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_check __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 /* 588 * Set the function pointers. 589 */ 590 if (enc_utf8) 591 { 592 mb_ptr2len_check = utfc_ptr2len_check; 593 mb_char2len = utf_char2len; 594 mb_char2bytes = utf_char2bytes; 595 mb_ptr2cells = utf_ptr2cells; 596 mb_char2cells = utf_char2cells; 597 mb_off2cells = utf_off2cells; 598 mb_ptr2char = utf_ptr2char; 599 mb_head_off = utf_head_off; 600 } 601 else if (enc_dbcs != 0) 602 { 603 mb_ptr2len_check = dbcs_ptr2len_check; 604 mb_char2len = dbcs_char2len; 605 mb_char2bytes = dbcs_char2bytes; 606 mb_ptr2cells = dbcs_ptr2cells; 607 mb_char2cells = dbcs_char2cells; 608 mb_off2cells = dbcs_off2cells; 609 mb_ptr2char = dbcs_ptr2char; 610 mb_head_off = dbcs_head_off; 611 } 612 else 613 { 614 mb_ptr2len_check = latin_ptr2len_check; 615 mb_char2len = latin_char2len; 616 mb_char2bytes = latin_char2bytes; 617 mb_ptr2cells = latin_ptr2cells; 618 mb_char2cells = latin_char2cells; 619 mb_off2cells = latin_off2cells; 620 mb_ptr2char = latin_ptr2char; 621 mb_head_off = latin_head_off; 622 } 623 624 /* 625 * Fill the mb_bytelen_tab[] for MB_BYTE2LEN(). 626 */ 627 #ifdef LEN_FROM_CONV 628 /* When 'encoding' is different from the current locale mblen() won't 629 * work. Use conversion to "utf-8" instead. */ 630 vimconv.vc_type = CONV_NONE; 631 if (enc_dbcs) 632 { 633 p = enc_locale(); 634 if (p == NULL || STRCMP(p, p_enc) != 0) 635 { 636 convert_setup(&vimconv, p_enc, (char_u *)"utf-8"); 637 vimconv.vc_fail = TRUE; 638 } 639 vim_free(p); 640 } 641 #endif 642 643 for (i = 0; i < 256; ++i) 644 { 645 /* Our own function to reliably check the length of UTF-8 characters, 646 * independent of mblen(). */ 647 if (enc_utf8) 648 n = utf8len_tab[i]; 649 else if (enc_dbcs == 0) 650 n = 1; 651 else 652 { 653 #if defined(WIN3264) || defined(WIN32UNIX) 654 /* enc_dbcs is set by setting 'fileencoding'. It becomes a Windows 655 * CodePage identifier, which we can pass directly in to Windows 656 * API */ 657 n = IsDBCSLeadByteEx(enc_dbcs, (BYTE)i) ? 2 : 1; 658 #else 659 # ifdef MACOS 660 /* 661 * if mblen() is not available, character which MSB is turned on 662 * are treated as leading byte character. (note : This assumption 663 * is not always true.) 664 */ 665 n = (i & 0x80) ? 2 : 1; 666 # else 667 char buf[MB_MAXBYTES]; 668 # ifdef X_LOCALE 669 # ifndef mblen 670 # define mblen _Xmblen 671 # endif 672 # endif 673 if (i == NUL) /* just in case mblen() can't handle "" */ 674 n = 1; 675 else 676 { 677 buf[0] = i; 678 buf[1] = 0; 679 #ifdef LEN_FROM_CONV 680 if (vimconv.vc_type != CONV_NONE) 681 { 682 /* 683 * string_convert() should fail when converting the first 684 * byte of a double-byte character. 685 */ 686 p = string_convert(&vimconv, (char_u *)buf, NULL); 687 if (p != NULL) 688 { 689 vim_free(p); 690 n = 1; 691 } 692 else 693 n = 2; 694 } 695 else 696 #endif 697 { 698 /* 699 * mblen() should return -1 for invalid (means the leading 700 * multibyte) character. However there are some platforms 701 * where mblen() returns 0 for invalid character. 702 * Therefore, following condition includes 0. 703 */ 704 (void)mblen(NULL, 0); /* First reset the state. */ 705 if (mblen(buf, (size_t)1) <= 0) 706 n = 2; 707 else 708 n = 1; 709 } 710 } 711 # endif 712 #endif 713 } 714 715 mb_bytelen_tab[i] = n; 716 } 717 718 #ifdef LEN_FROM_CONV 719 convert_setup(&vimconv, NULL, NULL); 720 #endif 721 722 /* The cell width depends on the type of multi-byte characters. */ 723 (void)init_chartab(); 724 725 /* When enc_utf8 is set or reset, (de)allocate ScreenLinesUC[] */ 726 screenalloc(FALSE); 727 728 /* When using Unicode, set default for 'fileencodings'. */ 729 if (enc_utf8 && !option_was_set((char_u *)"fencs")) 730 set_string_option_direct((char_u *)"fencs", -1, 731 (char_u *)"ucs-bom,utf-8,default,latin1", OPT_FREE); 732 #if defined(HAVE_BIND_TEXTDOMAIN_CODESET) && defined(FEAT_GETTEXT) 733 /* GNU gettext 0.10.37 supports this feature: set the codeset used for 734 * translated messages independently from the current locale. */ 735 (void)bind_textdomain_codeset(VIMPACKAGE, 736 enc_utf8 ? "utf-8" : (char *)p_enc); 737 #endif 738 739 #ifdef WIN32 740 /* When changing 'encoding' while starting up, then convert the command 741 * line arguments from the active codepage to 'encoding'. */ 742 if (starting != 0) 743 fix_arg_enc(); 744 #endif 745 746 #ifdef FEAT_AUTOCMD 747 /* Fire an autocommand to let people do custom font setup. This must be 748 * after Vim has been setup for the new encoding. */ 749 apply_autocmds(EVENT_ENCODINGCHANGED, NULL, (char_u *)"", FALSE, curbuf); 750 #endif 751 752 #ifdef FEAT_GUI_KDE 753 if (gui.in_use) 754 gui_mch_update_codec(); 755 #endif 756 757 #ifdef FEAT_SYN_HL 758 /* Need to reload spell dictionaries */ 759 spell_reload(); 760 #endif 761 762 return NULL; 763 } 764 765 /* 766 * Return the size of the BOM for the current buffer: 767 * 0 - no BOM 768 * 2 - UCS-2 or UTF-16 BOM 769 * 4 - UCS-4 BOM 770 * 3 - UTF-8 BOM 771 */ 772 int 773 bomb_size() 774 { 775 int n = 0; 776 777 if (curbuf->b_p_bomb && !curbuf->b_p_bin) 778 { 779 if (*curbuf->b_p_fenc == NUL) 780 { 781 if (enc_utf8) 782 { 783 if (enc_unicode != 0) 784 n = enc_unicode; 785 else 786 n = 3; 787 } 788 } 789 else if (STRCMP(curbuf->b_p_fenc, "utf-8") == 0) 790 n = 3; 791 else if (STRNCMP(curbuf->b_p_fenc, "ucs-2", 5) == 0 792 || STRNCMP(curbuf->b_p_fenc, "utf-16", 6) == 0) 793 n = 2; 794 else if (STRNCMP(curbuf->b_p_fenc, "ucs-4", 5) == 0) 795 n = 4; 796 } 797 return n; 798 } 799 800 /* 801 * Get class of pointer: 802 * 0 for blank or NUL 803 * 1 for punctuation 804 * 2 for an (ASCII) word character 805 * >2 for other word characters 806 */ 807 int 808 mb_get_class(p) 809 char_u *p; 810 { 811 if (MB_BYTE2LEN(p[0]) == 1) 812 { 813 if (p[0] == NUL || vim_iswhite(p[0])) 814 return 0; 815 if (vim_iswordc(p[0])) 816 return 2; 817 return 1; 818 } 819 if (enc_dbcs != 0 && p[0] != NUL && p[1] != NUL) 820 return dbcs_class(p[0], p[1]); 821 if (enc_utf8) 822 return utf_class(utf_ptr2char(p)); 823 return 0; 824 } 825 826 /* 827 * Get class of a double-byte character. This always returns 3 or bigger. 828 * TODO: Should return 1 for punctuation. 829 */ 830 int 831 dbcs_class(lead, trail) 832 unsigned lead; 833 unsigned trail; 834 { 835 switch (enc_dbcs) 836 { 837 /* please add classfy routine for your language in here */ 838 839 case DBCS_JPNU: /* ? */ 840 case DBCS_JPN: 841 { 842 /* JIS code classification */ 843 unsigned char lb = lead; 844 unsigned char tb = trail; 845 846 /* convert process code to JIS */ 847 # if defined(WIN3264) || defined(WIN32UNIX) || defined(MACOS) 848 /* process code is SJIS */ 849 if (lb <= 0x9f) 850 lb = (lb - 0x81) * 2 + 0x21; 851 else 852 lb = (lb - 0xc1) * 2 + 0x21; 853 if (tb <= 0x7e) 854 tb -= 0x1f; 855 else if (tb <= 0x9e) 856 tb -= 0x20; 857 else 858 { 859 tb -= 0x7e; 860 lb += 1; 861 } 862 # else 863 /* 864 * XXX: Code page identification can not use with all 865 * system! So, some other encoding information 866 * will be needed. 867 * In japanese: SJIS,EUC,UNICODE,(JIS) 868 * Note that JIS-code system don't use as 869 * process code in most system because it uses 870 * escape sequences(JIS is context depend encoding). 871 */ 872 /* assume process code is JAPANESE-EUC */ 873 lb &= 0x7f; 874 tb &= 0x7f; 875 # endif 876 /* exceptions */ 877 switch (lb << 8 | tb) 878 { 879 case 0x2121: /* ZENKAKU space */ 880 return 0; 881 case 0x2122: /* KU-TEN (Japanese comma) */ 882 case 0x2123: /* TOU-TEN (Japanese period) */ 883 case 0x2124: /* ZENKAKU comma */ 884 case 0x2125: /* ZENKAKU period */ 885 return 1; 886 case 0x213c: /* prolongedsound handled as KATAKANA */ 887 return 13; 888 } 889 /* sieved by KU code */ 890 switch (lb) 891 { 892 case 0x21: 893 case 0x22: 894 /* special symbols */ 895 return 10; 896 case 0x23: 897 /* alpha-numeric */ 898 return 11; 899 case 0x24: 900 /* hiragana */ 901 return 12; 902 case 0x25: 903 /* katakana */ 904 return 13; 905 case 0x26: 906 /* greek */ 907 return 14; 908 case 0x27: 909 /* russian */ 910 return 15; 911 case 0x28: 912 /* lines */ 913 return 16; 914 default: 915 /* kanji */ 916 return 17; 917 } 918 } 919 920 case DBCS_KORU: /* ? */ 921 case DBCS_KOR: 922 { 923 /* KS code classification */ 924 unsigned char c1 = lead; 925 unsigned char c2 = trail; 926 927 /* 928 * 20 : Hangul 929 * 21 : Hanja 930 * 22 : Symbols 931 * 23 : Alpha-numeric/Roman Letter (Full width) 932 * 24 : Hangul Letter(Alphabet) 933 * 25 : Roman Numeral/Greek Letter 934 * 26 : Box Drawings 935 * 27 : Unit Symbols 936 * 28 : Circled/Parenthesized Letter 937 * 29 : Hirigana/Katakana 938 * 30 : Cyrillic Letter 939 */ 940 941 if (c1 >= 0xB0 && c1 <= 0xC8) 942 /* Hangul */ 943 return 20; 944 #if defined(WIN3264) || defined(WIN32UNIX) 945 else if (c1 <= 0xA0 || c2 <= 0xA0) 946 /* Extended Hangul Region : MS UHC(Unified Hangul Code) */ 947 /* c1: 0x81-0xA0 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xFE 948 * c1: 0xA1-0xC6 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xA0 949 */ 950 return 20; 951 #endif 952 953 else if (c1 >= 0xCA && c1 <= 0xFD) 954 /* Hanja */ 955 return 21; 956 else switch (c1) 957 { 958 case 0xA1: 959 case 0xA2: 960 /* Symbols */ 961 return 22; 962 case 0xA3: 963 /* Alpha-numeric */ 964 return 23; 965 case 0xA4: 966 /* Hangul Letter(Alphabet) */ 967 return 24; 968 case 0xA5: 969 /* Roman Numeral/Greek Letter */ 970 return 25; 971 case 0xA6: 972 /* Box Drawings */ 973 return 26; 974 case 0xA7: 975 /* Unit Symbols */ 976 return 27; 977 case 0xA8: 978 case 0xA9: 979 if (c2 <= 0xAF) 980 return 25; /* Roman Letter */ 981 else if (c2 >= 0xF6) 982 return 22; /* Symbols */ 983 else 984 /* Circled/Parenthesized Letter */ 985 return 28; 986 case 0xAA: 987 case 0xAB: 988 /* Hirigana/Katakana */ 989 return 29; 990 case 0xAC: 991 /* Cyrillic Letter */ 992 return 30; 993 } 994 } 995 default: 996 break; 997 } 998 return 3; 999 } 1000 1001 /* 1002 * mb_char2len() function pointer. 1003 * Return length in bytes of character "c". 1004 * Returns 1 for a single-byte character. 1005 */ 1006 /* ARGSUSED */ 1007 int 1008 latin_char2len(c) 1009 int c; 1010 { 1011 return 1; 1012 } 1013 1014 static int 1015 dbcs_char2len(c) 1016 int c; 1017 { 1018 if (c >= 0x100) 1019 return 2; 1020 return 1; 1021 } 1022 1023 /* 1024 * mb_char2bytes() function pointer. 1025 * Convert a character to its bytes. 1026 * Returns the length in bytes. 1027 */ 1028 int 1029 latin_char2bytes(c, buf) 1030 int c; 1031 char_u *buf; 1032 { 1033 buf[0] = c; 1034 return 1; 1035 } 1036 1037 static int 1038 dbcs_char2bytes(c, buf) 1039 int c; 1040 char_u *buf; 1041 { 1042 if (c >= 0x100) 1043 { 1044 buf[0] = (unsigned)c >> 8; 1045 buf[1] = c; 1046 /* Never use a NUL byte, it causes lots of trouble. It's an invalid 1047 * character anyway. */ 1048 if (buf[1] == NUL) 1049 buf[1] = '\n'; 1050 return 2; 1051 } 1052 buf[0] = c; 1053 return 1; 1054 } 1055 1056 /* 1057 * mb_ptr2len_check() function pointer. 1058 * Get byte length of character at "*p" but stop at a NUL. 1059 * For UTF-8 this includes following composing characters. 1060 * Returns 0 when *p is NUL. 1061 * 1062 */ 1063 int 1064 latin_ptr2len_check(p) 1065 char_u *p; 1066 { 1067 return MB_BYTE2LEN(*p); 1068 } 1069 1070 static int 1071 dbcs_ptr2len_check(p) 1072 char_u *p; 1073 { 1074 int len; 1075 1076 /* Check if second byte is not missing. */ 1077 len = MB_BYTE2LEN(*p); 1078 if (len == 2 && p[1] == NUL) 1079 len = 1; 1080 return len; 1081 } 1082 1083 struct interval 1084 { 1085 unsigned short first; 1086 unsigned short last; 1087 }; 1088 static int intable __ARGS((struct interval *table, size_t size, int c)); 1089 1090 /* 1091 * Return TRUE if "c" is in "table[size / sizeof(struct interval)]". 1092 */ 1093 static int 1094 intable(table, size, c) 1095 struct interval *table; 1096 size_t size; 1097 int c; 1098 { 1099 int mid, bot, top; 1100 1101 /* first quick check for Latin1 etc. characters */ 1102 if (c < table[0].first) 1103 return FALSE; 1104 1105 /* binary search in table */ 1106 bot = 0; 1107 top = size / sizeof(struct interval) - 1; 1108 while (top >= bot) 1109 { 1110 mid = (bot + top) / 2; 1111 if (table[mid].last < c) 1112 bot = mid + 1; 1113 else if (table[mid].first > c) 1114 top = mid - 1; 1115 else 1116 return TRUE; 1117 } 1118 return FALSE; 1119 } 1120 1121 /* 1122 * For UTF-8 character "c" return 2 for a double-width character, 1 for others. 1123 * Returns 4 or 6 for an unprintable character. 1124 * Is only correct for characters >= 0x80. 1125 * When p_ambw is "double", return 2 for a character with East Asian Width 1126 * class 'A'(mbiguous). 1127 */ 1128 int 1129 utf_char2cells(c) 1130 int c; 1131 { 1132 /* sorted list of non-overlapping intervals of East Asian Ambiguous 1133 * characters, generated with: 1134 * "uniset +WIDTH-A -cat=Me -cat=Mn -cat=Cf c" */ 1135 static struct interval ambiguous[] = { 1136 {0x00A1, 0x00A1}, {0x00A4, 0x00A4}, {0x00A7, 0x00A8}, 1137 {0x00AA, 0x00AA}, {0x00AE, 0x00AE}, {0x00B0, 0x00B4}, 1138 {0x00B6, 0x00BA}, {0x00BC, 0x00BF}, {0x00C6, 0x00C6}, 1139 {0x00D0, 0x00D0}, {0x00D7, 0x00D8}, {0x00DE, 0x00E1}, 1140 {0x00E6, 0x00E6}, {0x00E8, 0x00EA}, {0x00EC, 0x00ED}, 1141 {0x00F0, 0x00F0}, {0x00F2, 0x00F3}, {0x00F7, 0x00FA}, 1142 {0x00FC, 0x00FC}, {0x00FE, 0x00FE}, {0x0101, 0x0101}, 1143 {0x0111, 0x0111}, {0x0113, 0x0113}, {0x011B, 0x011B}, 1144 {0x0126, 0x0127}, {0x012B, 0x012B}, {0x0131, 0x0133}, 1145 {0x0138, 0x0138}, {0x013F, 0x0142}, {0x0144, 0x0144}, 1146 {0x0148, 0x014B}, {0x014D, 0x014D}, {0x0152, 0x0153}, 1147 {0x0166, 0x0167}, {0x016B, 0x016B}, {0x01CE, 0x01CE}, 1148 {0x01D0, 0x01D0}, {0x01D2, 0x01D2}, {0x01D4, 0x01D4}, 1149 {0x01D6, 0x01D6}, {0x01D8, 0x01D8}, {0x01DA, 0x01DA}, 1150 {0x01DC, 0x01DC}, {0x0251, 0x0251}, {0x0261, 0x0261}, 1151 {0x02C4, 0x02C4}, {0x02C7, 0x02C7}, {0x02C9, 0x02CB}, 1152 {0x02CD, 0x02CD}, {0x02D0, 0x02D0}, {0x02D8, 0x02DB}, 1153 {0x02DD, 0x02DD}, {0x02DF, 0x02DF}, {0x0391, 0x03A1}, 1154 {0x03A3, 0x03A9}, {0x03B1, 0x03C1}, {0x03C3, 0x03C9}, 1155 {0x0401, 0x0401}, {0x0410, 0x044F}, {0x0451, 0x0451}, 1156 {0x2010, 0x2010}, {0x2013, 0x2016}, {0x2018, 0x2019}, 1157 {0x201C, 0x201D}, {0x2020, 0x2022}, {0x2024, 0x2027}, 1158 {0x2030, 0x2030}, {0x2032, 0x2033}, {0x2035, 0x2035}, 1159 {0x203B, 0x203B}, {0x203E, 0x203E}, {0x2074, 0x2074}, 1160 {0x207F, 0x207F}, {0x2081, 0x2084}, {0x20AC, 0x20AC}, 1161 {0x2103, 0x2103}, {0x2105, 0x2105}, {0x2109, 0x2109}, 1162 {0x2113, 0x2113}, {0x2116, 0x2116}, {0x2121, 0x2122}, 1163 {0x2126, 0x2126}, {0x212B, 0x212B}, {0x2153, 0x2154}, 1164 {0x215B, 0x215E}, {0x2160, 0x216B}, {0x2170, 0x2179}, 1165 {0x2190, 0x2199}, {0x21B8, 0x21B9}, {0x21D2, 0x21D2}, 1166 {0x21D4, 0x21D4}, {0x21E7, 0x21E7}, {0x2200, 0x2200}, 1167 {0x2202, 0x2203}, {0x2207, 0x2208}, {0x220B, 0x220B}, 1168 {0x220F, 0x220F}, {0x2211, 0x2211}, {0x2215, 0x2215}, 1169 {0x221A, 0x221A}, {0x221D, 0x2220}, {0x2223, 0x2223}, 1170 {0x2225, 0x2225}, {0x2227, 0x222C}, {0x222E, 0x222E}, 1171 {0x2234, 0x2237}, {0x223C, 0x223D}, {0x2248, 0x2248}, 1172 {0x224C, 0x224C}, {0x2252, 0x2252}, {0x2260, 0x2261}, 1173 {0x2264, 0x2267}, {0x226A, 0x226B}, {0x226E, 0x226F}, 1174 {0x2282, 0x2283}, {0x2286, 0x2287}, {0x2295, 0x2295}, 1175 {0x2299, 0x2299}, {0x22A5, 0x22A5}, {0x22BF, 0x22BF}, 1176 {0x2312, 0x2312}, {0x2460, 0x24E9}, {0x24EB, 0x254B}, 1177 {0x2550, 0x2573}, {0x2580, 0x258F}, {0x2592, 0x2595}, 1178 {0x25A0, 0x25A1}, {0x25A3, 0x25A9}, {0x25B2, 0x25B3}, 1179 {0x25B6, 0x25B7}, {0x25BC, 0x25BD}, {0x25C0, 0x25C1}, 1180 {0x25C6, 0x25C8}, {0x25CB, 0x25CB}, {0x25CE, 0x25D1}, 1181 {0x25E2, 0x25E5}, {0x25EF, 0x25EF}, {0x2605, 0x2606}, 1182 {0x2609, 0x2609}, {0x260E, 0x260F}, {0x2614, 0x2615}, 1183 {0x261C, 0x261C}, {0x261E, 0x261E}, {0x2640, 0x2640}, 1184 {0x2642, 0x2642}, {0x2660, 0x2661}, {0x2663, 0x2665}, 1185 {0x2667, 0x266A}, {0x266C, 0x266D}, {0x266F, 0x266F}, 1186 {0x273D, 0x273D}, {0x2776, 0x277F}, {0xE000, 0xF8FF}, 1187 {0xFFFD, 0xFFFD}, /* {0xF0000, 0xFFFFD}, {0x100000, 0x10FFFD} */ 1188 }; 1189 1190 if (c >= 0x100) 1191 { 1192 #ifdef USE_WCHAR_FUNCTIONS 1193 /* 1194 * Assume the library function wcwidth() works better than our own 1195 * stuff. It should return 1 for ambiguous width chars! 