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 = 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_GUI_KDE 757 if (gui.in_use) 758 gui_mch_update_codec(); 759 #endif 760 761 #ifdef FEAT_SYN_HL 762 /* Need to reload spell dictionaries */ 763 spell_reload(); 764 #endif 765 766 return NULL; 767 } 768 769 /* 770 * Return the size of the BOM for the current buffer: 771 * 0 - no BOM 772 * 2 - UCS-2 or UTF-16 BOM 773 * 4 - UCS-4 BOM 774 * 3 - UTF-8 BOM 775 */ 776 int 777 bomb_size() 778 { 779 int n = 0; 780 781 if (curbuf->b_p_bomb && !curbuf->b_p_bin) 782 { 783 if (*curbuf->b_p_fenc == NUL) 784 { 785 if (enc_utf8) 786 { 787 if (enc_unicode != 0) 788 n = enc_unicode; 789 else 790 n = 3; 791 } 792 } 793 else if (STRCMP(curbuf->b_p_fenc, "utf-8") == 0) 794 n = 3; 795 else if (STRNCMP(curbuf->b_p_fenc, "ucs-2", 5) == 0 796 || STRNCMP(curbuf->b_p_fenc, "utf-16", 6) == 0) 797 n = 2; 798 else if (STRNCMP(curbuf->b_p_fenc, "ucs-4", 5) == 0) 799 n = 4; 800 } 801 return n; 802 } 803 804 /* 805 * Get class of pointer: 806 * 0 for blank or NUL 807 * 1 for punctuation 808 * 2 for an (ASCII) word character 809 * >2 for other word characters 810 */ 811 int 812 mb_get_class(p) 813 char_u *p; 814 { 815 if (MB_BYTE2LEN(p[0]) == 1) 816 { 817 if (p[0] == NUL || vim_iswhite(p[0])) 818 return 0; 819 if (vim_iswordc(p[0])) 820 return 2; 821 return 1; 822 } 823 if (enc_dbcs != 0 && p[0] != NUL && p[1] != NUL) 824 return dbcs_class(p[0], p[1]); 825 if (enc_utf8) 826 return utf_class(utf_ptr2char(p)); 827 return 0; 828 } 829 830 /* 831 * Get class of a double-byte character. This always returns 3 or bigger. 832 * TODO: Should return 1 for punctuation. 833 */ 834 int 835 dbcs_class(lead, trail) 836 unsigned lead; 837 unsigned trail; 838 { 839 switch (enc_dbcs) 840 { 841 /* please add classfy routine for your language in here */ 842 843 case DBCS_JPNU: /* ? */ 844 case DBCS_JPN: 845 { 846 /* JIS code classification */ 847 unsigned char lb = lead; 848 unsigned char tb = trail; 849 850 /* convert process code to JIS */ 851 # if defined(WIN3264) || defined(WIN32UNIX) || defined(MACOS) 852 /* process code is SJIS */ 853 if (lb <= 0x9f) 854 lb = (lb - 0x81) * 2 + 0x21; 855 else 856 lb = (lb - 0xc1) * 2 + 0x21; 857 if (tb <= 0x7e) 858 tb -= 0x1f; 859 else if (tb <= 0x9e) 860 tb -= 0x20; 861 else 862 { 863 tb -= 0x7e; 864 lb += 1; 865 } 866 # else 867 /* 868 * XXX: Code page identification can not use with all 869 * system! So, some other encoding information 870 * will be needed. 871 * In japanese: SJIS,EUC,UNICODE,(JIS) 872 * Note that JIS-code system don't use as 873 * process code in most system because it uses 874 * escape sequences(JIS is context depend encoding). 875 */ 876 /* assume process code is JAPANESE-EUC */ 877 lb &= 0x7f; 878 tb &= 0x7f; 879 # endif 880 /* exceptions */ 881 switch (lb << 8 | tb) 882 { 883 case 0x2121: /* ZENKAKU space */ 884 return 0; 885 case 0x2122: /* KU-TEN (Japanese comma) */ 886 case 0x2123: /* TOU-TEN (Japanese period) */ 887 case 0x2124: /* ZENKAKU comma */ 888 case 0x2125: /* ZENKAKU period */ 889 return 1; 890 case 0x213c: /* prolongedsound handled as KATAKANA */ 891 return 13; 892 } 893 /* sieved by KU code */ 894 switch (lb) 895 { 896 case 0x21: 897 case 0x22: 898 /* special symbols */ 899 return 10; 900 case 0x23: 901 /* alpha-numeric */ 902 return 11; 903 case 0x24: 904 /* hiragana */ 905 return 12; 906 case 0x25: 907 /* katakana */ 908 return 13; 909 case 0x26: 910 /* greek */ 911 return 14; 912 case 0x27: 913 /* russian */ 914 return 15; 915 case 0x28: 916 /* lines */ 917 return 16; 918 default: 919 /* kanji */ 920 return 17; 921 } 922 } 923 924 case DBCS_KORU: /* ? */ 925 case DBCS_KOR: 926 { 927 /* KS code classification */ 928 unsigned char c1 = lead; 929 unsigned char c2 = trail; 930 931 /* 932 * 20 : Hangul 933 * 21 : Hanja 934 * 22 : Symbols 935 * 23 : Alpha-numeric/Roman Letter (Full width) 936 * 24 : Hangul Letter(Alphabet) 937 * 25 : Roman Numeral/Greek Letter 938 * 26 : Box Drawings 939 * 27 : Unit Symbols 940 * 28 : Circled/Parenthesized Letter 941 * 29 : Hirigana/Katakana 942 * 30 : Cyrillic Letter 943 */ 944 945 if (c1 >= 0xB0 && c1 <= 0xC8) 946 /* Hangul */ 947 return 20; 948 #if defined(WIN3264) || defined(WIN32UNIX) 949 else if (c1 <= 0xA0 || c2 <= 0xA0) 950 /* Extended Hangul Region : MS UHC(Unified Hangul Code) */ 951 /* c1: 0x81-0xA0 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xFE 952 * c1: 0xA1-0xC6 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xA0 953 */ 954 return 20; 955 #endif 956 957 else if (c1 >= 0xCA && c1 <= 0xFD) 958 /* Hanja */ 959 return 21; 960 else switch (c1) 961 { 962 case 0xA1: 963 case 0xA2: 964 /* Symbols */ 965 return 22; 966 case 0xA3: 967 /* Alpha-numeric */ 968 return 23; 969 case 0xA4: 970 /* Hangul Letter(Alphabet) */ 971 return 24; 972 case 0xA5: 973 /* Roman Numeral/Greek Letter */ 974 return 25; 975 case 0xA6: 976 /* Box Drawings */ 977 return 26; 978 case 0xA7: 979 /* Unit Symbols */ 980 return 27; 981 case 0xA8: 982 case 0xA9: 983 if (c2 <= 0xAF) 984 return 25; /* Roman Letter */ 985 else if (c2 >= 0xF6) 986 return 22; /* Symbols */ 987 else 988 /* Circled/Parenthesized Letter */ 989 return 28; 990 case 0xAA: 991 case 0xAB: 992 /* Hirigana/Katakana */ 993 return 29; 994 case 0xAC: 995 /* Cyrillic Letter */ 996 return 30; 997 } 998 } 999 default: 1000 break; 1001 } 1002 return 3; 1003 } 1004 1005 /* 1006 * mb_char2len() function pointer. 1007 * Return length in bytes of character "c". 1008 * Returns 1 for a single-byte character. 1009 */ 1010 /* ARGSUSED */ 1011 int 1012 latin_char2len(c) 1013 int c; 1014 { 1015 return 1; 1016 } 1017 1018 static int 1019 dbcs_char2len(c) 1020 int c; 1021 { 1022 if (c >= 0x100) 1023 return 2; 1024 return 1; 1025 } 1026 1027 /* 1028 * mb_char2bytes() function pointer. 1029 * Convert a character to its bytes. 1030 * Returns the length in bytes. 1031 */ 1032 int 1033 latin_char2bytes(c, buf) 1034 int c; 1035 char_u *buf; 1036 { 1037 buf[0] = c; 1038 return 1; 1039 } 1040 1041 static int 1042 dbcs_char2bytes(c, buf) 1043 int c; 1044 char_u *buf; 1045 { 1046 if (c >= 0x100) 1047 { 1048 buf[0] = (unsigned)c >> 8; 1049 buf[1] = c; 1050 /* Never use a NUL byte, it causes lots of trouble. It's an invalid 1051 * character anyway. */ 1052 if (buf[1] == NUL) 1053 buf[1] = '\n'; 1054 return 2; 1055 } 1056 buf[0] = c; 1057 return 1; 1058 } 1059 1060 /* 1061 * mb_ptr2len() function pointer. 1062 * Get byte length of character at "*p" but stop at a NUL. 1063 * For UTF-8 this includes following composing characters. 1064 * Returns 0 when *p is NUL. 1065 * 1066 */ 1067 int 1068 latin_ptr2len(p) 1069 char_u *p; 1070 { 1071 return MB_BYTE2LEN(*p); 1072 } 1073 1074 static int 1075 dbcs_ptr2len(p) 1076 char_u *p; 1077 { 1078 int len; 1079 1080 /* Check if second byte is not missing. */ 1081 len = MB_BYTE2LEN(*p); 1082 if (len == 2 && p[1] == NUL) 1083 len = 1; 1084 return len; 1085 } 1086 1087 struct interval 1088 { 1089 unsigned short first; 1090 unsigned short last; 1091 }; 1092 static int intable __ARGS((struct interval *table, size_t size, int c)); 1093 1094 /* 1095 * Return TRUE if "c" is in "table[size / sizeof(struct interval)]". 1096 */ 1097 static int 1098 intable(table, size, c) 1099 struct interval *table; 1100 size_t size; 1101 int c; 1102 { 1103 int mid, bot, top; 1104 1105 /* first quick check for Latin1 etc. characters */ 1106 if (c < table[0].first) 1107 return FALSE; 1108 1109 /* binary search in table */ 1110 bot = 0; 1111 top = size / sizeof(struct interval) - 1; 1112 while (top >= bot) 1113 { 1114 mid = (bot + top) / 2; 1115 if (table[mid].last < c) 1116 bot = mid + 1; 1117 else if (table[mid].first > c) 1118 top = mid - 1; 1119 else 1120 return TRUE; 1121 } 1122 return FALSE; 1123 } 1124 1125 /* 1126 * For UTF-8 character "c" return 2 for a double-width character, 1 for others. 1127 * Returns 4 or 6 for an unprintable character. 1128 * Is only correct for characters >= 0x80. 1129 * When p_ambw is "double", return 2 for a character with East Asian Width 1130 * class 'A'(mbiguous). 1131 */ 1132 int 1133 utf_char2cells(c) 1134 int c; 1135 { 1136 /* sorted list of non-overlapping intervals of East Asian Ambiguous 1137 * characters, generated with: 1138 * "uniset +WIDTH-A -cat=Me -cat=Mn -cat=Cf c" */ 1139 static struct interval ambiguous[] = { 1140 {0x00A1, 0x00A1}, {0x00A4, 0x00A4}, {0x00A7, 0x00A8}, 1141 {0x00AA, 0x00AA}, {0x00AE, 0x00AE}, {0x00B0, 0x00B4}, 1142 {0x00B6, 0x00BA}, {0x00BC, 0x00BF}, {0x00C6, 0x00C6}, 1143 {0x00D0, 0x00D0}, {0x00D7, 0x00D8}, {0x00DE, 0x00E1}, 1144 {0x00E6, 0x00E6}, {0x00E8, 0x00EA}, {0x00EC, 0x00ED}, 1145 {0x00F0, 0x00F0}, {0x00F2, 0x00F3}, {0x00F7, 0x00FA}, 1146 {0x00FC, 0x00FC}, {0x00FE, 0x00FE}, {0x0101, 0x0101}, 1147 {0x0111, 0x0111}, {0x0113, 0x0113}, {0x011B, 0x011B}, 1148 {0x0126, 0x0127}, {0x012B, 0x012B}, {0x0131, 0x0133}, 1149 {0x0138, 0x0138}, {0x013F, 0x0142}, {0x0144, 0x0144}, 1150 {0x0148, 0x014B}, {0x014D, 0x014D}, {0x0152, 0x0153}, 1151 {0x0166, 0x0167}, {0x016B, 0x016B}, {0x01CE, 0x01CE}, 1152 {0x01D0, 0x01D0}, {0x01D2, 0x01D2}, {0x01D4, 0x01D4}, 1153 {0x01D6, 0x01D6}, {0x01D8, 0x01D8}, {0x01DA, 0x01DA}, 1154 {0x01DC, 0x01DC}, {0x0251, 0x0251}, {0x0261, 0x0261}, 1155 {0x02C4, 0x02C4}, {0x02C7, 0x02C7}, {0x02C9, 0x02CB}, 1156 {0x02CD, 0x02CD}, {0x02D0, 0x02D0}, {0x02D8, 0x02DB}, 1157 {0x02DD, 0x02DD}, {0x02DF, 0x02DF}, {0x0391, 0x03A1}, 1158 {0x03A3, 0x03A9}, {0x03B1, 0x03C1}, {0x03C3, 0x03C9}, 1159 {0x0401, 0x0401}, {0x0410, 0x044F}, {0x0451, 0x0451}, 1160 {0x2010, 0x2010}, {0x2013, 0x2016}, {0x2018, 0x2019}, 1161 {0x201C, 0x201D}, {0x2020, 0x2022}, {0x2024, 0x2027}, 1162 {0x2030, 0x2030}, {0x2032, 0x2033}, {0x2035, 0x2035}, 1163 {0x203B, 0x203B}, {0x203E, 0x203E}, {0x2074, 0x2074}, 1164 {0x207F, 0x207F}, {0x2081, 0x2084}, {0x20AC, 0x20AC}, 1165 {0x2103, 0x2103}, {0x2105, 0x2105}, {0x2109, 0x2109}, 1166 {0x2113, 0x2113}, {0x2116, 0x2116}, {0x2121, 0x2122}, 1167 {0x2126, 0x2126}, {0x212B, 0x212B}, {0x2153, 0x2154}, 1168 {0x215B, 0x215E}, {0x2160, 0x216B}, {0x2170, 0x2179}, 1169 {0x2190, 0x2199}, {0x21B8, 0x21B9}, {0x21D2, 0x21D2}, 1170 {0x21D4, 0x21D4}, {0x21E7, 0x21E7}, {0x2200, 0x2200}, 1171 {0x2202, 0x2203}, {0x2207, 0x2208}, {0x220B, 0x220B}, 1172 {0x220F, 0x220F}, {0x2211, 0x2211}, {0x2215, 0x2215}, 1173 {0x221A, 0x221A}, {0x221D, 0x2220}, {0x2223, 0x2223}, 1174 {0x2225, 0x2225}, {0x2227, 0x222C}, {0x222E, 0x222E}, 1175 {0x2234, 0x2237}, {0x223C, 0x223D}, {0x2248, 0x2248}, 1176 {0x224C, 0x224C}, {0x2252, 0x2252}, {0x2260, 0x2261}, 1177 {0x2264, 0x2267}, {0x226A, 0x226B}, {0x226E, 0x226F}, 1178 {0x2282, 0x2283}, {0x2286, 0x2287}, {0x2295, 0x2295}, 1179 {0x2299, 0x2299}, {0x22A5, 0x22A5}, {0x22BF, 0x22BF}, 1180 {0x2312, 0x2312}, {0x2460, 0x24E9}, {0x24EB, 0x254B}, 1181 {0x2550, 0x2573}, {0x2580, 0x258F}, {0x2592, 0x2595}, 1182 {0x25A0, 0x25A1}, {0x25A3, 0x25A9}, {0x25B2, 0x25B3}, 1183 {0x25B6, 0x25B7}, {0x25BC, 0x25BD}, {0x25C0, 0x25C1}, 1184 {0x25C6, 0x25C8}, {0x25CB, 0x25CB}, {0x25CE, 0x25D1}, 1185 {0x25E2, 0x25E5}, {0x25EF, 0x25EF}, {0x2605, 0x2606}, 1186 {0x2609, 0x2609}, {0x260E, 0x260F}, {0x2614, 0x2615}, 1187 {0x261C, 0x261C}, {0x261E, 0x261E}, {0x2640, 0x2640}, 1188 {0x2642, 0x2642}, {0x2660, 0x2661}, {0x2663, 0x2665}, 1189 {0x2667, 0x266A}, {0x266C, 0x266D}, {0x266F, 0x266F}, 1190 {0x273D, 0x273D}, {0x2776, 0x277F}, {0xE000, 0xF8FF}, 1191 {0xFFFD, 0xFFFD}, /* {0xF0000, 0xFFFFD}, {0x100000, 0x10FFFD} */ 1192 }; 1193 1194 if (c >= 0x100) 1195 { 1196 #ifdef USE_WCHAR_FUNCTIONS 1197 /* 1198 * Assume the library function wcwidth() works better than our own 1199 * stuff. It should return 1 for ambiguous width chars! 1200 */ 1201 int n = wcwidth(c); 1202 1203 if (n < 0) 1204 return 6; /* unprintable, displays <xxxx> */ 1205 if (n > 1) 1206 return n; 1207 #else 1208 if (!utf_printable(c)) 1209 return 6; /* unprintable, displays <xxxx> */ 1210 if (c >= 0x1100 1211 && (c <= 0x115f /* Hangul Jamo */ 1212 || c == 0x2329 1213 || c == 0x232a 1214 || (c >= 0x2e80 && c <= 0xa4cf 1215 && c != 0x303f) /* CJK ... Yi */ 1216 || (c >= 0xac00 && c <= 0xd7a3) /* Hangul Syllables */ 1217 || (c >= 0xf900 && c <= 0xfaff) /* CJK Compatibility 1218 Ideographs */ 1219 || (c >= 0xfe30 && c <= 0xfe6f) /* CJK Compatibility Forms */ 1220 || (c >= 0xff00 && c <= 0xff60) /* Fullwidth Forms */ 1221 || (c >= 0xffe0 && c <= 0xffe6) 1222 || (c >= 0x20000 && c <= 0x2fffd) 1223 || (c >= 0x30000 && c <= 0x3fffd))) 1224 return 2; 1225 #endif 1226 } 1227 1228 /* Characters below 0x100 are influenced by 'isprint' option */ 1229 else if (c >= 0x80 && !vim_isprintc(c)) 1230 return 4; /* unprintable, displays <xx> */ 1231 1232 if (c >= 0x80 && *p_ambw == 'd' && intable(ambiguous, sizeof(ambiguous), c)) 1233 return 2; 1234 1235 return 1; 1236 } 1237 1238 /* 1239 * mb_ptr2cells() function pointer. 1240 * Return the number of display cells character at "*p" occupies. 1241 * This doesn't take care of unprintable characters, use ptr2cells() for that. 1242 */ 1243 /*ARGSUSED*/ 1244 int 1245 latin_ptr2cells(p) 1246 char_u *p; 1247 { 1248 return 1; 1249 } 1250 1251 int 1252 utf_ptr2cells(p) 1253 char_u *p; 1254 { 1255 int c; 1256 1257 /* Need to convert to a wide character. */ 1258 if (*p >= 0x80) 1259 { 1260 c = utf_ptr2char(p); 1261 /* An illegal byte is displayed as <xx>. */ 1262 if (utf_ptr2len(p) == 1 || c == NUL) 1263 return 4; 1264 /* If the char is ASCII it must be an overlong sequence. */ 1265 if (c < 0x80) 1266 return char2cells(c); 1267 return utf_char2cells(c); 1268 } 1269 return 1; 1270 } 1271 1272 int 1273 dbcs_ptr2cells(p) 1274 char_u *p; 1275 { 1276 /* Number of cells is equal to number of bytes, except for euc-jp when 1277 * the first byte is 0x8e. */ 1278 if (enc_dbcs == DBCS_JPNU && *p == 0x8e) 1279 return 1; 1280 return MB_BYTE2LEN(*p); 1281 } 1282 1283 /* 1284 * mb_char2cells() function pointer. 1285 * Return the number of display cells character "c" occupies. 1286 * Only takes care of multi-byte chars, not "^C" and such. 1287 */ 1288 /*ARGSUSED*/ 1289 int 1290 latin_char2cells(c) 1291 int c; 1292 { 1293 return 1; 1294 } 1295 1296 static int 1297 dbcs_char2cells(c) 1298 int c; 1299 { 1300 /* Number of cells is equal to number of bytes, except for euc-jp when 1301 * the first byte is 0x8e. */ 1302 if (enc_dbcs == DBCS_JPNU && ((unsigned)c >> 8) == 0x8e) 1303 return 1; 1304 /* use the first byte */ 1305 return MB_BYTE2LEN((unsigned)c >> 8); 1306 } 1307 1308 /* 1309 * mb_off2cells() function pointer. 1310 * Return number of display cells for char at ScreenLines[off]. 1311 * Caller must make sure "off" and "off + 1" are valid! 1312 */ 1313 /*ARGSUSED*/ 1314 int 1315 latin_off2cells(off) 1316 unsigned off; 1317 { 1318 return 1; 1319 } 1320 1321 int 1322 dbcs_off2cells(off) 1323 unsigned off; 1324 { 1325 /* Number of cells is equal to number of bytes, except for euc-jp when 1326 * the first byte is 0x8e. */ 1327 if (enc_dbcs == DBCS_JPNU && ScreenLines[off] == 0x8e) 1328 return 1; 1329 return MB_BYTE2LEN(ScreenLines[off]); 1330 } 1331 1332 int 1333 utf_off2cells(off) 1334 unsigned off; 1335 { 1336 return ScreenLines[off + 1] == 0 ? 2 : 1; 1337 } 1338 1339 /* 1340 * mb_ptr2char() function pointer. 1341 * Convert a byte sequence into a character. 1342 */ 1343 int 1344 latin_ptr2char(p) 1345 char_u *p; 1346 { 1347 return *p; 1348 } 1349 1350 static int 1351 dbcs_ptr2char(p) 1352 char_u *p; 1353 { 1354 if (MB_BYTE2LEN(*p) > 1 && p[1] != NUL) 1355 return (p[0] << 8) + p[1]; 1356 return *p; 1357 } 1358 1359 /* 1360 * Convert a UTF-8 byte sequence to a wide character. 1361 * If the sequence is illegal or truncated by a NUL the first byte is 1362 * returned. 1363 * Does not include composing characters, of course. 1364 */ 1365 int 1366 utf_ptr2char(p) 1367 char_u *p; 1368 { 1369 int len; 1370 1371 if (p[0] < 0x80) /* be quick for ASCII */ 1372 return p[0]; 1373 1374 len = utf8len_tab[p[0]]; 1375 if ((p[1] & 0xc0) == 0x80) 1376 { 1377 if (len == 2) 1378 return ((p[0] & 0x1f) << 6) + (p[1] & 0x3f); 1379 if ((p[2] & 0xc0) == 0x80) 1380 { 1381 if (len == 3) 1382 return ((p[0] & 0x0f) << 12) + ((p[1] & 0x3f) << 6) 1383 + (p[2] & 0x3f); 1384 if ((p[3] & 0xc0) == 0x80) 1385 { 1386 if (len == 4) 1387 return ((p[0] & 0x07) << 18) + ((p[1] & 0x3f) << 12) 1388 + ((p[2] & 0x3f) << 6) + (p[3] & 0x3f); 1389 if ((p[4] & 0xc0) == 0x80) 1390 { 1391 if (len == 5) 1392 return ((p[0] & 0x03) << 24) + ((p[1] & 0x3f) << 18) 1393 + ((p[2] & 0x3f) << 12) + ((p[3] & 0x3f) << 6) 1394 + (p[4] & 0x3f); 1395 if ((p[5] & 0xc0) == 0x80 && len == 6) 1396 return ((p[0] & 0x01) << 30) + ((p[1] & 0x3f) << 24) 1397 + ((p[2] & 0x3f) << 18) + ((p[3] & 0x3f) << 12) 1398 + ((p[4] & 0x3f) << 6) + (p[5] & 0x3f); 1399 } 1400 } 1401 } 1402 } 1403 /* Illegal value, just return the first byte */ 1404 return p[0]; 1405 } 1406 1407 /* 1408 * Get character at **pp and advance *pp to the next character. 1409 * Note: composing characters are skipped! 1410 */ 1411 int 1412 mb_ptr2char_adv(pp) 1413 char_u **pp; 1414 { 1415 int c; 1416 1417 c = (*mb_ptr2char)(*pp); 1418 *pp += (*mb_ptr2len)(*pp); 1419 return c; 1420 } 1421 1422 /* 1423 * Get character at **pp and advance *pp to the next character. 1424 * Note: composing characters are returned as separate characters. 1425 */ 1426 int 1427 mb_cptr2char_adv(pp) 1428 char_u **pp; 1429 { 1430 int c; 1431 1432 c = (*mb_ptr2char)(*pp); 1433 if (enc_utf8) 1434 *pp += utf_ptr2len(*pp); 1435 else 1436 *pp += (*mb_ptr2len)(*pp); 1437 return c; 1438 } 1439 1440 #if defined(FEAT_ARABIC) || defined(PROTO) 1441 /* 1442 * Check whether we are dealing with Arabic combining characters. 1443 * Note: these are NOT really composing characters! 1444 */ 1445 int 1446 arabic_combine(one, two) 1447 int one; /* first character */ 1448 int two; /* character just after "one" */ 1449 { 1450 if (one == a_LAM) 1451 return arabic_maycombine(two); 1452 return FALSE; 1453 } 1454 1455 /* 1456 * Check whether we are dealing with a character that could be regarded as an 1457 * Arabic combining character, need to check the character before this. 1458 */ 1459 int 1460 arabic_maycombine(two) 1461 int two; 1462 { 1463 if (p_arshape && !p_tbidi) 1464 return (two == a_ALEF_MADDA 1465 || two == a_ALEF_HAMZA_ABOVE 1466 || two == a_ALEF_HAMZA_BELOW 1467 || two == a_ALEF); 1468 return FALSE; 1469 } 1470 1471 /* 1472 * Check if the character pointed to by "p2" is a composing character when it 1473 * comes after "p1". For Arabic sometimes "ab" is replaced with "c", which 1474 * behaves like a composing character. 