NULL; 3937 3938 gtk_im_context_filter_keypress(xic, event); 3939 3940 /* For consistency, also send the corresponding release event. */ 3941 event->type = GDK_KEY_RELEASE; 3942 event->send_event = FALSE; 3943 gtk_im_context_filter_keypress(xic, event); 3944 3945 # ifdef HAVE_GTK_MULTIHEAD 3946 gdk_event_free((GdkEvent *)event); 3947 # else 3948 g_free(event); 3949 # endif 3950 } 3951 3952 void 3953 xim_reset(void) 3954 { 3955 if (xic != NULL) 3956 { 3957 /* 3958 * The third-party imhangul module (and maybe others too) ignores 3959 * gtk_im_context_reset() or at least doesn't reset the active state. 3960 * Thus sending imactivatekey would turn it off if it was on before, 3961 * which is clearly not what we want. Fortunately we can work around 3962 * that for imhangul by sending GDK_Escape, but I don't know if it 3963 * works with all IM modules that support an activation key :/ 3964 * 3965 * An alternative approach would be to destroy the IM context and 3966 * recreate it. But that means loading/unloading the IM module on 3967 * every mode switch, which causes a quite noticable delay even on 3968 * my rather fast box... 3969 * * 3970 * Moreover, there are some XIM which cannot respond to 3971 * im_synthesize_keypress(). we hope that they reset by 3972 * xim_shutdown(). 3973 */ 3974 if (im_activatekey_keyval != GDK_VoidSymbol && im_is_active) 3975 im_synthesize_keypress(GDK_Escape, 0U); 3976 3977 gtk_im_context_reset(xic); 3978 3979 /* 3980 * HACK for Ami: This sequence of function calls makes Ami handle 3981 * the IM reset gratiously, without breaking loads of other stuff. 3982 * It seems to force English mode as well, which is exactly what we 3983 * want because it makes the Ami status display work reliably. 3984 */ 3985 gtk_im_context_set_use_preedit(xic, FALSE); 3986 3987 if (p_imdisable) 3988 im_shutdown(); 3989 else 3990 { 3991 gtk_im_context_set_use_preedit(xic, TRUE); 3992 xim_set_focus(gui.in_focus); 3993 3994 if (im_activatekey_keyval != GDK_VoidSymbol) 3995 { 3996 if (im_is_active) 3997 { 3998 g_signal_handler_block(xic, im_commit_handler_id); 3999 im_synthesize_keypress(im_activatekey_keyval, 4000 im_activatekey_state); 4001 g_signal_handler_unblock(xic, im_commit_handler_id); 4002 } 4003 } 4004 else 4005 { 4006 im_shutdown(); 4007 xim_init(); 4008 xim_set_focus(gui.in_focus); 4009 } 4010 } 4011 } 4012 4013 preedit_start_col = MAXCOL; 4014 xim_has_preediting = FALSE; 4015 } 4016 4017 int 4018 xim_queue_key_press_event(GdkEventKey *event, int down) 4019 { 4020 if (down) 4021 { 4022 /* 4023 * Workaround GTK2 XIM 'feature' that always converts keypad keys to 4024 * chars., even when not part of an IM sequence (ref. feature of 4025 * gdk/gdkkeyuni.c). 4026 * Flag any keypad keys that might represent a single char. 4027 * If this (on its own - i.e., not part of an IM sequence) is 4028 * committed while we're processing one of these keys, we can ignore 4029 * that commit and go ahead & process it ourselves. That way we can 4030 * still distinguish keypad keys for use in mappings. 4031 */ 4032 switch (event->keyval) 4033 { 4034 case GDK_KP_Add: xim_expected_char = '+'; break; 4035 case GDK_KP_Subtract: xim_expected_char = '-'; break; 4036 case GDK_KP_Divide: xim_expected_char = '/'; break; 4037 case GDK_KP_Multiply: xim_expected_char = '*'; break; 4038 case GDK_KP_Decimal: xim_expected_char = '.'; break; 4039 case GDK_KP_Equal: xim_expected_char = '='; break; 4040 case GDK_KP_0: xim_expected_char = '0'; break; 4041 case GDK_KP_1: xim_expected_char = '1'; break; 4042 case GDK_KP_2: xim_expected_char = '2'; break; 4043 case GDK_KP_3: xim_expected_char = '3'; break; 4044 case GDK_KP_4: xim_expected_char = '4'; break; 4045 case GDK_KP_5: xim_expected_char = '5'; break; 4046 case GDK_KP_6: xim_expected_char = '6'; break; 4047 case GDK_KP_7: xim_expected_char = '7'; break; 4048 case GDK_KP_8: xim_expected_char = '8'; break; 4049 case GDK_KP_9: xim_expected_char = '9'; break; 4050 default: xim_expected_char = NUL; 4051 } 4052 xim_ignored_char = FALSE; 4053 } 4054 4055 /* 4056 * When typing fFtT, XIM may be activated. Thus it must pass 4057 * gtk_im_context_filter_keypress() in Normal mode. 4058 * And while doing :sh too. 4059 */ 4060 if (xic != NULL && !p_imdisable 4061 && (State & (INSERT | CMDLINE | NORMAL | EXTERNCMD)) != 0) 4062 { 4063 /* 4064 * Filter 'imactivatekey' and map it to CTRL-^. This way, Vim is 4065 * always aware of the current status of IM, and can even emulate 4066 * the activation key for modules that don't support one. 4067 */ 4068 if (event->keyval == im_activatekey_keyval 4069 && (event->state & im_activatekey_state) == im_activatekey_state) 4070 { 4071 unsigned int state_mask; 4072 4073 /* Require the state of the 3 most used modifiers to match exactly. 4074 * Otherwise e.g. <S-C-space> would be unusable for other purposes 4075 * if the IM activate key is <S-space>. */ 4076 state_mask = im_activatekey_state; 4077 state_mask |= ((int)GDK_SHIFT_MASK | (int)GDK_CONTROL_MASK 4078 | (int)GDK_MOD1_MASK); 4079 4080 if ((event->state & state_mask) != im_activatekey_state) 4081 return FALSE; 4082 4083 /* Don't send it a second time on GDK_KEY_RELEASE. */ 4084 if (event->type != GDK_KEY_PRESS) 4085 return TRUE; 4086 4087 if (map_to_exists_mode((char_u *)"", LANGMAP)) 4088 { 4089 im_set_active(FALSE); 4090 4091 /* ":lmap" mappings exists, toggle use of mappings. */ 4092 State ^= LANGMAP; 4093 if (State & LANGMAP) 4094 { 4095 curbuf->b_p_iminsert = B_IMODE_NONE; 4096 State &= ~LANGMAP; 4097 } 4098 else 4099 { 4100 curbuf->b_p_iminsert = B_IMODE_LMAP; 4101 State |= LANGMAP; 4102 } 4103 return TRUE; 4104 } 4105 4106 return gtk_im_context_filter_keypress(xic, event); 4107 } 4108 4109 /* Don't filter events through the IM context if IM isn't active 4110 * right now. Unlike with GTK+ 1.2 we cannot rely on the IM module 4111 * not doing anything before the activation key was sent. */ 4112 if (im_activatekey_keyval == GDK_VoidSymbol || im_is_active) 4113 { 4114 int imresult = gtk_im_context_filter_keypress(xic, event); 4115 4116 /* Some XIM send following sequence: 4117 * 1. preedited string. 4118 * 2. committed string. 4119 * 3. line changed key. 4120 * 4. preedited string. 4121 * 5. remove preedited string. 4122 * if 3, Vim can't move back the above line for 5. 