1196 */ 1197 int n = wcwidth(c); 1198 1199 if (n < 0) 1200 return 6; /* unprintable, displays <xxxx> */ 1201 if (n > 1) 1202 return n; 1203 #else 1204 if (!utf_printable(c)) 1205 return 6; /* unprintable, displays <xxxx> */ 1206 if (c >= 0x1100 1207 && (c <= 0x115f /* Hangul Jamo */ 1208 || c == 0x2329 1209 || c == 0x232a 1210 || (c >= 0x2e80 && c <= 0xa4cf 1211 && c != 0x303f) /* CJK ... Yi */ 1212 || (c >= 0xac00 && c <= 0xd7a3) /* Hangul Syllables */ 1213 || (c >= 0xf900 && c <= 0xfaff) /* CJK Compatibility 1214 Ideographs */ 1215 || (c >= 0xfe30 && c <= 0xfe6f) /* CJK Compatibility Forms */ 1216 || (c >= 0xff00 && c <= 0xff60) /* Fullwidth Forms */ 1217 || (c >= 0xffe0 && c <= 0xffe6) 1218 || (c >= 0x20000 && c <= 0x2fffd) 1219 || (c >= 0x30000 && c <= 0x3fffd))) 1220 return 2; 1221 #endif 1222 } 1223 1224 /* Characters below 0x100 are influenced by 'isprint' option */ 1225 else if (c >= 0x80 && !vim_isprintc(c)) 1226 return 4; /* unprintable, displays <xx> */ 1227 1228 if (c >= 0x80 && *p_ambw == 'd' && intable(ambiguous, sizeof(ambiguous), c)) 1229 return 2; 1230 1231 return 1; 1232 } 1233 1234 /* 1235 * mb_ptr2cells() function pointer. 1236 * Return the number of display cells character at "*p" occupies. 1237 * This doesn't take care of unprintable characters, use ptr2cells() for that. 1238 */ 1239 /*ARGSUSED*/ 1240 int 1241 latin_ptr2cells(p) 1242 char_u *p; 1243 { 1244 return 1; 1245 } 1246 1247 int 1248 utf_ptr2cells(p) 1249 char_u *p; 1250 { 1251 int c; 1252 1253 /* Need to convert to a wide character. */ 1254 if (*p >= 0x80) 1255 { 1256 c = utf_ptr2char(p); 1257 /* An illegal byte is displayed as <xx>. */ 1258 if (utf_ptr2len_check(p) == 1 || c == NUL) 1259 return 4; 1260 /* If the char is ASCII it must be an overlong sequence. */ 1261 if (c < 0x80) 1262 return char2cells(c); 1263 return utf_char2cells(c); 1264 } 1265 return 1; 1266 } 1267 1268 int 1269 dbcs_ptr2cells(p) 1270 char_u *p; 1271 { 1272 /* Number of cells is equal to number of bytes, except for euc-jp when 1273 * the first byte is 0x8e. */ 1274 if (enc_dbcs == DBCS_JPNU && *p == 0x8e) 1275 return 1; 1276 return MB_BYTE2LEN(*p); 1277 } 1278 1279 /* 1280 * mb_char2cells() function pointer. 1281 * Return the number of display cells character "c" occupies. 1282 * Only takes care of multi-byte chars, not "^C" and such. 1283 */ 1284 /*ARGSUSED*/ 1285 int 1286 latin_char2cells(c) 1287 int c; 1288 { 1289 return 1; 1290 } 1291 1292 static int 1293 dbcs_char2cells(c) 1294 int c; 1295 { 1296 /* Number of cells is equal to number of bytes, except for euc-jp when 1297 * the first byte is 0x8e. */ 1298 if (enc_dbcs == DBCS_JPNU && ((unsigned)c >> 8) == 0x8e) 1299 return 1; 1300 /* use the first byte */ 1301 return MB_BYTE2LEN((unsigned)c >> 8); 1302 } 1303 1304 /* 1305 * mb_off2cells() function pointer. 1306 * Return number of display cells for char at ScreenLines[off]. 1307 * Caller must make sure "off" and "off + 1" are valid! 1308 */ 1309 /*ARGSUSED*/ 1310 int 1311 latin_off2cells(off) 1312 unsigned off; 1313 { 1314 return 1; 1315 } 1316 1317 int 1318 dbcs_off2cells(off) 1319 unsigned off; 1320 { 1321 /* Number of cells is equal to number of bytes, except for euc-jp when 1322 * the first byte is 0x8e. */ 1323 if (enc_dbcs == DBCS_JPNU && ScreenLines[off] == 0x8e) 1324 return 1; 1325 return MB_BYTE2LEN(ScreenLines[off]); 1326 } 1327 1328 int 1329 utf_off2cells(off) 1330 unsigned off; 1331 { 1332 return ScreenLines[off + 1] == 0 ? 2 : 1; 1333 } 1334 1335 /* 1336 * mb_ptr2char() function pointer. 1337 * Convert a byte sequence into a character. 1338 */ 1339 int 1340 latin_ptr2char(p) 1341 char_u *p; 1342 { 1343 return *p; 1344 } 1345 1346 static int 1347 dbcs_ptr2char(p) 1348 char_u *p; 1349 { 1350 if (MB_BYTE2LEN(*p) > 1 && p[1] != NUL) 1351 return (p[0] << 8) + p[1]; 1352 return *p; 1353 } 1354 1355 /* 1356 * Convert a UTF-8 byte sequence to a wide character. 1357 * If the sequence is illegal or truncated by a NUL the first byte is 1358 * returned. 1359 * Does not include composing characters, of course. 1360 */ 1361 int 1362 utf_ptr2char(p) 1363 char_u *p; 1364 { 1365 int len; 1366 1367 if (p[0] < 0x80) /* be quick for ASCII */ 1368 return p[0]; 1369 1370 len = utf8len_tab[p[0]]; 1371 if ((p[1] & 0xc0) == 0x80) 1372 { 1373 if (len == 2) 1374 return ((p[0] & 0x1f) << 6) + (p[1] & 0x3f); 1375 if ((p[2] & 0xc0) == 0x80) 1376 { 1377 if (len == 3) 1378 return ((p[0] & 0x0f) << 12) + ((p[1] & 0x3f) << 6) 1379 + (p[2] & 0x3f); 1380 if ((p[3] & 0xc0) == 0x80) 1381 { 1382 if (len == 4) 1383 return ((p[0] & 0x07) << 18) + ((p[1] & 0x3f) << 12) 1384 + ((p[2] & 0x3f) << 6) + (p[3] & 0x3f); 1385 if ((p[4] & 0xc0) == 0x80) 1386 { 1387 if (len == 5) 1388 return ((p[0] & 0x03) << 24) + ((p[1] & 0x3f) << 18) 1389 + ((p[2] & 0x3f) << 12) + ((p[3] & 0x3f) << 6) 1390 + (p[4] & 0x3f); 1391 if ((p[5] & 0xc0) == 0x80 && len == 6) 1392 return ((p[0] & 0x01) << 30) + ((p[1] & 0x3f) << 24) 1393 + ((p[2] & 0x3f) << 18) + ((p[3] & 0x3f) << 12) 1394 + ((p[4] & 0x3f) << 6) + (p[5] & 0x3f); 1395 } 1396 } 1397 } 1398 } 1399 /* Illegal value, just return the first byte */ 1400 return p[0]; 1401 } 1402 1403 /* 1404 * Get character at **pp and advance *pp to the next character. 1405 * Note: composing characters are skipped! 1406 */ 1407 int 1408 mb_ptr2char_adv(pp) 1409 char_u **pp; 1410 { 1411 int c; 1412 1413 c = (*mb_ptr2char)(*pp); 1414 *pp += (*mb_ptr2len_check)(*pp); 1415 return c; 1416 } 1417 1418 #if defined(FEAT_ARABIC) || defined(PROTO) 1419 /* 1420 * Check whether we are dealing with Arabic combining characters. 1421 * Note: these are NOT really composing characters! 1422 */ 1423 int 1424 arabic_combine(one, two) 1425 int one; /* first character */ 1426 int two; /* character just after "one" */ 1427 { 1428 if (one == a_LAM) 1429 return arabic_maycombine(two); 1430 return FALSE; 1431 } 1432 1433 /* 1434 * Check whether we are dealing with a character that could be regarded as an 1435 * Arabic combining character, need to check the character before this. 1436 */ 1437 int 1438 arabic_maycombine(two) 1439 int two; 1440 { 1441 if (p_arshape && !p_tbidi) 1442 return (two == a_ALEF_MADDA 1443 || two == a_ALEF_HAMZA_ABOVE 1444 || two == a_ALEF_HAMZA_BELOW 1445 || two == a_ALEF); 1446 return FALSE; 1447 } 1448 1449 /* 1450 * Check if the character pointed to by "p2" is a composing character when it 1451 * comes after "p1". For Arabic sometimes "ab" is replaced with "c", which 1452 * behaves like a composing character. 1453 */ 1454 int 1455 utf_composinglike(p1, p2) 1456 char_u *p1; 1457 char_u *p2; 1458 { 1459 int c2; 1460 1461 c2 = utf_ptr2char(p2); 1462 if (utf_iscomposing(c2)) 1463 return TRUE; 1464 if (!arabic_maycombine(c2)) 1465 return FALSE; 1466 return arabic_combine(utf_ptr2char(p1), c2); 1467 } 1468 #endif 1469 1470 /* 1471 * Convert a UTF-8 byte string to a wide chararacter. Also get up to two 1472 * composing characters. 1473 */ 1474 int 1475 utfc_ptr2char(p, p1, p2) 1476 char_u *p; 1477 int *p1; /* return: first composing char or 0 */ 1478 int *p2; /* return: second composing char or 0 */ 1479 { 1480 int len; 1481 int c; 1482 int cc; 1483 1484 c = utf_ptr2char(p); 1485 len = utf_ptr2len_check(p); 1486 /* Only accept a composing char when the first char isn't illegal. */ 1487 if ((len > 1 || *p < 0x80) 1488 && p[len] >= 0x80 1489 && UTF_COMPOSINGLIKE(p, p + len)) 1490 { 1491 *p1 = utf_ptr2char(p + len); 1492 len += utf_ptr2len_check(p + len); 1493 if (p[len] >= 0x80 && utf_iscomposing(cc = utf_ptr2char(p + len))) 1494 *p2 = cc; 1495 else 1496 *p2 = 0; 1497 } 1498 else 1499 { 1500 *p1 = 0; 1501 *p2 = 0; 1502 } 1503 return c; 1504 } 1505 1506 /* 1507 * Convert a UTF-8 byte string to a wide chararacter. Also get up to two 1508 * composing characters. Use no more than p[maxlen]. 1509 */ 1510 int 1511 utfc_ptr2char_len(p, p1, p2, maxlen) 1512 char_u *p; 1513 int *p1; /* return: first composing char or 0 */ 1514 int *p2; /* return: second composing char or 0 */ 1515 int maxlen; 1516 { 1517 int len; 1518 int c; 1519 int cc; 1520 1521 c = utf_ptr2char(p); 1522 len = utf_ptr2len_check_len(p, maxlen); 1523 /* Only accept a composing char when the first char isn't illegal. */ 1524 if ((len > 1 || *p < 0x80) 1525 && len < maxlen 1526 && p[len] >= 0x80 1527 && UTF_COMPOSINGLIKE(p, p + len)) 1528 { 1529 *p1 = utf_ptr2char(p + len); 1530 len += utf_ptr2len_check_len(p + len, maxlen - len); 1531 if (len < maxlen 1532 && p[len] >= 0x80 1533 && utf_iscomposing(cc = utf_ptr2char(p + len))) 1534 *p2 = cc; 1535 else 1536 *p2 = 0; 1537 } 1538 else 1539 { 1540 *p1 = 0; 1541 *p2 = 0; 1542 } 1543 return c; 1544 } 1545 1546 /* 1547 * Convert the character at screen position "off" to a sequence of bytes. 1548 * Includes the composing characters. 1549 * "buf" must at least have the length MB_MAXBYTES. 1550 * Returns the produced number of bytes. 1551 */ 1552 int 1553 utfc_char2bytes(off, buf) 1554 int off; 1555 char_u *buf; 1556 { 1557 int len; 1558 1559 len = utf_char2bytes(ScreenLinesUC[off], buf); 1560 if (ScreenLinesC1[off] != 0) 1561 { 1562 len += utf_char2bytes(ScreenLinesC1[off], buf + len); 1563 if (ScreenLinesC2[off] != 0) 1564 len += utf_char2bytes(ScreenLinesC2[off], buf + len); 1565 } 1566 return len; 1567 } 1568 1569 /* 1570 * Get the length of a UTF-8 byte sequence, not including any following 1571 * composing characters. 1572 * Returns 0 for "". 1573 * Returns 1 for an illegal byte sequence. 1574 */ 1575 int 1576 utf_ptr2len_check(p) 1577 char_u *p; 1578 { 1579 int len; 1580 int i; 1581 1582 if (*p == NUL) 1583 return 0; 1584 len = utf8len_tab[*p]; 1585 for (i = 1; i < len; ++i) 1586 if ((p[i] & 0xc0) != 0x80) 1587 return 1; 1588 return len; 1589 } 1590 1591 /* 1592 * Return length of UTF-8 character, obtained from the first byte. 1593 * "b" must be between 0 and 255! 1594 */ 1595 int 1596 utf_byte2len(b) 1597 int b; 1598 { 1599 return utf8len_tab[b]; 1600 } 1601 1602 /* 1603 * Get the length of UTF-8 byte sequence "p[size]". Does not include any 1604 * following composing characters. 1605 * Returns 1 for "". 1606 * Returns 1 for an illegal byte sequence. 1607 * Returns number > "size" for an incomplete byte sequence. 1608 */ 1609 int 1610 utf_ptr2len_check_len(p, size) 1611 char_u *p; 1612 int size; 1613 { 1614 int len; 1615 int i; 1616 1617 if (*p == NUL) 1618 return 1; 1619 len = utf8len_tab[*p]; 1620 if (len > size) 1621 return len; /* incomplete byte sequence. */ 1622 for (i = 1; i < len; ++i) 1623 if ((p[i] & 0xc0) != 0x80) 1624 return 1; 1625 return len; 1626 } 1627 1628 /* 1629 * Return the number of bytes the UTF-8 encoding of the character at "p" takes. 1630 * This includes following composing characters. 1631 */ 1632 int 1633 utfc_ptr2len_check(p) 1634 char_u *p; 1635 { 1636 int len; 1637 int b0 = *p; 1638 #ifdef FEAT_ARABIC 1639 int prevlen; 1640 #endif 1641 1642 if (b0 == NUL) 1643 return 0; 1644 if (b0 < 0x80 && p[1] < 0x80) /* be quick for ASCII */ 1645 return 1; 1646 1647 /* Skip over first UTF-8 char, stopping at a NUL byte. */ 1648 len = utf_ptr2len_check(p); 1649 1650 /* Check for illegal byte. */ 1651 if (len == 1 && b0 >= 0x80) 1652 return 1; 1653 1654 /* 1655 * Check for composing characters. We can handle only the first two, but 1656 * skip all of them (otherwise the cursor would get stuck). 1657 */ 1658 #ifdef FEAT_ARABIC 1659 prevlen = 0; 1660 #endif 1661 for (;;) 1662 { 1663 if (p[len] < 0x80 || !UTF_COMPOSINGLIKE(p + prevlen, p + len)) 1664 return len; 1665 1666 /* Skip over composing char */ 1667 #ifdef FEAT_ARABIC 1668 prevlen = len; 1669 #endif 1670 len += utf_ptr2len_check(p + len); 1671 } 1672 } 1673 1674 /* 1675 * Return the number of bytes the UTF-8 encoding of the character at "p[size]" 1676 * takes. This includes following composing characters. 1677 * Returns 1 for an illegal char or an incomplete byte sequence. 1678 */ 1679 int 1680 utfc_ptr2len_check_len(p, size) 1681 char_u *p; 1682 int size; 1683 { 1684 int len; 1685 #ifdef FEAT_ARABIC 1686 int prevlen; 1687 #endif 1688 1689 if (*p == NUL) 1690 return 0; 1691 if (p[0] < 0x80 && (size == 1 || p[1] < 0x80)) /* be quick for ASCII */ 1692 return 1; 1693 1694 /* Skip over first UTF-8 char, stopping at a NUL byte. */ 1695 len = utf_ptr2len_check_len(p, size); 1696 1697 /* Check for illegal byte and incomplete byte sequence. */ 1698 if ((len == 1 && p[0] >= 0x80) || len > size) 1699 return 1; 1700 1701 /* 1702 * Check for composing characters. We can handle only the first two, but 1703 * skip all of them (otherwise the cursor would get stuck). 1704 */ 1705 #ifdef FEAT_ARABIC 1706 prevlen = 0; 1707 #endif 1708 while (len < size) 1709 { 1710 if (p[len] < 0x80 || !UTF_COMPOSINGLIKE(p + prevlen, p + len)) 1711 break; 1712 1713 /* Skip over composing char */ 1714 #ifdef FEAT_ARABIC 1715 prevlen = len; 1716 #endif 1717 len += utf_ptr2len_check_len(p + len, size - len); 1718 } 1719 return len; 1720 } 1721 1722 /* 1723 * Return the number of bytes the UTF-8 encoding of character "c" takes. 1724 * This does not include composing characters. 1725 */ 1726 int 1727 utf_char2len(c) 1728 int c; 1729 { 1730 if (c < 0x80) 1731 return 1; 1732 if (c < 0x800) 1733 return 2; 1734 if (c < 0x10000) 1735 return 3; 1736 if (c < 0x200000) 1737 return 4; 1738 if (c < 0x4000000) 1739 return 5; 1740 return 6; 1741 } 1742 1743 /* 1744 * Convert Unicode character "c" to UTF-8 string in "buf[]". 1745 * Returns the number of bytes. 1746 * This does not include composing characters. 