1475 */ 1476 int 1477 utf_composinglike(p1, p2) 1478 char_u *p1; 1479 char_u *p2; 1480 { 1481 int c2; 1482 1483 c2 = utf_ptr2char(p2); 1484 if (utf_iscomposing(c2)) 1485 return TRUE; 1486 if (!arabic_maycombine(c2)) 1487 return FALSE; 1488 return arabic_combine(utf_ptr2char(p1), c2); 1489 } 1490 #endif 1491 1492 /* 1493 * Convert a UTF-8 byte string to a wide chararacter. Also get up to two 1494 * composing characters. 1495 */ 1496 int 1497 utfc_ptr2char(p, p1, p2) 1498 char_u *p; 1499 int *p1; /* return: first composing char or 0 */ 1500 int *p2; /* return: second composing char or 0 */ 1501 { 1502 int len; 1503 int c; 1504 int cc; 1505 1506 c = utf_ptr2char(p); 1507 len = utf_ptr2len(p); 1508 /* Only accept a composing char when the first char isn't illegal. */ 1509 if ((len > 1 || *p < 0x80) 1510 && p[len] >= 0x80 1511 && UTF_COMPOSINGLIKE(p, p + len)) 1512 { 1513 *p1 = utf_ptr2char(p + len); 1514 len += utf_ptr2len(p + len); 1515 if (p[len] >= 0x80 && utf_iscomposing(cc = utf_ptr2char(p + len))) 1516 *p2 = cc; 1517 else 1518 *p2 = 0; 1519 } 1520 else 1521 { 1522 *p1 = 0; 1523 *p2 = 0; 1524 } 1525 return c; 1526 } 1527 1528 /* 1529 * Convert a UTF-8 byte string to a wide chararacter. Also get up to two 1530 * composing characters. Use no more than p[maxlen]. 1531 */ 1532 int 1533 utfc_ptr2char_len(p, p1, p2, maxlen) 1534 char_u *p; 1535 int *p1; /* return: first composing char or 0 */ 1536 int *p2; /* return: second composing char or 0 */ 1537 int maxlen; 1538 { 1539 int len; 1540 int c; 1541 int cc; 1542 1543 c = utf_ptr2char(p); 1544 len = utf_ptr2len_len(p, maxlen); 1545 /* Only accept a composing char when the first char isn't illegal. */ 1546 if ((len > 1 || *p < 0x80) 1547 && len < maxlen 1548 && p[len] >= 0x80 1549 && UTF_COMPOSINGLIKE(p, p + len)) 1550 { 1551 *p1 = utf_ptr2char(p + len); 1552 len += utf_ptr2len_len(p + len, maxlen - len); 1553 if (len < maxlen 1554 && p[len] >= 0x80 1555 && utf_iscomposing(cc = utf_ptr2char(p + len))) 1556 *p2 = cc; 1557 else 1558 *p2 = 0; 1559 } 1560 else 1561 { 1562 *p1 = 0; 1563 *p2 = 0; 1564 } 1565 return c; 1566 } 1567 1568 /* 1569 * Convert the character at screen position "off" to a sequence of bytes. 1570 * Includes the composing characters. 1571 * "buf" must at least have the length MB_MAXBYTES. 1572 * Returns the produced number of bytes. 1573 */ 1574 int 1575 utfc_char2bytes(off, buf) 1576 int off; 1577 char_u *buf; 1578 { 1579 int len; 1580 1581 len = utf_char2bytes(ScreenLinesUC[off], buf); 1582 if (ScreenLinesC1[off] != 0) 1583 { 1584 len += utf_char2bytes(ScreenLinesC1[off], buf + len); 1585 if (ScreenLinesC2[off] != 0) 1586 len += utf_char2bytes(ScreenLinesC2[off], buf + len); 1587 } 1588 return len; 1589 } 1590 1591 /* 1592 * Get the length of a UTF-8 byte sequence, not including any following 1593 * composing characters. 1594 * Returns 0 for "". 1595 * Returns 1 for an illegal byte sequence. 1596 */ 1597 int 1598 utf_ptr2len(p) 1599 char_u *p; 1600 { 1601 int len; 1602 int i; 1603 1604 if (*p == NUL) 1605 return 0; 1606 len = utf8len_tab[*p]; 1607 for (i = 1; i < len; ++i) 1608 if ((p[i] & 0xc0) != 0x80) 1609 return 1; 1610 return len; 1611 } 1612 1613 /* 1614 * Return length of UTF-8 character, obtained from the first byte. 1615 * "b" must be between 0 and 255! 1616 */ 1617 int 1618 utf_byte2len(b) 1619 int b; 1620 { 1621 return utf8len_tab[b]; 1622 } 1623 1624 /* 1625 * Get the length of UTF-8 byte sequence "p[size]". Does not include any 1626 * following composing characters. 1627 * Returns 1 for "". 1628 * Returns 1 for an illegal byte sequence. 1629 * Returns number > "size" for an incomplete byte sequence. 1630 */ 1631 int 1632 utf_ptr2len_len(p, size) 1633 char_u *p; 1634 int size; 1635 { 1636 int len; 1637 int i; 1638 1639 if (*p == NUL) 1640 return 1; 1641 len = utf8len_tab[*p]; 1642 if (len > size) 1643 return len; /* incomplete byte sequence. */ 1644 for (i = 1; i < len; ++i) 1645 if ((p[i] & 0xc0) != 0x80) 1646 return 1; 1647 return len; 1648 } 1649 1650 /* 1651 * Return the number of bytes the UTF-8 encoding of the character at "p" takes. 1652 * This includes following composing characters. 1653 */ 1654 int 1655 utfc_ptr2len(p) 1656 char_u *p; 1657 { 1658 int len; 1659 int b0 = *p; 1660 #ifdef FEAT_ARABIC 1661 int prevlen; 1662 #endif 1663 1664 if (b0 == NUL) 1665 return 0; 1666 if (b0 < 0x80 && p[1] < 0x80) /* be quick for ASCII */ 1667 return 1; 1668 1669 /* Skip over first UTF-8 char, stopping at a NUL byte. */ 1670 len = utf_ptr2len(p); 1671 1672 /* Check for illegal byte. */ 1673 if (len == 1 && b0 >= 0x80) 1674 return 1; 1675 1676 /* 1677 * Check for composing characters. We can handle only the first two, but 1678 * skip all of them (otherwise the cursor would get stuck). 1679 */ 1680 #ifdef FEAT_ARABIC 1681 prevlen = 0; 1682 #endif 1683 for (;;) 1684 { 1685 if (p[len] < 0x80 || !UTF_COMPOSINGLIKE(p + prevlen, p + len)) 1686 return len; 1687 1688 /* Skip over composing char */ 1689 #ifdef FEAT_ARABIC 1690 prevlen = len; 1691 #endif 1692 len += utf_ptr2len(p + len); 1693 } 1694 } 1695 1696 /* 1697 * Return the number of bytes the UTF-8 encoding of the character at "p[size]" 1698 * takes. This includes following composing characters. 1699 * Returns 1 for an illegal char or an incomplete byte sequence. 1700 */ 1701 int 1702 utfc_ptr2len_len(p, size) 1703 char_u *p; 1704 int size; 1705 { 1706 int len; 1707 #ifdef FEAT_ARABIC 1708 int prevlen; 1709 #endif 1710 1711 if (*p == NUL) 1712 return 0; 1713 if (p[0] < 0x80 && (size == 1 || p[1] < 0x80)) /* be quick for ASCII */ 1714 return 1; 1715 1716 /* Skip over first UTF-8 char, stopping at a NUL byte. */ 1717 len = utf_ptr2len_len(p, size); 1718 1719 /* Check for illegal byte and incomplete byte sequence. */ 1720 if ((len == 1 && p[0] >= 0x80) || len > size) 1721 return 1; 1722 1723 /* 1724 * Check for composing characters. We can handle only the first two, but 1725 * skip all of them (otherwise the cursor would get stuck). 1726 */ 1727 #ifdef FEAT_ARABIC 1728 prevlen = 0; 1729 #endif 1730 while (len < size) 1731 { 1732 if (p[len] < 0x80 || !UTF_COMPOSINGLIKE(p + prevlen, p + len)) 1733 break; 1734 1735 /* Skip over composing char */ 1736 #ifdef FEAT_ARABIC 1737 prevlen = len; 1738 #endif 1739 len += utf_ptr2len_len(p + len, size - len); 1740 } 1741 return len; 1742 } 1743 1744 /* 1745 * Return the number of bytes the UTF-8 encoding of character "c" takes. 1746 * This does not include composing characters. 1747 */ 1748 int 1749 utf_char2len(c) 1750 int c; 1751 { 1752 if (c < 0x80) 1753 return 1; 1754 if (c < 0x800) 1755 return 2; 1756 if (c < 0x10000) 1757 return 3; 1758 if (c < 0x200000) 1759 return 4; 1760 if (c < 0x4000000) 1761 return 5; 1762 return 6; 1763 } 1764 1765 /* 1766 * Convert Unicode character "c" to UTF-8 string in "buf[]". 1767 * Returns the number of bytes. 1768 * This does not include composing characters. 1769 */ 1770 int 1771 utf_char2bytes(c, buf) 1772 int c; 1773 char_u *buf; 1774 { 1775 if (c < 0x80) /* 7 bits */ 1776 { 1777 buf[0] = c; 1778 return 1; 1779 } 1780 if (c < 0x800) /* 11 bits */ 1781 { 1782 buf[0] = 0xc0 + ((unsigned)c >> 6); 1783 buf[1] = 0x80 + (c & 0x3f); 1784 return 2; 1785 } 1786 if (c < 0x10000) /* 16 bits */ 1787 { 1788 buf[0] = 0xe0 + ((unsigned)c >> 12); 1789 buf[1] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1790 buf[2] = 0x80 + (c & 0x3f); 1791 return 3; 1792 } 1793 if (c < 0x200000) /* 21 bits */ 1794 { 1795 buf[0] = 0xf0 + ((unsigned)c >> 18); 1796 buf[1] = 0x80 + (((unsigned)c >> 12) & 0x3f); 1797 buf[2] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1798 buf[3] = 0x80 + (c & 0x3f); 1799 return 4; 1800 } 1801 if (c < 0x4000000) /* 26 bits */ 1802 { 1803 buf[0] = 0xf8 + ((unsigned)c >> 24); 1804 buf[1] = 0x80 + (((unsigned)c >> 18) & 0x3f); 1805 buf[2] = 0x80 + (((unsigned)c >> 12) & 0x3f); 1806 buf[3] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1807 buf[4] = 0x80 + (c & 0x3f); 1808 return 5; 1809 } 1810 /* 31 bits */ 1811 buf[0] = 0xfc + ((unsigned)c >> 30); 1812 buf[1] = 0x80 + (((unsigned)c >> 24) & 0x3f); 1813 buf[2] = 0x80 + (((unsigned)c >> 18) & 0x3f); 1814 buf[3] = 0x80 + (((unsigned)c >> 12) & 0x3f); 1815 buf[4] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1816 buf[5] = 0x80 + (c & 0x3f); 1817 return 6; 1818 } 1819 1820 /* 1821 * Return TRUE if "c" is a composing UTF-8 character. This means it will be 1822 * drawn on top of the preceding character. 1823 * Based on code from Markus Kuhn. 1824 */ 1825 int 1826 utf_iscomposing(c) 1827 int c; 1828 { 1829 /* sorted list of non-overlapping intervals */ 1830 static struct interval combining[] = 1831 { 1832 {0x0300, 0x034f}, {0x0360, 0x036f}, {0x0483, 0x0486}, {0x0488, 0x0489}, 1833 {0x0591, 0x05a1}, {0x05a3, 0x05b9}, {0x05bb, 0x05bd}, {0x05bf, 0x05bf}, 1834 {0x05c1, 0x05c2}, {0x05c4, 0x05c4}, {0x0610, 0x0615}, {0x064b, 0x0658}, 1835 {0x0670, 0x0670}, {0x06d6, 0x06dc}, {0x06de, 0x06e4}, {0x06e7, 0x06e8}, 1836 {0x06ea, 0x06ed}, {0x0711, 0x0711}, {0x0730, 0x074a}, {0x07a6, 0x07b0}, 1837 {0x0901, 0x0903}, {0x093c, 0x093c}, {0x093e, 0x094d}, {0x0951, 0x0954}, 1838 {0x0962, 0x0963}, {0x0981, 0x0983}, {0x09bc, 0x09bc}, {0x09be, 0x09c4}, 1839 {0x09c7, 0x09c8}, {0x09cb, 0x09cd}, {0x09d7, 0x09d7}, {0x09e2, 0x09e3}, 1840 {0x0a01, 0x0a03}, {0x0a3c, 0x0a3c}, {0x0a3e, 0x0a42}, {0x0a47, 0x0a48}, 1841 {0x0a4b, 0x0a4d}, {0x0a70, 0x0a71}, {0x0a81, 0x0a83}, {0x0abc, 0x0abc}, 1842 {0x0abe, 0x0ac5}, {0x0ac7, 0x0ac9}, {0x0acb, 0x0acd}, {0x0ae2, 0x0ae3}, 1843 {0x0b01, 0x0b03}, {0x0b3c, 0x0b3c}, {0x0b3e, 0x0b43}, {0x0b47, 0x0b48}, 1844 {0x0b4b, 0x0b4d}, {0x0b56, 0x0b57}, {0x0b82, 0x0b82}, {0x0bbe, 0x0bc2}, 1845 {0x0bc6, 0x0bc8}, {0x0bca, 0x0bcd}, {0x0bd7, 0x0bd7}, {0x0c01, 0x0c03}, 1846 {0x0c3e, 0x0c44}, {0x0c46, 0x0c48}, {0x0c4a, 0x0c4d}, {0x0c55, 0x0c56}, 1847 {0x0c82, 0x0c83}, {0x0cbc, 0x0cbc}, {0x0cbe, 0x0cc4}, {0x0cc6, 0x0cc8}, 1848 {0x0cca, 0x0ccd}, {0x0cd5, 0x0cd6}, {0x0d02, 0x0d03}, {0x0d3e, 0x0d43}, 1849 {0x0d46, 0x0d48}, {0x0d4a, 0x0d4d}, {0x0d57, 0x0d57}, {0x0d82, 0x0d83}, 1850 {0x0dca, 0x0dca}, {0x0dcf, 0x0dd4}, {0x0dd6, 0x0dd6}, {0x0dd8, 0x0ddf}, 1851 {0x0df2, 0x0df3}, {0x0e31, 0x0e31}, {0x0e34, 0x0e3a}, {0x0e47, 0x0e4e}, 1852 {0x0eb1, 0x0eb1}, {0x0eb4, 0x0eb9}, {0x0ebb, 0x0ebc}, {0x0ec8, 0x0ecd}, 1853 {0x0f18, 0x0f19}, {0x0f35, 0x0f35}, {0x0f37, 0x0f37}, {0x0f39, 0x0f39}, 1854 {0x0f3e, 0x0f3f}, {0x0f71, 0x0f84}, {0x0f86, 0x0f87}, {0x0f90, 0x0f97}, 1855 {0x0f99, 0x0fbc}, {0x0fc6, 0x0fc6}, {0x102c, 0x1032}, {0x1036, 0x1039}, 1856 {0x1056, 0x1059}, {0x1712, 0x1714}, {0x1732, 0x1734}, {0x1752, 0x1753}, 1857 {0x1772, 0x1773}, {0x17b6, 0x17d3}, {0x17dd, 0x17dd}, {0x180b, 0x180d}, 1858 {0x18a9, 0x18a9}, {0x1920, 0x192b}, {0x1930, 0x193b}, {0x20d0, 0x20ea}, 1859 {0x302a, 0x302f}, {0x3099, 0x309a}, {0xfb1e, 0xfb1e}, {0xfe00, 0xfe0f}, 1860 {0xfe20, 0xfe23}, 1861 }; 1862 1863 return intable(combining, sizeof(combining), c); 1864 } 1865 1866 /* 1867 * Return TRUE for characters that can be displayed in a normal way. 1868 * Only for characters of 0x100 and above! 1869 */ 1870 int 1871 utf_printable(c) 1872 int c; 1873 { 1874 #ifdef USE_WCHAR_FUNCTIONS 1875 /* 1876 * Assume the iswprint() library function works better than our own stuff. 1877 */ 1878 return iswprint(c); 1879 #else 1880 /* Sorted list of non-overlapping intervals. 1881 * 0xd800-0xdfff is reserved for UTF-16, actually illegal. */ 1882 static struct interval nonprint[] = 1883 { 1884 {0x070f, 0x070f}, {0x180b, 0x180e}, {0x200b, 0x200f}, {0x202a, 0x202e}, 1885 {0x206a, 0x206f}, {0xd800, 0xdfff}, {0xfeff, 0xfeff}, {0xfff9, 0xfffb}, 1886 {0xfffe, 0xffff} 1887 }; 1888 1889 return !intable(nonprint, sizeof(nonprint), c); 1890 #endif 1891 } 1892 1893 /* 1894 * Get class of a Unicode character. 1895 * 0: white space 1896 * 1: punctuation 1897 * 2 or bigger: some class of word character. 1898 */ 1899 int 1900 utf_class(c) 1901 int c; 1902 { 1903 /* sorted list of non-overlapping intervals */ 1904 static struct clinterval 1905 { 1906 unsigned short first; 1907 unsigned short last; 1908 unsigned short class; 1909 } classes[] = 1910 { 1911 {0x037e, 0x037e, 1}, /* Greek question mark */ 1912 {0x0387, 0x0387, 1}, /* Greek ano teleia */ 1913 {0x055a, 0x055f, 1}, /* Armenian punctuation */ 1914 {0x0589, 0x0589, 1}, /* Armenian full stop */ 1915 {0x05be, 0x05be, 1}, 1916 {0x05c0, 0x05c0, 1}, 1917 {0x05c3, 0x05c3, 1}, 1918 {0x05f3, 0x05f4, 1}, 1919 {0x060c, 0x060c, 1}, 1920 {0x061b, 0x061b, 1}, 1921 {0x061f, 0x061f, 1}, 1922 {0x066a, 0x066d, 1}, 1923 {0x06d4, 0x06d4, 1}, 1924 {0x0700, 0x070d, 1}, /* Syriac punctuation */ 1925 {0x0964, 0x0965, 1}, 1926 {0x0970, 0x0970, 1}, 1927 {0x0df4, 0x0df4, 1}, 1928 {0x0e4f, 0x0e4f, 1}, 1929 {0x0e5a, 0x0e5b, 1}, 1930 {0x0f04, 0x0f12, 1}, 1931 {0x0f3a, 0x0f3d, 1}, 1932 {0x0f85, 0x0f85, 1}, 1933 {0x104a, 0x104f, 1}, /* Myanmar punctuation */ 1934 {0x10fb, 0x10fb, 1}, /* Georgian punctuation */ 1935 {0x1361, 0x1368, 1}, /* Ethiopic punctuation */ 1936 {0x166d, 0x166e, 1}, /* Canadian Syl. punctuation */ 1937 {0x1680, 0x1680, 0}, 1938 {0x169b, 0x169c, 1}, 1939 {0x16eb, 0x16ed, 1}, 1940 {0x1735, 0x1736, 1}, 1941 {0x17d4, 0x17dc, 1}, /* Khmer punctuation */ 1942 {0x1800, 0x180a, 1}, /* Mongolian punctuation */ 1943 {0x2000, 0x200b, 0}, /* spaces */ 1944 {0x200c, 0x2027, 1}, /* punctuation and symbols */ 1945 {0x2028, 0x2029, 0}, 1946 {0x202a, 0x202e, 1}, /* punctuation and symbols */ 1947 {0x202f, 0x202f, 0}, 1948 {0x2030, 0x205e, 1}, /* punctuation and symbols */ 1949 {0x205f, 0x205f, 0}, 1950 {0x2060, 0x27ff, 1}, /* punctuation and symbols */ 1951 {0x2070, 0x207f, 0x2070}, /* superscript */ 1952 {0x2080, 0x208f, 0x2080}, /* subscript */ 1953 {0x2983, 0x2998, 1}, 1954 {0x29d8, 0x29db, 1}, 1955 {0x29fc, 0x29fd, 1}, 1956 {0x3000, 0x3000, 0}, /* ideographic space */ 1957 {0x3001, 0x3020, 1}, /* ideographic punctuation */ 1958 {0x3030, 0x3030, 1}, 1959 {0x303d, 0x303d, 1}, 1960 {0x3040, 0x309f, 0x3040}, /* Hiragana */ 1961 {0x30a0, 0x30ff, 0x30a0}, /* Katakana */ 1962 {0x3300, 0x9fff, 0x4e00}, /* CJK Ideographs */ 1963 {0xac00, 0xd7a3, 0xac00}, /* Hangul Syllables */ 1964 {0xf900, 0xfaff, 0x4e00}, /* CJK Ideographs */ 1965 {0xfd3e, 0xfd3f, 1}, 1966 {0xfe30, 0xfe6b, 1}, /* punctuation forms */ 1967 {0xff00, 0xff0f, 1}, /* half/fullwidth ASCII */ 1968 {0xff1a, 0xff20, 1}, /* half/fullwidth ASCII */ 1969 {0xff3b, 0xff40, 1}, /* half/fullwidth ASCII */ 1970 {0xff5b, 0xff65, 1}, /* half/fullwidth ASCII */ 1971 }; 1972 int bot = 0; 1973 int top = sizeof(classes) / sizeof(struct clinterval) - 1; 1974 int mid; 1975 1976 /* First quick check for Latin1 characters, use 'iskeyword'. */ 1977 if (c < 0x100) 1978 { 1979 if (c == ' ' || c == '\t' || c == NUL || c == 0xa0) 1980 return 0; /* blank */ 1981 if (vim_iswordc(c)) 1982 return 2; /* word character */ 1983 return 1; /* punctuation */ 1984 } 1985 1986 /* binary search in table */ 1987 while (top >= bot) 1988 { 1989 mid = (bot + top) / 2; 1990 if (classes[mid].last < c) 1991 bot = mid + 1; 1992 else if (classes[mid].first > c) 1993 top = mid - 1; 1994 else 1995 return (int)classes[mid].class; 1996 } 1997 1998 /* most other characters are "word" characters */ 1999 return 2; 2000 } 2001 2002 /* 2003 * Code for Unicode case-dependent operations. Based on notes in 2004 * http://www.unicode.org/Public/UNIDATA/CaseFolding.txt 2005 * This code uses simple case folding, not full case folding. 2006 */ 2007 2008 /* 2009 * The following table is built by foldExtract.pl < CaseFolding.txt . 2010 * It must be in numeric order, because we use binary search on it. 2011 * An entry such as {0x41,0x5a,1,32} means that UCS-4 characters in the range 2012 * from 0x41 to 0x5a inclusive, stepping by 1, are folded by adding 32. 2013 */ 2014 2015 typedef struct 2016 { 2017 int rangeStart; 2018 int rangeEnd; 2019 int step; 2020 int offset; 2021 } convertStruct; 2022 2023 static convertStruct foldCase[] = 2024 { 2025 {0x41,0x5a,1,32}, {0xc0,0xd6,1,32}, {0xd8,0xde,1,32}, 2026 {0x100,0x12e,2,1}, {0x130,0x130,-1,-199}, {0x132,0x136,2,1}, 2027 {0x139,0x147,2,1}, {0x14a,0x176,2,1}, {0x178,0x178,-1,-121}, 2028 {0x179,0x17d,2,1}, {0x181,0x181,-1,210}, {0x182,0x184,2,1}, 2029 {0x186,0x186,-1,206}, {0x187,0x187,-1,1}, {0x189,0x18a,1,205}, 2030 {0x18b,0x18b,-1,1}, {0x18e,0x18e,-1,79}, {0x18f,0x18f,-1,202}, 2031 {0x190,0x190,-1,203}, {0x191,0x191,-1,1}, {0x193,0x193,-1,205}, 2032 {0x194,0x194,-1,207}, {0x196,0x196,-1,211}, {0x197,0x197,-1,209}, 2033 {0x198,0x198,-1,1}, {0x19c,0x19c,-1,211}, {0x19d,0x19d,-1,213}, 2034 {0x19f,0x19f,-1,214}, {0x1a0,0x1a4,2,1}, {0x1a6,0x1a6,-1,218}, 2035 {0x1a7,0x1a7,-1,1}, {0x1a9,0x1a9,-1,218}, {0x1ac,0x1ac,-1,1}, 2036 {0x1ae,0x1ae,-1,218}, {0x1af,0x1af,-1,1}, {0x1b1,0x1b2,1,217}, 2037 {0x1b3,0x1b5,2,1}, {0x1b7,0x1b7,-1,219}, {0x1b8,0x1bc,4,1}, 2038 {0x1c4,0x1c4,-1,2}, {0x1c5,0x1c5,-1,1}, {0x1c7,0x1c7,-1,2}, 2039 {0x1c8,0x1c8,-1,1}, {0x1ca,0x1ca,-1,2}, {0x1cb,0x1db,2,1}, 2040 {0x1de,0x1ee,2,1}, {0x1f1,0x1f1,-1,2}, {0x1f2,0x1f4,2,1}, 2041 {0x1f6,0x1f6,-1,-97}, {0x1f7,0x1f7,-1,-56}, {0x1f8,0x21e,2,1}, 2042 {0x220,0x220,-1,-130}, {0x222,0x232,2,1}, {0x386,0x386,-1,38}, 2043 {0x388,0x38a,1,37}, {0x38c,0x38c,-1,64}, {0x38e,0x38f,1,63}, 2044 {0x391,0x3a1,1,32}, {0x3a3,0x3ab,1,32}, {0x3d8,0x3ee,2,1}, 2045 {0x3f4,0x3f4,-1,-60}, {0x3f7,0x3f7,-1,1}, {0x3f9,0x3f9,-1,-7}, 2046 {0x3fa,0x3fa,-1,1}, {0x400,0x40f,1,80}, {0x410,0x42f,1,32}, 2047 {0x460,0x480,2,1}, {0x48a,0x4be,2,1}, {0x4c1,0x4cd,2,1}, 2048 {0x4d0,0x4f4,2,1}, {0x4f8,0x500,8,1}, {0x502,0x50e,2,1}, 2049 {0x531,0x556,1,48}, {0x1e00,0x1e94,2,1}, {0x1ea0,0x1ef8,2,1}, 2050 {0x1f08,0x1f0f,1,-8}, {0x1f18,0x1f1d,1,-8}, {0x1f28,0x1f2f,1,-8}, 2051 {0x1f38,0x1f3f,1,-8}, {0x1f48,0x1f4d,1,-8}, {0x1f59,0x1f5f,2,-8}, 2052 {0x1f68,0x1f6f,1,-8}, {0x1f88,0x1f8f,1,-8}, {0x1f98,0x1f9f,1,-8}, 2053 {0x1fa8,0x1faf,1,-8}, {0x1fb8,0x1fb9,1,-8}, {0x1fba,0x1fbb,1,-74}, 2054 {0x1fbc,0x1fbc,-1,-9}, {0x1fc8,0x1fcb,1,-86}, {0x1fcc,0x1fcc,-1,-9}, 2055 {0x1fd8,0x1fd9,1,-8}, {0x1fda,0x1fdb,1,-100}, {0x1fe8,0x1fe9,1,-8}, 2056 {0x1fea,0x1feb,1,-112}, {0x1fec,0x1fec,-1,-7}, {0x1ff8,0x1ff9,1,-128}, 2057 {0x1ffa,0x1ffb,1,-126}, {0x1ffc,0x1ffc,-1,-9}, {0x2126,0x2126,-1,-7517}, 2058 {0x212a,0x212a,-1,-8383}, {0x212b,0x212b,-1,-8262}, 2059 {0x2160,0x216f,1,16}, {0x24b6,0x24cf,1,26}, {0xff21,0xff3a,1,32}, 2060 {0x10400,0x10427,1,40} 2061 }; 2062 2063 static int utf_convert(int a, convertStruct table[], int tableSize); 2064 2065 /* 2066 * Generic conversion function for case operations. 