4123 * thus, this part should not parse the key. */ 4124 if (!imresult && preedit_start_col != MAXCOL 4125 && event->keyval == GDK_Return) 4126 { 4127 im_synthesize_keypress(GDK_Return, 0U); 4128 return FALSE; 4129 } 4130 4131 /* If XIM tried to commit a keypad key as a single char., 4132 * ignore it so we can use the keypad key 'raw', for mappings. */ 4133 if (xim_expected_char != NUL && xim_ignored_char) 4134 /* We had a keypad key, and XIM tried to thieve it */ 4135 return FALSE; 4136 4137 /* Normal processing */ 4138 return imresult; 4139 } 4140 } 4141 4142 return FALSE; 4143 } 4144 4145 int 4146 im_get_status(void) 4147 { 4148 return im_is_active; 4149 } 4150 4151 # else /* !HAVE_GTK2 */ 4152 4153 static int xim_is_active = FALSE; /* XIM should be active in the current 4154 mode */ 4155 static int xim_has_focus = FALSE; /* XIM is really being used for Vim */ 4156 #ifdef FEAT_GUI_X11 4157 static XIMStyle input_style; 4158 static int status_area_enabled = TRUE; 4159 #endif 4160 4161 #ifdef FEAT_GUI_GTK 4162 # ifdef WIN3264 4163 # include <gdk/gdkwin32.h> 4164 # else 4165 # include <gdk/gdkx.h> 4166 # endif 4167 #else 4168 # ifdef PROTO 4169 /* Define a few things to be able to generate prototypes while not configured 4170 * for GTK. */ 4171 # define GSList int 4172 # define gboolean int 4173 typedef int GdkEvent; 4174 typedef int GdkEventKey; 4175 # define GdkIC int 4176 # endif 4177 #endif 4178 4179 #if defined(FEAT_GUI_GTK) || defined(PROTO) 4180 static int preedit_buf_len = 0; 4181 static int xim_can_preediting INIT(= FALSE); /* XIM in showmode() */ 4182 static int xim_input_style; 4183 #ifndef FEAT_GUI_GTK 4184 # define gboolean int 4185 #endif 4186 static gboolean use_status_area = 0; 4187 4188 static int im_xim_str2keycode __ARGS((unsigned int *code, unsigned int *state)); 4189 static void im_xim_send_event_imactivate __ARGS((void)); 4190 4191 /* 4192 * Convert string to keycode and state for XKeyEvent. 4193 * When string is valid return OK, when invalid return FAIL. 4194 * 4195 * See 'imactivatekey' documentation for the format. 4196 */ 4197 static int 4198 im_xim_str2keycode(code, state) 4199 unsigned int *code; 4200 unsigned int *state; 4201 { 4202 int retval = OK; 4203 int len; 4204 unsigned keycode = 0, keystate = 0; 4205 Window window; 4206 Display *display; 4207 char_u *flag_end; 4208 char_u *str; 4209 4210 if (*p_imak != NUL) 4211 { 4212 len = STRLEN(p_imak); 4213 for (flag_end = p_imak + len - 1; 4214 flag_end > p_imak && *flag_end != '-'; --flag_end) 4215 ; 4216 4217 /* Parse modifier keys */ 4218 for (str = p_imak; str < flag_end; ++str) 4219 { 4220 switch (*str) 4221 { 4222 case 's': case 'S': 4223 keystate |= ShiftMask; 4224 break; 4225 case 'l': case 'L': 4226 keystate |= LockMask; 4227 break; 4228 case 'c': case 'C': 4229 keystate |= ControlMask; 4230 break; 4231 case '1': 4232 keystate |= Mod1Mask; 4233 break; 4234 case '2': 4235 keystate |= Mod2Mask; 4236 break; 4237 case '3': 4238 keystate |= Mod3Mask; 4239 break; 4240 case '4': 4241 keystate |= Mod4Mask; 4242 break; 4243 case '5': 4244 keystate |= Mod5Mask; 4245 break; 4246 case '-': 4247 break; 4248 default: 4249 retval = FAIL; 4250 } 4251 } 4252 if (*str == '-') 4253 ++str; 4254 4255 /* Get keycode from string. */ 4256 gui_get_x11_windis(&window, &display); 4257 if (display) 4258 keycode = XKeysymToKeycode(display, XStringToKeysym((char *)str)); 4259 if (keycode == 0) 4260 retval = FAIL; 4261 4262 if (code != NULL) 4263 *code = keycode; 4264 if (state != NULL) 4265 *state = keystate; 4266 } 4267 return retval; 4268 } 4269 4270 static void 4271 im_xim_send_event_imactivate() 4272 { 4273 /* Force turn on preedit state by symulate keypress event. 4274 * Keycode and state is specified by 'imactivatekey'. 4275 */ 4276 XKeyEvent ev; 4277 4278 gui_get_x11_windis(&ev.window, &ev.display); 4279 ev.root = RootWindow(ev.display, DefaultScreen(ev.display)); 4280 ev.subwindow = None; 4281 ev.time = CurrentTime; 4282 ev.x = 1; 4283 ev.y = 1; 4284 ev.x_root = 1; 4285 ev.y_root = 1; 4286 ev.same_screen = 1; 4287 ev.type = KeyPress; 4288 if (im_xim_str2keycode(&ev.keycode, &ev.state) == OK) 4289 XSendEvent(ev.display, ev.window, 1, KeyPressMask, (XEvent*)&ev); 4290 } 4291 4292 /* 4293 * Return TRUE if 'imactivatekey' has a valid value. 4294 */ 4295 int 4296 im_xim_isvalid_imactivate() 4297 { 4298 return im_xim_str2keycode(NULL, NULL) == OK; 4299 } 4300 #endif /* FEAT_GUI_GTK */ 4301 4302 /* 4303 * Switch using XIM on/off. This is used by the code that changes "State". 4304 */ 4305 void 4306 im_set_active(active) 4307 int active; 4308 { 4309 if (xic == NULL) 4310 return; 4311 4312 /* If 'imdisable' is set, XIM is never active. */ 4313 if (p_imdisable) 4314 active = FALSE; 4315 #if !defined (FEAT_GUI_GTK) 4316 else if (input_style & XIMPreeditPosition) 4317 /* There is a problem in switching XIM off when preediting is used, 4318 * and it is not clear how this can be solved. For now, keep XIM on 4319 * all the time, like it was done in Vim 5.8. */ 4320 active = TRUE; 4321 #endif 4322 4323 /* Remember the active state, it is needed when Vim gets keyboard focus. */ 4324 xim_is_active = active; 4325 4326 #ifdef FEAT_GUI_GTK 4327 /* When 'imactivatekey' has valid key-string, try to control XIM preedit 4328 * state. When 'imactivatekey' has no or invalid string, try old XIM 4329 * focus control. 4330 */ 4331 if (*p_imak != NUL) 4332 { 4333 /* BASIC STRATEGY: 4334 * Destroy old Input Context (XIC), and create new one. New XIC 4335 * would have a state of preedit that is off. When argument:active 4336 * is false, that's all. Else argument:active is true, send a key 4337 * event specified by 'imactivatekey' to activate XIM preedit state. 4338 */ 4339 4340 xim_is_active = TRUE; /* Disable old XIM focus control */ 4341 /* If we can monitor preedit state with preedit callback functions, 4342 * try least creation of new XIC. 4343 */ 4344 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS) 4345 { 4346 if (xim_can_preediting && !active) 4347 { 4348 /* Force turn off preedit state. With some IM 4349 * implementations, we cannot turn off preedit state by 4350 * symulate keypress event. It is why using such a method 4351 * that destroy old IC (input context), and create new one. 4352 * When create new IC, its preedit state is usually off. 4353 */ 4354 xim_reset(); 4355 xim_set_focus(FALSE); 4356 gdk_ic_destroy(xic); 4357 xim_init(); 4358 xim_can_preediting = FALSE; 4359 } 4360 else if (!xim_can_preediting && active) 4361 im_xim_send_event_imactivate(); 4362 } 4363 else 4364 { 4365 /* First, force destroy old IC, and create new one. It 4366 * symulates "turning off preedit state". 4367 */ 4368 xim_set_focus(FALSE); 4369 gdk_ic_destroy(xic); 4370 xim_init(); 4371 xim_can_preediting = FALSE; 4372 4373 /* 2nd, when requested to activate IM, symulate this by sending 4374 * the event. 