1747 */ 1748 int 1749 utf_char2bytes(c, buf) 1750 int c; 1751 char_u *buf; 1752 { 1753 if (c < 0x80) /* 7 bits */ 1754 { 1755 buf[0] = c; 1756 return 1; 1757 } 1758 if (c < 0x800) /* 11 bits */ 1759 { 1760 buf[0] = 0xc0 + ((unsigned)c >> 6); 1761 buf[1] = 0x80 + (c & 0x3f); 1762 return 2; 1763 } 1764 if (c < 0x10000) /* 16 bits */ 1765 { 1766 buf[0] = 0xe0 + ((unsigned)c >> 12); 1767 buf[1] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1768 buf[2] = 0x80 + (c & 0x3f); 1769 return 3; 1770 } 1771 if (c < 0x200000) /* 21 bits */ 1772 { 1773 buf[0] = 0xf0 + ((unsigned)c >> 18); 1774 buf[1] = 0x80 + (((unsigned)c >> 12) & 0x3f); 1775 buf[2] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1776 buf[3] = 0x80 + (c & 0x3f); 1777 return 4; 1778 } 1779 if (c < 0x4000000) /* 26 bits */ 1780 { 1781 buf[0] = 0xf8 + ((unsigned)c >> 24); 1782 buf[1] = 0x80 + (((unsigned)c >> 18) & 0x3f); 1783 buf[2] = 0x80 + (((unsigned)c >> 12) & 0x3f); 1784 buf[3] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1785 buf[4] = 0x80 + (c & 0x3f); 1786 return 5; 1787 } 1788 /* 31 bits */ 1789 buf[0] = 0xfc + ((unsigned)c >> 30); 1790 buf[1] = 0x80 + (((unsigned)c >> 24) & 0x3f); 1791 buf[2] = 0x80 + (((unsigned)c >> 18) & 0x3f); 1792 buf[3] = 0x80 + (((unsigned)c >> 12) & 0x3f); 1793 buf[4] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1794 buf[5] = 0x80 + (c & 0x3f); 1795 return 6; 1796 } 1797 1798 /* 1799 * Return TRUE if "c" is a composing UTF-8 character. This means it will be 1800 * drawn on top of the preceding character. 1801 * Based on code from Markus Kuhn. 1802 */ 1803 int 1804 utf_iscomposing(c) 1805 int c; 1806 { 1807 /* sorted list of non-overlapping intervals */ 1808 static struct interval combining[] = 1809 { 1810 {0x0300, 0x034f}, {0x0360, 0x036f}, {0x0483, 0x0486}, {0x0488, 0x0489}, 1811 {0x0591, 0x05a1}, {0x05a3, 0x05b9}, {0x05bb, 0x05bd}, {0x05bf, 0x05bf}, 1812 {0x05c1, 0x05c2}, {0x05c4, 0x05c4}, {0x0610, 0x0615}, {0x064b, 0x0658}, 1813 {0x0670, 0x0670}, {0x06d6, 0x06dc}, {0x06de, 0x06e4}, {0x06e7, 0x06e8}, 1814 {0x06ea, 0x06ed}, {0x0711, 0x0711}, {0x0730, 0x074a}, {0x07a6, 0x07b0}, 1815 {0x0901, 0x0903}, {0x093c, 0x093c}, {0x093e, 0x094d}, {0x0951, 0x0954}, 1816 {0x0962, 0x0963}, {0x0981, 0x0983}, {0x09bc, 0x09bc}, {0x09be, 0x09c4}, 1817 {0x09c7, 0x09c8}, {0x09cb, 0x09cd}, {0x09d7, 0x09d7}, {0x09e2, 0x09e3}, 1818 {0x0a01, 0x0a03}, {0x0a3c, 0x0a3c}, {0x0a3e, 0x0a42}, {0x0a47, 0x0a48}, 1819 {0x0a4b, 0x0a4d}, {0x0a70, 0x0a71}, {0x0a81, 0x0a83}, {0x0abc, 0x0abc}, 1820 {0x0abe, 0x0ac5}, {0x0ac7, 0x0ac9}, {0x0acb, 0x0acd}, {0x0ae2, 0x0ae3}, 1821 {0x0b01, 0x0b03}, {0x0b3c, 0x0b3c}, {0x0b3e, 0x0b43}, {0x0b47, 0x0b48}, 1822 {0x0b4b, 0x0b4d}, {0x0b56, 0x0b57}, {0x0b82, 0x0b82}, {0x0bbe, 0x0bc2}, 1823 {0x0bc6, 0x0bc8}, {0x0bca, 0x0bcd}, {0x0bd7, 0x0bd7}, {0x0c01, 0x0c03}, 1824 {0x0c3e, 0x0c44}, {0x0c46, 0x0c48}, {0x0c4a, 0x0c4d}, {0x0c55, 0x0c56}, 1825 {0x0c82, 0x0c83}, {0x0cbc, 0x0cbc}, {0x0cbe, 0x0cc4}, {0x0cc6, 0x0cc8}, 1826 {0x0cca, 0x0ccd}, {0x0cd5, 0x0cd6}, {0x0d02, 0x0d03}, {0x0d3e, 0x0d43}, 1827 {0x0d46, 0x0d48}, {0x0d4a, 0x0d4d}, {0x0d57, 0x0d57}, {0x0d82, 0x0d83}, 1828 {0x0dca, 0x0dca}, {0x0dcf, 0x0dd4}, {0x0dd6, 0x0dd6}, {0x0dd8, 0x0ddf}, 1829 {0x0df2, 0x0df3}, {0x0e31, 0x0e31}, {0x0e34, 0x0e3a}, {0x0e47, 0x0e4e}, 1830 {0x0eb1, 0x0eb1}, {0x0eb4, 0x0eb9}, {0x0ebb, 0x0ebc}, {0x0ec8, 0x0ecd}, 1831 {0x0f18, 0x0f19}, {0x0f35, 0x0f35}, {0x0f37, 0x0f37}, {0x0f39, 0x0f39}, 1832 {0x0f3e, 0x0f3f}, {0x0f71, 0x0f84}, {0x0f86, 0x0f87}, {0x0f90, 0x0f97}, 1833 {0x0f99, 0x0fbc}, {0x0fc6, 0x0fc6}, {0x102c, 0x1032}, {0x1036, 0x1039}, 1834 {0x1056, 0x1059}, {0x1712, 0x1714}, {0x1732, 0x1734}, {0x1752, 0x1753}, 1835 {0x1772, 0x1773}, {0x17b6, 0x17d3}, {0x17dd, 0x17dd}, {0x180b, 0x180d}, 1836 {0x18a9, 0x18a9}, {0x1920, 0x192b}, {0x1930, 0x193b}, {0x20d0, 0x20ea}, 1837 {0x302a, 0x302f}, {0x3099, 0x309a}, {0xfb1e, 0xfb1e}, {0xfe00, 0xfe0f}, 1838 {0xfe20, 0xfe23}, 1839 }; 1840 1841 return intable(combining, sizeof(combining), c); 1842 } 1843 1844 /* 1845 * Return TRUE for characters that can be displayed in a normal way. 1846 * Only for characters of 0x100 and above! 1847 */ 1848 int 1849 utf_printable(c) 1850 int c; 1851 { 1852 #ifdef USE_WCHAR_FUNCTIONS 1853 /* 1854 * Assume the iswprint() library function works better than our own stuff. 1855 */ 1856 return iswprint(c); 1857 #else 1858 /* Sorted list of non-overlapping intervals. 1859 * 0xd800-0xdfff is reserved for UTF-16, actually illegal. */ 1860 static struct interval nonprint[] = 1861 { 1862 {0x070f, 0x070f}, {0x180b, 0x180e}, {0x200b, 0x200f}, {0x202a, 0x202e}, 1863 {0x206a, 0x206f}, {0xd800, 0xdfff}, {0xfeff, 0xfeff}, {0xfff9, 0xfffb}, 1864 {0xfffe, 0xffff} 1865 }; 1866 1867 return !intable(nonprint, sizeof(nonprint), c); 1868 #endif 1869 } 1870 1871 /* 1872 * Get class of a Unicode character. 1873 * 0: white space 1874 * 1: punctuation 1875 * 2 or bigger: some class of word character. 1876 */ 1877 int 1878 utf_class(c) 1879 int c; 1880 { 1881 /* sorted list of non-overlapping intervals */ 1882 static struct clinterval 1883 { 1884 unsigned short first; 1885 unsigned short last; 1886 unsigned short class; 1887 } classes[] = 1888 { 1889 {0x037e, 0x037e, 1}, /* Greek question mark */ 1890 {0x0387, 0x0387, 1}, /* Greek ano teleia */ 1891 {0x055a, 0x055f, 1}, /* Armenian punctuation */ 1892 {0x0589, 0x0589, 1}, /* Armenian full stop */ 1893 {0x05be, 0x05be, 1}, 1894 {0x05c0, 0x05c0, 1}, 1895 {0x05c3, 0x05c3, 1}, 1896 {0x05f3, 0x05f4, 1}, 1897 {0x060c, 0x060c, 1}, 1898 {0x061b, 0x061b, 1}, 1899 {0x061f, 0x061f, 1}, 1900 {0x066a, 0x066d, 1}, 1901 {0x06d4, 0x06d4, 1}, 1902 {0x0700, 0x070d, 1}, /* Syriac punctuation */ 1903 {0x0964, 0x0965, 1}, 1904 {0x0970, 0x0970, 1}, 1905 {0x0df4, 0x0df4, 1}, 1906 {0x0e4f, 0x0e4f, 1}, 1907 {0x0e5a, 0x0e5b, 1}, 1908 {0x0f04, 0x0f12, 1}, 1909 {0x0f3a, 0x0f3d, 1}, 1910 {0x0f85, 0x0f85, 1}, 1911 {0x104a, 0x104f, 1}, /* Myanmar punctuation */ 1912 {0x10fb, 0x10fb, 1}, /* Georgian punctuation */ 1913 {0x1361, 0x1368, 1}, /* Ethiopic punctuation */ 1914 {0x166d, 0x166e, 1}, /* Canadian Syl. punctuation */ 1915 {0x1680, 0x1680, 0}, 1916 {0x169b, 0x169c, 1}, 1917 {0x16eb, 0x16ed, 1}, 1918 {0x1735, 0x1736, 1}, 1919 {0x17d4, 0x17dc, 1}, /* Khmer punctuation */ 1920 {0x1800, 0x180a, 1}, /* Mongolian punctuation */ 1921 {0x2000, 0x200b, 0}, /* spaces */ 1922 {0x200c, 0x2027, 1}, /* punctuation and symbols */ 1923 {0x2028, 0x2029, 0}, 1924 {0x202a, 0x202e, 1}, /* punctuation and symbols */ 1925 {0x202f, 0x202f, 0}, 1926 {0x2030, 0x205e, 1}, /* punctuation and symbols */ 1927 {0x205f, 0x205f, 0}, 1928 {0x2060, 0x27ff, 1}, /* punctuation and symbols */ 1929 {0x2070, 0x207f, 0x2070}, /* superscript */ 1930 {0x2080, 0x208f, 0x2080}, /* subscript */ 1931 {0x2983, 0x2998, 1}, 1932 {0x29d8, 0x29db, 1}, 1933 {0x29fc, 0x29fd, 1}, 1934 {0x3000, 0x3000, 0}, /* ideographic space */ 1935 {0x3001, 0x3020, 1}, /* ideographic punctuation */ 1936 {0x3030, 0x3030, 1}, 1937 {0x303d, 0x303d, 1}, 1938 {0x3040, 0x309f, 0x3040}, /* Hiragana */ 1939 {0x30a0, 0x30ff, 0x30a0}, /* Katakana */ 1940 {0x3300, 0x9fff, 0x4e00}, /* CJK Ideographs */ 1941 {0xac00, 0xd7a3, 0xac00}, /* Hangul Syllables */ 1942 {0xf900, 0xfaff, 0x4e00}, /* CJK Ideographs */ 1943 {0xfd3e, 0xfd3f, 1}, 1944 {0xfe30, 0xfe6b, 1}, /* punctuation forms */ 1945 {0xff00, 0xff0f, 1}, /* half/fullwidth ASCII */ 1946 {0xff1a, 0xff20, 1}, /* half/fullwidth ASCII */ 1947 {0xff3b, 0xff40, 1}, /* half/fullwidth ASCII */ 1948 {0xff5b, 0xff65, 1}, /* half/fullwidth ASCII */ 1949 }; 1950 int bot = 0; 1951 int top = sizeof(classes) / sizeof(struct clinterval) - 1; 1952 int mid; 1953 1954 /* First quick check for Latin1 characters, use 'iskeyword'. */ 1955 if (c < 0x100) 1956 { 1957 if (c == ' ' || c == '\t' || c == NUL || c == 0xa0) 1958 return 0; /* blank */ 1959 if (vim_iswordc(c)) 1960 return 2; /* word character */ 1961 return 1; /* punctuation */ 1962 } 1963 1964 /* binary search in table */ 1965 while (top >= bot) 1966 { 1967 mid = (bot + top) / 2; 1968 if (classes[mid].last < c) 1969 bot = mid + 1; 1970 else if (classes[mid].first > c) 1971 top = mid - 1; 1972 else 1973 return (int)classes[mid].class; 1974 } 1975 1976 /* most other characters are "word" characters */ 1977 return 2; 1978 } 1979 1980 /* 1981 * Code for Unicode case-dependent operations. Based on notes in 1982 * http://www.unicode.org/Public/UNIDATA/CaseFolding.txt 1983 * This code uses simple case folding, not full case folding. 1984 */ 1985 1986 /* 1987 * The following table is built by foldExtract.pl < CaseFolding.txt . 1988 * It must be in numeric order, because we use binary search on it. 1989 * An entry such as {0x41,0x5a,1,32} means that UCS-4 characters in the range 1990 * from 0x41 to 0x5a inclusive, stepping by 1, are folded by adding 32. 1991 */ 1992 1993 typedef struct 1994 { 1995 int rangeStart; 1996 int rangeEnd; 1997 int step; 1998 int offset; 1999 } convertStruct; 2000 2001 static convertStruct foldCase[] = 2002 { 2003 {0x41,0x5a,1,32}, {0xc0,0xd6,1,32}, {0xd8,0xde,1,32}, 2004 {0x100,0x12e,2,1}, {0x130,0x130,-1,-199}, {0x132,0x136,2,1}, 2005 {0x139,0x147,2,1}, {0x14a,0x176,2,1}, {0x178,0x178,-1,-121}, 2006 {0x179,0x17d,2,1}, {0x181,0x181,-1,210}, {0x182,0x184,2,1}, 2007 {0x186,0x186,-1,206}, {0x187,0x187,-1,1}, {0x189,0x18a,1,205}, 2008 {0x18b,0x18b,-1,1}, {0x18e,0x18e,-1,79}, {0x18f,0x18f,-1,202}, 2009 {0x190,0x190,-1,203}, {0x191,0x191,-1,1}, {0x193,0x193,-1,205}, 2010 {0x194,0x194,-1,207}, {0x196,0x196,-1,211}, {0x197,0x197,-1,209}, 2011 {0x198,0x198,-1,1}, {0x19c,0x19c,-1,211}, {0x19d,0x19d,-1,213}, 2012 {0x19f,0x19f,-1,214}, {0x1a0,0x1a4,2,1}, {0x1a6,0x1a6,-1,218}, 2013 {0x1a7,0x1a7,-1,1}, {0x1a9,0x1a9,-1,218}, {0x1ac,0x1ac,-1,1}, 2014 {0x1ae,0x1ae,-1,218}, {0x1af,0x1af,-1,1}, {0x1b1,0x1b2,1,217}, 2015 {0x1b3,0x1b5,2,1}, {0x1b7,0x1b7,-1,219}, {0x1b8,0x1bc,4,1}, 2016 {0x1c4,0x1c4,-1,2}, {0x1c5,0x1c5,-1,1}, {0x1c7,0x1c7,-1,2}, 2017 {0x1c8,0x1c8,-1,1}, {0x1ca,0x1ca,-1,2}, {0x1cb,0x1db,2,1}, 2018 {0x1de,0x1ee,2,1}, {0x1f1,0x1f1,-1,2}, {0x1f2,0x1f4,2,1}, 2019 {0x1f6,0x1f6,-1,-97}, {0x1f7,0x1f7,-1,-56}, {0x1f8,0x21e,2,1}, 2020 {0x220,0x220,-1,-130}, {0x222,0x232,2,1}, {0x386,0x386,-1,38}, 2021 {0x388,0x38a,1,37}, {0x38c,0x38c,-1,64}, {0x38e,0x38f,1,63}, 2022 {0x391,0x3a1,1,32}, {0x3a3,0x3ab,1,32}, {0x3d8,0x3ee,2,1}, 2023 {0x3f4,0x3f4,-1,-60}, {0x3f7,0x3f7,-1,1}, {0x3f9,0x3f9,-1,-7}, 2024 {0x3fa,0x3fa,-1,1}, {0x400,0x40f,1,80}, {0x410,0x42f,1,32}, 2025 {0x460,0x480,2,1}, {0x48a,0x4be,2,1}, {0x4c1,0x4cd,2,1}, 2026 {0x4d0,0x4f4,2,1}, {0x4f8,0x500,8,1}, {0x502,0x50e,2,1}, 2027 {0x531,0x556,1,48}, {0x1e00,0x1e94,2,1}, {0x1ea0,0x1ef8,2,1}, 2028 {0x1f08,0x1f0f,1,-8}, {0x1f18,0x1f1d,1,-8}, {0x1f28,0x1f2f,1,-8}, 2029 {0x1f38,0x1f3f,1,-8}, {0x1f48,0x1f4d,1,-8}, {0x1f59,0x1f5f,2,-8}, 2030 {0x1f68,0x1f6f,1,-8}, {0x1f88,0x1f8f,1,-8}, {0x1f98,0x1f9f,1,-8}, 2031 {0x1fa8,0x1faf,1,-8}, {0x1fb8,0x1fb9,1,-8}, {0x1fba,0x1fbb,1,-74}, 2032 {0x1fbc,0x1fbc,-1,-9}, {0x1fc8,0x1fcb,1,-86}, {0x1fcc,0x1fcc,-1,-9}, 2033 {0x1fd8,0x1fd9,1,-8}, {0x1fda,0x1fdb,1,-100}, {0x1fe8,0x1fe9,1,-8}, 2034 {0x1fea,0x1feb,1,-112}, {0x1fec,0x1fec,-1,-7}, {0x1ff8,0x1ff9,1,-128}, 2035 {0x1ffa,0x1ffb,1,-126}, {0x1ffc,0x1ffc,-1,-9}, {0x2126,0x2126,-1,-7517}, 2036 {0x212a,0x212a,-1,-8383}, {0x212b,0x212b,-1,-8262}, 2037 {0x2160,0x216f,1,16}, {0x24b6,0x24cf,1,26}, {0xff21,0xff3a,1,32}, 2038 {0x10400,0x10427,1,40} 2039 }; 2040 2041 static int utf_convert(int a, convertStruct table[], int tableSize); 2042 2043 /* 2044 * Generic conversion function for case operations. 2045 * Return the converted equivalent of "a", which is a UCS-4 character. Use 2046 * the given conversion "table". Uses binary search on "table". 2047 */ 2048 static int 2049 utf_convert(a, table, tableSize) 2050 int a; 2051 convertStruct table[]; 2052 int tableSize; 2053 { 2054 int start, mid, end; /* indices into table */ 2055 2056 start = 0; 2057 end = tableSize / sizeof(convertStruct); 2058 while (start < end) 2059 { 2060 /* need to search further */ 2061 mid = (end + start) /2; 2062 if (table[mid].rangeEnd < a) 2063 start = mid + 1; 2064 else 2065 end = mid; 2066 } 2067 if (table[start].rangeStart <= a && a <= table[start].rangeEnd 2068 && (a - table[start].rangeStart) % table[start].step == 0) 2069 return (a + table[start].offset); 2070 else 2071 return a; 2072 } 2073 2074 /* 2075 * Return the folded-case equivalent of "a", which is a UCS-4 character. Uses 2076 * simple case folding. 2077 */ 2078 int 2079 utf_fold(a) 2080 int a; 2081 { 2082 return utf_convert(a, foldCase, sizeof(foldCase)); 2083 } 2084 2085 /* 2086 * The following tables are built by upperLowerExtract.pl < UnicodeData.txt . 2087 * They must be in numeric order, because we use binary search on them. 2088 * An entry such as {0x41,0x5a,1,32} means that UCS-4 characters in the range 2089 * from 0x41 to 0x5a inclusive, stepping by 1, are switched to lower (for 2090 * example) by adding 32. 2091 */ 2092 static convertStruct toLower[] = 2093 { 2094 {0x41,0x5a,1,32}, {0xc0,0xd6,1,32}, {0xd8,0xde,1,32}, 2095 {0x100,0x12e,2,1}, {0x130,0x130,-1,-199}, {0x132,0x136,2,1}, 2096 {0x139,0x147,2,1}, {0x14a,0x176,2,1}, {0x178,0x178,-1,-121}, 2097 {0x179,0x17d,2,1}, {0x181,0x181,-1,210}, {0x182,0x184,2,1}, 2098 {0x186,0x186,-1,206}, {0x187,0x187,-1,1}, {0x189,0x18a,1,205}, 2099 {0x18b,0x18b,-1,1}, {0x18e,0x18e,-1,79}, {0x18f,0x18f,-1,202}, 2100 {0x190,0x190,-1,203}, {0x191,0x191,-1,1}, {0x193,0x193,-1,205}, 2101 {0x194,0x194,-1,207}, {0x196,0x196,-1,211}, {0x197,0x197,-1,209}, 2102 {0x198,0x198,-1,1}, {0x19c,0x19c,-1,211}, {0x19d,0x19d,-1,213}, 2103 {0x19f,0x19f,-1,214}, {0x1a0,0x1a4,2,1}, {0x1a6,0x1a6,-1,218}, 2104 {0x1a7,0x1a7,-1,1}, {0x1a9,0x1a9,-1,218}, {0x1ac,0x1ac,-1,1}, 2105 {0x1ae,0x1ae,-1,218}, {0x1af,0x1af,-1,1}, {0x1b1,0x1b2,1,217}, 2106 {0x1b3,0x1b5,2,1}, {0x1b7,0x1b7,-1,219}, {0x1b8,0x1bc,4,1}, 2107 {0x1c4,0x1ca,3,2}, {0x1cd,0x1db,2,1}, {0x1de,0x1ee,2,1}, 2108 {0x1f1,0x1f1,-1,2}, {0x1f4,0x1f4,-1,1}, {0x1f6,0x1f6,-1,-97}, 2109 {0x1f7,0x1f7,-1,-56}, {0x1f8,0x21e,2,1}, {0x220,0x220,-1,-130}, 2110 {0x222,0x232,2,1}, {0x386,0x386,-1,38}, {0x388,0x38a,1,37}, 2111 {0x38c,0x38c,-1,64}, {0x38e,0x38f,1,63}, {0x391,0x3a1,1,32}, 2112 {0x3a3,0x3ab,1,32}, {0x3d8,0x3ee,2,1}, {0x3f4,0x3f4,-1,-60}, 2113 {0x3f7,0x3f7,-1,1}, {0x3f9,0x3f9,-1,-7}, {0x3fa,0x3fa,-1,1}, 2114 {0x400,0x40f,1,80}, {0x410,0x42f,1,32}, {0x460,0x480,2,1}, 2115 {0x48a,0x4be,2,1}, {0x4c1,0x4cd,2,1}, {0x4d0,0x4f4,2,1}, 2116 {0x4f8,0x500,8,1}, {0x502,0x50e,2,1}, {0x531,0x556,1,48}, 2117 {0x1e00,0x1e94,2,1}, {0x1ea0,0x1ef8,2,1}, {0x1f08,0x1f0f,1,-8}, 2118 {0x1f18,0x1f1d,1,-8}, {0x1f28,0x1f2f,1,-8}, {0x1f38,0x1f3f,1,-8}, 2119 {0x1f48,0x1f4d,1,-8}, {0x1f59,0x1f5f,2,-8}, {0x1f68,0x1f6f,1,-8}, 2120 {0x1fb8,0x1fb9,1,-8}, {0x1fba,0x1fbb,1,-74}, {0x1fc8,0x1fcb,1,-86}, 2121 {0x1fd8,0x1fd9,1,-8}, {0x1fda,0x1fdb,1,-100}, {0x1fe8,0x1fe9,1,-8}, 2122 {0x1fea,0x1feb,1,-112}, {0x1fec,0x1fec,-1,-7}, {0x1ff8,0x1ff9,1,-128}, 2123 {0x1ffa,0x1ffb,1,-126}, {0x2126,0x2126,-1,-7517}, {0x212a,0x212a,-1,-8383}, 2124 {0x212b,0x212b,-1,-8262}, {0xff21,0xff3a,1,32}, {0x10400,0x10427,1,40} 2125 }; 2126 2127 static convertStruct toUpper[] = 2128 { 2129 {0x61,0x7a,1,-32}, {0xb5,0xb5,-1,743}, {0xe0,0xf6,1,-32}, 2130 {0xf8,0xfe,1,-32}, {0xff,0xff,-1,121}, {0x101,0x12f,2,-1}, 2131 {0x131,0x131,-1,-232}, {0x133,0x137,2,-1}, {0x13a,0x148,2,-1}, 2132 {0x14b,0x177,2,-1}, {0x17a,0x17e,2,-1}, {0x17f,0x17f,-1,-300}, 2133 {0x183,0x185,2,-1}, {0x188,0x18c,4,-1}, {0x192,0x192,-1,-1}, 2134 {0x195,0x195,-1,97}, {0x199,0x199,-1,-1}, {0x19e,0x19e,-1,130}, 2135 {0x1a1,0x1a5,2,-1}, {0x1a8,0x1ad,5,-1}, {0x1b0,0x1b4,4,-1}, 2136 {0x1b6,0x1b9,3,-1}, {0x1bd,0x1bd,-1,-1}, {0x1bf,0x1bf,-1,56}, 2137 {0x1c5,0x1c6,1,-1}, {0x1c8,0x1c9,1,-1}, {0x1cb,0x1cc,1,-1}, 2138 {0x1ce,0x1dc,2,-1}, {0x1dd,0x1dd,-1,-79}, {0x1df,0x1ef,2,-1}, 2139 {0x1f2,0x1f3,1,-1}, {0x1f5,0x1f9,4,-1}, {0x1fb,0x21f,2,-1}, 2140 {0x223,0x233,2,-1}, {0x253,0x253,-1,-210}, {0x254,0x254,-1,-206}, 2141 {0x256,0x257,1,-205}, {0x259,0x259,-1,-202}, {0x25b,0x25b,-1,-203}, 2142 {0x260,0x260,-1,-205}, {0x263,0x263,-1,-207}, {0x268,0x268,-1,-209}, 2143 {0x269,0x26f,6,-211}, {0x272,0x272,-1,-213}, {0x275,0x275,-1,-214}, 2144 {0x280,0x283,3,-218}, {0x288,0x288,-1,-218}, {0x28a,0x28b,1,-217}, 2145 {0x292,0x292,-1,-219}, {0x3ac,0x3ac,-1,-38}, {0x3ad,0x3af,1,-37}, 2146 {0x3b1,0x3c1,1,-32}, {0x3c2,0x3c2,-1,-31}, {0x3c3,0x3cb,1,-32}, 2147 {0x3cc,0x3cc,-1,-64}, {0x3cd,0x3ce,1,-63}, {0x3d0,0x3d0,-1,-62}, 2148 {0x3d1,0x3d1,-1,-57}, {0x3d5,0x3d5,-1,-47}, {0x3d6,0x3d6,-1,-54}, 2149 {0x3d9,0x3ef,2,-1}, {0x3f0,0x3f0,-1,-86}, {0x3f1,0x3f1,-1,-80}, 2150 {0x3f2,0x3f2,-1,7}, {0x3f5,0x3f5,-1,-96}, {0x3f8,0x3fb,3,-1}, 2151 {0x430,0x44f,1,-32}, {0x450,0x45f,1,-80}, {0x461,0x481,2,-1}, 2152 {0x48b,0x4bf,2,-1}, {0x4c2,0x4ce,2,-1}, {0x4d1,0x4f5,2,-1}, 2153 {0x4f9,0x501,8,-1}, {0x503,0x50f,2,-1}, {0x561,0x586,1,-48}, 2154 {0x1e01,0x1e95,2,-1}, {0x1e9b,0x1e9b,-1,-59}, {0x1ea1,0x1ef9,2,-1}, 2155 {0x1f00,0x1f07,1,8}, {0x1f10,0x1f15,1,8}, {0x1f20,0x1f27,1,8}, 2156 {0x1f30,0x1f37,1,8}, {0x1f40,0x1f45,1,8}, {0x1f51,0x1f57,2,8}, 2157 {0x1f60,0x1f67,1,8}, {0x1f70,0x1f71,1,74}, {0x1f72,0x1f75,1,86}, 2158 {0x1f76,0x1f77,1,100}, {0x1f78,0x1f79,1,128}, {0x1f7a,0x1f7b,1,112}, 2159 {0x1f7c,0x1f7d,1,126}, {0x1f80,0x1f87,1,8}, {0x1f90,0x1f97,1,8}, 2160 {0x1fa0,0x1fa7,1,8}, {0x1fb0,0x1fb1,1,8}, {0x1fb3,0x1fb3,-1,9}, 2161 {0x1fbe,0x1fbe,-1,-7205}, {0x1fc3,0x1fc3,-1,9}, {0x1fd0,0x1fd1,1,8}, 2162 {0x1fe0,0x1fe1,1,8}, {0x1fe5,0x1fe5,-1,7}, {0x1ff3,0x1ff3,-1,9}, 2163 {0xff41,0xff5a,1,-32}, {0x10428,0x1044f,1,-40} 2164 }; 2165 2166 /* 2167 * Return the upper-case equivalent of "a", which is a UCS-4 character. Use 2168 * simple case folding. 