2067 * Return the converted equivalent of "a", which is a UCS-4 character. Use 2068 * the given conversion "table". Uses binary search on "table". 2069 */ 2070 static int 2071 utf_convert(a, table, tableSize) 2072 int a; 2073 convertStruct table[]; 2074 int tableSize; 2075 { 2076 int start, mid, end; /* indices into table */ 2077 2078 start = 0; 2079 end = tableSize / sizeof(convertStruct); 2080 while (start < end) 2081 { 2082 /* need to search further */ 2083 mid = (end + start) /2; 2084 if (table[mid].rangeEnd < a) 2085 start = mid + 1; 2086 else 2087 end = mid; 2088 } 2089 if (table[start].rangeStart <= a && a <= table[start].rangeEnd 2090 && (a - table[start].rangeStart) % table[start].step == 0) 2091 return (a + table[start].offset); 2092 else 2093 return a; 2094 } 2095 2096 /* 2097 * Return the folded-case equivalent of "a", which is a UCS-4 character. Uses 2098 * simple case folding. 2099 */ 2100 int 2101 utf_fold(a) 2102 int a; 2103 { 2104 return utf_convert(a, foldCase, sizeof(foldCase)); 2105 } 2106 2107 /* 2108 * The following tables are built by upperLowerExtract.pl < UnicodeData.txt . 2109 * They must be in numeric order, because we use binary search on them. 2110 * An entry such as {0x41,0x5a,1,32} means that UCS-4 characters in the range 2111 * from 0x41 to 0x5a inclusive, stepping by 1, are switched to lower (for 2112 * example) by adding 32. 2113 */ 2114 static convertStruct toLower[] = 2115 { 2116 {0x41,0x5a,1,32}, {0xc0,0xd6,1,32}, {0xd8,0xde,1,32}, 2117 {0x100,0x12e,2,1}, {0x130,0x130,-1,-199}, {0x132,0x136,2,1}, 2118 {0x139,0x147,2,1}, {0x14a,0x176,2,1}, {0x178,0x178,-1,-121}, 2119 {0x179,0x17d,2,1}, {0x181,0x181,-1,210}, {0x182,0x184,2,1}, 2120 {0x186,0x186,-1,206}, {0x187,0x187,-1,1}, {0x189,0x18a,1,205}, 2121 {0x18b,0x18b,-1,1}, {0x18e,0x18e,-1,79}, {0x18f,0x18f,-1,202}, 2122 {0x190,0x190,-1,203}, {0x191,0x191,-1,1}, {0x193,0x193,-1,205}, 2123 {0x194,0x194,-1,207}, {0x196,0x196,-1,211}, {0x197,0x197,-1,209}, 2124 {0x198,0x198,-1,1}, {0x19c,0x19c,-1,211}, {0x19d,0x19d,-1,213}, 2125 {0x19f,0x19f,-1,214}, {0x1a0,0x1a4,2,1}, {0x1a6,0x1a6,-1,218}, 2126 {0x1a7,0x1a7,-1,1}, {0x1a9,0x1a9,-1,218}, {0x1ac,0x1ac,-1,1}, 2127 {0x1ae,0x1ae,-1,218}, {0x1af,0x1af,-1,1}, {0x1b1,0x1b2,1,217}, 2128 {0x1b3,0x1b5,2,1}, {0x1b7,0x1b7,-1,219}, {0x1b8,0x1bc,4,1}, 2129 {0x1c4,0x1ca,3,2}, {0x1cd,0x1db,2,1}, {0x1de,0x1ee,2,1}, 2130 {0x1f1,0x1f1,-1,2}, {0x1f4,0x1f4,-1,1}, {0x1f6,0x1f6,-1,-97}, 2131 {0x1f7,0x1f7,-1,-56}, {0x1f8,0x21e,2,1}, {0x220,0x220,-1,-130}, 2132 {0x222,0x232,2,1}, {0x386,0x386,-1,38}, {0x388,0x38a,1,37}, 2133 {0x38c,0x38c,-1,64}, {0x38e,0x38f,1,63}, {0x391,0x3a1,1,32}, 2134 {0x3a3,0x3ab,1,32}, {0x3d8,0x3ee,2,1}, {0x3f4,0x3f4,-1,-60}, 2135 {0x3f7,0x3f7,-1,1}, {0x3f9,0x3f9,-1,-7}, {0x3fa,0x3fa,-1,1}, 2136 {0x400,0x40f,1,80}, {0x410,0x42f,1,32}, {0x460,0x480,2,1}, 2137 {0x48a,0x4be,2,1}, {0x4c1,0x4cd,2,1}, {0x4d0,0x4f4,2,1}, 2138 {0x4f8,0x500,8,1}, {0x502,0x50e,2,1}, {0x531,0x556,1,48}, 2139 {0x1e00,0x1e94,2,1}, {0x1ea0,0x1ef8,2,1}, {0x1f08,0x1f0f,1,-8}, 2140 {0x1f18,0x1f1d,1,-8}, {0x1f28,0x1f2f,1,-8}, {0x1f38,0x1f3f,1,-8}, 2141 {0x1f48,0x1f4d,1,-8}, {0x1f59,0x1f5f,2,-8}, {0x1f68,0x1f6f,1,-8}, 2142 {0x1fb8,0x1fb9,1,-8}, {0x1fba,0x1fbb,1,-74}, {0x1fc8,0x1fcb,1,-86}, 2143 {0x1fd8,0x1fd9,1,-8}, {0x1fda,0x1fdb,1,-100}, {0x1fe8,0x1fe9,1,-8}, 2144 {0x1fea,0x1feb,1,-112}, {0x1fec,0x1fec,-1,-7}, {0x1ff8,0x1ff9,1,-128}, 2145 {0x1ffa,0x1ffb,1,-126}, {0x2126,0x2126,-1,-7517}, {0x212a,0x212a,-1,-8383}, 2146 {0x212b,0x212b,-1,-8262}, {0xff21,0xff3a,1,32}, {0x10400,0x10427,1,40} 2147 }; 2148 2149 static convertStruct toUpper[] = 2150 { 2151 {0x61,0x7a,1,-32}, {0xb5,0xb5,-1,743}, {0xe0,0xf6,1,-32}, 2152 {0xf8,0xfe,1,-32}, {0xff,0xff,-1,121}, {0x101,0x12f,2,-1}, 2153 {0x131,0x131,-1,-232}, {0x133,0x137,2,-1}, {0x13a,0x148,2,-1}, 2154 {0x14b,0x177,2,-1}, {0x17a,0x17e,2,-1}, {0x17f,0x17f,-1,-300}, 2155 {0x183,0x185,2,-1}, {0x188,0x18c,4,-1}, {0x192,0x192,-1,-1}, 2156 {0x195,0x195,-1,97}, {0x199,0x199,-1,-1}, {0x19e,0x19e,-1,130}, 2157 {0x1a1,0x1a5,2,-1}, {0x1a8,0x1ad,5,-1}, {0x1b0,0x1b4,4,-1}, 2158 {0x1b6,0x1b9,3,-1}, {0x1bd,0x1bd,-1,-1}, {0x1bf,0x1bf,-1,56}, 2159 {0x1c5,0x1c6,1,-1}, {0x1c8,0x1c9,1,-1}, {0x1cb,0x1cc,1,-1}, 2160 {0x1ce,0x1dc,2,-1}, {0x1dd,0x1dd,-1,-79}, {0x1df,0x1ef,2,-1}, 2161 {0x1f2,0x1f3,1,-1}, {0x1f5,0x1f9,4,-1}, {0x1fb,0x21f,2,-1}, 2162 {0x223,0x233,2,-1}, {0x253,0x253,-1,-210}, {0x254,0x254,-1,-206}, 2163 {0x256,0x257,1,-205}, {0x259,0x259,-1,-202}, {0x25b,0x25b,-1,-203}, 2164 {0x260,0x260,-1,-205}, {0x263,0x263,-1,-207}, {0x268,0x268,-1,-209}, 2165 {0x269,0x26f,6,-211}, {0x272,0x272,-1,-213}, {0x275,0x275,-1,-214}, 2166 {0x280,0x283,3,-218}, {0x288,0x288,-1,-218}, {0x28a,0x28b,1,-217}, 2167 {0x292,0x292,-1,-219}, {0x3ac,0x3ac,-1,-38}, {0x3ad,0x3af,1,-37}, 2168 {0x3b1,0x3c1,1,-32}, {0x3c2,0x3c2,-1,-31}, {0x3c3,0x3cb,1,-32}, 2169 {0x3cc,0x3cc,-1,-64}, {0x3cd,0x3ce,1,-63}, {0x3d0,0x3d0,-1,-62}, 2170 {0x3d1,0x3d1,-1,-57}, {0x3d5,0x3d5,-1,-47}, {0x3d6,0x3d6,-1,-54}, 2171 {0x3d9,0x3ef,2,-1}, {0x3f0,0x3f0,-1,-86}, {0x3f1,0x3f1,-1,-80}, 2172 {0x3f2,0x3f2,-1,7}, {0x3f5,0x3f5,-1,-96}, {0x3f8,0x3fb,3,-1}, 2173 {0x430,0x44f,1,-32}, {0x450,0x45f,1,-80}, {0x461,0x481,2,-1}, 2174 {0x48b,0x4bf,2,-1}, {0x4c2,0x4ce,2,-1}, {0x4d1,0x4f5,2,-1}, 2175 {0x4f9,0x501,8,-1}, {0x503,0x50f,2,-1}, {0x561,0x586,1,-48}, 2176 {0x1e01,0x1e95,2,-1}, {0x1e9b,0x1e9b,-1,-59}, {0x1ea1,0x1ef9,2,-1}, 2177 {0x1f00,0x1f07,1,8}, {0x1f10,0x1f15,1,8}, {0x1f20,0x1f27,1,8}, 2178 {0x1f30,0x1f37,1,8}, {0x1f40,0x1f45,1,8}, {0x1f51,0x1f57,2,8}, 2179 {0x1f60,0x1f67,1,8}, {0x1f70,0x1f71,1,74}, {0x1f72,0x1f75,1,86}, 2180 {0x1f76,0x1f77,1,100}, {0x1f78,0x1f79,1,128}, {0x1f7a,0x1f7b,1,112}, 2181 {0x1f7c,0x1f7d,1,126}, {0x1f80,0x1f87,1,8}, {0x1f90,0x1f97,1,8}, 2182 {0x1fa0,0x1fa7,1,8}, {0x1fb0,0x1fb1,1,8}, {0x1fb3,0x1fb3,-1,9}, 2183 {0x1fbe,0x1fbe,-1,-7205}, {0x1fc3,0x1fc3,-1,9}, {0x1fd0,0x1fd1,1,8}, 2184 {0x1fe0,0x1fe1,1,8}, {0x1fe5,0x1fe5,-1,7}, {0x1ff3,0x1ff3,-1,9}, 2185 {0xff41,0xff5a,1,-32}, {0x10428,0x1044f,1,-40} 2186 }; 2187 2188 /* 2189 * Return the upper-case equivalent of "a", which is a UCS-4 character. Use 2190 * simple case folding. 2191 */ 2192 int 2193 utf_toupper(a) 2194 int a; 2195 { 2196 /* If 'casemap' contains "keepascii" use ASCII style toupper(). */ 2197 if (a < 128 && (cmp_flags & CMP_KEEPASCII)) 2198 return TOUPPER_ASC(a); 2199 2200 #if defined(HAVE_TOWUPPER) && defined(__STDC__ISO_10646__) 2201 /* If towupper() is availble and handles Unicode, use it. */ 2202 if (!(cmp_flags & CMP_INTERNAL)) 2203 return towupper(a); 2204 #endif 2205 2206 /* For characters below 128 use locale sensitive toupper(). */ 2207 if (a < 128) 2208 return TOUPPER_LOC(a); 2209 2210 /* For any other characters use the above mapping table. */ 2211 return utf_convert(a, toUpper, sizeof(toUpper)); 2212 } 2213 2214 int 2215 utf_islower(a) 2216 int a; 2217 { 2218 return (utf_toupper(a) != a); 2219 } 2220 2221 /* 2222 * Return the lower-case equivalent of "a", which is a UCS-4 character. Use 2223 * simple case folding. 2224 */ 2225 int 2226 utf_tolower(a) 2227 int a; 2228 { 2229 /* If 'casemap' contains "keepascii" use ASCII style tolower(). */ 2230 if (a < 128 && (cmp_flags & CMP_KEEPASCII)) 2231 return TOLOWER_ASC(a); 2232 2233 #if defined(HAVE_TOWLOWER) && defined(__STDC__ISO_10646__) 2234 /* If towlower() is available and handles Unicode, use it. */ 2235 if (!(cmp_flags & CMP_INTERNAL)) 2236 return towlower(a); 2237 #endif 2238 2239 /* For characters below 128 use locale sensitive tolower(). */ 2240 if (a < 128) 2241 return TOLOWER_LOC(a); 2242 2243 /* For any other characters use the above mapping table. */ 2244 return utf_convert(a, toLower, sizeof(toLower)); 2245 } 2246 2247 int 2248 utf_isupper(a) 2249 int a; 2250 { 2251 return (utf_tolower(a) != a); 2252 } 2253 2254 /* 2255 * Version of strnicmp() that handles multi-byte characters. 2256 * Needed for Big5, Sjift-JIS and UTF-8 encoding. Other DBCS encodings can 2257 * probably use strnicmp(), because there are no ASCII characters in the 2258 * second byte. 2259 * Returns zero if s1 and s2 are equal (ignoring case), the difference between 2260 * two characters otherwise. 2261 */ 2262 int 2263 mb_strnicmp(s1, s2, nn) 2264 char_u *s1, *s2; 2265 size_t nn; 2266 { 2267 int i, j, l; 2268 int cdiff; 2269 int incomplete = FALSE; 2270 int n = nn; 2271 2272 for (i = 0; i < n; i += l) 2273 { 2274 if (s1[i] == NUL && s2[i] == NUL) /* both strings end */ 2275 return 0; 2276 if (enc_utf8) 2277 { 2278 l = utf_byte2len(s1[i]); 2279 if (l > n - i) 2280 { 2281 l = n - i; /* incomplete character */ 2282 incomplete = TRUE; 2283 } 2284 /* Check directly first, it's faster. */ 2285 for (j = 0; j < l; ++j) 2286 if (s1[i + j] != s2[i + j]) 2287 break; 2288 if (j < l) 2289 { 2290 /* If one of the two characters is incomplete return -1. */ 2291 if (incomplete || i + utf_byte2len(s2[i]) > n) 2292 return -1; 2293 cdiff = utf_fold(utf_ptr2char(s1 + i)) 2294 - utf_fold(utf_ptr2char(s2 + i)); 2295 if (cdiff != 0) 2296 return cdiff; 2297 } 2298 } 2299 else 2300 { 2301 l = (*mb_ptr2len)(s1 + i); 2302 if (l <= 1) 2303 { 2304 /* Single byte: first check normally, then with ignore case. */ 2305 if (s1[i] != s2[i]) 2306 { 2307 cdiff = TOLOWER_LOC(s1[i]) - TOLOWER_LOC(s2[i]); 2308 if (cdiff != 0) 2309 return cdiff; 2310 } 2311 } 2312 else 2313 { 2314 /* For non-Unicode multi-byte don't ignore case. */ 2315 if (l > n - i) 2316 l = n - i; 2317 cdiff = STRNCMP(s1 + i, s2 + i, l); 2318 if (cdiff != 0) 2319 return cdiff; 2320 } 2321 } 2322 } 2323 return 0; 2324 } 2325 2326 /* 2327 * "g8": show bytes of the UTF-8 char under the cursor. Doesn't matter what 2328 * 'encoding' has been set to. 2329 */ 2330 void 2331 show_utf8() 2332 { 2333 int len; 2334 int rlen = 0; 2335 char_u *line; 2336 int clen; 2337 int i; 2338 2339 /* Get the byte length of the char under the cursor, including composing 2340 * characters. */ 2341 line = ml_get_cursor(); 2342 len = utfc_ptr2len(line); 2343 if (len == 0) 2344 { 2345 MSG("NUL"); 2346 return; 2347 } 2348 2349 clen = 0; 2350 for (i = 0; i < len; ++i) 2351 { 2352 if (clen == 0) 2353 { 2354 /* start of (composing) character, get its length */ 2355 if (i > 0) 2356 { 2357 STRCPY(IObuff + rlen, "+ "); 2358 rlen += 2; 2359 } 2360 clen = utf_ptr2len(line + i); 2361 } 2362 sprintf((char *)IObuff + rlen, "%02x ", line[i]); 2363 --clen; 2364 rlen += STRLEN(IObuff + rlen); 2365 if (rlen > IOSIZE - 20) 2366 break; 2367 } 2368 2369 msg(IObuff); 2370 } 2371 2372 /* 2373 * mb_head_off() function pointer. 2374 * Return offset from "p" to the first byte of the character it points into. 2375 * Returns 0 when already at the first byte of a character. 2376 */ 2377 /*ARGSUSED*/ 2378 int 2379 latin_head_off(base, p) 2380 char_u *base; 2381 char_u *p; 2382 { 2383 return 0; 2384 } 2385 2386 int 2387 dbcs_head_off(base, p) 2388 char_u *base; 2389 char_u *p; 2390 { 2391 char_u *q; 2392 2393 /* It can't be a trailing byte when not using DBCS, at the start of the 2394 * string or the previous byte can't start a double-byte. */ 2395 if (p <= base || MB_BYTE2LEN(p[-1]) == 1) 2396 return 0; 2397 2398 /* This is slow: need to start at the base and go forward until the 2399 * byte we are looking for. Return 1 when we went past it, 0 otherwise. */ 2400 q = base; 2401 while (q < p) 2402 q += dbcs_ptr2len(q); 2403 return (q == p) ? 0 : 1; 2404 } 2405 2406 #if defined(FEAT_CLIPBOARD) || defined(FEAT_GUI) || defined(FEAT_RIGHTLEFT) \ 2407 || defined(PROTO) 2408 /* 2409 * Special version of dbcs_head_off() that works for ScreenLines[], where 2410 * single-width DBCS_JPNU characters are stored separately. 2411 */ 2412 int 2413 dbcs_screen_head_off(base, p) 2414 char_u *base; 2415 char_u *p; 2416 { 2417 char_u *q; 2418 2419 /* It can't be a trailing byte when not using DBCS, at the start of the 2420 * string or the previous byte can't start a double-byte. 2421 * For euc-jp an 0x8e byte in the previous cell always means we have a 2422 * lead byte in the current cell. */ 2423 if (p <= base 2424 || (enc_dbcs == DBCS_JPNU && p[-1] == 0x8e) 2425 || MB_BYTE2LEN(p[-1]) == 1) 2426 return 0; 2427 2428 /* This is slow: need to start at the base and go forward until the 2429 * byte we are looking for. Return 1 when we went past it, 0 otherwise. 2430 * For DBCS_JPNU look out for 0x8e, which means the second byte is not 2431 * stored as the next byte. */ 2432 q = base; 2433 while (q < p) 2434 { 2435 if (enc_dbcs == DBCS_JPNU && *q == 0x8e) 2436 ++q; 2437 else 2438 q += dbcs_ptr2len(q); 2439 } 2440 return (q == p) ? 0 : 1; 2441 } 2442 #endif 2443 2444 int 2445 utf_head_off(base, p) 2446 char_u *base; 2447 char_u *p; 2448 { 2449 char_u *q; 2450 char_u *s; 2451 int c; 2452 #ifdef FEAT_ARABIC 2453 char_u *j; 2454 #endif 2455 2456 if (*p < 0x80) /* be quick for ASCII */ 2457 return 0; 2458 2459 /* Skip backwards over trailing bytes: 10xx.xxxx 2460 * Skip backwards again if on a composing char. */ 2461 for (q = p; ; --q) 2462 { 2463 /* Move s to the last byte of this char. */ 2464 for (s = q; (s[1] & 0xc0) == 0x80; ++s) 2465 ; 2466 /* Move q to the first byte of this char. */ 2467 while (q > base && (*q & 0xc0) == 0x80) 2468 --q; 2469 /* Check for illegal sequence. Do allow an illegal byte after where we 2470 * started. */ 2471 if (utf8len_tab[*q] != (int)(s - q + 1) 2472 && utf8len_tab[*q] != (int)(p - q + 1)) 2473 return 0; 2474 2475 if (q <= base) 2476 break; 2477 2478 c = utf_ptr2char(q); 2479 if (utf_iscomposing(c)) 2480 continue; 2481 2482 #ifdef FEAT_ARABIC 2483 if (arabic_maycombine(c)) 2484 { 2485 /* Advance to get a sneak-peak at the next char */ 2486 j = q; 2487 --j; 2488 /* Move j to the first byte of this char. */ 2489 while (j > base && (*j & 0xc0) == 0x80) 2490 --j; 2491 if (arabic_combine(utf_ptr2char(j), c)) 2492 continue; 2493 } 2494 #endif 2495 break; 2496 } 2497 2498 return (int)(p - q); 2499 } 2500 2501 #if defined(FEAT_EVAL) || defined(PROTO) 2502 /* 2503 * Copy a character from "*fp" to "*tp" and advance the pointers. 2504 */ 2505 void 2506 mb_copy_char(fp, tp) 2507 char_u **fp; 2508 char_u **tp; 2509 { 2510 int l = (*mb_ptr2len)(*fp); 2511 2512 mch_memmove(*tp, *fp, (size_t)l); 2513 *tp += l; 2514 *fp += l; 2515 } 2516 #endif 2517 2518 /* 2519 * Return the offset from "p" to the first byte of a character. When "p" is 2520 * at the start of a character 0 is returned, otherwise the offset to the next 2521 * character. Can start anywhere in a stream of bytes. 2522 */ 2523 int 2524 mb_off_next(base, p) 2525 char_u *base; 2526 char_u *p; 2527 { 2528 int i; 2529 int j; 2530 2531 if (enc_utf8) 2532 { 2533 if (*p < 0x80) /* be quick for ASCII */ 2534 return 0; 2535 2536 /* Find the next character that isn't 10xx.xxxx */ 2537 for (i = 0; (p[i] & 0xc0) == 0x80; ++i) 2538 ; 2539 if (i > 0) 2540 { 2541 /* Check for illegal sequence. */ 2542 for (j = 0; p - j > base; ++j) 2543 if ((p[-j] & 0xc0) != 0x80) 2544 break; 2545 if (utf8len_tab[p[-j]] != i + j) 2546 return 0; 2547 } 2548 return i; 2549 } 2550 2551 /* Only need to check if we're on a trail byte, it doesn't matter if we 2552 * want the offset to the next or current character. */ 2553 return (*mb_head_off)(base, p); 2554 } 2555 2556 /* 2557 * Return the offset from "p" to the last byte of the character it points 2558 * into. Can start anywhere in a stream of bytes. 2559 */ 2560 int 2561 mb_tail_off(base, p) 2562 char_u *base; 2563 char_u *p; 2564 { 2565 int i; 2566 int j; 2567 2568 if (*p == NUL) 2569 return 0; 2570 2571 if (enc_utf8) 2572 { 2573 /* Find the last character that is 10xx.xxxx */ 2574 for (i = 0; (p[i + 1] & 0xc0) == 0x80; ++i) 2575 ; 2576 /* Check for illegal sequence. */ 2577 for (j = 0; p - j > base; ++j) 2578 if ((p[-j] & 0xc0) != 0x80) 2579 break; 2580 if (utf8len_tab[p[-j]] != i + j + 1) 2581 return 0; 2582 return i; 2583 } 2584 2585 /* It can't be the first byte if a double-byte when not using DBCS, at the 2586 * end of the string or the byte can't start a double-byte. */ 2587 if (enc_dbcs == 0 || p[1] == NUL || MB_BYTE2LEN(*p) == 1) 2588 return 0; 2589 2590 /* Return 1 when on the lead byte, 0 when on the tail byte. */ 2591 return 1 - dbcs_head_off(base, p); 2592 } 2593 2594 #if defined(HAVE_GTK2) || defined(PROTO) 2595 /* 2596 * Return TRUE if string "s" is a valid utf-8 string. 2597 * When "end" is NULL stop at the first NUL. 2598 * When "end" is positive stop there. 2599 */ 2600 int 2601 utf_valid_string(s, end) 2602 char_u *s; 2603 char_u *end; 2604 { 2605 int l; 2606 char_u *p = s; 2607 2608 while (end == NULL ? *p != NUL : p < end) 2609 { 2610 if ((*p & 0xc0) == 0x80) 2611 return FALSE; /* invalid lead byte */ 2612 l = utf8len_tab[*p]; 2613 if (end != NULL && p + l > end) 2614 return FALSE; /* incomplete byte sequence */ 2615 ++p; 2616 while (--l > 0) 2617 if ((*p++ & 0xc0) != 0x80) 2618 return FALSE; /* invalid trail byte */ 2619 } 2620 return TRUE; 2621 } 2622 #endif 2623 2624 #if defined(FEAT_GUI) || defined(PROTO) 2625 /* 2626 * Special version of mb_tail_off() for use in ScreenLines[]. 2627 */ 2628 int 2629 dbcs_screen_tail_off(base, p) 2630 char_u *base; 2631 char_u *p; 2632 { 2633 /* It can't be the first byte if a double-byte when not using DBCS, at the 2634 * end of the string or the byte can't start a double-byte. 2635 * For euc-jp an 0x8e byte always means we have a lead byte in the current 2636 * cell. */ 2637 if (*p == NUL || p[1] == NUL 2638 || (enc_dbcs == DBCS_JPNU && *p == 0x8e) 2639 || MB_BYTE2LEN(*p) == 1) 2640 return 0; 2641 2642 /* Return 1 when on the lead byte, 0 when on the tail byte. */ 2643 return 1 - dbcs_screen_head_off(base, p); 2644 } 2645 #endif 2646 2647 /* 2648 * If the cursor moves on an trail byte, set the cursor on the lead byte. 2649 * Thus it moves left if necessary. 2650 * Return TRUE when the cursor was adjusted. 2651 */ 2652 void 2653 mb_adjust_cursor() 2654 { 2655 mb_adjustpos(&curwin->w_cursor); 2656 } 2657 2658 /* 2659 * Adjust position "*lp" to point to the first byte of a multi-byte character. 