4375 */ 4376 if (active) 4377 { 4378 im_xim_send_event_imactivate(); 4379 xim_can_preediting = TRUE; 4380 } 4381 } 4382 } 4383 else 4384 { 4385 # ifndef XIMPreeditUnKnown 4386 /* X11R5 doesn't have these, it looks safe enough to define here. */ 4387 typedef unsigned long XIMPreeditState; 4388 # define XIMPreeditUnKnown 0L 4389 # define XIMPreeditEnable 1L 4390 # define XIMPreeditDisable (1L<<1) 4391 # define XNPreeditState "preeditState" 4392 # endif 4393 XIMPreeditState preedit_state = XIMPreeditUnKnown; 4394 XVaNestedList preedit_attr; 4395 XIC pxic; 4396 4397 preedit_attr = XVaCreateNestedList(0, 4398 XNPreeditState, &preedit_state, 4399 NULL); 4400 pxic = ((GdkICPrivate *)xic)->xic; 4401 4402 if (!XGetICValues(pxic, XNPreeditAttributes, preedit_attr, NULL)) 4403 { 4404 XFree(preedit_attr); 4405 preedit_attr = XVaCreateNestedList(0, 4406 XNPreeditState, 4407 active ? XIMPreeditEnable : XIMPreeditDisable, 4408 NULL); 4409 XSetICValues(pxic, XNPreeditAttributes, preedit_attr, NULL); 4410 xim_can_preediting = active; 4411 xim_is_active = active; 4412 } 4413 XFree(preedit_attr); 4414 } 4415 if (xim_input_style & XIMPreeditCallbacks) 4416 { 4417 preedit_buf_len = 0; 4418 init_preedit_start_col(); 4419 } 4420 #else 4421 # if 0 4422 /* When had tested kinput2 + canna + Athena GUI version with 4423 * 'imactivatekey' is "s-space", im_xim_send_event_imactivate() did not 4424 * work correctly. It just inserted one space. I don't know why we 4425 * couldn't switch state of XIM preediting. This is reason why these 4426 * codes are commented out. 4427 */ 4428 /* First, force destroy old IC, and create new one. It symulates 4429 * "turning off preedit state". 4430 */ 4431 xim_set_focus(FALSE); 4432 XDestroyIC(xic); 4433 xic = NULL; 4434 xim_init(); 4435 4436 /* 2nd, when requested to activate IM, symulate this by sending the 4437 * event. 4438 */ 4439 if (active) 4440 im_xim_send_event_imactivate(); 4441 # endif 4442 #endif 4443 xim_set_preedit(); 4444 } 4445 4446 /* 4447 * Adjust using XIM for gaining or losing keyboard focus. Also called when 4448 * "xim_is_active" changes. 4449 */ 4450 void 4451 xim_set_focus(focus) 4452 int focus; 4453 { 4454 if (xic == NULL) 4455 return; 4456 4457 /* 4458 * XIM only gets focus when the Vim window has keyboard focus and XIM has 4459 * been set active for the current mode. 4460 */ 4461 if (focus && xim_is_active) 4462 { 4463 if (!xim_has_focus) 4464 { 4465 xim_has_focus = TRUE; 4466 #ifdef FEAT_GUI_GTK 4467 gdk_im_begin(xic, gui.drawarea->window); 4468 #else 4469 XSetICFocus(xic); 4470 #endif 4471 } 4472 } 4473 else 4474 { 4475 if (xim_has_focus) 4476 { 4477 xim_has_focus = FALSE; 4478 #ifdef FEAT_GUI_GTK 4479 gdk_im_end(); 4480 #else 4481 XUnsetICFocus(xic); 4482 #endif 4483 } 4484 } 4485 } 4486 4487 /*ARGSUSED*/ 4488 void 4489 im_set_position(row, col) 4490 int row; 4491 int col; 4492 { 4493 xim_set_preedit(); 4494 } 4495 4496 /* 4497 * Set the XIM to the current cursor position. 4498 */ 4499 void 4500 xim_set_preedit() 4501 { 4502 if (xic == NULL) 4503 return; 4504 4505 xim_set_focus(TRUE); 4506 4507 #ifdef FEAT_GUI_GTK 4508 if (gdk_im_ready()) 4509 { 4510 int attrmask; 4511 GdkICAttr *attr; 4512 4513 if (!xic_attr) 4514 return; 4515 4516 attr = xic_attr; 4517 attrmask = 0; 4518 4519 # ifdef FEAT_XFONTSET 4520 if ((xim_input_style & (int)GDK_IM_PREEDIT_POSITION) 4521 && gui.fontset != NOFONTSET 4522 && gui.fontset->type == GDK_FONT_FONTSET) 4523 { 4524 if (!xim_has_focus) 4525 { 4526 if (attr->spot_location.y >= 0) 4527 { 4528 attr->spot_location.x = 0; 4529 attr->spot_location.y = -100; 4530 attrmask |= (int)GDK_IC_SPOT_LOCATION; 4531 } 4532 } 4533 else 4534 { 4535 gint width, height; 4536 4537 if (attr->spot_location.x != TEXT_X(gui.col) 4538 || attr->spot_location.y != TEXT_Y(gui.row)) 4539 { 4540 attr->spot_location.x = TEXT_X(gui.col); 4541 attr->spot_location.y = TEXT_Y(gui.row); 4542 attrmask |= (int)GDK_IC_SPOT_LOCATION; 4543 } 4544 4545 gdk_window_get_size(gui.drawarea->window, &width, &height); 4546 width -= 2 * gui.border_offset; 4547 height -= 2 * gui.border_offset; 4548 if (xim_input_style & (int)GDK_IM_STATUS_AREA) 4549 height -= gui.char_height; 4550 if (attr->preedit_area.width != width 4551 || attr->preedit_area.height != height) 4552 { 4553 attr->preedit_area.x = gui.border_offset; 4554 attr->preedit_area.y = gui.border_offset; 4555 attr->preedit_area.width = width; 4556 attr->preedit_area.height = height; 4557 attrmask |= (int)GDK_IC_PREEDIT_AREA; 4558 } 4559 4560 if (attr->preedit_fontset != gui.current_font) 4561 { 4562 attr->preedit_fontset = gui.current_font; 4563 attrmask |= (int)GDK_IC_PREEDIT_FONTSET; 4564 } 4565 } 4566 } 4567 # endif /* FEAT_XFONTSET */ 4568 4569 if (xim_fg_color == INVALCOLOR) 4570 { 4571 xim_fg_color = gui.def_norm_pixel; 4572 xim_bg_color = gui.def_back_pixel; 4573 } 4574 if (attr->preedit_foreground.pixel != xim_fg_color) 4575 { 4576 attr->preedit_foreground.pixel = xim_fg_color; 4577 attrmask |= (int)GDK_IC_PREEDIT_FOREGROUND; 4578 } 4579 if (attr->preedit_background.pixel != xim_bg_color) 4580 { 4581 attr->preedit_background.pixel = xim_bg_color; 4582 attrmask |= (int)GDK_IC_PREEDIT_BACKGROUND; 4583 } 4584 4585 if (attrmask != 0) 4586 gdk_ic_set_attr(xic, attr, (GdkICAttributesType)attrmask); 4587 } 4588 #else /* FEAT_GUI_GTK */ 4589 { 4590 XVaNestedList attr_list; 4591 XRectangle spot_area; 4592 XPoint over_spot; 4593 int line_space; 4594 4595 if (!xim_has_focus) 4596 { 4597 /* hide XIM cursor */ 4598 over_spot.x = 0; 4599 over_spot.y = -100; /* arbitrary invisible position */ 4600 attr_list = (XVaNestedList) XVaCreateNestedList(0, 4601 XNSpotLocation, 4602 &over_spot, 4603 NULL); 4604 XSetICValues(xic, XNPreeditAttributes, attr_list, NULL); 4605 XFree(attr_list); 4606 return; 4607 } 4608 4609 if (input_style & XIMPreeditPosition) 4610 { 4611 if (xim_fg_color == INVALCOLOR) 4612 { 4613 xim_fg_color = gui.def_norm_pixel; 4614 xim_bg_color = gui.def_back_pixel; 4615 } 4616 over_spot.x = TEXT_X(gui.col); 4617 over_spot.y = TEXT_Y(gui.row); 4618 spot_area.x = 0; 4619 spot_area.y = 0; 4620 spot_area.height = gui.char_height * Rows; 4621 spot_area.width = gui.char_width * Columns; 4622 line_space = gui.char_height; 4623 attr_list = (XVaNestedList) XVaCreateNestedList(0, 4624 XNSpotLocation, &over_spot, 4625 XNForeground, (Pixel) xim_fg_color, 4626 XNBackground, (Pixel) xim_bg_color, 4627 XNArea, &spot_area, 4628 XNLineSpace, line_space, 4629 NULL); 4630 if (XSetICValues(xic, XNPreeditAttributes, attr_list, NULL)) 4631 EMSG(_("E284: Cannot set IC values")); 4632 XFree(attr_list); 4633 } 4634 } 4635 #endif /* FEAT_GUI_GTK */ 4636 } 4637 4638 /* 4639 * Set up the status area. 4640 * 4641 * This should use a separate Widget, but that seems not possible, because 4642 * preedit_area and status_area should be set to the same window as for the 4643 * text input. Unfortunately this means the status area pollutes the text 4644 * window... 