2169 */ 2170 int 2171 utf_toupper(a) 2172 int a; 2173 { 2174 /* If 'casemap' contains "keepascii" use ASCII style toupper(). */ 2175 if (a < 128 && (cmp_flags & CMP_KEEPASCII)) 2176 return TOUPPER_ASC(a); 2177 2178 #if defined(HAVE_TOWUPPER) && defined(__STDC__ISO_10646__) 2179 /* If towupper() is availble and handles Unicode, use it. */ 2180 if (!(cmp_flags & CMP_INTERNAL)) 2181 return towupper(a); 2182 #endif 2183 2184 /* For characters below 128 use locale sensitive toupper(). */ 2185 if (a < 128) 2186 return TOUPPER_LOC(a); 2187 2188 /* For any other characters use the above mapping table. */ 2189 return utf_convert(a, toUpper, sizeof(toUpper)); 2190 } 2191 2192 int 2193 utf_islower(a) 2194 int a; 2195 { 2196 return (utf_toupper(a) != a); 2197 } 2198 2199 /* 2200 * Return the lower-case equivalent of "a", which is a UCS-4 character. Use 2201 * simple case folding. 2202 */ 2203 int 2204 utf_tolower(a) 2205 int a; 2206 { 2207 /* If 'casemap' contains "keepascii" use ASCII style tolower(). */ 2208 if (a < 128 && (cmp_flags & CMP_KEEPASCII)) 2209 return TOLOWER_ASC(a); 2210 2211 #if defined(HAVE_TOWLOWER) && defined(__STDC__ISO_10646__) 2212 /* If towlower() is available and handles Unicode, use it. */ 2213 if (!(cmp_flags & CMP_INTERNAL)) 2214 return towlower(a); 2215 #endif 2216 2217 /* For characters below 128 use locale sensitive tolower(). */ 2218 if (a < 128) 2219 return TOLOWER_LOC(a); 2220 2221 /* For any other characters use the above mapping table. */ 2222 return utf_convert(a, toLower, sizeof(toLower)); 2223 } 2224 2225 int 2226 utf_isupper(a) 2227 int a; 2228 { 2229 return (utf_tolower(a) != a); 2230 } 2231 2232 /* 2233 * Version of strnicmp() that handles multi-byte characters. 2234 * Needed for Big5, Sjift-JIS and UTF-8 encoding. Other DBCS encodings can 2235 * probably use strnicmp(), because there are no ASCII characters in the 2236 * second byte. 2237 * Returns zero if s1 and s2 are equal (ignoring case), the difference between 2238 * two characters otherwise. 2239 */ 2240 int 2241 mb_strnicmp(s1, s2, nn) 2242 char_u *s1, *s2; 2243 size_t nn; 2244 { 2245 int i, j, l; 2246 int cdiff; 2247 int incomplete = FALSE; 2248 int n = nn; 2249 2250 for (i = 0; i < n; i += l) 2251 { 2252 if (s1[i] == NUL && s2[i] == NUL) /* both strings end */ 2253 return 0; 2254 if (enc_utf8) 2255 { 2256 l = utf_byte2len(s1[i]); 2257 if (l > n - i) 2258 { 2259 l = n - i; /* incomplete character */ 2260 incomplete = TRUE; 2261 } 2262 /* Check directly first, it's faster. */ 2263 for (j = 0; j < l; ++j) 2264 if (s1[i + j] != s2[i + j]) 2265 break; 2266 if (j < l) 2267 { 2268 /* If one of the two characters is incomplete return -1. */ 2269 if (incomplete || i + utf_byte2len(s2[i]) > n) 2270 return -1; 2271 cdiff = utf_fold(utf_ptr2char(s1 + i)) 2272 - utf_fold(utf_ptr2char(s2 + i)); 2273 if (cdiff != 0) 2274 return cdiff; 2275 } 2276 } 2277 else 2278 { 2279 l = (*mb_ptr2len_check)(s1 + i); 2280 if (l <= 1) 2281 { 2282 /* Single byte: first check normally, then with ignore case. */ 2283 if (s1[i] != s2[i]) 2284 { 2285 cdiff = TOLOWER_LOC(s1[i]) - TOLOWER_LOC(s2[i]); 2286 if (cdiff != 0) 2287 return cdiff; 2288 } 2289 } 2290 else 2291 { 2292 /* For non-Unicode multi-byte don't ignore case. */ 2293 if (l > n - i) 2294 l = n - i; 2295 cdiff = STRNCMP(s1 + i, s2 + i, l); 2296 if (cdiff != 0) 2297 return cdiff; 2298 } 2299 } 2300 } 2301 return 0; 2302 } 2303 2304 /* 2305 * "g8": show bytes of the UTF-8 char under the cursor. Doesn't matter what 2306 * 'encoding' has been set to. 2307 */ 2308 void 2309 show_utf8() 2310 { 2311 int len; 2312 int rlen = 0; 2313 char_u *line; 2314 int clen; 2315 int i; 2316 2317 /* Get the byte length of the char under the cursor, including composing 2318 * characters. */ 2319 line = ml_get_cursor(); 2320 len = utfc_ptr2len_check(line); 2321 if (len == 0) 2322 { 2323 MSG("NUL"); 2324 return; 2325 } 2326 2327 clen = 0; 2328 for (i = 0; i < len; ++i) 2329 { 2330 if (clen == 0) 2331 { 2332 /* start of (composing) character, get its length */ 2333 if (i > 0) 2334 { 2335 STRCPY(IObuff + rlen, "+ "); 2336 rlen += 2; 2337 } 2338 clen = utf_ptr2len_check(line + i); 2339 } 2340 sprintf((char *)IObuff + rlen, "%02x ", line[i]); 2341 --clen; 2342 rlen += STRLEN(IObuff + rlen); 2343 if (rlen > IOSIZE - 20) 2344 break; 2345 } 2346 2347 msg(IObuff); 2348 } 2349 2350 /* 2351 * mb_head_off() function pointer. 2352 * Return offset from "p" to the first byte of the character it points into. 2353 * Returns 0 when already at the first byte of a character. 2354 */ 2355 /*ARGSUSED*/ 2356 int 2357 latin_head_off(base, p) 2358 char_u *base; 2359 char_u *p; 2360 { 2361 return 0; 2362 } 2363 2364 int 2365 dbcs_head_off(base, p) 2366 char_u *base; 2367 char_u *p; 2368 { 2369 char_u *q; 2370 2371 /* It can't be a trailing byte when not using DBCS, at the start of the 2372 * string or the previous byte can't start a double-byte. */ 2373 if (p <= base || MB_BYTE2LEN(p[-1]) == 1) 2374 return 0; 2375 2376 /* This is slow: need to start at the base and go forward until the 2377 * byte we are looking for. Return 1 when we went past it, 0 otherwise. */ 2378 q = base; 2379 while (q < p) 2380 q += dbcs_ptr2len_check(q); 2381 return (q == p) ? 0 : 1; 2382 } 2383 2384 #if defined(FEAT_CLIPBOARD) || defined(FEAT_GUI) || defined(FEAT_RIGHTLEFT) \ 2385 || defined(PROTO) 2386 /* 2387 * Special version of dbcs_head_off() that works for ScreenLines[], where 2388 * single-width DBCS_JPNU characters are stored separately. 2389 */ 2390 int 2391 dbcs_screen_head_off(base, p) 2392 char_u *base; 2393 char_u *p; 2394 { 2395 char_u *q; 2396 2397 /* It can't be a trailing byte when not using DBCS, at the start of the 2398 * string or the previous byte can't start a double-byte. 2399 * For euc-jp an 0x8e byte in the previous cell always means we have a 2400 * lead byte in the current cell. */ 2401 if (p <= base 2402 || (enc_dbcs == DBCS_JPNU && p[-1] == 0x8e) 2403 || MB_BYTE2LEN(p[-1]) == 1) 2404 return 0; 2405 2406 /* This is slow: need to start at the base and go forward until the 2407 * byte we are looking for. Return 1 when we went past it, 0 otherwise. 2408 * For DBCS_JPNU look out for 0x8e, which means the second byte is not 2409 * stored as the next byte. */ 2410 q = base; 2411 while (q < p) 2412 { 2413 if (enc_dbcs == DBCS_JPNU && *q == 0x8e) 2414 ++q; 2415 else 2416 q += dbcs_ptr2len_check(q); 2417 } 2418 return (q == p) ? 0 : 1; 2419 } 2420 #endif 2421 2422 int 2423 utf_head_off(base, p) 2424 char_u *base; 2425 char_u *p; 2426 { 2427 char_u *q; 2428 char_u *s; 2429 int c; 2430 #ifdef FEAT_ARABIC 2431 char_u *j; 2432 #endif 2433 2434 if (*p < 0x80) /* be quick for ASCII */ 2435 return 0; 2436 2437 /* Skip backwards over trailing bytes: 10xx.xxxx 2438 * Skip backwards again if on a composing char. */ 2439 for (q = p; ; --q) 2440 { 2441 /* Move s to the last byte of this char. */ 2442 for (s = q; (s[1] & 0xc0) == 0x80; ++s) 2443 ; 2444 /* Move q to the first byte of this char. */ 2445 while (q > base && (*q & 0xc0) == 0x80) 2446 --q; 2447 /* Check for illegal sequence. Do allow an illegal byte after where we 2448 * started. */ 2449 if (utf8len_tab[*q] != (int)(s - q + 1) 2450 && utf8len_tab[*q] != (int)(p - q + 1)) 2451 return 0; 2452 2453 if (q <= base) 2454 break; 2455 2456 c = utf_ptr2char(q); 2457 if (utf_iscomposing(c)) 2458 continue; 2459 2460 #ifdef FEAT_ARABIC 2461 if (arabic_maycombine(c)) 2462 { 2463 /* Advance to get a sneak-peak at the next char */ 2464 j = q; 2465 --j; 2466 /* Move j to the first byte of this char. */ 2467 while (j > base && (*j & 0xc0) == 0x80) 2468 --j; 2469 if (arabic_combine(utf_ptr2char(j), c)) 2470 continue; 2471 } 2472 #endif 2473 break; 2474 } 2475 2476 return (int)(p - q); 2477 } 2478 2479 #if defined(FEAT_EVAL) || defined(PROTO) 2480 /* 2481 * Copy a character from "*fp" to "*tp" and advance the pointers. 2482 */ 2483 void 2484 mb_copy_char(fp, tp) 2485 char_u **fp; 2486 char_u **tp; 2487 { 2488 int l = (*mb_ptr2len_check)(*fp); 2489 2490 mch_memmove(*tp, *fp, (size_t)l); 2491 *tp += l; 2492 *fp += l; 2493 } 2494 #endif 2495 2496 /* 2497 * Return the offset from "p" to the first byte of a character. When "p" is 2498 * at the start of a character 0 is returned, otherwise the offset to the next 2499 * character. Can start anywhere in a stream of bytes. 2500 */ 2501 int 2502 mb_off_next(base, p) 2503 char_u *base; 2504 char_u *p; 2505 { 2506 int i; 2507 int j; 2508 2509 if (enc_utf8) 2510 { 2511 if (*p < 0x80) /* be quick for ASCII */ 2512 return 0; 2513 2514 /* Find the next character that isn't 10xx.xxxx */ 2515 for (i = 0; (p[i] & 0xc0) == 0x80; ++i) 2516 ; 2517 if (i > 0) 2518 { 2519 /* Check for illegal sequence. */ 2520 for (j = 0; p - j > base; ++j) 2521 if ((p[-j] & 0xc0) != 0x80) 2522 break; 2523 if (utf8len_tab[p[-j]] != i + j) 2524 return 0; 2525 } 2526 return i; 2527 } 2528 2529 /* Only need to check if we're on a trail byte, it doesn't matter if we 2530 * want the offset to the next or current character. */ 2531 return (*mb_head_off)(base, p); 2532 } 2533 2534 /* 2535 * Return the offset from "p" to the last byte of the character it points 2536 * into. Can start anywhere in a stream of bytes. 2537 */ 2538 int 2539 mb_tail_off(base, p) 2540 char_u *base; 2541 char_u *p; 2542 { 2543 int i; 2544 int j; 2545 2546 if (*p == NUL) 2547 return 0; 2548 2549 if (enc_utf8) 2550 { 2551 /* Find the last character that is 10xx.xxxx */ 2552 for (i = 0; (p[i + 1] & 0xc0) == 0x80; ++i) 2553 ; 2554 /* Check for illegal sequence. */ 2555 for (j = 0; p - j > base; ++j) 2556 if ((p[-j] & 0xc0) != 0x80) 2557 break; 2558 if (utf8len_tab[p[-j]] != i + j + 1) 2559 return 0; 2560 return i; 2561 } 2562 2563 /* It can't be the first byte if a double-byte when not using DBCS, at the 2564 * end of the string or the byte can't start a double-byte. */ 2565 if (enc_dbcs == 0 || p[1] == NUL || MB_BYTE2LEN(*p) == 1) 2566 return 0; 2567 2568 /* Return 1 when on the lead byte, 0 when on the tail byte. */ 2569 return 1 - dbcs_head_off(base, p); 2570 } 2571 2572 #if defined(HAVE_GTK2) || defined(PROTO) 2573 /* 2574 * Return TRUE if string "s" is a valid utf-8 string. 2575 * When "end" is NULL stop at the first NUL. 2576 * When "end" is positive stop there. 2577 */ 2578 int 2579 utf_valid_string(s, end) 2580 char_u *s; 2581 char_u *end; 2582 { 2583 int l; 2584 char_u *p = s; 2585 2586 while (end == NULL ? *p != NUL : p < end) 2587 { 2588 if ((*p & 0xc0) == 0x80) 2589 return FALSE; /* invalid lead byte */ 2590 l = utf8len_tab[*p]; 2591 if (end != NULL && p + l > end) 2592 return FALSE; /* incomplete byte sequence */ 2593 ++p; 2594 while (--l > 0) 2595 if ((*p++ & 0xc0) != 0x80) 2596 return FALSE; /* invalid trail byte */ 2597 } 2598 return TRUE; 2599 } 2600 #endif 2601 2602 #if defined(FEAT_GUI) || defined(PROTO) 2603 /* 2604 * Special version of mb_tail_off() for use in ScreenLines[]. 2605 */ 2606 int 2607 dbcs_screen_tail_off(base, p) 2608 char_u *base; 2609 char_u *p; 2610 { 2611 /* It can't be the first byte if a double-byte when not using DBCS, at the 2612 * end of the string or the byte can't start a double-byte. 2613 * For euc-jp an 0x8e byte always means we have a lead byte in the current 2614 * cell. */ 2615 if (*p == NUL || p[1] == NUL 2616 || (enc_dbcs == DBCS_JPNU && *p == 0x8e) 2617 || MB_BYTE2LEN(*p) == 1) 2618 return 0; 2619 2620 /* Return 1 when on the lead byte, 0 when on the tail byte. */ 2621 return 1 - dbcs_screen_head_off(base, p); 2622 } 2623 #endif 2624 2625 /* 2626 * If the cursor moves on an trail byte, set the cursor on the lead byte. 2627 * Thus it moves left if necessary. 2628 * Return TRUE when the cursor was adjusted. 2629 */ 2630 void 2631 mb_adjust_cursor() 2632 { 2633 mb_adjustpos(&curwin->w_cursor); 2634 } 2635 2636 /* 2637 * Adjust position "*lp" to point to the first byte of a multi-byte character. 2638 * If it points to a tail byte it's moved backwards to the head byte. 2639 */ 2640 void 2641 mb_adjustpos(lp) 2642 pos_T *lp; 2643 { 2644 char_u *p; 2645 2646 if (lp->col > 0 2647 #ifdef FEAT_VIRTUALEDIT 2648 || lp->coladd > 1 2649 #endif 2650 ) 2651 { 2652 p = ml_get(lp->lnum); 2653 lp->col -= (*mb_head_off)(p, p + lp->col); 2654 #ifdef FEAT_VIRTUALEDIT 2655 /* Reset "coladd" when the cursor would be on the right half of a 2656 * double-wide character. */ 2657 if (lp->coladd == 1 2658 && p[lp->col] != TAB 2659 && vim_isprintc((*mb_ptr2char)(p + lp->col)) 2660 && ptr2cells(p + lp->col) > 1) 2661 lp->coladd = 0; 2662 #endif 2663 } 2664 } 2665 2666 /* 2667 * Return a pointer to the character before "*p", if there is one. 2668 */ 2669 char_u * 2670 mb_prevptr(line, p) 2671 char_u *line; /* start of the string */ 2672 char_u *p; 2673 { 2674 if (p > line) 2675 mb_ptr_back(line, p); 2676 return p; 2677 } 2678 2679 /* 2680 * Return the character length of "str". Each multi-byte character counts as 2681 * one. 2682 */ 2683 int 2684 mb_charlen(str) 2685 char_u *str; 2686 { 2687 int count; 2688 2689 if (str == NULL) 2690 return 0; 2691 2692 for (count = 0; *str != NUL; count++) 2693 str += (*mb_ptr2len_check)(str); 2694 2695 return count; 2696 } 2697 2698 /* 2699 * Try to un-escape a multi-byte character. 2700 * Used for the "to" and "from" part of a mapping. 2701 * Return the un-escaped string if it is a multi-byte character, and advance 2702 * "pp" to just after the bytes that formed it. 2703 * Return NULL if no multi-byte char was found. 