2660 * If it points to a tail byte it's moved backwards to the head byte. 2661 */ 2662 void 2663 mb_adjustpos(lp) 2664 pos_T *lp; 2665 { 2666 char_u *p; 2667 2668 if (lp->col > 0 2669 #ifdef FEAT_VIRTUALEDIT 2670 || lp->coladd > 1 2671 #endif 2672 ) 2673 { 2674 p = ml_get(lp->lnum); 2675 lp->col -= (*mb_head_off)(p, p + lp->col); 2676 #ifdef FEAT_VIRTUALEDIT 2677 /* Reset "coladd" when the cursor would be on the right half of a 2678 * double-wide character. */ 2679 if (lp->coladd == 1 2680 && p[lp->col] != TAB 2681 && vim_isprintc((*mb_ptr2char)(p + lp->col)) 2682 && ptr2cells(p + lp->col) > 1) 2683 lp->coladd = 0; 2684 #endif 2685 } 2686 } 2687 2688 /* 2689 * Return a pointer to the character before "*p", if there is one. 2690 */ 2691 char_u * 2692 mb_prevptr(line, p) 2693 char_u *line; /* start of the string */ 2694 char_u *p; 2695 { 2696 if (p > line) 2697 mb_ptr_back(line, p); 2698 return p; 2699 } 2700 2701 /* 2702 * Return the character length of "str". Each multi-byte character (with 2703 * following composing characters) counts as one. 2704 */ 2705 int 2706 mb_charlen(str) 2707 char_u *str; 2708 { 2709 char_u *p = str; 2710 int count; 2711 2712 if (p == NULL) 2713 return 0; 2714 2715 for (count = 0; *p != NUL; count++) 2716 p += (*mb_ptr2len)(p); 2717 2718 return count; 2719 } 2720 2721 #if defined(FEAT_SYN_HL) || defined(PROTO) 2722 /* 2723 * Like mb_charlen() but for a string with specified length. 2724 */ 2725 int 2726 mb_charlen_len(str, len) 2727 char_u *str; 2728 int len; 2729 { 2730 char_u *p = str; 2731 int count; 2732 2733 for (count = 0; *p != NUL && p < str + len; count++) 2734 p += (*mb_ptr2len)(p); 2735 2736 return count; 2737 } 2738 #endif 2739 2740 /* 2741 * Try to un-escape a multi-byte character. 2742 * Used for the "to" and "from" part of a mapping. 2743 * Return the un-escaped string if it is a multi-byte character, and advance 2744 * "pp" to just after the bytes that formed it. 2745 * Return NULL if no multi-byte char was found. 2746 */ 2747 char_u * 2748 mb_unescape(pp) 2749 char_u **pp; 2750 { 2751 static char_u buf[MB_MAXBYTES + 1]; 2752 int n, m = 0; 2753 char_u *str = *pp; 2754 2755 /* Must translate K_SPECIAL KS_SPECIAL KE_FILLER to K_SPECIAL and CSI 2756 * KS_EXTRA KE_CSI to CSI. */ 2757 for (n = 0; str[n] != NUL && m <= MB_MAXBYTES; ++n) 2758 { 2759 if (str[n] == K_SPECIAL 2760 && str[n + 1] == KS_SPECIAL 2761 && str[n + 2] == KE_FILLER) 2762 { 2763 buf[m++] = K_SPECIAL; 2764 n += 2; 2765 } 2766 # ifdef FEAT_GUI 2767 else if (str[n] == CSI 2768 && str[n + 1] == KS_EXTRA 2769 && str[n + 2] == (int)KE_CSI) 2770 { 2771 buf[m++] = CSI; 2772 n += 2; 2773 } 2774 # endif 2775 else if (str[n] == K_SPECIAL 2776 # ifdef FEAT_GUI 2777 || str[n] == CSI 2778 # endif 2779 ) 2780 break; /* a special key can't be a multibyte char */ 2781 else 2782 buf[m++] = str[n]; 2783 buf[m] = NUL; 2784 2785 /* Return a multi-byte character if it's found. An illegal sequence 2786 * will result in a 1 here. */ 2787 if ((*mb_ptr2len)(buf) > 1) 2788 { 2789 *pp = str + n + 1; 2790 return buf; 2791 } 2792 } 2793 return NULL; 2794 } 2795 2796 /* 2797 * Return TRUE if the character at "row"/"col" on the screen is the left side 2798 * of a double-width character. 2799 * Caller must make sure "row" and "col" are not invalid! 2800 */ 2801 int 2802 mb_lefthalve(row, col) 2803 int row; 2804 int col; 2805 { 2806 #ifdef FEAT_HANGULIN 2807 if (composing_hangul) 2808 return TRUE; 2809 #endif 2810 if (enc_dbcs != 0) 2811 return dbcs_off2cells(LineOffset[row] + col) > 1; 2812 if (enc_utf8) 2813 return (col + 1 < Columns 2814 && ScreenLines[LineOffset[row] + col + 1] == 0); 2815 return FALSE; 2816 } 2817 2818 # if defined(FEAT_CLIPBOARD) || defined(FEAT_GUI) || defined(FEAT_RIGHTLEFT) \ 2819 || defined(PROTO) 2820 /* 2821 * Correct a position on the screen, if it's the right halve of a double-wide 2822 * char move it to the left halve. Returns the corrected column. 2823 */ 2824 int 2825 mb_fix_col(col, row) 2826 int col; 2827 int row; 2828 { 2829 col = check_col(col); 2830 row = check_row(row); 2831 if (has_mbyte && ScreenLines != NULL && col > 0 2832 && ((enc_dbcs 2833 && ScreenLines[LineOffset[row] + col] != NUL 2834 && dbcs_screen_head_off(ScreenLines + LineOffset[row], 2835 ScreenLines + LineOffset[row] + col)) 2836 || (enc_utf8 && ScreenLines[LineOffset[row] + col] == 0))) 2837 --col; 2838 return col; 2839 } 2840 # endif 2841 #endif 2842 2843 #if defined(FEAT_MBYTE) || defined(FEAT_POSTSCRIPT) || defined(PROTO) 2844 static int enc_alias_search __ARGS((char_u *name)); 2845 2846 /* 2847 * Skip the Vim specific head of a 'encoding' name. 2848 */ 2849 char_u * 2850 enc_skip(p) 2851 char_u *p; 2852 { 2853 if (STRNCMP(p, "2byte-", 6) == 0) 2854 return p + 6; 2855 if (STRNCMP(p, "8bit-", 5) == 0) 2856 return p + 5; 2857 return p; 2858 } 2859 2860 /* 2861 * Find the canonical name for encoding "enc". 2862 * When the name isn't recognized, returns "enc" itself, but with all lower 2863 * case characters and '_' replaced with '-'. 2864 * Returns an allocated string. NULL for out-of-memory. 2865 */ 2866 char_u * 2867 enc_canonize(enc) 2868 char_u *enc; 2869 { 2870 char_u *r; 2871 char_u *p, *s; 2872 int i; 2873 2874 # ifdef FEAT_MBYTE 2875 if (STRCMP(enc, "default") == 0) 2876 { 2877 /* Use the default encoding as it's found by set_init_1(). */ 2878 r = get_encoding_default(); 2879 if (r == NULL) 2880 r = (char_u *)"latin1"; 2881 return vim_strsave(r); 2882 } 2883 # endif 2884 2885 /* copy "enc" to allocted memory, with room for two '-' */ 2886 r = alloc((unsigned)(STRLEN(enc) + 3)); 2887 if (r != NULL) 2888 { 2889 /* Make it all lower case and replace '_' with '-'. */ 2890 p = r; 2891 for (s = enc; *s != NUL; ++s) 2892 { 2893 if (*s == '_') 2894 *p++ = '-'; 2895 else 2896 *p++ = TOLOWER_ASC(*s); 2897 } 2898 *p = NUL; 2899 2900 /* Skip "2byte-" and "8bit-". */ 2901 p = enc_skip(r); 2902 2903 /* Change "microsoft-cp" to "cp". Used in some spell files. */ 2904 if (STRNCMP(p, "microsoft-cp", 12) == 0) 2905 mch_memmove(p, p + 10, STRLEN(p + 10) + 1); 2906 2907 /* "iso8859" -> "iso-8859" */ 2908 if (STRNCMP(p, "iso8859", 7) == 0) 2909 { 2910 mch_memmove(p + 4, p + 3, STRLEN(p + 2)); 2911 p[3] = '-'; 2912 } 2913 2914 /* "iso-8859n" -> "iso-8859-n" */ 2915 if (STRNCMP(p, "iso-8859", 8) == 0 && p[8] != '-') 2916 { 2917 mch_memmove(p + 9, p + 8, STRLEN(p + 7)); 2918 p[8] = '-'; 2919 } 2920 2921 /* "latin-N" -> "latinN" */ 2922 if (STRNCMP(p, "latin-", 6) == 0) 2923 mch_memmove(p + 5, p + 6, STRLEN(p + 5)); 2924 2925 if (enc_canon_search(p) >= 0) 2926 { 2927 /* canonical name can be used unmodified */ 2928 if (p != r) 2929 mch_memmove(r, p, STRLEN(p) + 1); 2930 } 2931 else if ((i = enc_alias_search(p)) >= 0) 2932 { 2933 /* alias recognized, get canonical name */ 2934 vim_free(r); 2935 r = vim_strsave((char_u *)enc_canon_table[i].name); 2936 } 2937 } 2938 return r; 2939 } 2940 2941 /* 2942 * Search for an encoding alias of "name". 2943 * Returns -1 when not found. 2944 */ 2945 static int 2946 enc_alias_search(name) 2947 char_u *name; 2948 { 2949 int i; 2950 2951 for (i = 0; enc_alias_table[i].name != NULL; ++i) 2952 if (STRCMP(name, enc_alias_table[i].name) == 0) 2953 return enc_alias_table[i].canon; 2954 return -1; 2955 } 2956 #endif 2957 2958 #if defined(FEAT_MBYTE) || defined(PROTO) 2959 2960 #ifdef HAVE_LANGINFO_H 2961 # include <langinfo.h> 2962 #endif 2963 2964 /* 2965 * Get the canonicalized encoding of the current locale. 2966 * Returns an allocated string when successful, NULL when not. 2967 */ 2968 char_u * 2969 enc_locale() 2970 { 2971 #ifndef WIN3264 2972 char *s; 2973 char *p; 2974 int i; 2975 #endif 2976 char buf[50]; 2977 #ifdef WIN3264 2978 long acp = GetACP(); 2979 2980 if (acp == 1200) 2981 STRCPY(buf, "ucs-2le"); 2982 else if (acp == 1252) /* cp1252 is used as latin1 */ 2983 STRCPY(buf, "latin1"); 2984 else 2985 sprintf(buf, "cp%ld", acp); 2986 #else 2987 # ifdef HAVE_NL_LANGINFO_CODESET 2988 if ((s = nl_langinfo(CODESET)) == NULL || *s == NUL) 2989 # endif 2990 # if defined(HAVE_LOCALE_H) || defined(X_LOCALE) 2991 if ((s = setlocale(LC_CTYPE, NULL)) == NULL || *s == NUL) 2992 # endif 2993 if ((s = getenv("LC_ALL")) == NULL || *s == NUL) 2994 if ((s = getenv("LC_CTYPE")) == NULL || *s == NUL) 2995 s = getenv("LANG"); 2996 2997 if (s == NULL || *s == NUL) 2998 return FAIL; 2999 3000 /* The most generic locale format is: 3001 * language[_territory][.codeset][@modifier][+special][,[sponsor][_revision]] 3002 * If there is a '.' remove the part before it. 3003 * if there is something after the codeset, remove it. 3004 * Make the name lowercase and replace '_' with '-'. 3005 * Exception: "ja_JP.EUC" == "euc-jp", "zh_CN.EUC" = "euc-cn", 3006 * "ko_KR.EUC" == "euc-kr" 3007 */ 3008 if ((p = (char *)vim_strchr((char_u *)s, '.')) != NULL) 3009 { 3010 if (p > s + 2 && STRNICMP(p + 1, "EUC", 3) == 0 3011 && !isalnum((int)p[4]) && p[4] != '-' && p[-3] == '_') 3012 { 3013 /* copy "XY.EUC" to "euc-XY" to buf[10] */ 3014 STRCPY(buf + 10, "euc-"); 3015 buf[14] = p[-2]; 3016 buf[15] = p[-1]; 3017 buf[16] = 0; 3018 s = buf + 10; 3019 } 3020 else 3021 s = p + 1; 3022 } 3023 for (i = 0; s[i] != NUL && i < sizeof(buf) - 1; ++i) 3024 { 3025 if (s[i] == '_' || s[i] == '-') 3026 buf[i] = '-'; 3027 else if (isalnum((int)s[i])) 3028 buf[i] = TOLOWER_ASC(s[i]); 3029 else 3030 break; 3031 } 3032 buf[i] = NUL; 3033 #endif 3034 3035 return enc_canonize((char_u *)buf); 3036 } 3037 3038 #if defined(WIN3264) || defined(PROTO) 3039 /* 3040 * Convert an encoding name to an MS-Windows codepage. 3041 * Returns zero if no codepage can be figured out. 3042 */ 3043 int 3044 encname2codepage(name) 3045 char_u *name; 3046 { 3047 int cp; 3048 char_u *p = name; 3049 int idx; 3050 3051 if (STRNCMP(p, "8bit-", 5) == 0) 3052 p += 5; 3053 else if (STRNCMP(p_enc, "2byte-", 6) == 0) 3054 p += 6; 3055 3056 if (p[0] == 'c' && p[1] == 'p') 3057 cp = atoi(p + 2); 3058 else if ((idx = enc_canon_search(p)) >= 0) 3059 cp = enc_canon_table[idx].codepage; 3060 else 3061 return 0; 3062 if (IsValidCodePage(cp)) 3063 return cp; 3064 return 0; 3065 } 3066 #endif 3067 3068 # if defined(USE_ICONV) || defined(PROTO) 3069 3070 static char_u *iconv_string __ARGS((vimconv_T *vcp, char_u *str, int slen, int *unconvlenp)); 3071 3072 /* 3073 * Call iconv_open() with a check if iconv() works properly (there are broken 3074 * versions). 3075 * Returns (void *)-1 if failed. 3076 * (should return iconv_t, but that causes problems with prototypes). 3077 */ 3078 void * 3079 my_iconv_open(to, from) 3080 char_u *to; 3081 char_u *from; 3082 { 3083 iconv_t fd; 3084 #define ICONV_TESTLEN 400 3085 char_u tobuf[ICONV_TESTLEN]; 3086 char *p; 3087 size_t tolen; 3088 static int iconv_ok = -1; 3089 3090 if (iconv_ok == FALSE) 3091 return (void *)-1; /* detected a broken iconv() previously */ 3092 3093 #ifdef DYNAMIC_ICONV 3094 /* Check if the iconv.dll can be found. */ 3095 if (!iconv_enabled(TRUE)) 3096 return (void *)-1; 3097 #endif 3098 3099 fd = iconv_open((char *)enc_skip(to), (char *)enc_skip(from)); 3100 3101 if (fd != (iconv_t)-1 && iconv_ok == -1) 3102 { 3103 /* 3104 * Do a dummy iconv() call to check if it actually works. There is a 3105 * version of iconv() on Linux that is broken. We can't ignore it, 3106 * because it's wide-spread. The symptoms are that after outputting 3107 * the initial shift state the "to" pointer is NULL and conversion 3108 * stops for no apparent reason after about 8160 characters. 3109 */ 3110 p = (char *)tobuf; 3111 tolen = ICONV_TESTLEN; 3112 (void)iconv(fd, NULL, NULL, &p, &tolen); 3113 if (p == NULL) 3114 { 3115 iconv_ok = FALSE; 3116 iconv_close(fd); 3117 fd = (iconv_t)-1; 3118 } 3119 else 3120 iconv_ok = TRUE; 3121 } 3122 3123 return (void *)fd; 3124 } 3125 3126 /* 3127 * Convert the string "str[slen]" with iconv(). 3128 * If "unconvlenp" is not NULL handle the string ending in an incomplete 3129 * sequence and set "*unconvlenp" to the length of it. 3130 * Returns the converted string in allocated memory. NULL for an error. 3131 */ 3132 static char_u * 3133 iconv_string(vcp, str, slen, unconvlenp) 3134 vimconv_T *vcp; 3135 char_u *str; 3136 int slen; 3137 int *unconvlenp; 3138 { 3139 const char *from; 3140 size_t fromlen; 3141 char *to; 3142 size_t tolen; 3143 size_t len = 0; 3144 size_t done = 0; 3145 char_u *result = NULL; 3146 char_u *p; 3147 int l; 3148 3149 from = (char *)str; 3150 fromlen = slen; 3151 for (;;) 3152 { 3153 if (len == 0 || ICONV_ERRNO == ICONV_E2BIG) 3154 { 3155 /* Allocate enough room for most conversions. When re-allocating 3156 * increase the buffer size. */ 3157 len = len + fromlen * 2 + 40; 3158 p = alloc((unsigned)len); 3159 if (p != NULL && done > 0) 3160 mch_memmove(p, result, done); 3161 vim_free(result); 3162 result = p; 3163 if (result == NULL) /* out of memory */ 3164 break; 3165 } 3166 3167 to = (char *)result + done; 3168 tolen = len - done - 2; 3169 /* Avoid a warning for systems with a wrong iconv() prototype by 3170 * casting the second argument to void *. */ 3171 if (iconv(vcp->vc_fd, (void *)&from, &fromlen, &to, &tolen) 3172 != (size_t)-1) 3173 { 3174 /* Finished, append a NUL. */ 3175 *to = NUL; 3176 break; 3177 } 3178 3179 /* Check both ICONV_EINVAL and EINVAL, because the dynamically loaded 3180 * iconv library may use one of them. */ 3181 if (!vcp->vc_fail && unconvlenp != NULL 3182 && (ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL)) 3183 { 3184 /* Handle an incomplete sequence at the end. */ 3185 *to = NUL; 3186 *unconvlenp = fromlen; 3187 break; 3188 } 3189 3190 /* Check both ICONV_EILSEQ and EILSEQ, because the dynamically loaded 3191 * iconv library may use one of them. */ 3192 else if (!vcp->vc_fail 3193 && (ICONV_ERRNO == ICONV_EILSEQ || ICONV_ERRNO == EILSEQ 3194 || ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL)) 3195 { 3196 /* Can't convert: insert a '?' and skip a character. This assumes 3197 * conversion from 'encoding' to something else. In other 3198 * situations we don't know what to skip anyway. */ 3199 *to++ = '?'; 3200 if ((*mb_ptr2cells)((char_u *)from) > 1) 3201 *to++ = '?'; 3202 if (enc_utf8) 3203 l = utfc_ptr2len_len((char_u *)from, fromlen); 3204 else 3205 { 3206 l = (*mb_ptr2len)((char_u *)from); 3207 if (l > (int)fromlen) 3208 l = fromlen; 3209 } 3210 from += l; 3211 fromlen -= l; 3212 } 3213 else if (ICONV_ERRNO != ICONV_E2BIG) 3214 { 3215 /* conversion failed */ 3216 vim_free(result); 3217 result = NULL; 3218 break; 3219 } 3220 /* Not enough room or skipping illegal sequence. */ 3221 done = to - (char *)result; 3222 } 3223 return result; 3224 } 3225 3226 # if defined(DYNAMIC_ICONV) || defined(PROTO) 3227 /* 3228 * Dynamically load the "iconv.dll" on Win32. 3229 */ 3230 3231 #ifndef DYNAMIC_ICONV /* just generating prototypes */ 3232 # define HINSTANCE int 3233 #endif 3234 static HINSTANCE hIconvDLL = 0; 3235 static HINSTANCE hMsvcrtDLL = 0; 3236 3237 # ifndef DYNAMIC_ICONV_DLL 3238 # define DYNAMIC_ICONV_DLL "iconv.dll" 3239 # define DYNAMIC_ICONV_DLL_ALT "libiconv.dll" 3240 # endif 3241 # ifndef DYNAMIC_MSVCRT_DLL 3242 # define DYNAMIC_MSVCRT_DLL "msvcrt.dll" 3243 # endif 3244 3245 /* 3246 * Try opening the iconv.dll and return TRUE if iconv() can be used. 3247 */ 3248 int 3249 iconv_enabled(verbose) 3250 int verbose; 3251 { 3252 if (hIconvDLL != 0 && hMsvcrtDLL != 0) 3253 return TRUE; 3254 hIconvDLL = LoadLibrary(DYNAMIC_ICONV_DLL); 3255 if (hIconvDLL == 0) /* sometimes it's called libiconv.dll */ 3256 hIconvDLL = LoadLibrary(DYNAMIC_ICONV_DLL_ALT); 3257 if (hIconvDLL != 0) 3258 hMsvcrtDLL = LoadLibrary(DYNAMIC_MSVCRT_DLL); 3259 if (hIconvDLL == 0 || hMsvcrtDLL == 0) 3260 { 3261 /* Only give the message when 'verbose' is set, otherwise it might be 3262 * done whenever a conversion is attempted. */ 3263 if (verbose && p_verbose > 0) 3264 { 3265 verbose_enter(); 3266 EMSG2(_(e_loadlib), 3267 hIconvDLL == 0 ? DYNAMIC_ICONV_DLL : DYNAMIC_MSVCRT_DLL); 3268 verbose_leave(); 3269 } 3270 iconv_end(); 3271 return FALSE; 3272 } 3273 3274 iconv = (void *)GetProcAddress(hIconvDLL, "libiconv"); 3275 iconv_open = (void *)GetProcAddress(hIconvDLL, "libiconv_open"); 3276 iconv_close = (void *)GetProcAddress(hIconvDLL, "libiconv_close"); 3277 iconvctl = (void *)GetProcAddress(hIconvDLL, "libiconvctl"); 3278 iconv_errno = (void *)GetProcAddress(hMsvcrtDLL, "_errno"); 3279 if (iconv == NULL || iconv_open == NULL || iconv_close == NULL 3280 || iconvctl == NULL || iconv_errno == NULL) 3281 { 3282 iconv_end(); 3283 if (verbose && p_verbose > 0) 3284 { 3285 verbose_enter(); 3286 EMSG2(_(e_loadfunc), "for libiconv"); 3287 verbose_leave(); 3288 } 3289 return FALSE; 3290 } 3291 return TRUE; 3292 } 3293 3294 void 3295 iconv_end() 3296 { 3297 /* Don't use iconv() when inputting or outputting characters. */ 3298 if (input_conv.vc_type == CONV_ICONV) 3299 convert_setup(&input_conv, NULL, NULL); 3300 if (output_conv.vc_type == CONV_ICONV) 3301 convert_setup(&output_conv, NULL, NULL); 3302 3303 if (hIconvDLL != 0) 3304 FreeLibrary(hIconvDLL); 3305 if (hMsvcrtDLL != 0) 3306 FreeLibrary(hMsvcrtDLL); 3307 hIconvDLL = 0; 3308 hMsvcrtDLL = 0; 3309 } 3310 # endif /* DYNAMIC_ICONV */ 3311 # endif /* USE_ICONV */ 3312 3313 #endif /* FEAT_MBYTE */ 3314 3315 #if defined(FEAT_XIM) || defined(PROTO) 3316 3317 # ifdef FEAT_GUI_GTK 3318 static int xim_has_preediting INIT(= FALSE); /* IM current status */ 3319 3320 /* 3321 * Set preedit_start_col to the current cursor position. 3322 */ 3323 static void 3324 init_preedit_start_col(void) 3325 { 3326 if (State & CMDLINE) 3327 preedit_start_col = cmdline_getvcol_cursor(); 3328 else if (curwin != NULL) 3329 getvcol(curwin, &curwin->w_cursor, &preedit_start_col, NULL, NULL); 3330 /* Prevent that preediting marks the buffer as changed. */ 3331 xim_changed_while_preediting = curbuf->b_changed; 3332 } 3333 # endif 3334 3335 # if defined(HAVE_GTK2) && !defined(PROTO) 3336 3337 static int im_is_active = FALSE; /* IM is enabled for current mode */ 3338 static int im_preedit_cursor = 0; /* cursor offset in characters */ 3339 static int im_preedit_trailing = 0; /* number of characters after cursor */ 3340 3341 static unsigned long im_commit_handler_id = 0; 3342 static unsigned int im_activatekey_keyval = GDK_VoidSymbol; 3343 static unsigned int im_activatekey_state = 0; 3344 3345 void 3346 im_set_active(int active) 3347 { 3348 int was_active; 3349 3350 was_active = !!im_is_active; 3351 im_is_active = (active && !p_imdisable); 3352 3353 if (im_is_active != was_active) 3354 xim_reset(); 3355 } 3356 3357 void 3358 xim_set_focus(int focus) 3359 { 3360 if (xic != NULL) 3361 { 3362 if (focus) 3363 gtk_im_context_focus_in(xic); 3364 else 3365 gtk_im_context_focus_out(xic); 3366 } 3367 } 3368 3369 void 3370 im_set_position(int row, int col) 3371 { 3372 if (xic != NULL) 3373 { 3374 GdkRectangle area; 3375 3376 area.x = FILL_X(col); 3377 area.y = FILL_Y(row); 3378 area.width = gui.char_width * (mb_lefthalve(row, col) ? 