4645 */ 4646 void 4647 xim_set_status_area() 4648 { 4649 if (xic == NULL) 4650 return; 4651 4652 #ifdef FEAT_GUI_GTK 4653 # if defined(FEAT_XFONTSET) 4654 if (use_status_area) 4655 { 4656 GdkICAttr *attr; 4657 int style; 4658 gint width, height; 4659 GtkWidget *widget; 4660 int attrmask; 4661 4662 if (!xic_attr) 4663 return; 4664 4665 attr = xic_attr; 4666 attrmask = 0; 4667 style = (int)gdk_ic_get_style(xic); 4668 if ((style & (int)GDK_IM_STATUS_MASK) == (int)GDK_IM_STATUS_AREA) 4669 { 4670 if (gui.fontset != NOFONTSET 4671 && gui.fontset->type == GDK_FONT_FONTSET) 4672 { 4673 widget = gui.mainwin; 4674 gdk_window_get_size(widget->window, &width, &height); 4675 4676 attrmask |= (int)GDK_IC_STATUS_AREA; 4677 attr->status_area.x = 0; 4678 attr->status_area.y = height - gui.char_height - 1; 4679 attr->status_area.width = width; 4680 attr->status_area.height = gui.char_height; 4681 } 4682 } 4683 if (attrmask != 0) 4684 gdk_ic_set_attr(xic, attr, (GdkICAttributesType)attrmask); 4685 } 4686 # endif 4687 #else 4688 { 4689 XVaNestedList preedit_list = 0, status_list = 0, list = 0; 4690 XRectangle pre_area, status_area; 4691 4692 if (input_style & XIMStatusArea) 4693 { 4694 if (input_style & XIMPreeditArea) 4695 { 4696 XRectangle *needed_rect; 4697 4698 /* to get status_area width */ 4699 status_list = XVaCreateNestedList(0, XNAreaNeeded, 4700 &needed_rect, NULL); 4701 XGetICValues(xic, XNStatusAttributes, status_list, NULL); 4702 XFree(status_list); 4703 4704 status_area.width = needed_rect->width; 4705 } 4706 else 4707 status_area.width = gui.char_width * Columns; 4708 4709 status_area.x = 0; 4710 status_area.y = gui.char_height * Rows + gui.border_offset; 4711 if (gui.which_scrollbars[SBAR_BOTTOM]) 4712 status_area.y += gui.scrollbar_height; 4713 #ifdef FEAT_MENU 4714 if (gui.menu_is_active) 4715 status_area.y += gui.menu_height; 4716 #endif 4717 status_area.height = gui.char_height; 4718 status_list = XVaCreateNestedList(0, XNArea, &status_area, NULL); 4719 } 4720 else 4721 { 4722 status_area.x = 0; 4723 status_area.y = gui.char_height * Rows + gui.border_offset; 4724 if (gui.which_scrollbars[SBAR_BOTTOM]) 4725 status_area.y += gui.scrollbar_height; 4726 #ifdef FEAT_MENU 4727 if (gui.menu_is_active) 4728 status_area.y += gui.menu_height; 4729 #endif 4730 status_area.width = 0; 4731 status_area.height = gui.char_height; 4732 } 4733 4734 if (input_style & XIMPreeditArea) /* off-the-spot */ 4735 { 4736 pre_area.x = status_area.x + status_area.width; 4737 pre_area.y = gui.char_height * Rows + gui.border_offset; 4738 pre_area.width = gui.char_width * Columns - pre_area.x; 4739 if (gui.which_scrollbars[SBAR_BOTTOM]) 4740 pre_area.y += gui.scrollbar_height; 4741 #ifdef FEAT_MENU 4742 if (gui.menu_is_active) 4743 pre_area.y += gui.menu_height; 4744 #endif 4745 pre_area.height = gui.char_height; 4746 preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL); 4747 } 4748 else if (input_style & XIMPreeditPosition) /* over-the-spot */ 4749 { 4750 pre_area.x = 0; 4751 pre_area.y = 0; 4752 pre_area.height = gui.char_height * Rows; 4753 pre_area.width = gui.char_width * Columns; 4754 preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL); 4755 } 4756 4757 if (preedit_list && status_list) 4758 list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list, 4759 XNStatusAttributes, status_list, NULL); 4760 else if (preedit_list) 4761 list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list, 4762 NULL); 4763 else if (status_list) 4764 list = XVaCreateNestedList(0, XNStatusAttributes, status_list, 4765 NULL); 4766 else 4767 list = NULL; 4768 4769 if (list) 4770 { 4771 XSetICValues(xic, XNVaNestedList, list, NULL); 4772 XFree(list); 4773 } 4774 if (status_list) 4775 XFree(status_list); 4776 if (preedit_list) 4777 XFree(preedit_list); 4778 } 4779 #endif 4780 } 4781 4782 #if defined(FEAT_GUI_X11) || defined(FEAT_GUI_GTK) 4783 static char e_xim[] = N_("E285: Failed to create input context"); 4784 #endif 4785 4786 #if defined(FEAT_GUI_X11) || defined(PROTO) 4787 # if defined(XtSpecificationRelease) && XtSpecificationRelease >= 6 && !defined(sun) 4788 # define USE_X11R6_XIM 4789 # endif 4790 4791 static int xim_real_init __ARGS((Window x11_window, Display *x11_display)); 4792 4793 4794 #ifdef USE_X11R6_XIM 4795 static void xim_instantiate_cb __ARGS((Display *display, XPointer client_data, XPointer call_data)); 4796 static void xim_destroy_cb __ARGS((XIM im, XPointer client_data, XPointer call_data)); 4797 4798 /*ARGSUSED*/ 4799 static void 4800 xim_instantiate_cb(display, client_data, call_data) 4801 Display *display; 4802 XPointer client_data; 4803 XPointer call_data; 4804 { 4805 Window x11_window; 4806 Display *x11_display; 4807 4808 #ifdef XIM_DEBUG 4809 xim_log("xim_instantiate_cb()\n"); 4810 #endif 4811 4812 gui_get_x11_windis(&x11_window, &x11_display); 4813 if (display != x11_display) 4814 return; 4815 4816 xim_real_init(x11_window, x11_display); 4817 gui_set_shellsize(FALSE, FALSE); 4818 if (xic != NULL) 4819 XUnregisterIMInstantiateCallback(x11_display, NULL, NULL, NULL, 4820 xim_instantiate_cb, NULL); 4821 } 4822 4823 /*ARGSUSED*/ 4824 static void 4825 xim_destroy_cb(im, client_data, call_data) 4826 XIM im; 4827 XPointer client_data; 4828 XPointer call_data; 4829 { 4830 Window x11_window; 4831 Display *x11_display; 4832 4833 #ifdef XIM_DEBUG 4834 xim_log("xim_destroy_cb()\n"); 4835 #endif 4836 gui_get_x11_windis(&x11_window, &x11_display); 4837 4838 xic = NULL; 4839 status_area_enabled = FALSE; 4840 4841 gui_set_shellsize(FALSE, FALSE); 4842 4843 XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL, 4844 xim_instantiate_cb, NULL); 4845 } 4846 #endif 4847 4848 void 4849 xim_init() 4850 { 4851 Window x11_window; 4852 Display *x11_display; 4853 4854 #ifdef XIM_DEBUG 4855 xim_log("xim_init()\n"); 4856 #endif 4857 4858 gui_get_x11_windis(&x11_window, &x11_display); 4859 4860 xic = NULL; 4861 4862 if (xim_real_init(x11_window, x11_display)) 4863 return; 4864 4865 gui_set_shellsize(FALSE, FALSE); 4866 4867 #ifdef USE_X11R6_XIM 4868 XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL, 4869 xim_instantiate_cb, NULL); 4870 #endif 4871 } 4872 4873 static int 4874 xim_real_init(x11_window, x11_display) 4875 Window x11_window; 4876 Display *x11_display; 4877 { 4878 int i; 4879 char *p, 4880 *s, 4881 *ns, 4882 *end, 4883 tmp[1024]; 4884 #define IMLEN_MAX 40 4885 char buf[IMLEN_MAX + 7]; 4886 XIM xim = NULL; 4887 XIMStyles *xim_styles; 4888 XIMStyle this_input_style = 0; 4889 Boolean found; 4890 XPoint over_spot; 4891 XVaNestedList preedit_list, status_list; 4892 4893 input_style = 0; 4894 status_area_enabled = FALSE; 4895 4896 if (xic != NULL) 4897 return FALSE; 4898 4899 if (gui.rsrc_input_method != NULL && *gui.rsrc_input_method != NUL) 4900 { 4901 strcpy(tmp, gui.rsrc_input_method); 4902 for (ns = s = tmp; ns != NULL && *s != NUL;) 4903 { 4904 s = (char *)skipwhite((char_u *)s); 4905 if (*s == NUL) 4906 break; 4907 if ((ns = end = strchr(s, ',')) == NULL) 4908 end = s + strlen(s); 4909 while (isspace(((char_u *)end)[-1])) 4910 end--; 4911 *end = NUL; 4912 4913 if (strlen(s) <= IMLEN_MAX) 4914 { 4915 strcpy(buf, "@im="); 4916 strcat(buf, s); 4917 if ((p = XSetLocaleModifiers(buf)) != NULL && *p != NUL 4918 && (xim = XOpenIM(x11_display, NULL, NULL, NULL)) 4919 != NULL) 4920 break; 4921 } 4922 4923 s = ns + 1; 4924 } 4925 } 4926 4927 if (xim == NULL && (p = XSetLocaleModifiers("")) != NULL && *p != NUL) 4928 xim = XOpenIM(x11_display, NULL, NULL, NULL); 4929 4930 /* This is supposed to be useful to obtain characters through 4931 * XmbLookupString() without really using a XIM. */ 4932 if (xim == NULL && (p = XSetLocaleModifiers("@im=none")) != NULL 4933 && *p != NUL) 4934 xim = XOpenIM(x11_display, NULL, NULL, NULL); 4935 4936 if (xim == NULL) 4937 { 4938 /* Only give this message when verbose is set, because too many people 4939 * got this message when they didn't want to use a XIM. */ 4940 if (p_verbose > 0) 4941 { 4942 verbose_enter(); 4943 EMSG(_("E286: Failed to open input method")); 4944 verbose_leave(); 4945 } 4946 return FALSE; 4947 } 4948 4949 #ifdef USE_X11R6_XIM 4950 { 4951 XIMCallback destroy_cb; 4952 4953 destroy_cb.callback = xim_destroy_cb; 4954 destroy_cb.client_data = NULL; 4955 if (XSetIMValues(xim, XNDestroyCallback, &destroy_cb, NULL)) 4956 EMSG(_("E287: Warning: Could not set destroy callback to IM")); 4957 } 4958 #endif 4959 4960 if (XGetIMValues(xim, XNQueryInputStyle, &xim_styles, NULL) || !xim_styles) 4961 { 4962 EMSG(_("E288: input method doesn't support any style")); 4963 XCloseIM(xim); 4964 return FALSE; 4965 } 4966 4967 found = False; 4968 strcpy(tmp, gui.rsrc_preedit_type_name); 4969 for (s = tmp; s && !found; ) 4970 { 4971 while (*s && isspace((unsigned char)*s)) 4972 s++; 4973 if (!