2704 */ 2705 char_u * 2706 mb_unescape(pp) 2707 char_u **pp; 2708 { 2709 static char_u buf[MB_MAXBYTES + 1]; 2710 int n, m = 0; 2711 char_u *str = *pp; 2712 2713 /* Must translate K_SPECIAL KS_SPECIAL KE_FILLER to K_SPECIAL and CSI 2714 * KS_EXTRA KE_CSI to CSI. */ 2715 for (n = 0; str[n] != NUL && m <= MB_MAXBYTES; ++n) 2716 { 2717 if (str[n] == K_SPECIAL 2718 && str[n + 1] == KS_SPECIAL 2719 && str[n + 2] == KE_FILLER) 2720 { 2721 buf[m++] = K_SPECIAL; 2722 n += 2; 2723 } 2724 # ifdef FEAT_GUI 2725 else if (str[n] == CSI 2726 && str[n + 1] == KS_EXTRA 2727 && str[n + 2] == (int)KE_CSI) 2728 { 2729 buf[m++] = CSI; 2730 n += 2; 2731 } 2732 # endif 2733 else if (str[n] == K_SPECIAL 2734 # ifdef FEAT_GUI 2735 || str[n] == CSI 2736 # endif 2737 ) 2738 break; /* a special key can't be a multibyte char */ 2739 else 2740 buf[m++] = str[n]; 2741 buf[m] = NUL; 2742 2743 /* Return a multi-byte character if it's found. An illegal sequence 2744 * will result in a 1 here. */ 2745 if ((*mb_ptr2len_check)(buf) > 1) 2746 { 2747 *pp = str + n + 1; 2748 return buf; 2749 } 2750 } 2751 return NULL; 2752 } 2753 2754 /* 2755 * Return TRUE if the character at "row"/"col" on the screen is the left side 2756 * of a double-width character. 2757 * Caller must make sure "row" and "col" are not invalid! 2758 */ 2759 int 2760 mb_lefthalve(row, col) 2761 int row; 2762 int col; 2763 { 2764 #ifdef FEAT_HANGULIN 2765 if (composing_hangul) 2766 return TRUE; 2767 #endif 2768 if (enc_dbcs != 0) 2769 return dbcs_off2cells(LineOffset[row] + col) > 1; 2770 if (enc_utf8) 2771 return (col + 1 < Columns 2772 && ScreenLines[LineOffset[row] + col + 1] == 0); 2773 return FALSE; 2774 } 2775 2776 # if defined(FEAT_CLIPBOARD) || defined(FEAT_GUI) || defined(FEAT_RIGHTLEFT) \ 2777 || defined(PROTO) 2778 /* 2779 * Correct a position on the screen, if it's the right halve of a double-wide 2780 * char move it to the left halve. Returns the corrected column. 2781 */ 2782 int 2783 mb_fix_col(col, row) 2784 int col; 2785 int row; 2786 { 2787 col = check_col(col); 2788 row = check_row(row); 2789 if (has_mbyte && ScreenLines != NULL && col > 0 2790 && ((enc_dbcs 2791 && ScreenLines[LineOffset[row] + col] != NUL 2792 && dbcs_screen_head_off(ScreenLines + LineOffset[row], 2793 ScreenLines + LineOffset[row] + col)) 2794 || (enc_utf8 && ScreenLines[LineOffset[row] + col] == 0))) 2795 --col; 2796 return col; 2797 } 2798 # endif 2799 #endif 2800 2801 #if defined(FEAT_MBYTE) || defined(FEAT_POSTSCRIPT) || defined(PROTO) 2802 static int enc_alias_search __ARGS((char_u *name)); 2803 2804 /* 2805 * Skip the Vim specific head of a 'encoding' name. 2806 */ 2807 char_u * 2808 enc_skip(p) 2809 char_u *p; 2810 { 2811 if (STRNCMP(p, "2byte-", 6) == 0) 2812 return p + 6; 2813 if (STRNCMP(p, "8bit-", 5) == 0) 2814 return p + 5; 2815 return p; 2816 } 2817 2818 /* 2819 * Find the canonical name for encoding "enc". 2820 * When the name isn't recognized, returns "enc" itself, but with all lower 2821 * case characters and '_' replaced with '-'. 2822 * Returns an allocated string. NULL for out-of-memory. 2823 */ 2824 char_u * 2825 enc_canonize(enc) 2826 char_u *enc; 2827 { 2828 char_u *r; 2829 char_u *p, *s; 2830 int i; 2831 2832 # ifdef FEAT_MBYTE 2833 if (STRCMP(enc, "default") == 0) 2834 { 2835 /* Use the default encoding as it's found by set_init_1(). */ 2836 r = get_encoding_default(); 2837 if (r == NULL) 2838 r = (char_u *)"latin1"; 2839 return vim_strsave(r); 2840 } 2841 # endif 2842 2843 /* copy "enc" to allocted memory, with room for two '-' */ 2844 r = alloc((unsigned)(STRLEN(enc) + 3)); 2845 if (r != NULL) 2846 { 2847 /* Make it all lower case and replace '_' with '-'. */ 2848 p = r; 2849 for (s = enc; *s != NUL; ++s) 2850 { 2851 if (*s == '_') 2852 *p++ = '-'; 2853 else 2854 *p++ = TOLOWER_ASC(*s); 2855 } 2856 *p = NUL; 2857 2858 /* Skip "2byte-" and "8bit-". */ 2859 p = enc_skip(r); 2860 2861 /* "iso8859" -> "iso-8859" */ 2862 if (STRNCMP(p, "iso8859", 7) == 0) 2863 { 2864 mch_memmove(p + 4, p + 3, STRLEN(p + 2)); 2865 p[3] = '-'; 2866 } 2867 2868 /* "iso-8859n" -> "iso-8859-n" */ 2869 if (STRNCMP(p, "iso-8859", 8) == 0 && p[8] != '-') 2870 { 2871 mch_memmove(p + 9, p + 8, STRLEN(p + 7)); 2872 p[8] = '-'; 2873 } 2874 2875 /* "latin-N" -> "latinN" */ 2876 if (STRNCMP(p, "latin-", 6) == 0) 2877 mch_memmove(p + 5, p + 6, STRLEN(p + 5)); 2878 2879 if (enc_canon_search(p) >= 0) 2880 { 2881 /* canonical name can be used unmodified */ 2882 if (p != r) 2883 mch_memmove(r, p, STRLEN(p) + 1); 2884 } 2885 else if ((i = enc_alias_search(p)) >= 0) 2886 { 2887 /* alias recognized, get canonical name */ 2888 vim_free(r); 2889 r = vim_strsave((char_u *)enc_canon_table[i].name); 2890 } 2891 } 2892 return r; 2893 } 2894 2895 /* 2896 * Search for an encoding alias of "name". 2897 * Returns -1 when not found. 2898 */ 2899 static int 2900 enc_alias_search(name) 2901 char_u *name; 2902 { 2903 int i; 2904 2905 for (i = 0; enc_alias_table[i].name != NULL; ++i) 2906 if (STRCMP(name, enc_alias_table[i].name) == 0) 2907 return enc_alias_table[i].canon; 2908 return -1; 2909 } 2910 #endif 2911 2912 #if defined(FEAT_MBYTE) || defined(PROTO) 2913 2914 #ifdef HAVE_LANGINFO_H 2915 # include <langinfo.h> 2916 #endif 2917 2918 /* 2919 * Get the canonicalized encoding of the current locale. 2920 * Returns an allocated string when successful, NULL when not. 2921 */ 2922 char_u * 2923 enc_locale() 2924 { 2925 #ifndef WIN3264 2926 char *s; 2927 char *p; 2928 int i; 2929 #endif 2930 char buf[50]; 2931 #ifdef WIN3264 2932 long acp = GetACP(); 2933 2934 if (acp == 1200) 2935 STRCPY(buf, "ucs-2le"); 2936 else if (acp == 1252) /* cp1252 is used as latin1 */ 2937 STRCPY(buf, "latin1"); 2938 else 2939 sprintf(buf, "cp%ld", acp); 2940 #else 2941 # ifdef HAVE_NL_LANGINFO_CODESET 2942 if ((s = nl_langinfo(CODESET)) == NULL || *s == NUL) 2943 # endif 2944 # ifdef MACOS 2945 s = "utf-8"; 2946 # else 2947 # if defined(HAVE_LOCALE_H) || defined(X_LOCALE) 2948 if ((s = setlocale(LC_CTYPE, NULL)) == NULL || *s == NUL) 2949 # endif 2950 if ((s = getenv("LC_ALL")) == NULL || *s == NUL) 2951 if ((s = getenv("LC_CTYPE")) == NULL || *s == NUL) 2952 s = getenv("LANG"); 2953 # endif 2954 2955 if (s == NULL || *s == NUL) 2956 return FAIL; 2957 2958 /* The most generic locale format is: 2959 * language[_territory][.codeset][@modifier][+special][,[sponsor][_revision]] 2960 * If there is a '.' remove the part before it. 2961 * if there is something after the codeset, remove it. 2962 * Make the name lowercase and replace '_' with '-'. 2963 * Exception: "ja_JP.EUC" == "euc-jp", "zh_CN.EUC" = "euc-cn", 2964 * "ko_KR.EUC" == "euc-kr" 2965 */ 2966 if ((p = (char *)vim_strchr((char_u *)s, '.')) != NULL) 2967 { 2968 if (p > s + 2 && STRNICMP(p + 1, "EUC", 3) == 0 2969 && !isalnum((int)p[4]) && p[4] != '-' && p[-3] == '_') 2970 { 2971 /* copy "XY.EUC" to "euc-XY" to buf[10] */ 2972 STRCPY(buf + 10, "euc-"); 2973 buf[14] = p[-2]; 2974 buf[15] = p[-1]; 2975 buf[16] = 0; 2976 s = buf + 10; 2977 } 2978 else 2979 s = p + 1; 2980 } 2981 for (i = 0; s[i] != NUL && i < sizeof(buf) - 1; ++i) 2982 { 2983 if (s[i] == '_' || s[i] == '-') 2984 buf[i] = '-'; 2985 else if (isalnum((int)s[i])) 2986 buf[i] = TOLOWER_ASC(s[i]); 2987 else 2988 break; 2989 } 2990 buf[i] = NUL; 2991 #endif 2992 2993 return enc_canonize((char_u *)buf); 2994 } 2995 2996 #if defined(WIN3264) || defined(PROTO) 2997 /* 2998 * Convert an encoding name to an MS-Windows codepage. 2999 * Returns zero if no codepage can be figured out. 3000 */ 3001 int 3002 encname2codepage(name) 3003 char_u *name; 3004 { 3005 int cp; 3006 char_u *p = name; 3007 int idx; 3008 3009 if (STRNCMP(p, "8bit-", 5) == 0) 3010 p += 5; 3011 else if (STRNCMP(p_enc, "2byte-", 6) == 0) 3012 p += 6; 3013 3014 if (p[0] == 'c' && p[1] == 'p') 3015 cp = atoi(p + 2); 3016 else if ((idx = enc_canon_search(p)) >= 0) 3017 cp = enc_canon_table[idx].codepage; 3018 else 3019 return 0; 3020 if (IsValidCodePage(cp)) 3021 return cp; 3022 return 0; 3023 } 3024 #endif 3025 3026 # if defined(USE_ICONV) || defined(PROTO) 3027 3028 static char_u *iconv_string __ARGS((vimconv_T *vcp, char_u *str, int slen, int *unconvlenp)); 3029 3030 /* 3031 * Call iconv_open() with a check if iconv() works properly (there are broken 3032 * versions). 3033 * Returns (void *)-1 if failed. 3034 * (should return iconv_t, but that causes problems with prototypes). 3035 */ 3036 void * 3037 my_iconv_open(to, from) 3038 char_u *to; 3039 char_u *from; 3040 { 3041 iconv_t fd; 3042 #define ICONV_TESTLEN 400 3043 char_u tobuf[ICONV_TESTLEN]; 3044 char *p; 3045 size_t tolen; 3046 static int iconv_ok = -1; 3047 3048 if (iconv_ok == FALSE) 3049 return (void *)-1; /* detected a broken iconv() previously */ 3050 3051 #ifdef DYNAMIC_ICONV 3052 /* Check if the iconv.dll can be found. */ 3053 if (!iconv_enabled(TRUE)) 3054 return (void *)-1; 3055 #endif 3056 3057 fd = iconv_open((char *)enc_skip(to), (char *)enc_skip(from)); 3058 3059 if (fd != (iconv_t)-1 && iconv_ok == -1) 3060 { 3061 /* 3062 * Do a dummy iconv() call to check if it actually works. There is a 3063 * version of iconv() on Linux that is broken. We can't ignore it, 3064 * because it's wide-spread. The symptoms are that after outputting 3065 * the initial shift state the "to" pointer is NULL and conversion 3066 * stops for no apparent reason after about 8160 characters. 3067 */ 3068 p = (char *)tobuf; 3069 tolen = ICONV_TESTLEN; 3070 (void)iconv(fd, NULL, NULL, &p, &tolen); 3071 if (p == NULL) 3072 { 3073 iconv_ok = FALSE; 3074 iconv_close(fd); 3075 fd = (iconv_t)-1; 3076 } 3077 else 3078 iconv_ok = TRUE; 3079 } 3080 3081 return (void *)fd; 3082 } 3083 3084 /* 3085 * Convert the string "str[slen]" with iconv(). 3086 * If "unconvlenp" is not NULL handle the string ending in an incomplete 3087 * sequence and set "*unconvlenp" to the length of it. 3088 * Returns the converted string in allocated memory. NULL for an error. 3089 */ 3090 static char_u * 3091 iconv_string(vcp, str, slen, unconvlenp) 3092 vimconv_T *vcp; 3093 char_u *str; 3094 int slen; 3095 int *unconvlenp; 3096 { 3097 const char *from; 3098 size_t fromlen; 3099 char *to; 3100 size_t tolen; 3101 size_t len = 0; 3102 size_t done = 0; 3103 char_u *result = NULL; 3104 char_u *p; 3105 int l; 3106 3107 from = (char *)str; 3108 fromlen = slen; 3109 for (;;) 3110 { 3111 if (len == 0 || ICONV_ERRNO == ICONV_E2BIG) 3112 { 3113 /* Allocate enough room for most conversions. When re-allocating 3114 * increase the buffer size. */ 3115 len = len + fromlen * 2 + 40; 3116 p = alloc((unsigned)len); 3117 if (p != NULL && done > 0) 3118 mch_memmove(p, result, done); 3119 vim_free(result); 3120 result = p; 3121 if (result == NULL) /* out of memory */ 3122 break; 3123 } 3124 3125 to = (char *)result + done; 3126 tolen = len - done - 2; 3127 /* Avoid a warning for systems with a wrong iconv() prototype by 3128 * casting the second argument to void *. */ 3129 if (iconv(vcp->vc_fd, (void *)&from, &fromlen, &to, &tolen) 3130 != (size_t)-1) 3131 { 3132 /* Finished, append a NUL. */ 3133 *to = NUL; 3134 break; 3135 } 3136 3137 /* Check both ICONV_EINVAL and EINVAL, because the dynamically loaded 3138 * iconv library may use one of them. */ 3139 if (!vcp->vc_fail && unconvlenp != NULL 3140 && (ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL)) 3141 { 3142 /* Handle an incomplete sequence at the end. */ 3143 *to = NUL; 3144 *unconvlenp = fromlen; 3145 break; 3146 } 3147 3148 /* Check both ICONV_EILSEQ and EILSEQ, because the dynamically loaded 3149 * iconv library may use one of them. */ 3150 else if (!vcp->vc_fail 3151 && (ICONV_ERRNO == ICONV_EILSEQ || ICONV_ERRNO == EILSEQ 3152 || ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL)) 3153 { 3154 /* Can't convert: insert a '?' and skip a character. This assumes 3155 * conversion from 'encoding' to something else. In other 3156 * situations we don't know what to skip anyway. */ 3157 *to++ = '?'; 3158 if ((*mb_ptr2cells)((char_u *)from) > 1) 3159 *to++ = '?'; 3160 if (enc_utf8) 3161 l = utfc_ptr2len_check_len((char_u *)from, fromlen); 3162 else 3163 { 3164 l = (*mb_ptr2len_check)((char_u *)from); 3165 if (l > (int)fromlen) 3166 l = fromlen; 3167 } 3168 from += l; 3169 fromlen -= l; 3170 } 3171 else if (ICONV_ERRNO != ICONV_E2BIG) 3172 { 3173 /* conversion failed */ 3174 vim_free(result); 3175 result = NULL; 3176 break; 3177 } 3178 /* Not enough room or skipping illegal sequence. */ 3179 done = to - (char *)result; 3180 } 3181 return result; 3182 } 3183 3184 # if defined(DYNAMIC_ICONV) || defined(PROTO) 3185 /* 3186 * Dynamically load the "iconv.dll" on Win32. 3187 */ 3188 3189 #ifndef DYNAMIC_ICONV /* just generating prototypes */ 3190 # define HINSTANCE int 3191 #endif 3192 static HINSTANCE hIconvDLL = 0; 3193 static HINSTANCE hMsvcrtDLL = 0; 3194 3195 # ifndef DYNAMIC_ICONV_DLL 3196 # define DYNAMIC_ICONV_DLL "iconv.dll" 3197 # define DYNAMIC_ICONV_DLL_ALT "libiconv.dll" 3198 # endif 3199 # ifndef DYNAMIC_MSVCRT_DLL 3200 # define DYNAMIC_MSVCRT_DLL "msvcrt.dll" 3201 # endif 3202 3203 /* 3204 * Try opening the iconv.dll and return TRUE if iconv() can be used. 3205 */ 3206 int 3207 iconv_enabled(verbose) 3208 int verbose; 3209 { 3210 if (hIconvDLL != 0 && hMsvcrtDLL != 0) 3211 return TRUE; 3212 hIconvDLL = LoadLibrary(DYNAMIC_ICONV_DLL); 3213 if (hIconvDLL == 0) /* sometimes it's called libiconv.dll */ 3214 hIconvDLL = LoadLibrary(DYNAMIC_ICONV_DLL_ALT); 3215 if (hIconvDLL != 0) 3216 hMsvcrtDLL = LoadLibrary(DYNAMIC_MSVCRT_DLL); 3217 if (hIconvDLL == 0 || hMsvcrtDLL == 0) 3218 { 3219 /* Only give the message when 'verbose' is set, otherwise it might be 3220 * done whenever a conversion is attempted. */ 3221 if (verbose && p_verbose > 0) 3222 { 3223 verbose_enter(); 3224 EMSG2(_(e_loadlib), 3225 hIconvDLL == 0 ? DYNAMIC_ICONV_DLL : DYNAMIC_MSVCRT_DLL); 3226 verbose_leave(); 3227 } 3228 iconv_end(); 3229 return FALSE; 3230 } 3231 3232 iconv = (void *)GetProcAddress(hIconvDLL, "libiconv"); 3233 iconv_open = (void *)GetProcAddress(hIconvDLL, "libiconv_open"); 3234 iconv_close = (void *)GetProcAddress(hIconvDLL, "libiconv_close"); 3235 iconvctl = (void *)GetProcAddress(hIconvDLL, "libiconvctl"); 3236 iconv_errno = (void *)GetProcAddress(hMsvcrtDLL, "_errno"); 3237 if (iconv == NULL || iconv_open == NULL || iconv_close == NULL 3238 || iconvctl == NULL || iconv_errno == NULL) 3239 { 3240 iconv_end(); 3241 if (verbose && p_verbose > 0) 3242 { 3243 verbose_enter(); 3244 EMSG2(_(e_loadfunc), "for libiconv"); 3245 verbose_leave(); 3246 } 3247 return FALSE; 3248 } 3249 return TRUE; 3250 } 3251 3252 void 3253 iconv_end() 3254 { 3255 /* Don't use iconv() when inputting or outputting characters. */ 3256 if (input_conv.vc_type == CONV_ICONV) 3257 convert_setup(&input_conv, NULL, NULL); 3258 if (output_conv.vc_type == CONV_ICONV) 3259 convert_setup(&output_conv, NULL, NULL); 3260 3261 if (hIconvDLL != 0) 3262 FreeLibrary(hIconvDLL); 3263 if (hMsvcrtDLL != 0) 3264 FreeLibrary(hMsvcrtDLL); 3265 hIconvDLL = 0; 3266 hMsvcrtDLL = 0; 3267 } 3268 # endif /* DYNAMIC_ICONV */ 3269 # endif /* USE_ICONV */ 3270 3271 #endif /* FEAT_MBYTE */ 3272 3273 #if defined(FEAT_XIM) || defined(PROTO) 3274 3275 # ifdef FEAT_GUI_GTK 3276 static int xim_has_preediting INIT(= FALSE); /* IM current status */ 3277 3278 /* 3279 * Set preedit_start_col to the current cursor position. 