2 : 1); 3379 area.height = gui.char_height; 3380 3381 gtk_im_context_set_cursor_location(xic, &area); 3382 } 3383 } 3384 3385 # if 0 || defined(PROTO) /* apparently only used in gui_x11.c */ 3386 void 3387 xim_set_preedit(void) 3388 { 3389 im_set_position(gui.row, gui.col); 3390 } 3391 # endif 3392 3393 static void 3394 im_add_to_input(char_u *str, int len) 3395 { 3396 /* Convert from 'termencoding' (always "utf-8") to 'encoding' */ 3397 if (input_conv.vc_type != CONV_NONE) 3398 { 3399 str = string_convert(&input_conv, str, &len); 3400 g_return_if_fail(str != NULL); 3401 } 3402 3403 add_to_input_buf_csi(str, len); 3404 3405 if (input_conv.vc_type != CONV_NONE) 3406 vim_free(str); 3407 3408 if (p_mh) /* blank out the pointer if necessary */ 3409 gui_mch_mousehide(TRUE); 3410 } 3411 3412 static void 3413 im_delete_preedit(void) 3414 { 3415 char_u bskey[] = {CSI, 'k', 'b'}; 3416 char_u delkey[] = {CSI, 'k', 'D'}; 3417 3418 if (State & NORMAL) 3419 { 3420 im_preedit_cursor = 0; 3421 return; 3422 } 3423 for (; im_preedit_cursor > 0; --im_preedit_cursor) 3424 add_to_input_buf(bskey, (int)sizeof(bskey)); 3425 3426 for (; im_preedit_trailing > 0; --im_preedit_trailing) 3427 add_to_input_buf(delkey, (int)sizeof(delkey)); 3428 } 3429 3430 static void 3431 im_correct_cursor(int num_move_back) 3432 { 3433 char_u backkey[] = {CSI, 'k', 'l'}; 3434 3435 if (State & NORMAL) 3436 return; 3437 # ifdef FEAT_RIGHTLEFT 3438 if ((State & CMDLINE) == 0 && curwin != NULL && curwin->w_p_rl) 3439 backkey[2] = 'r'; 3440 # endif 3441 for (; num_move_back > 0; --num_move_back) 3442 add_to_input_buf(backkey, (int)sizeof(backkey)); 3443 } 3444 3445 static int xim_expected_char = NUL; 3446 static int xim_ignored_char = FALSE; 3447 3448 /* 3449 * Update the mode and cursor while in an IM callback. 3450 */ 3451 static void 3452 im_show_info(void) 3453 { 3454 int old_vgetc_busy; 3455 old_vgetc_busy = vgetc_busy; 3456 vgetc_busy = TRUE; 3457 showmode(); 3458 vgetc_busy = old_vgetc_busy; 3459 setcursor(); 3460 out_flush(); 3461 } 3462 3463 /* 3464 * Callback invoked when the user finished preediting. 3465 * Put the final string into the input buffer. 3466 */ 3467 /*ARGSUSED0*/ 3468 static void 3469 im_commit_cb(GtkIMContext *context, const gchar *str, gpointer data) 3470 { 3471 int slen = (int)STRLEN(str); 3472 int add_to_input = TRUE; 3473 int clen; 3474 int len = slen; 3475 int commit_with_preedit = TRUE; 3476 char_u *im_str, *p; 3477 3478 #ifdef XIM_DEBUG 3479 xim_log("im_commit_cb(): %s\n", str); 3480 #endif 3481 3482 /* The imhangul module doesn't reset the preedit string before 3483 * committing. Call im_delete_preedit() to work around that. */ 3484 im_delete_preedit(); 3485 3486 /* Indicate that preediting has finished. */ 3487 if (preedit_start_col == MAXCOL) 3488 { 3489 init_preedit_start_col(); 3490 commit_with_preedit = FALSE; 3491 } 3492 3493 /* The thing which setting "preedit_start_col" to MAXCOL means that 3494 * "preedit_start_col" will be set forcely when calling 3495 * preedit_changed_cb() next time. 3496 * "preedit_start_col" should not reset with MAXCOL on this part. Vim 3497 * is simulating the preediting by using add_to_input_str(). when 3498 * preedit begin immediately before committed, the typebuf is not 3499 * flushed to screen, then it can't get correct "preedit_start_col". 3500 * Thus, it should calculate the cells by adding cells of the committed 3501 * string. */ 3502 if (input_conv.vc_type != CONV_NONE) 3503 { 3504 im_str = string_convert(&input_conv, (char_u *)str, &len); 3505 g_return_if_fail(im_str != NULL); 3506 } 3507 else 3508 im_str = (char_u *)str; 3509 clen = 0; 3510 for (p = im_str; p < im_str + len; p += (*mb_ptr2len)(p)) 3511 clen += (*mb_ptr2cells)(p); 3512 if (input_conv.vc_type != CONV_NONE) 3513 vim_free(im_str); 3514 preedit_start_col += clen; 3515 3516 /* Is this a single character that matches a keypad key that's just 3517 * been pressed? If so, we don't want it to be entered as such - let 3518 * us carry on processing the raw keycode so that it may be used in 3519 * mappings as <kSomething>. */ 3520 if (xim_expected_char != NUL) 3521 { 3522 /* We're currently processing a keypad or other special key */ 3523 if (slen == 1 && str[0] == xim_expected_char) 3524 { 3525 /* It's a match - don't do it here */ 3526 xim_ignored_char = TRUE; 3527 add_to_input = FALSE; 3528 } 3529 else 3530 { 3531 /* Not a match */ 3532 xim_ignored_char = FALSE; 3533 } 3534 } 3535 3536 if (add_to_input) 3537 im_add_to_input((char_u *)str, slen); 3538 3539 /* Inserting chars while "im_is_active" is set does not cause a change of 3540 * buffer. When the chars are committed the buffer must be marked as 3541 * changed. */ 3542 if (!commit_with_preedit) 3543 preedit_start_col = MAXCOL; 3544 3545 /* This flag is used in changed() at next call. */ 3546 xim_changed_while_preediting = TRUE; 3547 3548 if (gtk_main_level() > 0) 3549 gtk_main_quit(); 3550 } 3551 3552 /* 3553 * Callback invoked after start to the preedit. 3554 */ 3555 /*ARGSUSED*/ 3556 static void 3557 im_preedit_start_cb(GtkIMContext *context, gpointer data) 3558 { 3559 #ifdef XIM_DEBUG 3560 xim_log("im_preedit_start_cb()\n"); 3561 #endif 3562 3563 im_is_active = TRUE; 3564 gui_update_cursor(TRUE, FALSE); 3565 } 3566 3567 /* 3568 * Callback invoked after end to the preedit. 3569 */ 3570 /*ARGSUSED*/ 3571 static void 3572 im_preedit_end_cb(GtkIMContext *context, gpointer data) 3573 { 3574 #ifdef XIM_DEBUG 3575 xim_log("im_preedit_end_cb()\n"); 3576 #endif 3577 im_delete_preedit(); 3578 3579 /* Indicate that preediting has finished */ 3580 preedit_start_col = MAXCOL; 3581 xim_has_preediting = FALSE; 3582 3583 im_is_active = FALSE; 3584 gui_update_cursor(TRUE, FALSE); 3585 im_show_info(); 3586 } 3587 3588 /* 3589 * Callback invoked after changes to the preedit string. If the preedit 3590 * string was empty before, remember the preedit start column so we know 3591 * where to apply feedback attributes. Delete the previous preedit string 3592 * if there was one, save the new preedit cursor offset, and put the new 3593 * string into the input buffer. 3594 * 3595 * TODO: The pragmatic "put into input buffer" approach used here has 3596 * several fundamental problems: 3597 * 3598 * - The characters in the preedit string are subject to remapping. 3599 * That's broken, only the finally committed string should be remapped. 3600 * 3601 * - There is a race condition involved: The retrieved value for the 3602 * current cursor position will be wrong if any unprocessed characters 3603 * are still queued in the input buffer. 3604 * 3605 * - Due to the lack of synchronization between the file buffer in memory 3606 * and any typed characters, it's practically impossible to implement the 3607 * "retrieve_surrounding" and "delete_surrounding" signals reliably. IM 3608 * modules for languages such as Thai are likely to rely on this feature 3609 * for proper operation. 3610 * 3611 * Conclusions: I think support for preediting needs to be moved to the 3612 * core parts of Vim. Ideally, until it has been committed, the preediting 3613 * string should only be displayed and not affect the buffer content at all. 3614 * The question how to deal with the synchronization issue still remains. 3615 * Circumventing the input buffer is probably not desirable. Anyway, I think 3616 * implementing "retrieve_surrounding" is the only hard problem. 3617 * 3618 * One way to solve all of this in a clean manner would be to queue all key 3619 * press/release events "as is" in the input buffer, and apply the IM filtering 3620 * at the receiving end of the queue. This, however, would have a rather large 3621 * impact on the code base. If there is an easy way to force processing of all 3622 * remaining input from within the "retrieve_surrounding" signal handler, this 3623 * might not be necessary. Gotta ask on vim-dev for opinions. 3624 */ 3625 /*ARGSUSED1*/ 3626 static void 3627 im_preedit_changed_cb(GtkIMContext *context, gpointer data) 3628 { 3629 char *preedit_string = NULL; 3630 int cursor_index = 0; 3631 int num_move_back = 0; 3632 char_u *str; 3633 char_u *p; 3634 int i; 3635 3636 gtk_im_context_get_preedit_string(context, 3637 &preedit_string, NULL, 3638 &cursor_index); 3639 3640 #ifdef XIM_DEBUG 3641 xim_log("im_preedit_changed_cb(): %s\n", preedit_string); 3642 #endif 3643 3644 g_return_if_fail(preedit_string != NULL); /* just in case */ 3645 3646 /* If preedit_start_col is MAXCOL set it to the current cursor position. */ 3647 if (preedit_start_col == MAXCOL && preedit_string[0] != '\0') 3648 { 3649 xim_has_preediting = TRUE; 3650 3651 /* Urgh, this breaks if the input buffer isn't empty now */ 3652 init_preedit_start_col(); 3653 } 3654 else if (cursor_index == 0 && preedit_string[0] == '\0') 3655 { 3656 if (preedit_start_col == MAXCOL) 3657 xim_has_preediting = FALSE; 3658 3659 /* If at the start position (after typing backspace) 3660 * preedit_start_col must be reset. */ 3661 preedit_start_col = MAXCOL; 3662 } 3663 3664 im_delete_preedit(); 3665 3666 /* 3667 * Compute the end of the preediting area: "preedit_end_col". 3668 * According to the documentation of gtk_im_context_get_preedit_string(), 3669 * the cursor_pos output argument returns the offset in bytes. This is 3670 * unfortunately not true -- real life shows the offset is in characters, 3671 * and the GTK+ source code agrees with me. Will file a bug later. 3672 */ 3673 if (preedit_start_col != MAXCOL) 3674 preedit_end_col = preedit_start_col; 3675 str = (char_u *)preedit_string; 3676 for (p = str, i = 0; *p != NUL; p += utf_byte2len(*p), ++i) 3677 { 3678 int is_composing; 3679 3680 is_composing = ((*p & 0x80) != 0 && utf_iscomposing(utf_ptr2char(p))); 3681 /* 3682 * These offsets are used as counters when generating <BS> and <Del> 3683 * to delete the preedit string. So don't count composing characters 3684 * unless 'delcombine' is enabled. 3685 */ 3686 if (!is_composing || p_deco) 3687 { 3688 if (i < cursor_index) 3689 ++im_preedit_cursor; 3690 else 3691 ++im_preedit_trailing; 3692 } 3693 if (!is_composing && i >= cursor_index) 3694 { 3695 /* This is essentially the same as im_preedit_trailing, except 3696 * composing characters are not counted even if p_deco is set. */ 3697 ++num_move_back; 3698 } 3699 if (preedit_start_col != MAXCOL) 3700 preedit_end_col += utf_ptr2cells(p); 3701 } 3702 3703 if (p > str) 3704 { 3705 im_add_to_input(str, (int)(p - str)); 3706 im_correct_cursor(num_move_back); 3707 } 3708 3709 g_free(preedit_string); 3710 3711 if (gtk_main_level() > 0) 3712 gtk_main_quit(); 3713 } 3714 3715 /* 3716 * Translate the Pango attributes at iter to Vim highlighting attributes. 3717 * Ignore attributes not supported by Vim highlighting. This shouldn't have 3718 * too much impact -- right now we handle even more attributes than necessary 3719 * for the IM modules I tested with. 3720 */ 3721 static int 3722 translate_pango_attributes(PangoAttrIterator *iter) 3723 { 3724 PangoAttribute *attr; 3725 int char_attr = HL_NORMAL; 3726 3727 attr = pango_attr_iterator_get(iter, PANGO_ATTR_UNDERLINE); 3728 if (attr != NULL && ((PangoAttrInt *)attr)->value 3729 != (int)PANGO_UNDERLINE_NONE) 3730 char_attr |= HL_UNDERLINE; 3731 3732 attr = pango_attr_iterator_get(iter, PANGO_ATTR_WEIGHT); 3733 if (attr != NULL && ((PangoAttrInt *)attr)->value >= (int)PANGO_WEIGHT_BOLD) 3734 char_attr |= HL_BOLD; 3735 3736 attr = pango_attr_iterator_get(iter, PANGO_ATTR_STYLE); 3737 if (attr != NULL && ((PangoAttrInt *)attr)->value 3738 != (int)PANGO_STYLE_NORMAL) 3739 char_attr |= HL_ITALIC; 3740 3741 attr = pango_attr_iterator_get(iter, PANGO_ATTR_BACKGROUND); 3742 if (attr != NULL) 3743 { 3744 const PangoColor *color = &((PangoAttrColor *)attr)->color; 3745 3746 /* Assume inverse if black background is requested */ 3747 if ((color->red | color->green | color->blue) == 0) 3748 char_attr |= HL_INVERSE; 3749 } 3750 3751 return char_attr; 3752 } 3753 3754 /* 3755 * Retrieve the highlighting attributes at column col in the preedit string. 3756 * Return -1 if not in preediting mode or if col is out of range. 3757 */ 3758 int 3759 im_get_feedback_attr(int col) 3760 { 3761 char *preedit_string = NULL; 3762 PangoAttrList *attr_list = NULL; 3763 int char_attr = -1; 3764 3765 if (xic == NULL) 3766 return char_attr; 3767 3768 gtk_im_context_get_preedit_string(xic, &preedit_string, &attr_list, NULL); 3769 3770 if (preedit_string != NULL && attr_list != NULL) 3771 { 3772 int index; 3773 3774 /* Get the byte index as used by PangoAttrIterator */ 3775 for (index = 0; col > 0 && preedit_string[index] != '\0'; --col) 3776 index += utfc_ptr2len((char_u *)preedit_string + index); 3777 3778 if (preedit_string[index] != '\0') 3779 { 3780 PangoAttrIterator *iter; 3781 int start, end; 3782 3783 char_attr = HL_NORMAL; 3784 iter = pango_attr_list_get_iterator(attr_list); 3785 3786 /* Extract all relevant attributes from the list. */ 3787 do 3788 { 3789 pango_attr_iterator_range(iter, &start, &end); 3790 3791 if (index >= start && index < end) 3792 char_attr |= translate_pango_attributes(iter); 3793 } 3794 while (pango_attr_iterator_next(iter)); 3795 3796 pango_attr_iterator_destroy(iter); 3797 } 3798 } 3799 3800 if (attr_list != NULL) 3801 pango_attr_list_unref(attr_list); 3802 g_free(preedit_string); 3803 3804 return char_attr; 3805 } 3806 3807 void 3808 xim_init(void) 3809 { 3810 #ifdef XIM_DEBUG 3811 xim_log("xim_init()\n"); 3812 #endif 3813 3814 g_return_if_fail(gui.drawarea != NULL); 3815 g_return_if_fail(gui.drawarea->window != NULL); 3816 3817 xic = gtk_im_multicontext_new(); 3818 g_object_ref(xic); 3819 3820 im_commit_handler_id = g_signal_connect(G_OBJECT(xic), "commit", 3821 G_CALLBACK(&im_commit_cb), NULL); 3822 g_signal_connect(G_OBJECT(xic), "preedit_changed", 3823 G_CALLBACK(&im_preedit_changed_cb), NULL); 3824 g_signal_connect(G_OBJECT(xic), "preedit_start", 3825 G_CALLBACK(&im_preedit_start_cb), NULL); 3826 g_signal_connect(G_OBJECT(xic), "preedit_end", 3827 G_CALLBACK(&im_preedit_end_cb), NULL); 3828 3829 gtk_im_context_set_client_window(xic, gui.drawarea->window); 3830 } 3831 3832 void 3833 im_shutdown(void) 3834 { 3835 #ifdef XIM_DEBUG 3836 xim_log("im_shutdown()\n"); 3837 #endif 3838 3839 if (xic != NULL) 3840 { 3841 gtk_im_context_focus_out(xic); 3842 g_object_unref(xic); 3843 xic = NULL; 3844 } 3845 im_is_active = FALSE; 3846 im_commit_handler_id = 0; 3847 preedit_start_col = MAXCOL; 3848 xim_has_preediting = FALSE; 3849 } 3850 3851 /* 3852 * Convert the string argument to keyval and state for GdkEventKey. 3853 * If str is valid return TRUE, otherwise FALSE. 3854 * 3855 * See 'imactivatekey' for documentation of the format. 3856 */ 3857 static int 3858 im_string_to_keyval(const char *str, unsigned int *keyval, unsigned int *state) 3859 { 3860 const char *mods_end; 3861 unsigned tmp_keyval; 3862 unsigned tmp_state = 0; 3863 3864 mods_end = strrchr(str, '-'); 3865 mods_end = (mods_end != NULL) ? mods_end + 1 : str; 3866 3867 /* Parse modifier keys */ 3868 while (str < mods_end) 3869 switch (*str++) 3870 { 3871 case '-': break; 3872 case 'S': case 's': tmp_state |= (unsigned)GDK_SHIFT_MASK; break; 3873 case 'L': case 'l': tmp_state |= (unsigned)GDK_LOCK_MASK; break; 3874 case 'C': case 'c': tmp_state |= (unsigned)GDK_CONTROL_MASK;break; 3875 case '1': tmp_state |= (unsigned)GDK_MOD1_MASK; break; 3876 case '2': tmp_state |= (unsigned)GDK_MOD2_MASK; break; 3877 case '3': tmp_state |= (unsigned)GDK_MOD3_MASK; break; 3878 case '4': tmp_state |= (unsigned)GDK_MOD4_MASK; break; 3879 case '5': tmp_state |= (unsigned)GDK_MOD5_MASK; break; 3880 default: 3881 return FALSE; 3882 } 3883 3884 tmp_keyval = gdk_keyval_from_name(str); 3885 3886 if (tmp_keyval == 0 || tmp_keyval == GDK_VoidSymbol) 3887 return FALSE; 3888 3889 if (keyval != NULL) 3890 *keyval = tmp_keyval; 3891 if (state != NULL) 3892 *state = tmp_state; 3893 3894 return TRUE; 3895 } 3896 3897 /* 3898 * Return TRUE if p_imak is valid, otherwise FALSE. As a special case, an 3899 * empty string is also regarded as valid. 3900 * 3901 * Note: The numerical key value of p_imak is cached if it was valid; thus 3902 * boldly assuming im_xim_isvalid_imactivate() will always be called whenever 3903 * 'imak' changes. This is currently the case but not obvious -- should 3904 * probably rename the function for clarity. 3905 */ 3906 int 3907 im_xim_isvalid_imactivate(void) 3908 { 3909 if (p_imak[0] == NUL) 3910 { 3911 im_activatekey_keyval = GDK_VoidSymbol; 3912 im_activatekey_state = 0; 3913 return TRUE; 3914 } 3915 3916 return im_string_to_keyval((const char *)p_imak, 3917 &im_activatekey_keyval, 3918 &im_activatekey_state); 3919 } 3920 3921 static void 3922 im_synthesize_keypress(unsigned int keyval, unsigned int state) 3923 { 3924 GdkEventKey *event; 3925 3926 # ifdef HAVE_GTK_MULTIHEAD 3927 event = (GdkEventKey *)gdk_event_new(GDK_KEY_PRESS); 3928 g_object_ref(gui.drawarea->window); /* unreffed by gdk_event_free() */ 3929 # else 3930 event = (GdkEventKey *)g_malloc0((gulong)sizeof(GdkEvent)); 3931 event->type = GDK_KEY_PRESS; 3932 # endif 3933 event->window = gui.drawarea->window; 3934 event->send_event = TRUE; 3935 event->time = GDK_CURRENT_TIME; 3936 event->state = state; 3937 event->keyval = keyval; 3938 event->hardware_keycode = /* needed for XIM */ 3939 XKeysymToKeycode(GDK_WINDOW_XDISPLAY(event->window), (KeySym)keyval); 3940 event->length = 0; 3941 event->string = NULL; 3942 3943 gtk_im_context_filter_keypress(xic, event); 3944 3945 /* For consistency, also send the corresponding release event. */ 3946 event->type = GDK_KEY_RELEASE; 3947 event->send_event = FALSE; 3948 gtk_im_context_filter_keypress(xic, event); 3949 3950 # ifdef HAVE_GTK_MULTIHEAD 3951 gdk_event_free((GdkEvent *)event); 3952 # else 3953 g_free(event); 3954 # endif 3955 } 3956 3957 void 3958 xim_reset(void) 3959 { 3960 if (xic != NULL) 3961 { 3962 /* 3963 * The third-party imhangul module (and maybe others too) ignores 3964 * gtk_im_context_reset() or at least doesn't reset the active state. 3965 * Thus sending imactivatekey would turn it off if it was on before, 3966 * which is clearly not what we want. Fortunately we can work around 3967 * that for imhangul by sending GDK_Escape, but I don't know if it 3968 * works with all IM modules that support an activation key :/ 3969 * 3970 * An alternative approach would be to destroy the IM context and 3971 * recreate it. But that means loading/unloading the IM module on 3972 * every mode switch, which causes a quite noticable delay even on 3973 * my rather fast box... 3974 * * 3975 * Moreover, there are some XIM which cannot respond to 3976 * im_synthesize_keypress(). we hope that they reset by 3977 * xim_shutdown(). 