*s) 4974 break; 4975 if ((ns = end = strchr(s, ',')) != 0) 4976 ns++; 4977 else 4978 end = s + strlen(s); 4979 while (isspace((unsigned char)*end)) 4980 end--; 4981 *end = '\0'; 4982 4983 if (!strcmp(s, "OverTheSpot")) 4984 this_input_style = (XIMPreeditPosition | XIMStatusArea); 4985 else if (!strcmp(s, "OffTheSpot")) 4986 this_input_style = (XIMPreeditArea | XIMStatusArea); 4987 else if (!strcmp(s, "Root")) 4988 this_input_style = (XIMPreeditNothing | XIMStatusNothing); 4989 4990 for (i = 0; (unsigned short)i < xim_styles->count_styles; i++) 4991 { 4992 if (this_input_style == xim_styles->supported_styles[i]) 4993 { 4994 found = True; 4995 break; 4996 } 4997 } 4998 if (!found) 4999 for (i = 0; (unsigned short)i < xim_styles->count_styles; i++) 5000 { 5001 if ((xim_styles->supported_styles[i] & this_input_style) 5002 == (this_input_style & ~XIMStatusArea)) 5003 { 5004 this_input_style &= ~XIMStatusArea; 5005 found = True; 5006 break; 5007 } 5008 } 5009 5010 s = ns; 5011 } 5012 XFree(xim_styles); 5013 5014 if (!found) 5015 { 5016 /* Only give this message when verbose is set, because too many people 5017 * got this message when they didn't want to use a XIM. */ 5018 if (p_verbose > 0) 5019 { 5020 verbose_enter(); 5021 EMSG(_("E289: input method doesn't support my preedit type")); 5022 verbose_leave(); 5023 } 5024 XCloseIM(xim); 5025 return FALSE; 5026 } 5027 5028 over_spot.x = TEXT_X(gui.col); 5029 over_spot.y = TEXT_Y(gui.row); 5030 input_style = this_input_style; 5031 5032 /* A crash was reported when trying to pass gui.norm_font as XNFontSet, 5033 * thus that has been removed. Hopefully the default works... */ 5034 #ifdef FEAT_XFONTSET 5035 if (gui.fontset != NOFONTSET) 5036 { 5037 preedit_list = XVaCreateNestedList(0, 5038 XNSpotLocation, &over_spot, 5039 XNForeground, (Pixel)gui.def_norm_pixel, 5040 XNBackground, (Pixel)gui.def_back_pixel, 5041 XNFontSet, (XFontSet)gui.fontset, 5042 NULL); 5043 status_list = XVaCreateNestedList(0, 5044 XNForeground, (Pixel)gui.def_norm_pixel, 5045 XNBackground, (Pixel)gui.def_back_pixel, 5046 XNFontSet, (XFontSet)gui.fontset, 5047 NULL); 5048 } 5049 else 5050 #endif 5051 { 5052 preedit_list = XVaCreateNestedList(0, 5053 XNSpotLocation, &over_spot, 5054 XNForeground, (Pixel)gui.def_norm_pixel, 5055 XNBackground, (Pixel)gui.def_back_pixel, 5056 NULL); 5057 status_list = XVaCreateNestedList(0, 5058 XNForeground, (Pixel)gui.def_norm_pixel, 5059 XNBackground, (Pixel)gui.def_back_pixel, 5060 NULL); 5061 } 5062 5063 xic = XCreateIC(xim, 5064 XNInputStyle, input_style, 5065 XNClientWindow, x11_window, 5066 XNFocusWindow, gui.wid, 5067 XNPreeditAttributes, preedit_list, 5068 XNStatusAttributes, status_list, 5069 NULL); 5070 XFree(status_list); 5071 XFree(preedit_list); 5072 if (xic != NULL) 5073 { 5074 if (input_style & XIMStatusArea) 5075 { 5076 xim_set_status_area(); 5077 status_area_enabled = TRUE; 5078 } 5079 else 5080 gui_set_shellsize(FALSE, FALSE); 5081 } 5082 else 5083 { 5084 EMSG(_(e_xim)); 5085 XCloseIM(xim); 5086 return FALSE; 5087 } 5088 5089 return TRUE; 5090 } 5091 5092 #endif /* FEAT_GUI_X11 */ 5093 5094 #if defined(FEAT_GUI_GTK) || defined(PROTO) 5095 5096 # ifdef FEAT_XFONTSET 5097 static char e_overthespot[] = N_("E290: over-the-spot style requires fontset"); 5098 # endif 5099 5100 # ifdef PROTO 5101 typedef int GdkIC; 5102 # endif 5103 5104 void 5105 xim_decide_input_style() 5106 { 5107 /* GDK_IM_STATUS_CALLBACKS was disabled, enabled it to allow Japanese 5108 * OverTheSpot. */ 5109 int supported_style = (int)GDK_IM_PREEDIT_NONE | 5110 (int)GDK_IM_PREEDIT_NOTHING | 5111 (int)GDK_IM_PREEDIT_POSITION | 5112 (int)GDK_IM_PREEDIT_CALLBACKS | 5113 (int)GDK_IM_STATUS_CALLBACKS | 5114 (int)GDK_IM_STATUS_AREA | 5115 (int)GDK_IM_STATUS_NONE | 5116 (int)GDK_IM_STATUS_NOTHING; 5117 5118 #ifdef XIM_DEBUG 5119 xim_log("xim_decide_input_style()\n"); 5120 #endif 5121 5122 if (!gdk_im_ready()) 5123 xim_input_style = 0; 5124 else 5125 { 5126 if (gtk_major_version > 1 5127 || (gtk_major_version == 1 5128 && (gtk_minor_version > 2 5129 || (gtk_minor_version == 2 && gtk_micro_version >= 3)))) 5130 use_status_area = TRUE; 5131 else 5132 { 5133 EMSG(_("E291: Your GTK+ is older than 1.2.3. Status area disabled")); 5134 use_status_area = FALSE; 5135 } 5136 #ifdef FEAT_XFONTSET 5137 if (gui.fontset == NOFONTSET || gui.fontset->type != GDK_FONT_FONTSET) 5138 #endif 5139 supported_style &= ~((int)GDK_IM_PREEDIT_POSITION 5140 | (int)GDK_IM_STATUS_AREA); 5141 if (!use_status_area) 5142 supported_style &= ~(int)GDK_IM_STATUS_AREA; 5143 xim_input_style = (int)gdk_im_decide_style((GdkIMStyle)supported_style); 5144 } 5145 } 5146 5147 /*ARGSUSED*/ 5148 static void 5149 preedit_start_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5150 { 5151 #ifdef XIM_DEBUG 5152 xim_log("xim_decide_input_style()\n"); 5153 #endif 5154 5155 draw_feedback = NULL; 5156 xim_can_preediting = TRUE; 5157 xim_has_preediting = TRUE; 5158 gui_update_cursor(TRUE, FALSE); 5159 if (showmode() > 0) 5160 { 5161 setcursor(); 5162 out_flush(); 5163 } 5164 } 5165 5166 static void 5167 xim_back_delete(int n) 5168 { 5169 char_u str[3]; 5170 5171 str[0] = CSI; 5172 str[1] = 'k'; 5173 str[2] = 'b'; 5174 while (n-- > 0) 5175 add_to_input_buf(str, 3); 5176 } 5177 5178 static GSList *key_press_event_queue = NULL; 5179 static gboolean processing_queued_event = FALSE; 5180 5181 /*ARGSUSED*/ 5182 static void 5183 preedit_draw_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5184 { 5185 XIMPreeditDrawCallbackStruct *draw_data; 5186 XIMText *text; 5187 char *src; 5188 GSList *event_queue; 5189 5190 #ifdef XIM_DEBUG 5191 xim_log("preedit_draw_cbproc()\n"); 5192 #endif 5193 5194 draw_data = (XIMPreeditDrawCallbackStruct *) call_data; 5195 text = (XIMText *) draw_data->text; 5196 5197 if ((text == NULL && draw_data->chg_length == preedit_buf_len) 5198 || preedit_buf_len == 0) 5199 { 5200 init_preedit_start_col(); 5201 vim_free(draw_feedback); 5202 draw_feedback = NULL; 5203 } 5204 if (draw_data->chg_length > 0) 5205 { 