3280 */ 3281 static void 3282 init_preedit_start_col(void) 3283 { 3284 if (State & CMDLINE) 3285 preedit_start_col = cmdline_getvcol_cursor(); 3286 else if (curwin != NULL) 3287 getvcol(curwin, &curwin->w_cursor, &preedit_start_col, NULL, NULL); 3288 /* Prevent that preediting marks the buffer as changed. */ 3289 xim_changed_while_preediting = curbuf->b_changed; 3290 } 3291 # endif 3292 3293 # if defined(HAVE_GTK2) && !defined(PROTO) 3294 3295 static int im_is_active = FALSE; /* IM is enabled for current mode */ 3296 static int im_preedit_cursor = 0; /* cursor offset in characters */ 3297 static int im_preedit_trailing = 0; /* number of characters after cursor */ 3298 3299 static unsigned long im_commit_handler_id = 0; 3300 static unsigned int im_activatekey_keyval = GDK_VoidSymbol; 3301 static unsigned int im_activatekey_state = 0; 3302 3303 void 3304 im_set_active(int active) 3305 { 3306 int was_active; 3307 3308 was_active = !!im_is_active; 3309 im_is_active = (active && !p_imdisable); 3310 3311 if (im_is_active != was_active) 3312 xim_reset(); 3313 } 3314 3315 void 3316 xim_set_focus(int focus) 3317 { 3318 if (xic != NULL) 3319 { 3320 if (focus) 3321 gtk_im_context_focus_in(xic); 3322 else 3323 gtk_im_context_focus_out(xic); 3324 } 3325 } 3326 3327 void 3328 im_set_position(int row, int col) 3329 { 3330 if (xic != NULL) 3331 { 3332 GdkRectangle area; 3333 3334 area.x = FILL_X(col); 3335 area.y = FILL_Y(row); 3336 area.width = gui.char_width * (mb_lefthalve(row, col) ? 2 : 1); 3337 area.height = gui.char_height; 3338 3339 gtk_im_context_set_cursor_location(xic, &area); 3340 } 3341 } 3342 3343 # if 0 || defined(PROTO) /* apparently only used in gui_x11.c */ 3344 void 3345 xim_set_preedit(void) 3346 { 3347 im_set_position(gui.row, gui.col); 3348 } 3349 # endif 3350 3351 static void 3352 im_add_to_input(char_u *str, int len) 3353 { 3354 /* Convert from 'termencoding' (always "utf-8") to 'encoding' */ 3355 if (input_conv.vc_type != CONV_NONE) 3356 { 3357 str = string_convert(&input_conv, str, &len); 3358 g_return_if_fail(str != NULL); 3359 } 3360 3361 add_to_input_buf_csi(str, len); 3362 3363 if (input_conv.vc_type != CONV_NONE) 3364 vim_free(str); 3365 3366 if (p_mh) /* blank out the pointer if necessary */ 3367 gui_mch_mousehide(TRUE); 3368 } 3369 3370 static void 3371 im_delete_preedit(void) 3372 { 3373 char_u bskey[] = {CSI, 'k', 'b'}; 3374 char_u delkey[] = {CSI, 'k', 'D'}; 3375 3376 if (State & NORMAL) 3377 { 3378 im_preedit_cursor = 0; 3379 return; 3380 } 3381 for (; im_preedit_cursor > 0; --im_preedit_cursor) 3382 add_to_input_buf(bskey, (int)sizeof(bskey)); 3383 3384 for (; im_preedit_trailing > 0; --im_preedit_trailing) 3385 add_to_input_buf(delkey, (int)sizeof(delkey)); 3386 } 3387 3388 static void 3389 im_correct_cursor(int num_move_back) 3390 { 3391 char_u backkey[] = {CSI, 'k', 'l'}; 3392 3393 if (State & NORMAL) 3394 return; 3395 # ifdef FEAT_RIGHTLEFT 3396 if ((State & CMDLINE) == 0 && curwin != NULL && curwin->w_p_rl) 3397 backkey[2] = 'r'; 3398 # endif 3399 for (; num_move_back > 0; --num_move_back) 3400 add_to_input_buf(backkey, (int)sizeof(backkey)); 3401 } 3402 3403 static int xim_expected_char = NUL; 3404 static int xim_ignored_char = FALSE; 3405 3406 /* 3407 * Update the mode and cursor while in an IM callback. 3408 */ 3409 static void 3410 im_show_info(void) 3411 { 3412 int old_vgetc_busy; 3413 old_vgetc_busy = vgetc_busy; 3414 vgetc_busy = TRUE; 3415 showmode(); 3416 vgetc_busy = old_vgetc_busy; 3417 setcursor(); 3418 out_flush(); 3419 } 3420 3421 /* 3422 * Callback invoked when the user finished preediting. 3423 * Put the final string into the input buffer. 3424 */ 3425 /*ARGSUSED0*/ 3426 static void 3427 im_commit_cb(GtkIMContext *context, const gchar *str, gpointer data) 3428 { 3429 int slen = (int)STRLEN(str); 3430 int add_to_input = TRUE; 3431 int clen; 3432 int len = slen; 3433 int commit_with_preedit = TRUE; 3434 char_u *im_str, *p; 3435 3436 #ifdef XIM_DEBUG 3437 xim_log("im_commit_cb(): %s\n", str); 3438 #endif 3439 3440 /* The imhangul module doesn't reset the preedit string before 3441 * committing. Call im_delete_preedit() to work around that. */ 3442 im_delete_preedit(); 3443 3444 /* Indicate that preediting has finished. */ 3445 if (preedit_start_col == MAXCOL) 3446 { 3447 init_preedit_start_col(); 3448 commit_with_preedit = FALSE; 3449 } 3450 3451 /* The thing which setting "preedit_start_col" to MAXCOL means that 3452 * "preedit_start_col" will be set forcely when calling 3453 * preedit_changed_cb() next time. 3454 * "preedit_start_col" should not reset with MAXCOL on this part. Vim 3455 * is simulating the preediting by using add_to_input_str(). when 3456 * preedit begin immediately before committed, the typebuf is not 3457 * flushed to screen, then it can't get correct "preedit_start_col". 3458 * Thus, it should calculate the cells by adding cells of the committed 3459 * string. */ 3460 if (input_conv.vc_type != CONV_NONE) 3461 { 3462 im_str = string_convert(&input_conv, (char_u *)str, &len); 3463 g_return_if_fail(im_str != NULL); 3464 } 3465 else 3466 im_str = (char_u *)str; 3467 clen = 0; 3468 for (p = im_str; p < im_str + len; p += (*mb_ptr2len_check)(p)) 3469 clen += (*mb_ptr2cells)(p); 3470 if (input_conv.vc_type != CONV_NONE) 3471 vim_free(im_str); 3472 preedit_start_col += clen; 3473 3474 /* Is this a single character that matches a keypad key that's just 3475 * been pressed? If so, we don't want it to be entered as such - let 3476 * us carry on processing the raw keycode so that it may be used in 3477 * mappings as <kSomething>. */ 3478 if (xim_expected_char != NUL) 3479 { 3480 /* We're currently processing a keypad or other special key */ 3481 if (slen == 1 && str[0] == xim_expected_char) 3482 { 3483 /* It's a match - don't do it here */ 3484 xim_ignored_char = TRUE; 3485 add_to_input = FALSE; 3486 } 3487 else 3488 { 3489 /* Not a match */ 3490 xim_ignored_char = FALSE; 3491 } 3492 } 3493 3494 if (add_to_input) 3495 im_add_to_input((char_u *)str, slen); 3496 3497 /* Inserting chars while "im_is_active" is set does not cause a change of 3498 * buffer. When the chars are committed the buffer must be marked as 3499 * changed. */ 3500 if (!commit_with_preedit) 3501 preedit_start_col = MAXCOL; 3502 3503 /* This flag is used in changed() at next call. */ 3504 xim_changed_while_preediting = TRUE; 3505 3506 if (gtk_main_level() > 0) 3507 gtk_main_quit(); 3508 } 3509 3510 /* 3511 * Callback invoked after start to the preedit. 3512 */ 3513 /*ARGSUSED*/ 3514 static void 3515 im_preedit_start_cb(GtkIMContext *context, gpointer data) 3516 { 3517 #ifdef XIM_DEBUG 3518 xim_log("im_preedit_start_cb()\n"); 3519 #endif 3520 3521 im_is_active = TRUE; 3522 gui_update_cursor(TRUE, FALSE); 3523 } 3524 3525 /* 3526 * Callback invoked after end to the preedit. 3527 */ 3528 /*ARGSUSED*/ 3529 static void 3530 im_preedit_end_cb(GtkIMContext *context, gpointer data) 3531 { 3532 #ifdef XIM_DEBUG 3533 xim_log("im_preedit_end_cb()\n"); 3534 #endif 3535 im_delete_preedit(); 3536 3537 /* Indicate that preediting has finished */ 3538 preedit_start_col = MAXCOL; 3539 xim_has_preediting = FALSE; 3540 3541 im_is_active = FALSE; 3542 gui_update_cursor(TRUE, FALSE); 3543 im_show_info(); 3544 } 3545 3546 /* 3547 * Callback invoked after changes to the preedit string. If the preedit 3548 * string was empty before, remember the preedit start column so we know 3549 * where to apply feedback attributes. Delete the previous preedit string 3550 * if there was one, save the new preedit cursor offset, and put the new 3551 * string into the input buffer. 3552 * 3553 * TODO: The pragmatic "put into input buffer" approach used here has 3554 * several fundamental problems: 3555 * 3556 * - The characters in the preedit string are subject to remapping. 3557 * That's broken, only the finally committed string should be remapped. 3558 * 3559 * - There is a race condition involved: The retrieved value for the 3560 * current cursor position will be wrong if any unprocessed characters 3561 * are still queued in the input buffer. 3562 * 3563 * - Due to the lack of synchronization between the file buffer in memory 3564 * and any typed characters, it's practically impossible to implement the 3565 * "retrieve_surrounding" and "delete_surrounding" signals reliably. IM 3566 * modules for languages such as Thai are likely to rely on this feature 3567 * for proper operation. 3568 * 3569 * Conclusions: I think support for preediting needs to be moved to the 3570 * core parts of Vim. Ideally, until it has been committed, the preediting 3571 * string should only be displayed and not affect the buffer content at all. 3572 * The question how to deal with the synchronization issue still remains. 3573 * Circumventing the input buffer is probably not desirable. Anyway, I think 3574 * implementing "retrieve_surrounding" is the only hard problem. 3575 * 3576 * One way to solve all of this in a clean manner would be to queue all key 3577 * press/release events "as is" in the input buffer, and apply the IM filtering 3578 * at the receiving end of the queue. This, however, would have a rather large 3579 * impact on the code base. If there is an easy way to force processing of all 3580 * remaining input from within the "retrieve_surrounding" signal handler, this 3581 * might not be necessary. Gotta ask on vim-dev for opinions. 3582 */ 3583 /*ARGSUSED1*/ 3584 static void 3585 im_preedit_changed_cb(GtkIMContext *context, gpointer data) 3586 { 3587 char *preedit_string = NULL; 3588 int cursor_index = 0; 3589 int num_move_back = 0; 3590 char_u *str; 3591 char_u *p; 3592 int i; 3593 3594 gtk_im_context_get_preedit_string(context, 3595 &preedit_string, NULL, 3596 &cursor_index); 3597 3598 #ifdef XIM_DEBUG 3599 xim_log("im_preedit_changed_cb(): %s\n", preedit_string); 3600 #endif 3601 3602 g_return_if_fail(preedit_string != NULL); /* just in case */ 3603 3604 /* If preedit_start_col is MAXCOL set it to the current cursor position. */ 3605 if (preedit_start_col == MAXCOL && preedit_string[0] != '\0') 3606 { 3607 xim_has_preediting = TRUE; 3608 3609 /* Urgh, this breaks if the input buffer isn't empty now */ 3610 init_preedit_start_col(); 3611 } 3612 else if (cursor_index == 0 && preedit_string[0] == '\0') 3613 { 3614 if (preedit_start_col == MAXCOL) 3615 xim_has_preediting = FALSE; 3616 3617 /* If at the start position (after typing backspace) 3618 * preedit_start_col must be reset. */ 3619 preedit_start_col = MAXCOL; 3620 } 3621 3622 im_delete_preedit(); 3623 3624 /* 3625 * Compute the end of the preediting area: "preedit_end_col". 3626 * According to the documentation of gtk_im_context_get_preedit_string(), 3627 * the cursor_pos output argument returns the offset in bytes. This is 3628 * unfortunately not true -- real life shows the offset is in characters, 3629 * and the GTK+ source code agrees with me. Will file a bug later. 3630 */ 3631 if (preedit_start_col != MAXCOL) 3632 preedit_end_col = preedit_start_col; 3633 str = (char_u *)preedit_string; 3634 for (p = str, i = 0; *p != NUL; p += utf_byte2len(*p), ++i) 3635 { 3636 int is_composing; 3637 3638 is_composing = ((*p & 0x80) != 0 && utf_iscomposing(utf_ptr2char(p))); 3639 /* 3640 * These offsets are used as counters when generating <BS> and <Del> 3641 * to delete the preedit string. So don't count composing characters 3642 * unless 'delcombine' is enabled. 3643 */ 3644 if (!is_composing || p_deco) 3645 { 3646 if (i < cursor_index) 3647 ++im_preedit_cursor; 3648 else 3649 ++im_preedit_trailing; 3650 } 3651 if (!is_composing && i >= cursor_index) 3652 { 3653 /* This is essentially the same as im_preedit_trailing, except 3654 * composing characters are not counted even if p_deco is set. */ 3655 ++num_move_back; 3656 } 3657 if (preedit_start_col != MAXCOL) 3658 preedit_end_col += utf_ptr2cells(p); 3659 } 3660 3661 if (p > str) 3662 { 3663 im_add_to_input(str, (int)(p - str)); 3664 im_correct_cursor(num_move_back); 3665 } 3666 3667 g_free(preedit_string); 3668 3669 if (gtk_main_level() > 0) 3670 gtk_main_quit(); 3671 } 3672 3673 /* 3674 * Translate the Pango attributes at iter to Vim highlighting attributes. 3675 * Ignore attributes not supported by Vim highlighting. This shouldn't have 3676 * too much impact -- right now we handle even more attributes than necessary 3677 * for the IM modules I tested with. 3678 */ 3679 static int 3680 translate_pango_attributes(PangoAttrIterator *iter) 3681 { 3682 PangoAttribute *attr; 3683 int char_attr = HL_NORMAL; 3684 3685 attr = pango_attr_iterator_get(iter, PANGO_ATTR_UNDERLINE); 3686 if (attr != NULL && ((PangoAttrInt *)attr)->value 3687 != (int)PANGO_UNDERLINE_NONE) 3688 char_attr |= HL_UNDERLINE; 3689 3690 attr = pango_attr_iterator_get(iter, PANGO_ATTR_WEIGHT); 3691 if (attr != NULL && ((PangoAttrInt *)attr)->value >= (int)PANGO_WEIGHT_BOLD) 3692 char_attr |= HL_BOLD; 3693 3694 attr = pango_attr_iterator_get(iter, PANGO_ATTR_STYLE); 3695 if (attr != NULL && ((PangoAttrInt *)attr)->value 3696 != (int)PANGO_STYLE_NORMAL) 3697 char_attr |= HL_ITALIC; 3698 3699 attr = pango_attr_iterator_get(iter, PANGO_ATTR_BACKGROUND); 3700 if (attr != NULL) 3701 { 3702 const PangoColor *color = &((PangoAttrColor *)attr)->color; 3703 3704 /* Assume inverse if black background is requested */ 3705 if ((color->red | color->green | color->blue) == 0) 3706 char_attr |= HL_INVERSE; 3707 } 3708 3709 return char_attr; 3710 } 3711 3712 /* 3713 * Retrieve the highlighting attributes at column col in the preedit string. 3714 * Return -1 if not in preediting mode or if col is out of range. 3715 */ 3716 int 3717 im_get_feedback_attr(int col) 3718 { 3719 char *preedit_string = NULL; 3720 PangoAttrList *attr_list = NULL; 3721 int char_attr = -1; 3722 3723 if (xic == NULL) 3724 return char_attr; 3725 3726 gtk_im_context_get_preedit_string(xic, &preedit_string, &attr_list, NULL); 3727 3728 if (preedit_string != NULL && attr_list != NULL) 3729 { 3730 int index; 3731 3732 /* Get the byte index as used by PangoAttrIterator */ 3733 for (index = 0; col > 0 && preedit_string[index] != '\0'; --col) 3734 index += utfc_ptr2len_check((char_u *)preedit_string + index); 3735 3736 if (preedit_string[index] != '\0') 3737 { 3738 PangoAttrIterator *iter; 3739 int start, end; 3740 3741 char_attr = HL_NORMAL; 3742 iter = pango_attr_list_get_iterator(attr_list); 3743 3744 /* Extract all relevant attributes from the list. */ 3745 do 3746 { 3747 pango_attr_iterator_range(iter, &start, &end); 3748 3749 if (index >= start && index < end) 3750 char_attr |= translate_pango_attributes(iter); 3751 } 3752 while (pango_attr_iterator_next(iter)); 3753 3754 pango_attr_iterator_destroy(iter); 3755 } 3756 } 3757 3758 if (attr_list != NULL) 3759 pango_attr_list_unref(attr_list); 3760 g_free(preedit_string); 3761 3762 return char_attr; 3763 } 3764 3765 void 3766 xim_init(void) 3767 { 3768 #ifdef XIM_DEBUG 3769 xim_log("xim_init()\n"); 3770 #endif 3771 3772 g_return_if_fail(gui.drawarea != NULL); 3773 g_return_if_fail(gui.drawarea->window != NULL); 3774 3775 xic = gtk_im_multicontext_new(); 3776 g_object_ref(xic); 3777 3778 im_commit_handler_id = g_signal_connect(G_OBJECT(xic), "commit", 3779 G_CALLBACK(&im_commit_cb), NULL); 3780 g_signal_connect(G_OBJECT(xic), "preedit_changed", 3781 G_CALLBACK(&im_preedit_changed_cb), NULL); 3782 g_signal_connect(G_OBJECT(xic), "preedit_start", 3783 G_CALLBACK(&im_preedit_start_cb), NULL); 3784 g_signal_connect(G_OBJECT(xic), "preedit_end", 3785 G_CALLBACK(&im_preedit_end_cb), NULL); 3786 3787 gtk_im_context_set_client_window(xic, gui.drawarea->window); 3788 } 3789 3790 void 3791 im_shutdown(void) 3792 { 3793 #ifdef XIM_DEBUG 3794 xim_log("im_shutdown()\n"); 3795 #endif 3796 3797 if (xic != NULL) 3798 { 3799 gtk_im_context_focus_out(xic); 3800 g_object_unref(xic); 3801 xic = NULL; 3802 } 3803 im_is_active = FALSE; 3804 im_commit_handler_id = 0; 3805 preedit_start_col = MAXCOL; 3806 xim_has_preediting = FALSE; 3807 } 3808 3809 /* 3810 * Convert the string argument to keyval and state for GdkEventKey. 