3978 */ 3979 if (im_activatekey_keyval != GDK_VoidSymbol && im_is_active) 3980 im_synthesize_keypress(GDK_Escape, 0U); 3981 3982 gtk_im_context_reset(xic); 3983 3984 /* 3985 * HACK for Ami: This sequence of function calls makes Ami handle 3986 * the IM reset gratiously, without breaking loads of other stuff. 3987 * It seems to force English mode as well, which is exactly what we 3988 * want because it makes the Ami status display work reliably. 3989 */ 3990 gtk_im_context_set_use_preedit(xic, FALSE); 3991 3992 if (p_imdisable) 3993 im_shutdown(); 3994 else 3995 { 3996 gtk_im_context_set_use_preedit(xic, TRUE); 3997 xim_set_focus(gui.in_focus); 3998 3999 if (im_activatekey_keyval != GDK_VoidSymbol) 4000 { 4001 if (im_is_active) 4002 { 4003 g_signal_handler_block(xic, im_commit_handler_id); 4004 im_synthesize_keypress(im_activatekey_keyval, 4005 im_activatekey_state); 4006 g_signal_handler_unblock(xic, im_commit_handler_id); 4007 } 4008 } 4009 else 4010 { 4011 im_shutdown(); 4012 xim_init(); 4013 xim_set_focus(gui.in_focus); 4014 } 4015 } 4016 } 4017 4018 preedit_start_col = MAXCOL; 4019 xim_has_preediting = FALSE; 4020 } 4021 4022 int 4023 xim_queue_key_press_event(GdkEventKey *event, int down) 4024 { 4025 if (down) 4026 { 4027 /* 4028 * Workaround GTK2 XIM 'feature' that always converts keypad keys to 4029 * chars., even when not part of an IM sequence (ref. feature of 4030 * gdk/gdkkeyuni.c). 4031 * Flag any keypad keys that might represent a single char. 4032 * If this (on its own - i.e., not part of an IM sequence) is 4033 * committed while we're processing one of these keys, we can ignore 4034 * that commit and go ahead & process it ourselves. That way we can 4035 * still distinguish keypad keys for use in mappings. 4036 */ 4037 switch (event->keyval) 4038 { 4039 case GDK_KP_Add: xim_expected_char = '+'; break; 4040 case GDK_KP_Subtract: xim_expected_char = '-'; break; 4041 case GDK_KP_Divide: xim_expected_char = '/'; break; 4042 case GDK_KP_Multiply: xim_expected_char = '*'; break; 4043 case GDK_KP_Decimal: xim_expected_char = '.'; break; 4044 case GDK_KP_Equal: xim_expected_char = '='; break; 4045 case GDK_KP_0: xim_expected_char = '0'; break; 4046 case GDK_KP_1: xim_expected_char = '1'; break; 4047 case GDK_KP_2: xim_expected_char = '2'; break; 4048 case GDK_KP_3: xim_expected_char = '3'; break; 4049 case GDK_KP_4: xim_expected_char = '4'; break; 4050 case GDK_KP_5: xim_expected_char = '5'; break; 4051 case GDK_KP_6: xim_expected_char = '6'; break; 4052 case GDK_KP_7: xim_expected_char = '7'; break; 4053 case GDK_KP_8: xim_expected_char = '8'; break; 4054 case GDK_KP_9: xim_expected_char = '9'; break; 4055 default: xim_expected_char = NUL; 4056 } 4057 xim_ignored_char = FALSE; 4058 } 4059 4060 /* 4061 * When typing fFtT, XIM may be activated. Thus it must pass 4062 * gtk_im_context_filter_keypress() in Normal mode. 4063 * And while doing :sh too. 4064 */ 4065 if (xic != NULL && !p_imdisable 4066 && (State & (INSERT | CMDLINE | NORMAL | EXTERNCMD)) != 0) 4067 { 4068 /* 4069 * Filter 'imactivatekey' and map it to CTRL-^. This way, Vim is 4070 * always aware of the current status of IM, and can even emulate 4071 * the activation key for modules that don't support one. 4072 */ 4073 if (event->keyval == im_activatekey_keyval 4074 && (event->state & im_activatekey_state) == im_activatekey_state) 4075 { 4076 unsigned int state_mask; 4077 4078 /* Require the state of the 3 most used modifiers to match exactly. 4079 * Otherwise e.g. <S-C-space> would be unusable for other purposes 4080 * if the IM activate key is <S-space>. */ 4081 state_mask = im_activatekey_state; 4082 state_mask |= ((int)GDK_SHIFT_MASK | (int)GDK_CONTROL_MASK 4083 | (int)GDK_MOD1_MASK); 4084 4085 if ((event->state & state_mask) != im_activatekey_state) 4086 return FALSE; 4087 4088 /* Don't send it a second time on GDK_KEY_RELEASE. */ 4089 if (event->type != GDK_KEY_PRESS) 4090 return TRUE; 4091 4092 if (map_to_exists_mode((char_u *)"", LANGMAP)) 4093 { 4094 im_set_active(FALSE); 4095 4096 /* ":lmap" mappings exists, toggle use of mappings. */ 4097 State ^= LANGMAP; 4098 if (State & LANGMAP) 4099 { 4100 curbuf->b_p_iminsert = B_IMODE_NONE; 4101 State &= ~LANGMAP; 4102 } 4103 else 4104 { 4105 curbuf->b_p_iminsert = B_IMODE_LMAP; 4106 State |= LANGMAP; 4107 } 4108 return TRUE; 4109 } 4110 4111 return gtk_im_context_filter_keypress(xic, event); 4112 } 4113 4114 /* Don't filter events through the IM context if IM isn't active 4115 * right now. Unlike with GTK+ 1.2 we cannot rely on the IM module 4116 * not doing anything before the activation key was sent. */ 4117 if (im_activatekey_keyval == GDK_VoidSymbol || im_is_active) 4118 { 4119 int imresult = gtk_im_context_filter_keypress(xic, event); 4120 4121 /* Some XIM send following sequence: 4122 * 1. preedited string. 4123 * 2. committed string. 4124 * 3. line changed key. 4125 * 4. preedited string. 4126 * 5. remove preedited string. 4127 * if 3, Vim can't move back the above line for 5. 4128 * thus, this part should not parse the key. */ 4129 if (!imresult && preedit_start_col != MAXCOL 4130 && event->keyval == GDK_Return) 4131 { 4132 im_synthesize_keypress(GDK_Return, 0U); 4133 return FALSE; 4134 } 4135 4136 /* If XIM tried to commit a keypad key as a single char., 4137 * ignore it so we can use the keypad key 'raw', for mappings. */ 4138 if (xim_expected_char != NUL && xim_ignored_char) 4139 /* We had a keypad key, and XIM tried to thieve it */ 4140 return FALSE; 4141 4142 /* Normal processing */ 4143 return imresult; 4144 } 4145 } 4146 4147 return FALSE; 4148 } 4149 4150 int 4151 im_get_status(void) 4152 { 4153 return im_is_active; 4154 } 4155 4156 # else /* !HAVE_GTK2 */ 4157 4158 static int xim_is_active = FALSE; /* XIM should be active in the current 4159 mode */ 4160 static int xim_has_focus = FALSE; /* XIM is really being used for Vim */ 4161 #ifdef FEAT_GUI_X11 4162 static XIMStyle input_style; 4163 static int status_area_enabled = TRUE; 4164 #endif 4165 4166 #ifdef FEAT_GUI_GTK 4167 # ifdef WIN3264 4168 # include <gdk/gdkwin32.h> 4169 # else 4170 # include <gdk/gdkx.h> 4171 # endif 4172 #else 4173 # ifdef PROTO 4174 /* Define a few things to be able to generate prototypes while not configured 4175 * for GTK. */ 4176 # define GSList int 4177 # define gboolean int 4178 typedef int GdkEvent; 4179 typedef int GdkEventKey; 4180 # define GdkIC int 4181 # endif 4182 #endif 4183 4184 #if defined(FEAT_GUI_GTK) || defined(PROTO) || defined(FEAT_GUI_KDE) 4185 static int preedit_buf_len = 0; 4186 static int xim_can_preediting INIT(= FALSE); /* XIM in showmode() */ 4187 static int xim_input_style; 4188 #ifndef FEAT_GUI_GTK 4189 # define gboolean int 4190 #endif 4191 static gboolean use_status_area = 0; 4192 4193 static int im_xim_str2keycode __ARGS((unsigned int *code, unsigned int *state)); 4194 static void im_xim_send_event_imactivate __ARGS((void)); 4195 4196 /* 4197 * Convert string to keycode and state for XKeyEvent. 4198 * When string is valid return OK, when invalid return FAIL. 4199 * 4200 * See 'imactivatekey' documentation for the format. 4201 */ 4202 static int 4203 im_xim_str2keycode(code, state) 4204 unsigned int *code; 4205 unsigned int *state; 4206 { 4207 int retval = OK; 4208 int len; 4209 unsigned keycode = 0, keystate = 0; 4210 Window window; 4211 Display *display; 4212 char_u *flag_end; 4213 char_u *str; 4214 4215 if (*p_imak != NUL) 4216 { 4217 len = STRLEN(p_imak); 4218 for (flag_end = p_imak + len - 1; 4219 flag_end > p_imak && *flag_end != '-'; --flag_end) 4220 ; 4221 4222 /* Parse modifier keys */ 4223 for (str = p_imak; str < flag_end; ++str) 4224 { 4225 switch (*str) 4226 { 4227 case 's': case 'S': 4228 keystate |= ShiftMask; 4229 break; 4230 case 'l': case 'L': 4231 keystate |= LockMask; 4232 break; 4233 case 'c': case 'C': 4234 keystate |= ControlMask; 4235 break; 4236 case '1': 4237 keystate |= Mod1Mask; 4238 break; 4239 case '2': 4240 keystate |= Mod2Mask; 4241 break; 4242 case '3': 4243 keystate |= Mod3Mask; 4244 break; 4245 case '4': 4246 keystate |= Mod4Mask; 4247 break; 4248 case '5': 4249 keystate |= Mod5Mask; 4250 break; 4251 case '-': 4252 break; 4253 default: 4254 retval = FAIL; 4255 } 4256 } 4257 if (*str == '-') 4258 ++str; 4259 4260 /* Get keycode from string. */ 4261 gui_get_x11_windis(&window, &display); 4262 if (display) 4263 keycode = XKeysymToKeycode(display, XStringToKeysym((char *)str)); 4264 if (keycode == 0) 4265 retval = FAIL; 4266 4267 if (code != NULL) 4268 *code = keycode; 4269 if (state != NULL) 4270 *state = keystate; 4271 } 4272 return retval; 4273 } 4274 4275 static void 4276 im_xim_send_event_imactivate() 4277 { 4278 /* Force turn on preedit state by symulate keypress event. 4279 * Keycode and state is specified by 'imactivatekey'. 4280 */ 4281 XKeyEvent ev; 4282 4283 gui_get_x11_windis(&ev.window, &ev.display); 4284 ev.root = RootWindow(ev.display, DefaultScreen(ev.display)); 4285 ev.subwindow = None; 4286 ev.time = CurrentTime; 4287 ev.x = 1; 4288 ev.y = 1; 4289 ev.x_root = 1; 4290 ev.y_root = 1; 4291 ev.same_screen = 1; 4292 ev.type = KeyPress; 4293 if (im_xim_str2keycode(&ev.keycode, &ev.state) == OK) 4294 XSendEvent(ev.display, ev.window, 1, KeyPressMask, (XEvent*)&ev); 4295 } 4296 4297 /* 4298 * Return TRUE if 'imactivatekey' has a valid value. 4299 */ 4300 int 4301 im_xim_isvalid_imactivate() 4302 { 4303 return im_xim_str2keycode(NULL, NULL) == OK; 4304 } 4305 #endif /* FEAT_GUI_GTK */ 4306 4307 /* 4308 * Switch using XIM on/off. This is used by the code that changes "State". 4309 */ 4310 void 4311 im_set_active(active) 4312 int active; 4313 { 4314 if (xic == NULL) 4315 return; 4316 4317 /* If 'imdisable' is set, XIM is never active. */ 4318 if (p_imdisable) 4319 active = FALSE; 4320 #if !defined (FEAT_GUI_GTK) && !defined (FEAT_GUI_KDE) 4321 else if (input_style & XIMPreeditPosition) 4322 /* There is a problem in switching XIM off when preediting is used, 4323 * and it is not clear how this can be solved. For now, keep XIM on 4324 * all the time, like it was done in Vim 5.8. */ 4325 active = TRUE; 4326 #endif 4327 4328 /* Remember the active state, it is needed when Vim gets keyboard focus. */ 4329 xim_is_active = active; 4330 4331 #ifdef FEAT_GUI_GTK 4332 /* When 'imactivatekey' has valid key-string, try to control XIM preedit 4333 * state. When 'imactivatekey' has no or invalid string, try old XIM 4334 * focus control. 4335 */ 4336 if (*p_imak != NUL) 4337 { 4338 /* BASIC STRATEGY: 4339 * Destroy old Input Context (XIC), and create new one. New XIC 4340 * would have a state of preedit that is off. When argument:active 4341 * is false, that's all. Else argument:active is true, send a key 4342 * event specified by 'imactivatekey' to activate XIM preedit state. 4343 */ 4344 4345 xim_is_active = TRUE; /* Disable old XIM focus control */ 4346 /* If we can monitor preedit state with preedit callback functions, 4347 * try least creation of new XIC. 4348 */ 4349 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS) 4350 { 4351 if (xim_can_preediting && !active) 4352 { 4353 /* Force turn off preedit state. With some IM 4354 * implementations, we cannot turn off preedit state by 4355 * symulate keypress event. It is why using such a method 4356 * that destroy old IC (input context), and create new one. 4357 * When create new IC, its preedit state is usually off. 4358 */ 4359 xim_reset(); 4360 xim_set_focus(FALSE); 4361 gdk_ic_destroy(xic); 4362 xim_init(); 4363 xim_can_preediting = FALSE; 4364 } 4365 else if (!xim_can_preediting && active) 4366 im_xim_send_event_imactivate(); 4367 } 4368 else 4369 { 4370 /* First, force destroy old IC, and create new one. It 4371 * symulates "turning off preedit state". 4372 */ 4373 xim_set_focus(FALSE); 4374 gdk_ic_destroy(xic); 4375 xim_init(); 4376 xim_can_preediting = FALSE; 4377 4378 /* 2nd, when requested to activate IM, symulate this by sending 4379 * the event. 4380 */ 4381 if (active) 4382 { 4383 im_xim_send_event_imactivate(); 4384 xim_can_preediting = TRUE; 4385 } 4386 } 4387 } 4388 else 4389 { 4390 # ifndef XIMPreeditUnKnown 4391 /* X11R5 doesn't have these, it looks safe enough to define here. */ 4392 typedef unsigned long XIMPreeditState; 4393 # define XIMPreeditUnKnown 0L 4394 # define XIMPreeditEnable 1L 4395 # define XIMPreeditDisable (1L<<1) 4396 # define XNPreeditState "preeditState" 4397 # endif 4398 XIMPreeditState preedit_state = XIMPreeditUnKnown; 4399 XVaNestedList preedit_attr; 4400 XIC pxic; 4401 4402 preedit_attr = XVaCreateNestedList(0, 4403 XNPreeditState, &preedit_state, 4404 NULL); 4405 pxic = ((GdkICPrivate *)xic)->xic; 4406 4407 if (!XGetICValues(pxic, XNPreeditAttributes, preedit_attr, NULL)) 4408 { 4409 XFree(preedit_attr); 4410 preedit_attr = XVaCreateNestedList(0, 4411 XNPreeditState, 4412 active ? XIMPreeditEnable : XIMPreeditDisable, 4413 NULL); 4414 XSetICValues(pxic, XNPreeditAttributes, preedit_attr, NULL); 4415 xim_can_preediting = active; 4416 xim_is_active = active; 4417 } 4418 XFree(preedit_attr); 4419 } 4420 if (xim_input_style & XIMPreeditCallbacks) 4421 { 4422 preedit_buf_len = 0; 4423 init_preedit_start_col(); 4424 } 4425 #else 4426 # if 0 4427 /* When had tested kinput2 + canna + Athena GUI version with 4428 * 'imactivatekey' is "s-space", im_xim_send_event_imactivate() did not 4429 * work correctly. It just inserted one space. I don't know why we 4430 * couldn't switch state of XIM preediting. This is reason why these 4431 * codes are commented out. 4432 */ 4433 /* First, force destroy old IC, and create new one. It symulates 4434 * "turning off preedit state". 4435 */ 4436 xim_set_focus(FALSE); 4437 XDestroyIC(xic); 4438 xic = NULL; 4439 xim_init(); 4440 4441 /* 2nd, when requested to activate IM, symulate this by sending the 4442 * event. 4443 */ 4444 if (active) 4445 im_xim_send_event_imactivate(); 4446 # endif 4447 #endif 4448 xim_set_preedit(); 4449 } 4450 4451 /* 4452 * Adjust using XIM for gaining or losing keyboard focus. Also called when 4453 * "xim_is_active" changes. 4454 */ 4455 void 4456 xim_set_focus(focus) 4457 int focus; 4458 { 4459 if (xic == NULL) 4460 return; 4461 4462 /* 4463 * XIM only gets focus when the Vim window has keyboard focus and XIM has 4464 * been set active for the current mode. 4465 */ 4466 if (focus && xim_is_active) 4467 { 4468 if (!xim_has_focus) 4469 { 4470 xim_has_focus = TRUE; 4471 #ifdef FEAT_GUI_GTK 4472 gdk_im_begin(xic, gui.drawarea->window); 4473 #else 4474 XSetICFocus(xic); 4475 #endif 4476 } 4477 } 4478 else 4479 { 4480 if (xim_has_focus) 4481 { 4482 xim_has_focus = FALSE; 4483 #ifdef FEAT_GUI_GTK 4484 gdk_im_end(); 4485 #else 4486 XUnsetICFocus(xic); 4487 #endif 4488 } 4489 } 4490 } 4491 4492 #ifndef FEAT_GUI_KDE 4493 /*ARGSUSED*/ 4494 void 4495 im_set_position(row, col) 4496 int row; 4497 int col; 4498 { 4499 xim_set_preedit(); 4500 } 4501 #endif 4502 4503 /* 4504 * Set the XIM to the current cursor position. 4505 */ 4506 void 4507 xim_set_preedit() 4508 { 4509 if (xic == NULL) 4510 return; 4511 4512 xim_set_focus(TRUE); 4513 4514 #ifdef FEAT_GUI_GTK 4515 if (gdk_im_ready()) 4516 { 4517 int attrmask; 4518 GdkICAttr *attr; 4519 4520 if (!xic_attr) 4521 return; 4522 4523 attr = xic_attr; 4524 attrmask = 0; 4525 4526 # ifdef FEAT_XFONTSET 4527 if ((xim_input_style & (int)GDK_IM_PREEDIT_POSITION) 4528 && gui.fontset != NOFONTSET 4529 && gui.fontset->type == GDK_FONT_FONTSET) 4530 { 4531 if (!xim_has_focus) 4532 { 4533 if (attr->spot_location.y >= 0) 4534 { 4535 attr->spot_location.x = 0; 4536 attr->spot_location.y = -100; 4537 attrmask |= (int)GDK_IC_SPOT_LOCATION; 4538 } 4539 } 4540 else 4541 { 4542 gint width, height; 4543 4544 if (attr->spot_location.x != TEXT_X(gui.col) 4545 || attr->spot_location.y != TEXT_Y(gui.row)) 4546 { 4547 attr->spot_location.x = TEXT_X(gui.col); 4548 attr->spot_location.y = TEXT_Y(gui.row); 4549 attrmask |= (int)GDK_IC_SPOT_LOCATION; 4550 } 4551 4552 gdk_window_get_size(gui.drawarea->window, &width, &height); 4553 width -= 2 * gui.border_offset; 4554 height -= 2 * gui.border_offset; 4555 if (xim_input_style & (int)GDK_IM_STATUS_AREA) 4556 height -= gui.char_height; 4557 if (attr->preedit_area.width != width 4558 || attr->preedit_area.height != height) 4559 { 4560 attr->preedit_area.x = gui.border_offset; 4561 attr->preedit_area.y = gui.border_offset; 4562 attr->preedit_area.width = width; 4563 attr->preedit_area.height = height; 4564 attrmask |= (int)GDK_IC_PREEDIT_AREA; 4565 } 4566 4567 if (attr->preedit_fontset != gui.current_font) 4568 { 4569 attr->preedit_fontset = gui.current_font; 4570 attrmask |= (int)GDK_IC_PREEDIT_FONTSET; 4571 } 4572 } 4573 } 4574 # endif /* FEAT_XFONTSET */ 4575 4576 if (xim_fg_color == INVALCOLOR) 4577 { 4578 xim_fg_color = gui.def_norm_pixel; 4579 xim_bg_color = gui.def_back_pixel; 4580 } 4581 if (attr->preedit_foreground.pixel != xim_fg_color) 4582 { 4583 attr->preedit_foreground.pixel = xim_fg_color; 4584 attrmask |= (int)GDK_IC_PREEDIT_FOREGROUND; 4585 } 4586 if (attr->preedit_background.pixel != xim_bg_color) 4587 { 4588 attr->preedit_background.pixel = xim_bg_color; 4589 attrmask |= (int)GDK_IC_PREEDIT_BACKGROUND; 4590 } 4591 4592 if (attrmask != 0) 4593 gdk_ic_set_attr(xic, attr, (GdkICAttributesType)attrmask); 4594 } 4595 #else /* FEAT_GUI_GTK */ 4596 # ifdef FEAT_GUI_KDE 4597 # else 4598 { 4599 XVaNestedList attr_list; 4600 XRectangle spot_area; 4601 XPoint over_spot; 4602 int line_space; 4603 4604 if (!xim_has_focus) 4605 { 4606 /* hide XIM cursor */ 4607 over_spot.x = 0; 4608 over_spot.y = -100; /* arbitrary invisible position */ 4609 attr_list = (XVaNestedList) XVaCreateNestedList(0, 4610 XNSpotLocation, 4611 &over_spot, 4612 NULL); 4613 XSetICValues(xic, XNPreeditAttributes, attr_list, NULL); 4614 XFree(attr_list); 4615 return; 4616 } 4617 4618 if (input_style & XIMPreeditPosition) 4619 { 4620 if (xim_fg_color == INVALCOLOR) 4621 { 4622 xim_fg_color = gui.def_norm_pixel; 4623 xim_bg_color = gui.def_back_pixel; 4624 } 4625 over_spot.x = TEXT_X(gui.col); 4626 over_spot.y = TEXT_Y(gui.row); 4627 spot_area.x = 0; 4628 spot_area.y = 0; 4629 spot_area.height = gui.char_height * Rows; 4630 spot_area.width = gui.char_width * Columns; 4631 line_space = gui.char_height; 4632 attr_list = (XVaNestedList) XVaCreateNestedList(0, 4633 XNSpotLocation, &over_spot, 4634 XNForeground, (Pixel) xim_fg_color, 4635 XNBackground, (Pixel) xim_bg_color, 4636 XNArea, &spot_area, 4637 XNLineSpace, line_space, 4638 NULL); 4639 if (XSetICValues(xic, XNPreeditAttributes, attr_list, NULL)) 4640 EMSG(_("E284: Cannot set IC values")); 4641 XFree(attr_list); 4642 } 4643 } 4644 # endif /* FEAT_GUI_KDE */ 4645 #endif /* FEAT_GUI_GTK */ 4646 } 4647 4648 /* 4649 * Set up the status area. 4650 * 4651 * This should use a separate Widget, but that seems not possible, because 4652 * preedit_area and status_area should be set to the same window as for the 4653 * text input. Unfortunately this means the status area pollutes the text 4654 * window... 