5206 int bs_cnt; 5207 5208 if (draw_data->chg_length > preedit_buf_len) 5209 bs_cnt = preedit_buf_len; 5210 else 5211 bs_cnt = draw_data->chg_length; 5212 xim_back_delete(bs_cnt); 5213 preedit_buf_len -= bs_cnt; 5214 } 5215 if (text != NULL) 5216 { 5217 int len; 5218 #ifdef FEAT_MBYTE 5219 char_u *buf = NULL; 5220 unsigned int nfeedback = 0; 5221 #endif 5222 char_u *ptr; 5223 5224 src = text->string.multi_byte; 5225 if (src != NULL && !text->encoding_is_wchar) 5226 { 5227 len = strlen(src); 5228 ptr = (char_u *)src; 5229 /* Avoid the enter for decision */ 5230 if (*ptr == '\n') 5231 return; 5232 5233 #ifdef FEAT_MBYTE 5234 if (input_conv.vc_type != CONV_NONE 5235 && (buf = string_convert(&input_conv, 5236 (char_u *)src, &len)) != NULL) 5237 { 5238 /* Converted from 'termencoding' to 'encoding'. */ 5239 add_to_input_buf_csi(buf, len); 5240 ptr = buf; 5241 } 5242 else 5243 #endif 5244 add_to_input_buf_csi((char_u *)src, len); 5245 /* Add count of character to preedit_buf_len */ 5246 while (*ptr != NUL) 5247 { 5248 #ifdef FEAT_MBYTE 5249 if (draw_data->text->feedback != NULL) 5250 { 5251 if (draw_feedback == NULL) 5252 draw_feedback = (char *)alloc(draw_data->chg_first 5253 + text->length); 5254 else 5255 draw_feedback = realloc(draw_feedback, 5256 draw_data->chg_first + text->length); 5257 if (draw_feedback != NULL) 5258 { 5259 draw_feedback[nfeedback + draw_data->chg_first] 5260 = draw_data->text->feedback[nfeedback]; 5261 nfeedback++; 5262 } 5263 } 5264 if (has_mbyte) 5265 ptr += (*mb_ptr2len)(ptr); 5266 else 5267 #endif 5268 ptr++; 5269 preedit_buf_len++; 5270 } 5271 #ifdef FEAT_MBYTE 5272 vim_free(buf); 5273 #endif 5274 preedit_end_col = MAXCOL; 5275 } 5276 } 5277 if (text != NULL || draw_data->chg_length > 0) 5278 { 5279 event_queue = key_press_event_queue; 5280 processing_queued_event = TRUE; 5281 while (event_queue != NULL && processing_queued_event) 5282 { 5283 GdkEvent *ev = event_queue->data; 5284 5285 gboolean *ret; 5286 gtk_signal_emit_by_name((GtkObject*)gui.mainwin, "key_press_event", 5287 ev, &ret); 5288 gdk_event_free(ev); 5289 event_queue = event_queue->next; 5290 } 5291 processing_queued_event = FALSE; 5292 if (key_press_event_queue) 5293 { 5294 g_slist_free(key_press_event_queue); 5295 key_press_event_queue = NULL; 5296 } 5297 } 5298 if (gtk_main_level() > 0) 5299 gtk_main_quit(); 5300 } 5301 5302 /* 5303 * Retrieve the highlighting attributes at column col in the preedit string. 5304 * Return -1 if not in preediting mode or if col is out of range. 5305 */ 5306 int 5307 im_get_feedback_attr(int col) 5308 { 5309 if (draw_feedback != NULL && col < preedit_buf_len) 5310 { 5311 if (draw_feedback[col] & XIMReverse) 5312 return HL_INVERSE; 5313 else if (draw_feedback[col] & XIMUnderline) 5314 return HL_UNDERLINE; 5315 else 5316 return hl_attr(HLF_V); 5317 } 5318 5319 return -1; 5320 } 5321 5322 /*ARGSUSED*/ 5323 static void 5324 preedit_caret_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5325 { 5326 #ifdef XIM_DEBUG 5327 xim_log("preedit_caret_cbproc()\n"); 5328 #endif 5329 } 5330 5331 /*ARGSUSED*/ 5332 static void 5333 preedit_done_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5334 { 5335 #ifdef XIM_DEBUG 5336 xim_log("preedit_done_cbproc()\n"); 5337 #endif 5338 5339 vim_free(draw_feedback); 5340 draw_feedback = NULL; 5341 xim_can_preediting = FALSE; 5342 xim_has_preediting = FALSE; 5343 gui_update_cursor(TRUE, FALSE); 5344 if (showmode() > 0) 5345 { 5346 setcursor(); 5347 out_flush(); 5348 } 5349 } 5350 5351 void 5352 xim_reset(void) 5353 { 5354 char *text; 5355 5356 #ifdef XIM_DEBUG 5357 xim_log("xim_reset()\n"); 5358 #endif 5359 5360 if (xic != NULL) 5361 { 5362 text = XmbResetIC(((GdkICPrivate *)xic)->xic); 5363 if (text != NULL && !(xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS)) 5364 add_to_input_buf_csi((char_u *)text, strlen(text)); 5365 else 5366 preedit_buf_len = 0; 5367 if (text != NULL) 5368 XFree(text); 5369 } 5370 } 5371 5372 /*ARGSUSED*/ 5373 int 5374 xim_queue_key_press_event(GdkEventKey *event, int down) 5375 { 5376 #ifdef XIM_DEBUG 5377 xim_log("xim_queue_key_press_event()\n"); 5378 #endif 5379 5380 if (preedit_buf_len <= 0) 5381 return FALSE; 5382 if (processing_queued_event) 5383 processing_queued_event = FALSE; 5384 5385 key_press_event_queue = g_slist_append(key_press_event_queue, 5386 gdk_event_copy((GdkEvent *)event)); 5387 return TRUE; 5388 } 5389 5390 /*ARGSUSED*/ 5391 static void 5392 preedit_callback_setup(GdkIC *ic) 5393 { 5394 XIC xxic; 5395 XVaNestedList preedit_attr; 5396 XIMCallback preedit_start_cb; 5397 XIMCallback preedit_draw_cb; 5398 XIMCallback preedit_caret_cb; 5399 XIMCallback preedit_done_cb; 5400 5401 xxic = ((GdkICPrivate*)xic)->xic; 5402 preedit_start_cb.callback = (XIMProc)preedit_start_cbproc; 5403 preedit_draw_cb.callback = (XIMProc)preedit_draw_cbproc; 5404 preedit_caret_cb.callback = (XIMProc)preedit_caret_cbproc; 5405 preedit_done_cb.callback = (XIMProc)preedit_done_cbproc; 5406 preedit_attr 5407 = XVaCreateNestedList (0, 5408 XNPreeditStartCallback, &preedit_start_cb, 5409 XNPreeditDrawCallback, &preedit_draw_cb, 5410 XNPreeditCaretCallback, &preedit_caret_cb, 5411 XNPreeditDoneCallback, &preedit_done_cb, 5412 0); 5413 XSetICValues (xxic, XNPreeditAttributes, preedit_attr, 0); 5414 XFree(preedit_attr); 5415 } 5416 5417 /*ARGSUSED*/ 5418 static void 5419 reset_state_setup(GdkIC *ic) 5420 { 5421 #ifdef USE_X11R6_XIM 5422 /* don't change the input context when we call reset */ 5423 XSetICValues(((GdkICPrivate*)ic)->xic, XNResetState, XIMPreserveState, 0); 5424 #endif 5425 } 5426 5427 void 5428 xim_init(void) 5429 { 5430 #ifdef XIM_DEBUG 5431 xim_log("xim_init()\n"); 5432 #endif 5433 5434 xic = NULL; 5435 xic_attr = NULL; 5436 5437 if (!gdk_im_ready()) 5438 { 5439 if (p_verbose > 0) 5440 { 5441 verbose_enter(); 5442 EMSG(_("E292: Input Method Server is not running")); 5443 verbose_leave(); 5444 } 5445 return; 5446 } 5447 if ((xic_attr = gdk_ic_attr_new()) != NULL) 5448 { 5449 #ifdef FEAT_XFONTSET 5450 gint width, height; 5451 #endif 5452 int mask; 5453 GdkColormap *colormap; 5454 GdkICAttr *attr = xic_attr; 5455 int attrmask = (int)GDK_IC_ALL_REQ; 5456 GtkWidget *widget = gui.drawarea; 5457 5458 attr->style = (GdkIMStyle)xim_input_style; 5459 attr->client_window = gui.mainwin->window; 5460 5461 if ((colormap = gtk_widget_get_colormap(widget)) != 5462 gtk_widget_get_default_colormap()) 5463 { 5464 attrmask |= (int)GDK_IC_PREEDIT_COLORMAP; 5465 attr->preedit_colormap = colormap; 5466 } 5467 attrmask |= (int)GDK_IC_PREEDIT_FOREGROUND; 5468 attrmask |= (int)GDK_IC_PREEDIT_BACKGROUND; 5469 attr->preedit_foreground = widget->style->fg[GTK_STATE_NORMAL]; 5470 attr->preedit_background = widget->style->base[GTK_STATE_NORMAL]; 5471 5472 #ifdef FEAT_XFONTSET 5473 if ((xim_input_style & (int)GDK_IM_PREEDIT_MASK) 5474 == (int)GDK_IM_PREEDIT_POSITION) 5475 { 5476 if (gui.fontset == NOFONTSET 5477 || gui.fontset->type != GDK_FONT_FONTSET) 5478 { 5479 EMSG(_(e_overthespot)); 