3811 * If str is valid return TRUE, otherwise FALSE. 3812 * 3813 * See 'imactivatekey' for documentation of the format. 3814 */ 3815 static int 3816 im_string_to_keyval(const char *str, unsigned int *keyval, unsigned int *state) 3817 { 3818 const char *mods_end; 3819 unsigned tmp_keyval; 3820 unsigned tmp_state = 0; 3821 3822 mods_end = strrchr(str, '-'); 3823 mods_end = (mods_end != NULL) ? mods_end + 1 : str; 3824 3825 /* Parse modifier keys */ 3826 while (str < mods_end) 3827 switch (*str++) 3828 { 3829 case '-': break; 3830 case 'S': case 's': tmp_state |= (unsigned)GDK_SHIFT_MASK; break; 3831 case 'L': case 'l': tmp_state |= (unsigned)GDK_LOCK_MASK; break; 3832 case 'C': case 'c': tmp_state |= (unsigned)GDK_CONTROL_MASK;break; 3833 case '1': tmp_state |= (unsigned)GDK_MOD1_MASK; break; 3834 case '2': tmp_state |= (unsigned)GDK_MOD2_MASK; break; 3835 case '3': tmp_state |= (unsigned)GDK_MOD3_MASK; break; 3836 case '4': tmp_state |= (unsigned)GDK_MOD4_MASK; break; 3837 case '5': tmp_state |= (unsigned)GDK_MOD5_MASK; break; 3838 default: 3839 return FALSE; 3840 } 3841 3842 tmp_keyval = gdk_keyval_from_name(str); 3843 3844 if (tmp_keyval == 0 || tmp_keyval == GDK_VoidSymbol) 3845 return FALSE; 3846 3847 if (keyval != NULL) 3848 *keyval = tmp_keyval; 3849 if (state != NULL) 3850 *state = tmp_state; 3851 3852 return TRUE; 3853 } 3854 3855 /* 3856 * Return TRUE if p_imak is valid, otherwise FALSE. As a special case, an 3857 * empty string is also regarded as valid. 3858 * 3859 * Note: The numerical key value of p_imak is cached if it was valid; thus 3860 * boldly assuming im_xim_isvalid_imactivate() will always be called whenever 3861 * 'imak' changes. This is currently the case but not obvious -- should 3862 * probably rename the function for clarity. 3863 */ 3864 int 3865 im_xim_isvalid_imactivate(void) 3866 { 3867 if (p_imak[0] == NUL) 3868 { 3869 im_activatekey_keyval = GDK_VoidSymbol; 3870 im_activatekey_state = 0; 3871 return TRUE; 3872 } 3873 3874 return im_string_to_keyval((const char *)p_imak, 3875 &im_activatekey_keyval, 3876 &im_activatekey_state); 3877 } 3878 3879 static void 3880 im_synthesize_keypress(unsigned int keyval, unsigned int state) 3881 { 3882 GdkEventKey *event; 3883 3884 # ifdef HAVE_GTK_MULTIHEAD 3885 event = (GdkEventKey *)gdk_event_new(GDK_KEY_PRESS); 3886 g_object_ref(gui.drawarea->window); /* unreffed by gdk_event_free() */ 3887 # else 3888 event = (GdkEventKey *)g_malloc0((gulong)sizeof(GdkEvent)); 3889 event->type = GDK_KEY_PRESS; 3890 # endif 3891 event->window = gui.drawarea->window; 3892 event->send_event = TRUE; 3893 event->time = GDK_CURRENT_TIME; 3894 event->state = state; 3895 event->keyval = keyval; 3896 event->hardware_keycode = /* needed for XIM */ 3897 XKeysymToKeycode(GDK_WINDOW_XDISPLAY(event->window), (KeySym)keyval); 3898 event->length = 0; 3899 event->string = NULL; 3900 3901 gtk_im_context_filter_keypress(xic, event); 3902 3903 /* For consistency, also send the corresponding release event. */ 3904 event->type = GDK_KEY_RELEASE; 3905 event->send_event = FALSE; 3906 gtk_im_context_filter_keypress(xic, event); 3907 3908 # ifdef HAVE_GTK_MULTIHEAD 3909 gdk_event_free((GdkEvent *)event); 3910 # else 3911 g_free(event); 3912 # endif 3913 } 3914 3915 void 3916 xim_reset(void) 3917 { 3918 if (xic != NULL) 3919 { 3920 /* 3921 * The third-party imhangul module (and maybe others too) ignores 3922 * gtk_im_context_reset() or at least doesn't reset the active state. 3923 * Thus sending imactivatekey would turn it off if it was on before, 3924 * which is clearly not what we want. Fortunately we can work around 3925 * that for imhangul by sending GDK_Escape, but I don't know if it 3926 * works with all IM modules that support an activation key :/ 3927 * 3928 * An alternative approach would be to destroy the IM context and 3929 * recreate it. But that means loading/unloading the IM module on 3930 * every mode switch, which causes a quite noticable delay even on 3931 * my rather fast box... 3932 * * 3933 * Moreover, there are some XIM which cannot respond to 3934 * im_synthesize_keypress(). we hope that they reset by 3935 * xim_shutdown(). 3936 */ 3937 if (im_activatekey_keyval != GDK_VoidSymbol && im_is_active) 3938 im_synthesize_keypress(GDK_Escape, 0U); 3939 3940 gtk_im_context_reset(xic); 3941 3942 /* 3943 * HACK for Ami: This sequence of function calls makes Ami handle 3944 * the IM reset gratiously, without breaking loads of other stuff. 3945 * It seems to force English mode as well, which is exactly what we 3946 * want because it makes the Ami status display work reliably. 3947 */ 3948 gtk_im_context_set_use_preedit(xic, FALSE); 3949 3950 if (p_imdisable) 3951 im_shutdown(); 3952 else 3953 { 3954 gtk_im_context_set_use_preedit(xic, TRUE); 3955 xim_set_focus(gui.in_focus); 3956 3957 if (im_activatekey_keyval != GDK_VoidSymbol) 3958 { 3959 if (im_is_active) 3960 { 3961 g_signal_handler_block(xic, im_commit_handler_id); 3962 im_synthesize_keypress(im_activatekey_keyval, 3963 im_activatekey_state); 3964 g_signal_handler_unblock(xic, im_commit_handler_id); 3965 } 3966 } 3967 else 3968 { 3969 im_shutdown(); 3970 xim_init(); 3971 xim_set_focus(gui.in_focus); 3972 } 3973 } 3974 } 3975 3976 preedit_start_col = MAXCOL; 3977 xim_has_preediting = FALSE; 3978 } 3979 3980 int 3981 xim_queue_key_press_event(GdkEventKey *event, int down) 3982 { 3983 if (down) 3984 { 3985 /* 3986 * Workaround GTK2 XIM 'feature' that always converts keypad keys to 3987 * chars., even when not part of an IM sequence (ref. feature of 3988 * gdk/gdkkeyuni.c). 3989 * Flag any keypad keys that might represent a single char. 3990 * If this (on its own - i.e., not part of an IM sequence) is 3991 * committed while we're processing one of these keys, we can ignore 3992 * that commit and go ahead & process it ourselves. That way we can 3993 * still distinguish keypad keys for use in mappings. 3994 */ 3995 switch (event->keyval) 3996 { 3997 case GDK_KP_Add: xim_expected_char = '+'; break; 3998 case GDK_KP_Subtract: xim_expected_char = '-'; break; 3999 case GDK_KP_Divide: xim_expected_char = '/'; break; 4000 case GDK_KP_Multiply: xim_expected_char = '*'; break; 4001 case GDK_KP_Decimal: xim_expected_char = '.'; break; 4002 case GDK_KP_Equal: xim_expected_char = '='; break; 4003 case GDK_KP_0: xim_expected_char = '0'; break; 4004 case GDK_KP_1: xim_expected_char = '1'; break; 4005 case GDK_KP_2: xim_expected_char = '2'; break; 4006 case GDK_KP_3: xim_expected_char = '3'; break; 4007 case GDK_KP_4: xim_expected_char = '4'; break; 4008 case GDK_KP_5: xim_expected_char = '5'; break; 4009 case GDK_KP_6: xim_expected_char = '6'; break; 4010 case GDK_KP_7: xim_expected_char = '7'; break; 4011 case GDK_KP_8: xim_expected_char = '8'; break; 4012 case GDK_KP_9: xim_expected_char = '9'; break; 4013 default: xim_expected_char = NUL; 4014 } 4015 xim_ignored_char = FALSE; 4016 } 4017 4018 /* 4019 * When typing fFtT, XIM may be activated. Thus it must pass 4020 * gtk_im_context_filter_keypress() in Normal mode. 4021 * And while doing :sh too. 4022 */ 4023 if (xic != NULL && !p_imdisable 4024 && (State & (INSERT | CMDLINE | NORMAL | EXTERNCMD)) != 0) 4025 { 4026 /* 4027 * Filter 'imactivatekey' and map it to CTRL-^. This way, Vim is 4028 * always aware of the current status of IM, and can even emulate 4029 * the activation key for modules that don't support one. 4030 */ 4031 if (event->keyval == im_activatekey_keyval 4032 && (event->state & im_activatekey_state) == im_activatekey_state) 4033 { 4034 unsigned int state_mask; 4035 4036 /* Require the state of the 3 most used modifiers to match exactly. 4037 * Otherwise e.g. <S-C-space> would be unusable for other purposes 4038 * if the IM activate key is <S-space>. */ 4039 state_mask = im_activatekey_state; 4040 state_mask |= ((int)GDK_SHIFT_MASK | (int)GDK_CONTROL_MASK 4041 | (int)GDK_MOD1_MASK); 4042 4043 if ((event->state & state_mask) != im_activatekey_state) 4044 return FALSE; 4045 4046 /* Don't send it a second time on GDK_KEY_RELEASE. */ 4047 if (event->type != GDK_KEY_PRESS) 4048 return TRUE; 4049 4050 if (map_to_exists_mode((char_u *)"", LANGMAP)) 4051 { 4052 im_set_active(FALSE); 4053 4054 /* ":lmap" mappings exists, toggle use of mappings. */ 4055 State ^= LANGMAP; 4056 if (State & LANGMAP) 4057 { 4058 curbuf->b_p_iminsert = B_IMODE_NONE; 4059 State &= ~LANGMAP; 4060 } 4061 else 4062 { 4063 curbuf->b_p_iminsert = B_IMODE_LMAP; 4064 State |= LANGMAP; 4065 } 4066 return TRUE; 4067 } 4068 4069 return gtk_im_context_filter_keypress(xic, event); 4070 } 4071 4072 /* Don't filter events through the IM context if IM isn't active 4073 * right now. Unlike with GTK+ 1.2 we cannot rely on the IM module 4074 * not doing anything before the activation key was sent. */ 4075 if (im_activatekey_keyval == GDK_VoidSymbol || im_is_active) 4076 { 4077 int imresult = gtk_im_context_filter_keypress(xic, event); 4078 4079 /* Some XIM send following sequence: 4080 * 1. preedited string. 4081 * 2. committed string. 4082 * 3. line changed key. 4083 * 4. preedited string. 4084 * 5. remove preedited string. 4085 * if 3, Vim can't move back the above line for 5. 4086 * thus, this part should not parse the key. */ 4087 if (!imresult && preedit_start_col != MAXCOL 4088 && event->keyval == GDK_Return) 4089 { 4090 im_synthesize_keypress(GDK_Return, 0U); 4091 return FALSE; 4092 } 4093 4094 /* If XIM tried to commit a keypad key as a single char., 4095 * ignore it so we can use the keypad key 'raw', for mappings. */ 4096 if (xim_expected_char != NUL && xim_ignored_char) 4097 /* We had a keypad key, and XIM tried to thieve it */ 4098 return FALSE; 4099 4100 /* Normal processing */ 4101 return imresult; 4102 } 4103 } 4104 4105 return FALSE; 4106 } 4107 4108 int 4109 im_get_status(void) 4110 { 4111 return im_is_active; 4112 } 4113 4114 # else /* !HAVE_GTK2 */ 4115 4116 static int xim_is_active = FALSE; /* XIM should be active in the current 4117 mode */ 4118 static int xim_has_focus = FALSE; /* XIM is really being used for Vim */ 4119 #ifdef FEAT_GUI_X11 4120 static XIMStyle input_style; 4121 static int status_area_enabled = TRUE; 4122 #endif 4123 4124 #ifdef FEAT_GUI_GTK 4125 # ifdef WIN3264 4126 # include <gdk/gdkwin32.h> 4127 # else 4128 # include <gdk/gdkx.h> 4129 # endif 4130 #else 4131 # ifdef PROTO 4132 /* Define a few things to be able to generate prototypes while not configured 4133 * for GTK. */ 4134 # define GSList int 4135 # define gboolean int 4136 typedef int GdkEvent; 4137 typedef int GdkEventKey; 4138 # define GdkIC int 4139 # endif 4140 #endif 4141 4142 #if defined(FEAT_GUI_GTK) || defined(PROTO) || defined(FEAT_GUI_KDE) 4143 static int preedit_buf_len = 0; 4144 static int xim_can_preediting INIT(= FALSE); /* XIM in showmode() */ 4145 static int xim_input_style; 4146 #ifndef FEAT_GUI_GTK 4147 # define gboolean int 4148 #endif 4149 static gboolean use_status_area = 0; 4150 4151 static int im_xim_str2keycode __ARGS((unsigned int *code, unsigned int *state)); 4152 static void im_xim_send_event_imactivate __ARGS((void)); 4153 4154 /* 4155 * Convert string to keycode and state for XKeyEvent. 4156 * When string is valid return OK, when invalid return FAIL. 4157 * 4158 * See 'imactivatekey' documentation for the format. 4159 */ 4160 static int 4161 im_xim_str2keycode(code, state) 4162 unsigned int *code; 4163 unsigned int *state; 4164 { 4165 int retval = OK; 4166 int len; 4167 unsigned keycode = 0, keystate = 0; 4168 Window window; 4169 Display *display; 4170 char_u *flag_end; 4171 char_u *str; 4172 4173 if (*p_imak != NUL) 4174 { 4175 len = STRLEN(p_imak); 4176 for (flag_end = p_imak + len - 1; 4177 flag_end > p_imak && *flag_end != '-'; --flag_end) 4178 ; 4179 4180 /* Parse modifier keys */ 4181 for (str = p_imak; str < flag_end; ++str) 4182 { 4183 switch (*str) 4184 { 4185 case 's': case 'S': 4186 keystate |= ShiftMask; 4187 break; 4188 case 'l': case 'L': 4189 keystate |= LockMask; 4190 break; 4191 case 'c': case 'C': 4192 keystate |= ControlMask; 4193 break; 4194 case '1': 4195 keystate |= Mod1Mask; 4196 break; 4197 case '2': 4198 keystate |= Mod2Mask; 4199 break; 4200 case '3': 4201 keystate |= Mod3Mask; 4202 break; 4203 case '4': 4204 keystate |= Mod4Mask; 4205 break; 4206 case '5': 4207 keystate |= Mod5Mask; 4208 break; 4209 case '-': 4210 break; 4211 default: 4212 retval = FAIL; 4213 } 4214 } 4215 if (*str == '-') 4216 ++str; 4217 4218 /* Get keycode from string. */ 4219 gui_get_x11_windis(&window, &display); 4220 if (display) 4221 keycode = XKeysymToKeycode(display, XStringToKeysym((char *)str)); 4222 if (keycode == 0) 4223 retval = FAIL; 4224 4225 if (code != NULL) 4226 *code = keycode; 4227 if (state != NULL) 4228 *state = keystate; 4229 } 4230 return retval; 4231 } 4232 4233 static void 4234 im_xim_send_event_imactivate() 4235 { 4236 /* Force turn on preedit state by symulate keypress event. 4237 * Keycode and state is specified by 'imactivatekey'. 4238 */ 4239 XKeyEvent ev; 4240 4241 gui_get_x11_windis(&ev.window, &ev.display); 4242 ev.root = RootWindow(ev.display, DefaultScreen(ev.display)); 4243 ev.subwindow = None; 4244 ev.time = CurrentTime; 4245 ev.x = 1; 4246 ev.y = 1; 4247 ev.x_root = 1; 4248 ev.y_root = 1; 4249 ev.same_screen = 1; 4250 ev.type = KeyPress; 4251 if (im_xim_str2keycode(&ev.keycode, &ev.state) == OK) 4252 XSendEvent(ev.display, ev.window, 1, KeyPressMask, (XEvent*)&ev); 4253 } 4254 4255 /* 4256 * Return TRUE if 'imactivatekey' has a valid value. 4257 */ 4258 int 4259 im_xim_isvalid_imactivate() 4260 { 4261 return im_xim_str2keycode(NULL, NULL) == OK; 4262 } 4263 #endif /* FEAT_GUI_GTK */ 4264 4265 /* 4266 * Switch using XIM on/off. This is used by the code that changes "State". 4267 */ 4268 void 4269 im_set_active(active) 4270 int active; 4271 { 4272 if (xic == NULL) 4273 return; 4274 4275 /* If 'imdisable' is set, XIM is never active. */ 4276 if (p_imdisable) 4277 active = FALSE; 4278 #if !defined (FEAT_GUI_GTK) && !defined (FEAT_GUI_KDE) 4279 else if (input_style & XIMPreeditPosition) 4280 /* There is a problem in switching XIM off when preediting is used, 4281 * and it is not clear how this can be solved. For now, keep XIM on 4282 * all the time, like it was done in Vim 5.8. */ 4283 active = TRUE; 4284 #endif 4285 4286 /* Remember the active state, it is needed when Vim gets keyboard focus. */ 4287 xim_is_active = active; 4288 4289 #ifdef FEAT_GUI_GTK 4290 /* When 'imactivatekey' has valid key-string, try to control XIM preedit 4291 * state. When 'imactivatekey' has no or invalid string, try old XIM 4292 * focus control. 