4655 */ 4656 void 4657 xim_set_status_area() 4658 { 4659 if (xic == NULL) 4660 return; 4661 4662 #ifdef FEAT_GUI_GTK 4663 # if defined(FEAT_XFONTSET) 4664 if (use_status_area) 4665 { 4666 GdkICAttr *attr; 4667 int style; 4668 gint width, height; 4669 GtkWidget *widget; 4670 int attrmask; 4671 4672 if (!xic_attr) 4673 return; 4674 4675 attr = xic_attr; 4676 attrmask = 0; 4677 style = (int)gdk_ic_get_style(xic); 4678 if ((style & (int)GDK_IM_STATUS_MASK) == (int)GDK_IM_STATUS_AREA) 4679 { 4680 if (gui.fontset != NOFONTSET 4681 && gui.fontset->type == GDK_FONT_FONTSET) 4682 { 4683 widget = gui.mainwin; 4684 gdk_window_get_size(widget->window, &width, &height); 4685 4686 attrmask |= (int)GDK_IC_STATUS_AREA; 4687 attr->status_area.x = 0; 4688 attr->status_area.y = height - gui.char_height - 1; 4689 attr->status_area.width = width; 4690 attr->status_area.height = gui.char_height; 4691 } 4692 } 4693 if (attrmask != 0) 4694 gdk_ic_set_attr(xic, attr, (GdkICAttributesType)attrmask); 4695 } 4696 # endif 4697 #else 4698 # ifdef FEAT_GUI_KDE 4699 # else 4700 { 4701 XVaNestedList preedit_list = 0, status_list = 0, list = 0; 4702 XRectangle pre_area, status_area; 4703 4704 if (input_style & XIMStatusArea) 4705 { 4706 if (input_style & XIMPreeditArea) 4707 { 4708 XRectangle *needed_rect; 4709 4710 /* to get status_area width */ 4711 status_list = XVaCreateNestedList(0, XNAreaNeeded, 4712 &needed_rect, NULL); 4713 XGetICValues(xic, XNStatusAttributes, status_list, NULL); 4714 XFree(status_list); 4715 4716 status_area.width = needed_rect->width; 4717 } 4718 else 4719 status_area.width = gui.char_width * Columns; 4720 4721 status_area.x = 0; 4722 status_area.y = gui.char_height * Rows + gui.border_offset; 4723 if (gui.which_scrollbars[SBAR_BOTTOM]) 4724 status_area.y += gui.scrollbar_height; 4725 #ifdef FEAT_MENU 4726 if (gui.menu_is_active) 4727 status_area.y += gui.menu_height; 4728 #endif 4729 status_area.height = gui.char_height; 4730 status_list = XVaCreateNestedList(0, XNArea, &status_area, NULL); 4731 } 4732 else 4733 { 4734 status_area.x = 0; 4735 status_area.y = gui.char_height * Rows + gui.border_offset; 4736 if (gui.which_scrollbars[SBAR_BOTTOM]) 4737 status_area.y += gui.scrollbar_height; 4738 #ifdef FEAT_MENU 4739 if (gui.menu_is_active) 4740 status_area.y += gui.menu_height; 4741 #endif 4742 status_area.width = 0; 4743 status_area.height = gui.char_height; 4744 } 4745 4746 if (input_style & XIMPreeditArea) /* off-the-spot */ 4747 { 4748 pre_area.x = status_area.x + status_area.width; 4749 pre_area.y = gui.char_height * Rows + gui.border_offset; 4750 pre_area.width = gui.char_width * Columns - pre_area.x; 4751 if (gui.which_scrollbars[SBAR_BOTTOM]) 4752 pre_area.y += gui.scrollbar_height; 4753 #ifdef FEAT_MENU 4754 if (gui.menu_is_active) 4755 pre_area.y += gui.menu_height; 4756 #endif 4757 pre_area.height = gui.char_height; 4758 preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL); 4759 } 4760 else if (input_style & XIMPreeditPosition) /* over-the-spot */ 4761 { 4762 pre_area.x = 0; 4763 pre_area.y = 0; 4764 pre_area.height = gui.char_height * Rows; 4765 pre_area.width = gui.char_width * Columns; 4766 preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL); 4767 } 4768 4769 if (preedit_list && status_list) 4770 list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list, 4771 XNStatusAttributes, status_list, NULL); 4772 else if (preedit_list) 4773 list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list, 4774 NULL); 4775 else if (status_list) 4776 list = XVaCreateNestedList(0, XNStatusAttributes, status_list, 4777 NULL); 4778 else 4779 list = NULL; 4780 4781 if (list) 4782 { 4783 XSetICValues(xic, XNVaNestedList, list, NULL); 4784 XFree(list); 4785 } 4786 if (status_list) 4787 XFree(status_list); 4788 if (preedit_list) 4789 XFree(preedit_list); 4790 } 4791 # endif /* FEAT_GUI_KDE */ 4792 #endif 4793 } 4794 4795 #if defined(FEAT_GUI_X11) || defined(FEAT_GUI_GTK) || defined(FEAT_GUI_KDE) 4796 static char e_xim[] = N_("E285: Failed to create input context"); 4797 #endif 4798 4799 #if defined(FEAT_GUI_X11) || defined(PROTO) 4800 # if defined(XtSpecificationRelease) && XtSpecificationRelease >= 6 && !defined(sun) 4801 # define USE_X11R6_XIM 4802 # endif 4803 4804 static int xim_real_init __ARGS((Window x11_window, Display *x11_display)); 4805 4806 4807 #ifdef USE_X11R6_XIM 4808 static void xim_instantiate_cb __ARGS((Display *display, XPointer client_data, XPointer call_data)); 4809 static void xim_destroy_cb __ARGS((XIM im, XPointer client_data, XPointer call_data)); 4810 4811 /*ARGSUSED*/ 4812 static void 4813 xim_instantiate_cb(display, client_data, call_data) 4814 Display *display; 4815 XPointer client_data; 4816 XPointer call_data; 4817 { 4818 Window x11_window; 4819 Display *x11_display; 4820 4821 #ifdef XIM_DEBUG 4822 xim_log("xim_instantiate_cb()\n"); 4823 #endif 4824 4825 gui_get_x11_windis(&x11_window, &x11_display); 4826 if (display != x11_display) 4827 return; 4828 4829 xim_real_init(x11_window, x11_display); 4830 gui_set_shellsize(FALSE, FALSE); 4831 if (xic != NULL) 4832 XUnregisterIMInstantiateCallback(x11_display, NULL, NULL, NULL, 4833 xim_instantiate_cb, NULL); 4834 } 4835 4836 /*ARGSUSED*/ 4837 static void 4838 xim_destroy_cb(im, client_data, call_data) 4839 XIM im; 4840 XPointer client_data; 4841 XPointer call_data; 4842 { 4843 Window x11_window; 4844 Display *x11_display; 4845 4846 #ifdef XIM_DEBUG 4847 xim_log("xim_destroy_cb()\n"); 4848 #endif 4849 gui_get_x11_windis(&x11_window, &x11_display); 4850 4851 xic = NULL; 4852 status_area_enabled = FALSE; 4853 4854 gui_set_shellsize(FALSE, FALSE); 4855 4856 XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL, 4857 xim_instantiate_cb, NULL); 4858 } 4859 #endif 4860 4861 void 4862 xim_init() 4863 { 4864 Window x11_window; 4865 Display *x11_display; 4866 4867 #ifdef XIM_DEBUG 4868 xim_log("xim_init()\n"); 4869 #endif 4870 4871 gui_get_x11_windis(&x11_window, &x11_display); 4872 4873 xic = NULL; 4874 4875 if (xim_real_init(x11_window, x11_display)) 4876 return; 4877 4878 gui_set_shellsize(FALSE, FALSE); 4879 4880 #ifdef USE_X11R6_XIM 4881 XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL, 4882 xim_instantiate_cb, NULL); 4883 #endif 4884 } 4885 4886 static int 4887 xim_real_init(x11_window, x11_display) 4888 Window x11_window; 4889 Display *x11_display; 4890 { 4891 int i; 4892 char *p, 4893 *s, 4894 *ns, 4895 *end, 4896 tmp[1024]; 4897 #define IMLEN_MAX 40 4898 char buf[IMLEN_MAX + 7]; 4899 XIM xim = NULL; 4900 XIMStyles *xim_styles; 4901 XIMStyle this_input_style = 0; 4902 Boolean found; 4903 XPoint over_spot; 4904 XVaNestedList preedit_list, status_list; 4905 4906 input_style = 0; 4907 status_area_enabled = FALSE; 4908 4909 if (xic != NULL) 4910 return FALSE; 4911 4912 if (gui.rsrc_input_method != NULL && *gui.rsrc_input_method != NUL) 4913 { 4914 strcpy(tmp, gui.rsrc_input_method); 4915 for (ns = s = tmp; ns != NULL && *s != NUL;) 4916 { 4917 s = (char *)skipwhite((char_u *)s); 4918 if (*s == NUL) 4919 break; 4920 if ((ns = end = strchr(s, ',')) == NULL) 4921 end = s + strlen(s); 4922 while (isspace(((char_u *)end)[-1])) 4923 end--; 4924 *end = NUL; 4925 4926 if (strlen(s) <= IMLEN_MAX) 4927 { 4928 strcpy(buf, "@im="); 4929 strcat(buf, s); 4930 if ((p = XSetLocaleModifiers(buf)) != NULL && *p != NUL 4931 && (xim = XOpenIM(x11_display, NULL, NULL, NULL)) 4932 != NULL) 4933 break; 4934 } 4935 4936 s = ns + 1; 4937 } 4938 } 4939 4940 if (xim == NULL && (p = XSetLocaleModifiers("")) != NULL && *p != NUL) 4941 xim = XOpenIM(x11_display, NULL, NULL, NULL); 4942 4943 /* This is supposed to be useful to obtain characters through 4944 * XmbLookupString() without really using a XIM. */ 4945 if (xim == NULL && (p = XSetLocaleModifiers("@im=none")) != NULL 4946 && *p != NUL) 4947 xim = XOpenIM(x11_display, NULL, NULL, NULL); 4948 4949 if (xim == NULL) 4950 { 4951 /* Only give this message when verbose is set, because too many people 4952 * got this message when they didn't want to use a XIM. */ 4953 if (p_verbose > 0) 4954 { 4955 verbose_enter(); 4956 EMSG(_("E286: Failed to open input method")); 4957 verbose_leave(); 4958 } 4959 return FALSE; 4960 } 4961 4962 #ifdef USE_X11R6_XIM 4963 { 4964 XIMCallback destroy_cb; 4965 4966 destroy_cb.callback = xim_destroy_cb; 4967 destroy_cb.client_data = NULL; 4968 if (XSetIMValues(xim, XNDestroyCallback, &destroy_cb, NULL)) 4969 EMSG(_("E287: Warning: Could not set destroy callback to IM")); 4970 } 4971 #endif 4972 4973 if (XGetIMValues(xim, XNQueryInputStyle, &xim_styles, NULL) || !xim_styles) 4974 { 4975 EMSG(_("E288: input method doesn't support any style")); 4976 XCloseIM(xim); 4977 return FALSE; 4978 } 4979 4980 found = False; 4981 strcpy(tmp, gui.rsrc_preedit_type_name); 4982 for (s = tmp; s && !found; ) 4983 { 4984 while (*s && isspace((unsigned char)*s)) 4985 s++; 4986 if (!*s) 4987 break; 4988 if ((ns = end = strchr(s, ',')) != 0) 4989 ns++; 4990 else 4991 end = s + strlen(s); 4992 while (isspace((unsigned char)*end)) 4993 end--; 4994 *end = '\0'; 4995 4996 if (!strcmp(s, "OverTheSpot")) 4997 this_input_style = (XIMPreeditPosition | XIMStatusArea); 4998 else if (!strcmp(s, "OffTheSpot")) 4999 this_input_style = (XIMPreeditArea | XIMStatusArea); 5000 else if (!strcmp(s, "Root")) 5001 this_input_style = (XIMPreeditNothing | XIMStatusNothing); 5002 5003 for (i = 0; (unsigned short)i < xim_styles->count_styles; i++) 5004 { 5005 if (this_input_style == xim_styles->supported_styles[i]) 5006 { 5007 found = True; 5008 break; 5009 } 5010 } 5011 if (!found) 5012 for (i = 0; (unsigned short)i < xim_styles->count_styles; i++) 5013 { 5014 if ((xim_styles->supported_styles[i] & this_input_style) 5015 == (this_input_style & ~XIMStatusArea)) 5016 { 5017 this_input_style &= ~XIMStatusArea; 5018 found = True; 5019 break; 5020 } 5021 } 5022 5023 s = ns; 5024 } 5025 XFree(xim_styles); 5026 5027 if (!found) 5028 { 5029 /* Only give this message when verbose is set, because too many people 5030 * got this message when they didn't want to use a XIM. */ 5031 if (p_verbose > 0) 5032 { 5033 verbose_enter(); 5034 EMSG(_("E289: input method doesn't support my preedit type")); 5035 verbose_leave(); 5036 } 5037 XCloseIM(xim); 5038 return FALSE; 5039 } 5040 5041 over_spot.x = TEXT_X(gui.col); 5042 over_spot.y = TEXT_Y(gui.row); 5043 input_style = this_input_style; 5044 5045 /* A crash was reported when trying to pass gui.norm_font as XNFontSet, 5046 * thus that has been removed. Hopefully the default works... */ 5047 #ifdef FEAT_XFONTSET 5048 if (gui.fontset != NOFONTSET) 5049 { 5050 preedit_list = XVaCreateNestedList(0, 5051 XNSpotLocation, &over_spot, 5052 XNForeground, (Pixel)gui.def_norm_pixel, 5053 XNBackground, (Pixel)gui.def_back_pixel, 5054 XNFontSet, (XFontSet)gui.fontset, 5055 NULL); 5056 status_list = XVaCreateNestedList(0, 5057 XNForeground, (Pixel)gui.def_norm_pixel, 5058 XNBackground, (Pixel)gui.def_back_pixel, 5059 XNFontSet, (XFontSet)gui.fontset, 5060 NULL); 5061 } 5062 else 5063 #endif 5064 { 5065 preedit_list = XVaCreateNestedList(0, 5066 XNSpotLocation, &over_spot, 5067 XNForeground, (Pixel)gui.def_norm_pixel, 5068 XNBackground, (Pixel)gui.def_back_pixel, 5069 NULL); 5070 status_list = XVaCreateNestedList(0, 5071 XNForeground, (Pixel)gui.def_norm_pixel, 5072 XNBackground, (Pixel)gui.def_back_pixel, 5073 NULL); 5074 } 5075 5076 xic = XCreateIC(xim, 5077 XNInputStyle, input_style, 5078 XNClientWindow, x11_window, 5079 XNFocusWindow, gui.wid, 5080 XNPreeditAttributes, preedit_list, 5081 XNStatusAttributes, status_list, 5082 NULL); 5083 XFree(status_list); 5084 XFree(preedit_list); 5085 if (xic != NULL) 5086 { 5087 if (input_style & XIMStatusArea) 5088 { 5089 xim_set_status_area(); 5090 status_area_enabled = TRUE; 5091 } 5092 else 5093 gui_set_shellsize(FALSE, FALSE); 5094 } 5095 else 5096 { 5097 EMSG(_(e_xim)); 5098 XCloseIM(xim); 5099 return FALSE; 5100 } 5101 5102 return TRUE; 5103 } 5104 5105 #endif /* FEAT_GUI_X11 */ 5106 5107 #if defined(FEAT_GUI_GTK) || defined(PROTO) 5108 5109 # ifdef FEAT_XFONTSET 5110 static char e_overthespot[] = N_("E290: over-the-spot style requires fontset"); 5111 # endif 5112 5113 # ifdef PROTO 5114 typedef int GdkIC; 5115 # endif 5116 5117 void 5118 xim_decide_input_style() 5119 { 5120 /* GDK_IM_STATUS_CALLBACKS was disabled, enabled it to allow Japanese 5121 * OverTheSpot. */ 5122 int supported_style = (int)GDK_IM_PREEDIT_NONE | 5123 (int)GDK_IM_PREEDIT_NOTHING | 5124 (int)GDK_IM_PREEDIT_POSITION | 5125 (int)GDK_IM_PREEDIT_CALLBACKS | 5126 (int)GDK_IM_STATUS_CALLBACKS | 5127 (int)GDK_IM_STATUS_AREA | 5128 (int)GDK_IM_STATUS_NONE | 5129 (int)GDK_IM_STATUS_NOTHING; 5130 5131 #ifdef XIM_DEBUG 5132 xim_log("xim_decide_input_style()\n"); 5133 #endif 5134 5135 if (!gdk_im_ready()) 5136 xim_input_style = 0; 5137 else 5138 { 5139 if (gtk_major_version > 1 5140 || (gtk_major_version == 1 5141 && (gtk_minor_version > 2 5142 || (gtk_minor_version == 2 && gtk_micro_version >= 3)))) 5143 use_status_area = TRUE; 5144 else 5145 { 5146 EMSG(_("E291: Your GTK+ is older than 1.2.3. Status area disabled")); 5147 use_status_area = FALSE; 5148 } 5149 #ifdef FEAT_XFONTSET 5150 if (gui.fontset == NOFONTSET || gui.fontset->type != GDK_FONT_FONTSET) 5151 #endif 5152 supported_style &= ~((int)GDK_IM_PREEDIT_POSITION 5153 | (int)GDK_IM_STATUS_AREA); 5154 if (!use_status_area) 5155 supported_style &= ~(int)GDK_IM_STATUS_AREA; 5156 xim_input_style = (int)gdk_im_decide_style((GdkIMStyle)supported_style); 5157 } 5158 } 5159 5160 /*ARGSUSED*/ 5161 static void 5162 preedit_start_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5163 { 5164 #ifdef XIM_DEBUG 5165 xim_log("xim_decide_input_style()\n"); 5166 #endif 5167 5168 draw_feedback = NULL; 5169 xim_can_preediting = TRUE; 5170 xim_has_preediting = TRUE; 5171 gui_update_cursor(TRUE, FALSE); 5172 if (showmode() > 0) 5173 { 5174 setcursor(); 5175 out_flush(); 5176 } 5177 } 5178 5179 static void 5180 xim_back_delete(int n) 5181 { 5182 char_u str[3]; 5183 5184 str[0] = CSI; 5185 str[1] = 'k'; 5186 str[2] = 'b'; 5187 while (n-- > 0) 5188 add_to_input_buf(str, 3); 5189 } 5190 5191 static GSList *key_press_event_queue = NULL; 5192 static gboolean processing_queued_event = FALSE; 5193 5194 /*ARGSUSED*/ 5195 static void 5196 preedit_draw_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5197 { 5198 XIMPreeditDrawCallbackStruct *draw_data; 5199 XIMText *text; 5200 char *src; 5201 GSList *event_queue; 5202 5203 #ifdef XIM_DEBUG 5204 xim_log("preedit_draw_cbproc()\n"); 5205 #endif 5206 5207 draw_data = (XIMPreeditDrawCallbackStruct *) call_data; 5208 text = (XIMText *) draw_data->text; 5209 5210 if ((text == NULL && draw_data->chg_length == preedit_buf_len) 5211 || preedit_buf_len == 0) 5212 { 5213 init_preedit_start_col(); 5214 vim_free(draw_feedback); 5215 draw_feedback = NULL; 5216 } 5217 if (draw_data->chg_length > 0) 5218 { 5219 int bs_cnt; 5220 5221 if (draw_data->chg_length > preedit_buf_len) 5222 bs_cnt = preedit_buf_len; 5223 else 5224 bs_cnt = draw_data->chg_length; 5225 xim_back_delete(bs_cnt); 5226 preedit_buf_len -= bs_cnt; 5227 } 5228 if (text != NULL) 5229 { 5230 int len; 5231 #ifdef FEAT_MBYTE 5232 char_u *buf = NULL; 5233 unsigned int nfeedback = 0; 5234 #endif 5235 char_u *ptr; 5236 5237 src = text->string.multi_byte; 5238 if (src != NULL && !text->encoding_is_wchar) 5239 { 5240 len = strlen(src); 5241 ptr = (char_u *)src; 5242 /* Avoid the enter for decision */ 5243 if (*ptr == '\n') 5244 return; 5245 5246 #ifdef FEAT_MBYTE 5247 if (input_conv.vc_type != CONV_NONE 5248 && (buf = string_convert(&input_conv, 5249 (char_u *)src, &len)) != NULL) 5250 { 5251 /* Converted from 'termencoding' to 'encoding'. */ 5252 add_to_input_buf_csi(buf, len); 5253 ptr = buf; 5254 } 5255 else 5256 #endif 5257 add_to_input_buf_csi((char_u *)src, len); 5258 /* Add count of character to preedit_buf_len */ 5259 while (*ptr != NUL) 5260 { 5261 #ifdef FEAT_MBYTE 5262 if (draw_data->text->feedback != NULL) 5263 { 5264 if (draw_feedback == NULL) 5265 draw_feedback = (char *)alloc(draw_data->chg_first 5266 + text->length); 5267 else 5268 draw_feedback = realloc(draw_feedback, 5269 draw_data->chg_first + text->length); 5270 if (draw_feedback != NULL) 5271 { 5272 draw_feedback[nfeedback + draw_data->chg_first] 5273 = draw_data->text->feedback[nfeedback]; 5274 nfeedback++; 5275 } 5276 } 5277 if (has_mbyte) 5278 ptr += (*mb_ptr2len)(ptr); 5279 else 5280 #endif 5281 ptr++; 5282 preedit_buf_len++; 5283 } 5284 #ifdef FEAT_MBYTE 5285 vim_free(buf); 5286 #endif 5287 preedit_end_col = MAXCOL; 5288 } 5289 } 5290 if (text != NULL || draw_data->chg_length > 0) 5291 { 5292 event_queue = key_press_event_queue; 5293 processing_queued_event = TRUE; 5294 while (event_queue != NULL && processing_queued_event) 5295 { 5296 GdkEvent *ev = event_queue->data; 5297 5298 gboolean *ret; 5299 gtk_signal_emit_by_name((GtkObject*)gui.mainwin, "key_press_event", 5300 ev, &ret); 5301 gdk_event_free(ev); 5302 event_queue = event_queue->next; 5303 } 5304 processing_queued_event = FALSE; 5305 if (key_press_event_queue) 5306 { 5307 g_slist_free(key_press_event_queue); 5308 key_press_event_queue = NULL; 5309 } 5310 } 5311 if (gtk_main_level() > 0) 5312 gtk_main_quit(); 5313 } 5314 5315 /* 5316 * Retrieve the highlighting attributes at column col in the preedit string. 5317 * Return -1 if not in preediting mode or if col is out of range. 