5480 } 5481 else 5482 { 5483 gdk_window_get_size(widget->window, &width, &height); 5484 5485 attrmask |= (int)GDK_IC_PREEDIT_POSITION_REQ; 5486 attr->spot_location.x = TEXT_X(0); 5487 attr->spot_location.y = TEXT_Y(0); 5488 attr->preedit_area.x = gui.border_offset; 5489 attr->preedit_area.y = gui.border_offset; 5490 attr->preedit_area.width = width - 2*gui.border_offset; 5491 attr->preedit_area.height = height - 2*gui.border_offset; 5492 attr->preedit_fontset = gui.fontset; 5493 } 5494 } 5495 5496 if ((xim_input_style & (int)GDK_IM_STATUS_MASK) 5497 == (int)GDK_IM_STATUS_AREA) 5498 { 5499 if (gui.fontset == NOFONTSET 5500 || gui.fontset->type != GDK_FONT_FONTSET) 5501 { 5502 EMSG(_(e_overthespot)); 5503 } 5504 else 5505 { 5506 gdk_window_get_size(gui.mainwin->window, &width, &height); 5507 attrmask |= (int)GDK_IC_STATUS_AREA_REQ; 5508 attr->status_area.x = 0; 5509 attr->status_area.y = height - gui.char_height - 1; 5510 attr->status_area.width = width; 5511 attr->status_area.height = gui.char_height; 5512 attr->status_fontset = gui.fontset; 5513 } 5514 } 5515 else if ((xim_input_style & (int)GDK_IM_STATUS_MASK) 5516 == (int)GDK_IM_STATUS_CALLBACKS) 5517 { 5518 /* FIXME */ 5519 } 5520 #endif 5521 5522 xic = gdk_ic_new(attr, (GdkICAttributesType)attrmask); 5523 5524 if (xic == NULL) 5525 EMSG(_(e_xim)); 5526 else 5527 { 5528 mask = (int)gdk_window_get_events(widget->window); 5529 mask |= (int)gdk_ic_get_events(xic); 5530 gdk_window_set_events(widget->window, (GdkEventMask)mask); 5531 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS) 5532 preedit_callback_setup(xic); 5533 reset_state_setup(xic); 5534 } 5535 } 5536 } 5537 5538 void 5539 im_shutdown(void) 5540 { 5541 #ifdef XIM_DEBUG 5542 xim_log("im_shutdown()\n"); 5543 #endif 5544 5545 if (xic != NULL) 5546 { 5547 gdk_im_end(); 5548 gdk_ic_destroy(xic); 5549 xic = NULL; 5550 } 5551 xim_is_active = FALSE; 5552 xim_can_preediting = FALSE; 5553 preedit_start_col = MAXCOL; 5554 xim_has_preediting = FALSE; 5555 } 5556 5557 #endif /* FEAT_GUI_GTK */ 5558 5559 int 5560 xim_get_status_area_height() 5561 { 5562 #ifdef FEAT_GUI_GTK 5563 if (xim_input_style & (int)GDK_IM_STATUS_AREA) 5564 return gui.char_height; 5565 #else 5566 if (status_area_enabled) 5567 return gui.char_height; 5568 #endif 5569 return 0; 5570 } 5571 5572 /* 5573 * Get IM status. When IM is on, return TRUE. Else return FALSE. 5574 * FIXME: This doesn't work correctly: Having focus doesn't always mean XIM is 5575 * active, when not having focus XIM may still be active (e.g., when using a 5576 * tear-off menu item). 5577 */ 5578 int 5579 im_get_status() 5580 { 5581 # ifdef FEAT_GUI_GTK 5582 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS) 5583 return xim_can_preediting; 5584 # endif 5585 return xim_has_focus; 5586 } 5587 5588 # endif /* !HAVE_GTK2 */ 5589 5590 # if defined(FEAT_GUI_GTK) || defined(PROTO) 5591 int 5592 im_is_preediting() 5593 { 5594 return xim_has_preediting; 5595 } 5596 # endif 5597 #endif /* FEAT_XIM */ 5598 5599 #if defined(FEAT_MBYTE) || defined(PROTO) 5600 5601 /* 5602 * Setup "vcp" for conversion from "from" to "to". 5603 * The names must have been made canonical with enc_canonize(). 5604 * vcp->vc_type must have been initialized to CONV_NONE. 5605 * Note: cannot be used for conversion from/to ucs-2 and ucs-4 (will use utf-8 5606 * instead). 5607 * Afterwards invoke with "from" and "to" equal to NULL to cleanup. 5608 * Return FAIL when conversion is not supported, OK otherwise. 5609 */ 5610 int 5611 convert_setup(vcp, from, to) 5612 vimconv_T *vcp; 5613 char_u *from; 5614 char_u *to; 5615 { 5616 int from_prop; 5617 int to_prop; 5618 5619 /* Reset to no conversion. */ 5620 # ifdef USE_ICONV 5621 if (vcp->vc_type == CONV_ICONV && vcp->vc_fd != (iconv_t)-1) 5622 iconv_close(vcp->vc_fd); 5623 # endif 5624 vcp->vc_type = CONV_NONE; 5625 vcp->vc_factor = 1; 5626 vcp->vc_fail = FALSE; 5627 5628 /* No conversion when one of the names is empty or they are equal. */ 5629 if (from == NULL || *from == NUL || to == NULL || *to == NUL 5630 || STRCMP(from, to) == 0) 5631 return OK; 5632 5633 from_prop = enc_canon_props(from); 5634 to_prop = enc_canon_props(to); 5635 if ((from_prop & ENC_LATIN1) && (to_prop & ENC_UNICODE)) 5636 { 5637 /* Internal latin1 -> utf-8 conversion. */ 5638 vcp->vc_type = CONV_TO_UTF8; 5639 vcp->vc_factor = 2; /* up to twice as long */ 5640 } 5641 else if ((from_prop & ENC_LATIN9) && (to_prop & ENC_UNICODE)) 5642 { 5643 /* Internal latin9 -> utf-8 conversion. */ 5644 vcp->vc_type = CONV_9_TO_UTF8; 5645 vcp->vc_factor = 3; /* up to three as long (euro sign) */ 5646 } 5647 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_LATIN1)) 5648 { 5649 /* Internal utf-8 -> latin1 conversion. */ 5650 vcp->vc_type = CONV_TO_LATIN1; 5651 } 5652 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_LATIN9)) 5653 { 5654 /* Internal utf-8 -> latin9 conversion. */ 5655 vcp->vc_type = CONV_TO_LATIN9; 5656 } 5657 #ifdef WIN3264 5658 /* Win32-specific codepage <-> codepage conversion without iconv. */ 5659 else if (((from_prop & ENC_UNICODE) || encname2codepage(from) > 0) 5660 && ((to_prop & ENC_UNICODE) || encname2codepage(to) > 0)) 5661 { 5662 vcp->vc_type = CONV_CODEPAGE; 5663 vcp->vc_factor = 2; /* up to twice as long */ 5664 vcp->vc_cpfrom = (from_prop & ENC_UNICODE) ? 0 : encname2codepage(from); 5665 vcp->vc_cpto = (to_prop & ENC_UNICODE) ? 0 : encname2codepage(to); 5666 } 5667 #endif 5668 #ifdef MACOS_X 5669 else if ((from_prop & ENC_MACROMAN) && (to_prop & ENC_LATIN1)) 5670 { 5671 vcp->vc_type = CONV_MAC_LATIN1; 5672 } 5673 else if ((from_prop & ENC_MACROMAN) && (to_prop & ENC_UNICODE)) 5674 { 5675 vcp->vc_type = CONV_MAC_UTF8; 5676 vcp->vc_factor = 2; /* up to twice as long */ 5677 } 5678 else if ((from_prop & ENC_LATIN1) && (to_prop & ENC_MACROMAN)) 5679 { 5680 vcp->vc_type = CONV_LATIN1_MAC; 5681 } 5682 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_MACROMAN)) 5683 { 5684 vcp->vc_type = CONV_UTF8_MAC; 5685 } 5686 #endif 5687 # ifdef USE_ICONV 5688 else 5689 { 5690 /* Use iconv() for conversion. */ 5691 vcp->vc_fd = (iconv_t)my_iconv_open( 5692 (to_prop & ENC_UNICODE) ? (char_u *)"utf-8" : to, 5693 (from_prop & ENC_UNICODE) ? (char_u *)"utf-8" : from); 5694 if (vcp->vc_fd != (iconv_t)-1) 5695 { 5696 vcp->vc_type = CONV_ICONV; 5697 vcp->vc_factor = 4; /* could be longer too... */ 5698 } 5699 } 5700 # endif 5701 if (vcp->vc_type == CONV_NONE) 5702 return FAIL; 5703 5704 return OK; 5705 } 5706 5707 #if defined(FEAT_GUI) || defined(AMIGA) || defined(WIN3264) \ 5708 || defined(MSDOS) || defined(PROTO) 5709 /* 5710 * Do conversion on typed input characters in-place. 5711 * The input and output are not NUL terminated! 5712 * Returns the length after conversion. 