5318 */ 5319 int 5320 im_get_feedback_attr(int col) 5321 { 5322 if (draw_feedback != NULL && col < preedit_buf_len) 5323 { 5324 if (draw_feedback[col] & XIMReverse) 5325 return HL_INVERSE; 5326 else if (draw_feedback[col] & XIMUnderline) 5327 return HL_UNDERLINE; 5328 else 5329 return hl_attr(HLF_V); 5330 } 5331 5332 return -1; 5333 } 5334 5335 /*ARGSUSED*/ 5336 static void 5337 preedit_caret_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5338 { 5339 #ifdef XIM_DEBUG 5340 xim_log("preedit_caret_cbproc()\n"); 5341 #endif 5342 } 5343 5344 /*ARGSUSED*/ 5345 static void 5346 preedit_done_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5347 { 5348 #ifdef XIM_DEBUG 5349 xim_log("preedit_done_cbproc()\n"); 5350 #endif 5351 5352 vim_free(draw_feedback); 5353 draw_feedback = NULL; 5354 xim_can_preediting = FALSE; 5355 xim_has_preediting = FALSE; 5356 gui_update_cursor(TRUE, FALSE); 5357 if (showmode() > 0) 5358 { 5359 setcursor(); 5360 out_flush(); 5361 } 5362 } 5363 5364 void 5365 xim_reset(void) 5366 { 5367 char *text; 5368 5369 #ifdef XIM_DEBUG 5370 xim_log("xim_reset()\n"); 5371 #endif 5372 5373 if (xic != NULL) 5374 { 5375 text = XmbResetIC(((GdkICPrivate *)xic)->xic); 5376 if (text != NULL && !(xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS)) 5377 add_to_input_buf_csi((char_u *)text, strlen(text)); 5378 else 5379 preedit_buf_len = 0; 5380 if (text != NULL) 5381 XFree(text); 5382 } 5383 } 5384 5385 /*ARGSUSED*/ 5386 int 5387 xim_queue_key_press_event(GdkEventKey *event, int down) 5388 { 5389 #ifdef XIM_DEBUG 5390 xim_log("xim_queue_key_press_event()\n"); 5391 #endif 5392 5393 if (preedit_buf_len <= 0) 5394 return FALSE; 5395 if (processing_queued_event) 5396 processing_queued_event = FALSE; 5397 5398 key_press_event_queue = g_slist_append(key_press_event_queue, 5399 gdk_event_copy((GdkEvent *)event)); 5400 return TRUE; 5401 } 5402 5403 /*ARGSUSED*/ 5404 static void 5405 preedit_callback_setup(GdkIC *ic) 5406 { 5407 XIC xxic; 5408 XVaNestedList preedit_attr; 5409 XIMCallback preedit_start_cb; 5410 XIMCallback preedit_draw_cb; 5411 XIMCallback preedit_caret_cb; 5412 XIMCallback preedit_done_cb; 5413 5414 xxic = ((GdkICPrivate*)xic)->xic; 5415 preedit_start_cb.callback = (XIMProc)preedit_start_cbproc; 5416 preedit_draw_cb.callback = (XIMProc)preedit_draw_cbproc; 5417 preedit_caret_cb.callback = (XIMProc)preedit_caret_cbproc; 5418 preedit_done_cb.callback = (XIMProc)preedit_done_cbproc; 5419 preedit_attr 5420 = XVaCreateNestedList (0, 5421 XNPreeditStartCallback, &preedit_start_cb, 5422 XNPreeditDrawCallback, &preedit_draw_cb, 5423 XNPreeditCaretCallback, &preedit_caret_cb, 5424 XNPreeditDoneCallback, &preedit_done_cb, 5425 0); 5426 XSetICValues (xxic, XNPreeditAttributes, preedit_attr, 0); 5427 XFree(preedit_attr); 5428 } 5429 5430 /*ARGSUSED*/ 5431 static void 5432 reset_state_setup(GdkIC *ic) 5433 { 5434 #ifdef USE_X11R6_XIM 5435 /* don't change the input context when we call reset */ 5436 XSetICValues(((GdkICPrivate*)ic)->xic, XNResetState, XIMPreserveState, 0); 5437 #endif 5438 } 5439 5440 void 5441 xim_init(void) 5442 { 5443 #ifdef XIM_DEBUG 5444 xim_log("xim_init()\n"); 5445 #endif 5446 5447 xic = NULL; 5448 xic_attr = NULL; 5449 5450 if (!gdk_im_ready()) 5451 { 5452 if (p_verbose > 0) 5453 { 5454 verbose_enter(); 5455 EMSG(_("E292: Input Method Server is not running")); 5456 verbose_leave(); 5457 } 5458 return; 5459 } 5460 if ((xic_attr = gdk_ic_attr_new()) != NULL) 5461 { 5462 #ifdef FEAT_XFONTSET 5463 gint width, height; 5464 #endif 5465 int mask; 5466 GdkColormap *colormap; 5467 GdkICAttr *attr = xic_attr; 5468 int attrmask = (int)GDK_IC_ALL_REQ; 5469 GtkWidget *widget = gui.drawarea; 5470 5471 attr->style = (GdkIMStyle)xim_input_style; 5472 attr->client_window = gui.mainwin->window; 5473 5474 if ((colormap = gtk_widget_get_colormap(widget)) != 5475 gtk_widget_get_default_colormap()) 5476 { 5477 attrmask |= (int)GDK_IC_PREEDIT_COLORMAP; 5478 attr->preedit_colormap = colormap; 5479 } 5480 attrmask |= (int)GDK_IC_PREEDIT_FOREGROUND; 5481 attrmask |= (int)GDK_IC_PREEDIT_BACKGROUND; 5482 attr->preedit_foreground = widget->style->fg[GTK_STATE_NORMAL]; 5483 attr->preedit_background = widget->style->base[GTK_STATE_NORMAL]; 5484 5485 #ifdef FEAT_XFONTSET 5486 if ((xim_input_style & (int)GDK_IM_PREEDIT_MASK) 5487 == (int)GDK_IM_PREEDIT_POSITION) 5488 { 5489 if (gui.fontset == NOFONTSET 5490 || gui.fontset->type != GDK_FONT_FONTSET) 5491 { 5492 EMSG(_(e_overthespot)); 5493 } 5494 else 5495 { 5496 gdk_window_get_size(widget->window, &width, &height); 5497 5498 attrmask |= (int)GDK_IC_PREEDIT_POSITION_REQ; 5499 attr->spot_location.x = TEXT_X(0); 5500 attr->spot_location.y = TEXT_Y(0); 5501 attr->preedit_area.x = gui.border_offset; 5502 attr->preedit_area.y = gui.border_offset; 5503 attr->preedit_area.width = width - 2*gui.border_offset; 5504 attr->preedit_area.height = height - 2*gui.border_offset; 5505 attr->preedit_fontset = gui.fontset; 5506 } 5507 } 5508 5509 if ((xim_input_style & (int)GDK_IM_STATUS_MASK) 5510 == (int)GDK_IM_STATUS_AREA) 5511 { 5512 if (gui.fontset == NOFONTSET 5513 || gui.fontset->type != GDK_FONT_FONTSET) 5514 { 5515 EMSG(_(e_overthespot)); 5516 } 5517 else 5518 { 5519 gdk_window_get_size(gui.mainwin->window, &width, &height); 5520 attrmask |= (int)GDK_IC_STATUS_AREA_REQ; 5521 attr->status_area.x = 0; 5522 attr->status_area.y = height - gui.char_height - 1; 5523 attr->status_area.width = width; 5524 attr->status_area.height = gui.char_height; 5525 attr->status_fontset = gui.fontset; 5526 } 5527 } 5528 else if ((xim_input_style & (int)GDK_IM_STATUS_MASK) 5529 == (int)GDK_IM_STATUS_CALLBACKS) 5530 { 5531 /* FIXME */ 5532 } 5533 #endif 5534 5535 xic = gdk_ic_new(attr, (GdkICAttributesType)attrmask); 5536 5537 if (xic == NULL) 5538 EMSG(_(e_xim)); 5539 else 5540 { 5541 mask = (int)gdk_window_get_events(widget->window); 5542 mask |= (int)gdk_ic_get_events(xic); 5543 gdk_window_set_events(widget->window, (GdkEventMask)mask); 5544 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS) 5545 preedit_callback_setup(xic); 5546 reset_state_setup(xic); 5547 } 5548 } 5549 } 5550 5551 void 5552 im_shutdown(void) 5553 { 5554 #ifdef XIM_DEBUG 5555 xim_log("im_shutdown()\n"); 5556 #endif 5557 5558 if (xic != NULL) 5559 { 5560 gdk_im_end(); 5561 gdk_ic_destroy(xic); 5562 xic = NULL; 5563 } 5564 xim_is_active = FALSE; 5565 xim_can_preediting = FALSE; 5566 preedit_start_col = MAXCOL; 5567 xim_has_preediting = FALSE; 5568 } 5569 5570 #endif /* FEAT_GUI_GTK */ 5571 5572 int 5573 xim_get_status_area_height() 5574 { 5575 #ifdef FEAT_GUI_GTK 5576 if (xim_input_style & (int)GDK_IM_STATUS_AREA) 5577 return gui.char_height; 5578 #else 5579 # if defined FEAT_GUI_KDE 5580 /* always return zero? */ 5581 # else 5582 if (status_area_enabled) 5583 return gui.char_height; 5584 # endif 5585 #endif 5586 return 0; 5587 } 5588 5589 /* 5590 * Get IM status. When IM is on, return TRUE. Else return FALSE. 5591 * FIXME: This doesn't work correctly: Having focus doesn't always mean XIM is 5592 * active, when not having focus XIM may still be active (e.g., when using a 5593 * tear-off menu item). 5594 */ 5595 int 5596 im_get_status() 5597 { 5598 # ifdef FEAT_GUI_GTK 5599 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS) 5600 return xim_can_preediting; 5601 # endif 5602 # ifdef FEAT_GUI_KDE 5603 if (preedit_start_col != MAXCOL) 5604 return TRUE; 5605 # endif 5606 return xim_has_focus; 5607 } 5608 5609 # endif /* !HAVE_GTK2 */ 5610 5611 # if defined(FEAT_GUI_GTK) || defined(PROTO) 5612 int 5613 im_is_preediting() 5614 { 5615 return xim_has_preediting; 5616 } 5617 # endif 5618 #endif /* FEAT_XIM */ 5619 5620 #if defined(FEAT_MBYTE) || defined(PROTO) 5621 5622 /* 5623 * Setup "vcp" for conversion from "from" to "to". 5624 * The names must have been made canonical with enc_canonize(). 5625 * vcp->vc_type must have been initialized to CONV_NONE. 5626 * Note: cannot be used for conversion from/to ucs-2 and ucs-4 (will use utf-8 5627 * instead). 5628 * Afterwards invoke with "from" and "to" equal to NULL to cleanup. 5629 * Return FAIL when conversion is not supported, OK otherwise. 5630 */ 5631 int 5632 convert_setup(vcp, from, to) 5633 vimconv_T *vcp; 5634 char_u *from; 5635 char_u *to; 5636 { 5637 int from_prop; 5638 int to_prop; 5639 5640 /* Reset to no conversion. */ 5641 # ifdef USE_ICONV 5642 if (vcp->vc_type == CONV_ICONV && vcp->vc_fd != (iconv_t)-1) 5643 iconv_close(vcp->vc_fd); 5644 # endif 5645 vcp->vc_type = CONV_NONE; 5646 vcp->vc_factor = 1; 5647 vcp->vc_fail = FALSE; 5648 5649 /* No conversion when one of the names is empty or they are equal. */ 5650 if (from == NULL || *from == NUL || to == NULL || *to == NUL 5651 || STRCMP(from, to) == 0) 5652 return OK; 5653 5654 from_prop = enc_canon_props(from); 5655 to_prop = enc_canon_props(to); 5656 if ((from_prop & ENC_LATIN1) && (to_prop & ENC_UNICODE)) 5657 { 5658 /* Internal latin1 -> utf-8 conversion. */ 5659 vcp->vc_type = CONV_TO_UTF8; 5660 vcp->vc_factor = 2; /* up to twice as long */ 5661 } 5662 else if ((from_prop & ENC_LATIN9) && (to_prop & ENC_UNICODE)) 5663 { 5664 /* Internal latin9 -> utf-8 conversion. */ 5665 vcp->vc_type = CONV_9_TO_UTF8; 5666 vcp->vc_factor = 3; /* up to three as long (euro sign) */ 5667 } 5668 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_LATIN1)) 5669 { 5670 /* Internal utf-8 -> latin1 conversion. */ 5671 vcp->vc_type = CONV_TO_LATIN1; 5672 } 5673 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_LATIN9)) 5674 { 5675 /* Internal utf-8 -> latin9 conversion. */ 5676 vcp->vc_type = CONV_TO_LATIN9; 5677 } 5678 #ifdef WIN3264 5679 /* Win32-specific codepage <-> codepage conversion without iconv. */ 5680 else if (((from_prop & ENC_UNICODE) || encname2codepage(from) > 0) 5681 && ((to_prop & ENC_UNICODE) || encname2codepage(to) > 0)) 5682 { 5683 vcp->vc_type = CONV_CODEPAGE; 5684 vcp->vc_factor = 2; /* up to twice as long */ 5685 vcp->vc_cpfrom = (from_prop & ENC_UNICODE) ? 0 : encname2codepage(from); 5686 vcp->vc_cpto = (to_prop & ENC_UNICODE) ? 0 : encname2codepage(to); 5687 } 5688 #endif 5689 #ifdef MACOS_X 5690 else if ((from_prop & ENC_MACROMAN) && (to_prop & ENC_LATIN1)) 5691 { 5692 vcp->vc_type = CONV_MAC_LATIN1; 5693 } 5694 else if ((from_prop & ENC_MACROMAN) && (to_prop & ENC_UNICODE)) 5695 { 5696 vcp->vc_type = CONV_MAC_UTF8; 5697 vcp->vc_factor = 2; /* up to twice as long */ 5698 } 5699 else if ((from_prop & ENC_LATIN1) && (to_prop & ENC_MACROMAN)) 5700 { 5701 vcp->vc_type = CONV_LATIN1_MAC; 5702 } 5703 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_MACROMAN)) 5704 { 5705 vcp->vc_type = CONV_UTF8_MAC; 5706 } 5707 #endif 5708 # ifdef USE_ICONV 5709 else 5710 { 5711 /* Use iconv() for conversion. */ 5712 vcp->vc_fd = (iconv_t)my_iconv_open( 5713 (to_prop & ENC_UNICODE) ? (char_u *)"utf-8" : to, 5714 (from_prop & ENC_UNICODE) ? (char_u *)"utf-8" : from); 5715 if (vcp->vc_fd != (iconv_t)-1) 5716 { 5717 vcp->vc_type = CONV_ICONV; 5718 vcp->vc_factor = 4; /* could be longer too... */ 5719 } 5720 } 5721 # endif 5722 if (vcp->vc_type == CONV_NONE) 5723 return FAIL; 5724 5725 return OK; 5726 } 5727 5728 #if defined(FEAT_GUI) || defined(AMIGA) || defined(WIN3264) \ 5729 || defined(MSDOS) || defined(PROTO) 5730 /* 5731 * Do conversion on typed input characters in-place. 5732 * The input and output are not NUL terminated! 5733 * Returns the length after conversion. 5734 */ 5735 int 5736 convert_input(ptr, len, maxlen) 5737 char_u *ptr; 5738 int len; 5739 int maxlen; 5740 { 5741 return convert_input_safe(ptr, len, maxlen, NULL, NULL); 5742 } 5743 #endif 5744 5745 /* 5746 * Like convert_input(), but when there is an incomplete byte sequence at the 5747 * end return that as an allocated string in "restp" and set "*restlenp" to 5748 * the length. If "restp" is NULL it is not used. 5749 */ 5750 int 5751 convert_input_safe(ptr, len, maxlen, restp, restlenp) 5752 char_u *ptr; 5753 int len; 5754 int maxlen; 5755 char_u **restp; 5756 int *restlenp; 5757 { 5758 char_u *d; 5759 int dlen = len; 5760 int unconvertlen = 0; 5761 5762 d = string_convert_ext(&input_conv, ptr, &dlen, 5763 restp == NULL ? NULL : &unconvertlen); 5764 if (d != NULL) 5765 { 5766 if (dlen <= maxlen) 5767 { 5768 if (unconvertlen > 0) 5769 { 5770 /* Move the unconverted characters to allocated memory. */ 5771 *restp = alloc(unconvertlen); 5772 if (*restp != NULL) 5773 mch_memmove(*restp, ptr + len - unconvertlen, unconvertlen); 5774 *restlenp = unconvertlen; 5775 } 5776 mch_memmove(ptr, d, dlen); 5777 } 5778 else 5779 /* result is too long, keep the unconverted text (the caller must 5780 * have done something wrong!) */ 5781 dlen = len; 5782 vim_free(d); 5783 } 5784 return dlen; 5785 } 5786 5787 /* 5788 * Convert text "ptr[*lenp]" according to "vcp". 5789 * Returns the result in allocated memory and sets "*lenp". 5790 * When "lenp" is NULL, use NUL terminated strings. 5791 * Illegal chars are often changed to "?", unless vcp->vc_fail is set. 5792 * When something goes wrong, NULL is returned and "*lenp" is unchanged. 5793 */ 5794 char_u * 5795 string_convert(vcp, ptr, lenp) 5796 vimconv_T *vcp; 5797 char_u *ptr; 5798 int *lenp; 5799 { 5800 return string_convert_ext(vcp, ptr, lenp, NULL); 5801 } 5802 5803 /* 5804 * Like string_convert(), but when "unconvlenp" is not NULL and there are is 5805 * an incomplete sequence at the end it is not converted and "*unconvlenp" is 5806 * set to the number of remaining bytes. 5807 */ 5808 char_u * 5809 string_convert_ext(vcp, ptr, lenp, unconvlenp) 5810 vimconv_T *vcp; 5811 char_u *ptr; 5812 int *lenp; 5813 int *unconvlenp; 5814 { 5815 char_u *retval = NULL; 5816 char_u *d; 5817 int len; 5818 int i; 5819 int l; 5820 int c; 5821 5822 if (lenp == NULL) 5823 len = (int)STRLEN(ptr); 5824 else 5825 len = *lenp; 5826 if (len == 0) 5827 return vim_strsave((char_u *)""); 5828 5829 switch (vcp->vc_type) 5830 { 5831 case CONV_TO_UTF8: /* latin1 to utf-8 conversion */ 5832 retval = alloc(len * 2 + 1); 5833 if (retval == NULL) 5834 break; 5835 d = retval; 5836 for (i = 0; i < len; ++i) 5837 { 5838 c = ptr[i]; 5839 if (c < 0x80) 5840 *d++ = c; 5841 else 5842 { 5843 *d++ = 0xc0 + ((unsigned)c >> 6); 5844 *d++ = 0x80 + (c & 0x3f); 5845 } 5846 } 5847 *d = NUL; 5848 if (lenp != NULL) 5849 *lenp = (int)(d - retval); 5850 break; 5851 5852 case CONV_9_TO_UTF8: /* latin9 to utf-8 conversion */ 5853 retval = alloc(len * 3 + 1); 5854 if (retval == NULL) 5855 break; 5856 d = retval; 5857 for (i = 0; i < len; ++i) 5858 { 5859 c = ptr[i]; 5860 switch (c) 5861 { 5862 case 0xa4: c = 0x20ac; break; /* euro */ 5863 case 0xa6: c = 0x0160; break; /* S hat */ 5864 case 0xa8: c = 0x0161; break; /* S -hat */ 5865 case 0xb4: c = 0x017d; break; /* Z hat */ 5866 case 0xb8: c = 0x017e; break; /* Z -hat */ 5867 case 0xbc: c = 0x0152; break; /* OE */ 5868 case 0xbd: c = 0x0153; break; /* oe */ 5869 case 0xbe: c = 0x0178; break; /* Y */ 5870 } 5871 d += utf_char2bytes(c, d); 5872 } 5873 *d = NUL; 5874 if (lenp != NULL) 5875 *lenp = (int)(d - retval); 5876 break; 5877 5878 case CONV_TO_LATIN1: /* utf-8 to latin1 conversion */ 5879 case CONV_TO_LATIN9: /* utf-8 to latin9 conversion */ 5880 retval = alloc(len + 1); 5881 if (retval == NULL) 5882 break; 5883 d = retval; 5884 for (i = 0; i < len; ++i) 5885 { 5886 l = utf_ptr2len(ptr + i); 5887 if (l == 0) 5888 *d++ = NUL; 5889 else if (l == 1) 5890 { 5891 if (unconvlenp != NULL && utf8len_tab[ptr[i]] > len - i) 5892 { 5893 /* Incomplete sequence at the end. */ 5894 *unconvlenp = len - i; 5895 break; 5896 } 5897 *d++ = ptr[i]; 5898 } 5899 else 5900 { 5901 c = utf_ptr2char(ptr + i); 5902 if (vcp->vc_type == CONV_TO_LATIN9) 5903 switch (c) 5904 { 5905 case 0x20ac: c = 0xa4; break; /* euro */ 5906 case 0x0160: c = 0xa6; break; /* S hat */ 5907 case 0x0161: c = 0xa8; break; /* S -hat */ 5908 case 0x017d: c = 0xb4; break; /* Z hat */ 5909 case 0x017e: c = 0xb8; break; /* Z -hat */ 5910 case 0x0152: c = 0xbc; break; /* OE */ 5911 case 0x0153: c = 0xbd; break; /* oe */ 5912 case 0x0178: c = 0xbe; break; /* Y */ 5913 case 0xa4: 5914 case 0xa6: 5915 case 0xa8: 5916 case 0xb4: 5917 case 0xb8: 5918 case 0xbc: 5919 case 0xbd: 5920 case 0xbe: c = 0x100; break; /* not in latin9 */ 5921 } 5922 if (!utf_iscomposing(c)) /* skip composing chars */ 5923 { 5924 if (c < 0x100) 5925 *d++ = c; 5926 else if (vcp->vc_fail) 5927 { 5928 vim_free(retval); 5929 return NULL; 5930 } 5931 else 5932 { 5933 *d++ = 0xbf; 5934 if (utf_char2cells(c) > 1) 5935 *d++ = '?'; 5936 } 5937 } 5938 i += l - 1; 5939 } 5940 } 5941 *d = NUL; 5942 if (lenp != NULL) 5943 *lenp = (int)(d - retval); 5944 break; 5945 5946 # ifdef MACOS_X 5947 case CONV_MAC_LATIN1: 5948 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 5949 'm', 'l', unconvlenp); 5950 break; 5951 5952 case CONV_LATIN1_MAC: 5953 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 5954 'l', 'm', unconvlenp); 5955 break; 5956 5957 case CONV_MAC_UTF8: 5958 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 5959 'm', 'u', unconvlenp); 5960 break; 5961 5962 case CONV_UTF8_MAC: 5963 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 5964 'u', 'm', unconvlenp); 5965 break; 5966 # endif 5967 5968 # ifdef USE_ICONV 5969 case CONV_ICONV: /* conversion with output_conv.vc_fd */ 5970 retval = iconv_string(vcp, ptr, len, unconvlenp); 5971 if (retval != NULL && lenp != NULL) 5972 *lenp = (int)STRLEN(retval); 5973 break; 5974 # endif 5975 # ifdef WIN3264 5976 case CONV_CODEPAGE: /* codepage -> codepage */ 5977 { 5978 int retlen; 5979 int tmp_len; 5980 short_u *tmp; 5981 5982 /* 1. codepage/UTF-8 -> ucs-2. */ 5983 if (vcp->vc_cpfrom == 0) 5984 tmp_len = utf8_to_ucs2(ptr, len, NULL, NULL); 5985 else 5986 tmp_len = MultiByteToWideChar(vcp->vc_cpfrom, 0, 5987 ptr, len, 0, 0); 5988 tmp = (short_u *)alloc(sizeof(short_u) * tmp_len); 5989 if (tmp == NULL) 5990 break; 5991 if (vcp->vc_cpfrom == 0) 5992 utf8_to_ucs2(ptr, len, tmp, unconvlenp); 5993 else 5994 MultiByteToWideChar(vcp->vc_cpfrom, 0, ptr, len, tmp, tmp_len); 5995 5996 /* 2. ucs-2 -> codepage/UTF-8. */ 5997 if (vcp->vc_cpto == 0) 5998 retlen = ucs2_to_utf8(tmp, tmp_len, NULL); 5999 else 6000 retlen = WideCharToMultiByte(vcp->vc_cpto, 0, 6001 tmp, tmp_len, 0, 0, 0, 0); 6002 retval = alloc(retlen + 1); 6003 if (retval != NULL) 6004 { 6005 if (vcp->vc_cpto == 0) 6006 ucs2_to_utf8(tmp, tmp_len, retval); 6007 else 6008 WideCharToMultiByte(vcp->vc_cpto, 0, 6009 tmp, tmp_len, retval, retlen, 0, 0); 6010 retval[retlen] = NUL; 6011 if (lenp != NULL) 6012 *lenp = retlen; 6013 } 6014 vim_free(tmp); 6015 break; 6016 } 6017 # endif 6018 } 6019 6020 return retval; 6021 } 6022 #endif 6023