5713 */ 5714 int 5715 convert_input(ptr, len, maxlen) 5716 char_u *ptr; 5717 int len; 5718 int maxlen; 5719 { 5720 return convert_input_safe(ptr, len, maxlen, NULL, NULL); 5721 } 5722 #endif 5723 5724 /* 5725 * Like convert_input(), but when there is an incomplete byte sequence at the 5726 * end return that as an allocated string in "restp" and set "*restlenp" to 5727 * the length. If "restp" is NULL it is not used. 5728 */ 5729 int 5730 convert_input_safe(ptr, len, maxlen, restp, restlenp) 5731 char_u *ptr; 5732 int len; 5733 int maxlen; 5734 char_u **restp; 5735 int *restlenp; 5736 { 5737 char_u *d; 5738 int dlen = len; 5739 int unconvertlen = 0; 5740 5741 d = string_convert_ext(&input_conv, ptr, &dlen, 5742 restp == NULL ? NULL : &unconvertlen); 5743 if (d != NULL) 5744 { 5745 if (dlen <= maxlen) 5746 { 5747 if (unconvertlen > 0) 5748 { 5749 /* Move the unconverted characters to allocated memory. */ 5750 *restp = alloc(unconvertlen); 5751 if (*restp != NULL) 5752 mch_memmove(*restp, ptr + len - unconvertlen, unconvertlen); 5753 *restlenp = unconvertlen; 5754 } 5755 mch_memmove(ptr, d, dlen); 5756 } 5757 else 5758 /* result is too long, keep the unconverted text (the caller must 5759 * have done something wrong!) */ 5760 dlen = len; 5761 vim_free(d); 5762 } 5763 return dlen; 5764 } 5765 5766 /* 5767 * Convert text "ptr[*lenp]" according to "vcp". 5768 * Returns the result in allocated memory and sets "*lenp". 5769 * When "lenp" is NULL, use NUL terminated strings. 5770 * Illegal chars are often changed to "?", unless vcp->vc_fail is set. 5771 * When something goes wrong, NULL is returned and "*lenp" is unchanged. 5772 */ 5773 char_u * 5774 string_convert(vcp, ptr, lenp) 5775 vimconv_T *vcp; 5776 char_u *ptr; 5777 int *lenp; 5778 { 5779 return string_convert_ext(vcp, ptr, lenp, NULL); 5780 } 5781 5782 /* 5783 * Like string_convert(), but when "unconvlenp" is not NULL and there are is 5784 * an incomplete sequence at the end it is not converted and "*unconvlenp" is 5785 * set to the number of remaining bytes. 5786 */ 5787 char_u * 5788 string_convert_ext(vcp, ptr, lenp, unconvlenp) 5789 vimconv_T *vcp; 5790 char_u *ptr; 5791 int *lenp; 5792 int *unconvlenp; 5793 { 5794 char_u *retval = NULL; 5795 char_u *d; 5796 int len; 5797 int i; 5798 int l; 5799 int c; 5800 5801 if (lenp == NULL) 5802 len = (int)STRLEN(ptr); 5803 else 5804 len = *lenp; 5805 if (len == 0) 5806 return vim_strsave((char_u *)""); 5807 5808 switch (vcp->vc_type) 5809 { 5810 case CONV_TO_UTF8: /* latin1 to utf-8 conversion */ 5811 retval = alloc(len * 2 + 1); 5812 if (retval == NULL) 5813 break; 5814 d = retval; 5815 for (i = 0; i < len; ++i) 5816 { 5817 c = ptr[i]; 5818 if (c < 0x80) 5819 *d++ = c; 5820 else 5821 { 5822 *d++ = 0xc0 + ((unsigned)c >> 6); 5823 *d++ = 0x80 + (c & 0x3f); 5824 } 5825 } 5826 *d = NUL; 5827 if (lenp != NULL) 5828 *lenp = (int)(d - retval); 5829 break; 5830 5831 case CONV_9_TO_UTF8: /* latin9 to utf-8 conversion */ 5832 retval = alloc(len * 3 + 1); 5833 if (retval == NULL) 5834 break; 5835 d = retval; 5836 for (i = 0; i < len; ++i) 5837 { 5838 c = ptr[i]; 5839 switch (c) 5840 { 5841 case 0xa4: c = 0x20ac; break; /* euro */ 5842 case 0xa6: c = 0x0160; break; /* S hat */ 5843 case 0xa8: c = 0x0161; break; /* S -hat */ 5844 case 0xb4: c = 0x017d; break; /* Z hat */ 5845 case 0xb8: c = 0x017e; break; /* Z -hat */ 5846 case 0xbc: c = 0x0152; break; /* OE */ 5847 case 0xbd: c = 0x0153; break; /* oe */ 5848 case 0xbe: c = 0x0178; break; /* Y */ 5849 } 5850 d += utf_char2bytes(c, d); 5851 } 5852 *d = NUL; 5853 if (lenp != NULL) 5854 *lenp = (int)(d - retval); 5855 break; 5856 5857 case CONV_TO_LATIN1: /* utf-8 to latin1 conversion */ 5858 case CONV_TO_LATIN9: /* utf-8 to latin9 conversion */ 5859 retval = alloc(len + 1); 5860 if (retval == NULL) 5861 break; 5862 d = retval; 5863 for (i = 0; i < len; ++i) 5864 { 5865 l = utf_ptr2len(ptr + i); 5866 if (l == 0) 5867 *d++ = NUL; 5868 else if (l == 1) 5869 { 5870 if (unconvlenp != NULL && utf8len_tab[ptr[i]] > len - i) 5871 { 5872 /* Incomplete sequence at the end. */ 5873 *unconvlenp = len - i; 5874 break; 5875 } 5876 *d++ = ptr[i]; 5877 } 5878 else 5879 { 5880 c = utf_ptr2char(ptr + i); 5881 if (vcp->vc_type == CONV_TO_LATIN9) 5882 switch (c) 5883 { 5884 case 0x20ac: c = 0xa4; break; /* euro */ 5885 case 0x0160: c = 0xa6; break; /* S hat */ 5886 case 0x0161: c = 0xa8; break; /* S -hat */ 5887 case 0x017d: c = 0xb4; break; /* Z hat */ 5888 case 0x017e: c = 0xb8; break; /* Z -hat */ 5889 case 0x0152: c = 0xbc; break; /* OE */ 5890 case 0x0153: c = 0xbd; break; /* oe */ 5891 case 0x0178: c = 0xbe; break; /* Y */ 5892 case 0xa4: 5893 case 0xa6: 5894 case 0xa8: 5895 case 0xb4: 5896 case 0xb8: 5897 case 0xbc: 5898 case 0xbd: 5899 case 0xbe: c = 0x100; break; /* not in latin9 */ 5900 } 5901 if (!utf_iscomposing(c)) /* skip composing chars */ 5902 { 5903 if (c < 0x100) 5904 *d++ = c; 5905 else if (vcp->vc_fail) 5906 { 5907 vim_free(retval); 5908 return NULL; 5909 } 5910 else 5911 { 5912 *d++ = 0xbf; 5913 if (utf_char2cells(c) > 1) 5914 *d++ = '?'; 5915 } 5916 } 5917 i += l - 1; 5918 } 5919 } 5920 *d = NUL; 5921 if (lenp != NULL) 5922 *lenp = (int)(d - retval); 5923 break; 5924 5925 # ifdef MACOS_X 5926 case CONV_MAC_LATIN1: 5927 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 5928 'm', 'l', unconvlenp); 5929 break; 5930 5931 case CONV_LATIN1_MAC: 5932 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 5933 'l', 'm', unconvlenp); 5934 break; 5935 5936 case CONV_MAC_UTF8: 5937 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 5938 'm', 'u', unconvlenp); 5939 break; 5940 5941 case CONV_UTF8_MAC: 5942 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 5943 'u', 'm', unconvlenp); 5944 break; 5945 # endif 5946 5947 # ifdef USE_ICONV 5948 case CONV_ICONV: /* conversion with output_conv.vc_fd */ 5949 retval = iconv_string(vcp, ptr, len, unconvlenp); 5950 if (retval != NULL && lenp != NULL) 5951 *lenp = (int)STRLEN(retval); 5952 break; 5953 # endif 5954 # ifdef WIN3264 5955 case CONV_CODEPAGE: /* codepage -> codepage */ 5956 { 5957 int retlen; 5958 int tmp_len; 5959 short_u *tmp; 5960 5961 /* 1. codepage/UTF-8 -> ucs-2. */ 5962 if (vcp->vc_cpfrom == 0) 5963 tmp_len = utf8_to_ucs2(ptr, len, NULL, NULL); 5964 else 5965 tmp_len = MultiByteToWideChar(vcp->vc_cpfrom, 0, 5966 ptr, len, 0, 0); 5967 tmp = (short_u *)alloc(sizeof(short_u) * tmp_len); 5968 if (tmp == NULL) 5969 break; 5970 if (vcp->vc_cpfrom == 0) 5971 utf8_to_ucs2(ptr, len, tmp, unconvlenp); 5972 else 5973 MultiByteToWideChar(vcp->vc_cpfrom, 0, ptr, len, tmp, tmp_len); 5974 5975 /* 2. ucs-2 -> codepage/UTF-8. */ 5976 if (vcp->vc_cpto == 0) 5977 retlen = ucs2_to_utf8(tmp, tmp_len, NULL); 5978 else 5979 retlen = WideCharToMultiByte(vcp->vc_cpto, 0, 5980 tmp, tmp_len, 0, 0, 0, 0); 5981 retval = alloc(retlen + 1); 5982 if (retval != NULL) 5983 { 5984 if (vcp->vc_cpto == 0) 5985 ucs2_to_utf8(tmp, tmp_len, retval); 5986 else 5987 WideCharToMultiByte(vcp->vc_cpto, 0, 5988 tmp, tmp_len, retval, retlen, 0, 0); 5989 retval[retlen] = NUL; 5990 if (lenp != NULL) 5991 *lenp = retlen; 5992 } 5993 vim_free(tmp); 5994 break; 5995 } 5996 # endif 5997 } 5998 5999 return retval; 6000 } 6001 #endif 6002