4293 */ 4294 if (*p_imak != NUL) 4295 { 4296 /* BASIC STRATEGY: 4297 * Destroy old Input Context (XIC), and create new one. New XIC 4298 * would have a state of preedit that is off. When argument:active 4299 * is false, that's all. Else argument:active is true, send a key 4300 * event specified by 'imactivatekey' to activate XIM preedit state. 4301 */ 4302 4303 xim_is_active = TRUE; /* Disable old XIM focus control */ 4304 /* If we can monitor preedit state with preedit callback functions, 4305 * try least creation of new XIC. 4306 */ 4307 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS) 4308 { 4309 if (xim_can_preediting && !active) 4310 { 4311 /* Force turn off preedit state. With some IM 4312 * implementations, we cannot turn off preedit state by 4313 * symulate keypress event. It is why using such a method 4314 * that destroy old IC (input context), and create new one. 4315 * When create new IC, its preedit state is usually off. 4316 */ 4317 xim_reset(); 4318 xim_set_focus(FALSE); 4319 gdk_ic_destroy(xic); 4320 xim_init(); 4321 xim_can_preediting = FALSE; 4322 } 4323 else if (!xim_can_preediting && active) 4324 im_xim_send_event_imactivate(); 4325 } 4326 else 4327 { 4328 /* First, force destroy old IC, and create new one. It 4329 * symulates "turning off preedit state". 4330 */ 4331 xim_set_focus(FALSE); 4332 gdk_ic_destroy(xic); 4333 xim_init(); 4334 xim_can_preediting = FALSE; 4335 4336 /* 2nd, when requested to activate IM, symulate this by sending 4337 * the event. 4338 */ 4339 if (active) 4340 { 4341 im_xim_send_event_imactivate(); 4342 xim_can_preediting = TRUE; 4343 } 4344 } 4345 } 4346 else 4347 { 4348 # ifndef XIMPreeditUnKnown 4349 /* X11R5 doesn't have these, it looks safe enough to define here. */ 4350 typedef unsigned long XIMPreeditState; 4351 # define XIMPreeditUnKnown 0L 4352 # define XIMPreeditEnable 1L 4353 # define XIMPreeditDisable (1L<<1) 4354 # define XNPreeditState "preeditState" 4355 # endif 4356 XIMPreeditState preedit_state = XIMPreeditUnKnown; 4357 XVaNestedList preedit_attr; 4358 XIC pxic; 4359 4360 preedit_attr = XVaCreateNestedList(0, 4361 XNPreeditState, &preedit_state, 4362 NULL); 4363 pxic = ((GdkICPrivate *)xic)->xic; 4364 4365 if (!XGetICValues(pxic, XNPreeditAttributes, preedit_attr, NULL)) 4366 { 4367 XFree(preedit_attr); 4368 preedit_attr = XVaCreateNestedList(0, 4369 XNPreeditState, 4370 active ? XIMPreeditEnable : XIMPreeditDisable, 4371 NULL); 4372 XSetICValues(pxic, XNPreeditAttributes, preedit_attr, NULL); 4373 xim_can_preediting = active; 4374 xim_is_active = active; 4375 } 4376 XFree(preedit_attr); 4377 } 4378 if (xim_input_style & XIMPreeditCallbacks) 4379 { 4380 preedit_buf_len = 0; 4381 init_preedit_start_col(); 4382 } 4383 #else 4384 # if 0 4385 /* When had tested kinput2 + canna + Athena GUI version with 4386 * 'imactivatekey' is "s-space", im_xim_send_event_imactivate() did not 4387 * work correctly. It just inserted one space. I don't know why we 4388 * couldn't switch state of XIM preediting. This is reason why these 4389 * codes are commented out. 4390 */ 4391 /* First, force destroy old IC, and create new one. It symulates 4392 * "turning off preedit state". 4393 */ 4394 xim_set_focus(FALSE); 4395 XDestroyIC(xic); 4396 xic = NULL; 4397 xim_init(); 4398 4399 /* 2nd, when requested to activate IM, symulate this by sending the 4400 * event. 4401 */ 4402 if (active) 4403 im_xim_send_event_imactivate(); 4404 # endif 4405 #endif 4406 xim_set_preedit(); 4407 } 4408 4409 /* 4410 * Adjust using XIM for gaining or losing keyboard focus. Also called when 4411 * "xim_is_active" changes. 4412 */ 4413 void 4414 xim_set_focus(focus) 4415 int focus; 4416 { 4417 if (xic == NULL) 4418 return; 4419 4420 /* 4421 * XIM only gets focus when the Vim window has keyboard focus and XIM has 4422 * been set active for the current mode. 4423 */ 4424 if (focus && xim_is_active) 4425 { 4426 if (!xim_has_focus) 4427 { 4428 xim_has_focus = TRUE; 4429 #ifdef FEAT_GUI_GTK 4430 gdk_im_begin(xic, gui.drawarea->window); 4431 #else 4432 XSetICFocus(xic); 4433 #endif 4434 } 4435 } 4436 else 4437 { 4438 if (xim_has_focus) 4439 { 4440 xim_has_focus = FALSE; 4441 #ifdef FEAT_GUI_GTK 4442 gdk_im_end(); 4443 #else 4444 XUnsetICFocus(xic); 4445 #endif 4446 } 4447 } 4448 } 4449 4450 #ifndef FEAT_GUI_KDE 4451 /*ARGSUSED*/ 4452 void 4453 im_set_position(row, col) 4454 int row; 4455 int col; 4456 { 4457 xim_set_preedit(); 4458 } 4459 #endif 4460 4461 /* 4462 * Set the XIM to the current cursor position. 4463 */ 4464 void 4465 xim_set_preedit() 4466 { 4467 if (xic == NULL) 4468 return; 4469 4470 xim_set_focus(TRUE); 4471 4472 #ifdef FEAT_GUI_GTK 4473 if (gdk_im_ready()) 4474 { 4475 int attrmask; 4476 GdkICAttr *attr; 4477 4478 if (!xic_attr) 4479 return; 4480 4481 attr = xic_attr; 4482 attrmask = 0; 4483 4484 # ifdef FEAT_XFONTSET 4485 if ((xim_input_style & (int)GDK_IM_PREEDIT_POSITION) 4486 && gui.fontset != NOFONTSET 4487 && gui.fontset->type == GDK_FONT_FONTSET) 4488 { 4489 if (!xim_has_focus) 4490 { 4491 if (attr->spot_location.y >= 0) 4492 { 4493 attr->spot_location.x = 0; 4494 attr->spot_location.y = -100; 4495 attrmask |= (int)GDK_IC_SPOT_LOCATION; 4496 } 4497 } 4498 else 4499 { 4500 gint width, height; 4501 4502 if (attr->spot_location.x != TEXT_X(gui.col) 4503 || attr->spot_location.y != TEXT_Y(gui.row)) 4504 { 4505 attr->spot_location.x = TEXT_X(gui.col); 4506 attr->spot_location.y = TEXT_Y(gui.row); 4507 attrmask |= (int)GDK_IC_SPOT_LOCATION; 4508 } 4509 4510 gdk_window_get_size(gui.drawarea->window, &width, &height); 4511 width -= 2 * gui.border_offset; 4512 height -= 2 * gui.border_offset; 4513 if (xim_input_style & (int)GDK_IM_STATUS_AREA) 4514 height -= gui.char_height; 4515 if (attr->preedit_area.width != width 4516 || attr->preedit_area.height != height) 4517 { 4518 attr->preedit_area.x = gui.border_offset; 4519 attr->preedit_area.y = gui.border_offset; 4520 attr->preedit_area.width = width; 4521 attr->preedit_area.height = height; 4522 attrmask |= (int)GDK_IC_PREEDIT_AREA; 4523 } 4524 4525 if (attr->preedit_fontset != gui.current_font) 4526 { 4527 attr->preedit_fontset = gui.current_font; 4528 attrmask |= (int)GDK_IC_PREEDIT_FONTSET; 4529 } 4530 } 4531 } 4532 # endif /* FEAT_XFONTSET */ 4533 4534 if (xim_fg_color == INVALCOLOR) 4535 { 4536 xim_fg_color = gui.def_norm_pixel; 4537 xim_bg_color = gui.def_back_pixel; 4538 } 4539 if (attr->preedit_foreground.pixel != xim_fg_color) 4540 { 4541 attr->preedit_foreground.pixel = xim_fg_color; 4542 attrmask |= (int)GDK_IC_PREEDIT_FOREGROUND; 4543 } 4544 if (attr->preedit_background.pixel != xim_bg_color) 4545 { 4546 attr->preedit_background.pixel = xim_bg_color; 4547 attrmask |= (int)GDK_IC_PREEDIT_BACKGROUND; 4548 } 4549 4550 if (attrmask != 0) 4551 gdk_ic_set_attr(xic, attr, (GdkICAttributesType)attrmask); 4552 } 4553 #else /* FEAT_GUI_GTK */ 4554 # ifdef FEAT_GUI_KDE 4555 # else 4556 { 4557 XVaNestedList attr_list; 4558 XRectangle spot_area; 4559 XPoint over_spot; 4560 int line_space; 4561 4562 if (!xim_has_focus) 4563 { 4564 /* hide XIM cursor */ 4565 over_spot.x = 0; 4566 over_spot.y = -100; /* arbitrary invisible position */ 4567 attr_list = (XVaNestedList) XVaCreateNestedList(0, 4568 XNSpotLocation, 4569 &over_spot, 4570 NULL); 4571 XSetICValues(xic, XNPreeditAttributes, attr_list, NULL); 4572 XFree(attr_list); 4573 return; 4574 } 4575 4576 if (input_style & XIMPreeditPosition) 4577 { 4578 if (xim_fg_color == INVALCOLOR) 4579 { 4580 xim_fg_color = gui.def_norm_pixel; 4581 xim_bg_color = gui.def_back_pixel; 4582 } 4583 over_spot.x = TEXT_X(gui.col); 4584 over_spot.y = TEXT_Y(gui.row); 4585 spot_area.x = 0; 4586 spot_area.y = 0; 4587 spot_area.height = gui.char_height * Rows; 4588 spot_area.width = gui.char_width * Columns; 4589 line_space = gui.char_height; 4590 attr_list = (XVaNestedList) XVaCreateNestedList(0, 4591 XNSpotLocation, &over_spot, 4592 XNForeground, (Pixel) xim_fg_color, 4593 XNBackground, (Pixel) xim_bg_color, 4594 XNArea, &spot_area, 4595 XNLineSpace, line_space, 4596 NULL); 4597 if (XSetICValues(xic, XNPreeditAttributes, attr_list, NULL)) 4598 EMSG(_("E284: Cannot set IC values")); 4599 XFree(attr_list); 4600 } 4601 } 4602 # endif /* FEAT_GUI_KDE */ 4603 #endif /* FEAT_GUI_GTK */ 4604 } 4605 4606 /* 4607 * Set up the status area. 4608 * 4609 * This should use a separate Widget, but that seems not possible, because 4610 * preedit_area and status_area should be set to the same window as for the 4611 * text input. Unfortunately this means the status area pollutes the text 4612 * window... 4613 */ 4614 void 4615 xim_set_status_area() 4616 { 4617 if (xic == NULL) 4618 return; 4619 4620 #ifdef FEAT_GUI_GTK 4621 # if defined(FEAT_XFONTSET) 4622 if (use_status_area) 4623 { 4624 GdkICAttr *attr; 4625 int style; 4626 gint width, height; 4627 GtkWidget *widget; 4628 int attrmask; 4629 4630 if (!xic_attr) 4631 return; 4632 4633 attr = xic_attr; 4634 attrmask = 0; 4635 style = (int)gdk_ic_get_style(xic); 4636 if ((style & (int)GDK_IM_STATUS_MASK) == (int)GDK_IM_STATUS_AREA) 4637 { 4638 if (gui.fontset != NOFONTSET 4639 && gui.fontset->type == GDK_FONT_FONTSET) 4640 { 4641 widget = gui.mainwin; 4642 gdk_window_get_size(widget->window, &width, &height); 4643 4644 attrmask |= (int)GDK_IC_STATUS_AREA; 4645 attr->status_area.x = 0; 4646 attr->status_area.y = height - gui.char_height - 1; 4647 attr->status_area.width = width; 4648 attr->status_area.height = gui.char_height; 4649 } 4650 } 4651 if (attrmask != 0) 4652 gdk_ic_set_attr(xic, attr, (GdkICAttributesType)attrmask); 4653 } 4654 # endif 4655 #else 4656 # ifdef FEAT_GUI_KDE 4657 # else 4658 { 4659 XVaNestedList preedit_list = 0, status_list = 0, list = 0; 4660 XRectangle pre_area, status_area; 4661 4662 if (input_style & XIMStatusArea) 4663 { 4664 if (input_style & XIMPreeditArea) 4665 { 4666 XRectangle *needed_rect; 4667 4668 /* to get status_area width */ 4669 status_list = XVaCreateNestedList(0, XNAreaNeeded, 4670 &needed_rect, NULL); 4671 XGetICValues(xic, XNStatusAttributes, status_list, NULL); 4672 XFree(status_list); 4673 4674 status_area.width = needed_rect->width; 4675 } 4676 else 4677 status_area.width = gui.char_width * Columns; 4678 4679 status_area.x = 0; 4680 status_area.y = gui.char_height * Rows + gui.border_offset; 4681 if (gui.which_scrollbars[SBAR_BOTTOM]) 4682 status_area.y += gui.scrollbar_height; 4683 #ifdef FEAT_MENU 4684 if (gui.menu_is_active) 4685 status_area.y += gui.menu_height; 4686 #endif 4687 status_area.height = gui.char_height; 4688 status_list = XVaCreateNestedList(0, XNArea, &status_area, NULL); 4689 } 4690 else 4691 { 4692 status_area.x = 0; 4693 status_area.y = gui.char_height * Rows + gui.border_offset; 4694 if (gui.which_scrollbars[SBAR_BOTTOM]) 4695 status_area.y += gui.scrollbar_height; 4696 #ifdef FEAT_MENU 4697 if (gui.menu_is_active) 4698 status_area.y += gui.menu_height; 4699 #endif 4700 status_area.width = 0; 4701 status_area.height = gui.char_height; 4702 } 4703 4704 if (input_style & XIMPreeditArea) /* off-the-spot */ 4705 { 4706 pre_area.x = status_area.x + status_area.width; 4707 pre_area.y = gui.char_height * Rows + gui.border_offset; 4708 pre_area.width = gui.char_width * Columns - pre_area.x; 4709 if (gui.which_scrollbars[SBAR_BOTTOM]) 4710 pre_area.y += gui.scrollbar_height; 4711 #ifdef FEAT_MENU 4712 if (gui.menu_is_active) 4713 pre_area.y += gui.menu_height; 4714 #endif 4715 pre_area.height = gui.char_height; 4716 preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL); 4717 } 4718 else if (input_style & XIMPreeditPosition) /* over-the-spot */ 4719 { 4720 pre_area.x = 0; 4721 pre_area.y = 0; 4722 pre_area.height = gui.char_height * Rows; 4723 pre_area.width = gui.char_width * Columns; 4724 preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL); 4725 } 4726 4727 if (preedit_list && status_list) 4728 list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list, 4729 XNStatusAttributes, status_list, NULL); 4730 else if (preedit_list) 4731 list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list, 4732 NULL); 4733 else if (status_list) 4734 list = XVaCreateNestedList(0, XNStatusAttributes, status_list, 4735 NULL); 4736 else 4737 list = NULL; 4738 4739 if (list) 4740 { 4741 XSetICValues(xic, XNVaNestedList, list, NULL); 4742 XFree(list); 4743 } 4744 if (status_list) 4745 XFree(status_list); 4746 if (preedit_list) 4747 XFree(preedit_list); 4748 } 4749 # endif /* FEAT_GUI_KDE */ 4750 #endif 4751 } 4752 4753 #if defined(FEAT_GUI_X11) || defined(FEAT_GUI_GTK) || defined(FEAT_GUI_KDE) 4754 static char e_xim[] = N_("E285: Failed to create input context"); 4755 #endif 4756 4757 #if defined(FEAT_GUI_X11) || defined(PROTO) 4758 # if defined(XtSpecificationRelease) && XtSpecificationRelease >= 6 && !defined(sun) 4759 # define USE_X11R6_XIM 4760 # endif 4761 4762 static int xim_real_init __ARGS((Window x11_window, Display *x11_display)); 4763 4764 4765 #ifdef USE_X11R6_XIM 4766 static void xim_instantiate_cb __ARGS((Display *display, XPointer client_data, XPointer call_data)); 4767 static void xim_destroy_cb __ARGS((XIM im, XPointer client_data, XPointer call_data)); 4768 4769 /*ARGSUSED*/ 4770 static void 4771 xim_instantiate_cb(display, client_data, call_data) 4772 Display *display; 4773 XPointer client_data; 4774 XPointer call_data; 4775 { 4776 Window x11_window; 4777 Display *x11_display; 4778 4779 #ifdef XIM_DEBUG 4780 xim_log("xim_instantiate_cb()\n"); 4781 #endif 4782 4783 gui_get_x11_windis(&x11_window, &x11_display); 4784 if (display != x11_display) 4785 return; 4786 4787 xim_real_init(x11_window, x11_display); 4788 gui_set_shellsize(FALSE, FALSE); 4789 if (xic != NULL) 4790 XUnregisterIMInstantiateCallback(x11_display, NULL, NULL, NULL, 4791 xim_instantiate_cb, NULL); 4792 } 4793 4794 /*ARGSUSED*/ 4795 static void 4796 xim_destroy_cb(im, client_data, call_data) 4797 XIM im; 4798 XPointer client_data; 4799 XPointer call_data; 4800 { 4801 Window x11_window; 4802 Display *x11_display; 4803 4804 #ifdef XIM_DEBUG 4805 xim_log("xim_destroy_cb()\n"); 4806 #endif 4807 gui_get_x11_windis(&x11_window, &x11_display); 4808 4809 xic = NULL; 4810 status_area_enabled = FALSE; 4811 4812 gui_set_shellsize(FALSE, FALSE); 4813 4814 XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL, 4815 xim_instantiate_cb, NULL); 4816 } 4817 #endif 4818 4819 void 4820 xim_init() 4821 { 4822 Window x11_window; 4823 Display *x11_display; 4824 4825 #ifdef XIM_DEBUG 4826 xim_log("xim_init()\n"); 4827 #endif 4828 4829 gui_get_x11_windis(&x11_window, &x11_display); 4830 4831 xic = NULL; 4832 4833 if (xim_real_init(x11_window, x11_display)) 4834 return; 4835 4836 gui_set_shellsize(FALSE, FALSE); 4837 4838 #ifdef USE_X11R6_XIM 4839 XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL, 4840 xim_instantiate_cb, NULL); 4841 #endif 4842 } 4843 4844 static int 4845 xim_real_init(x11_window, x11_display) 4846 Window x11_window; 4847 Display *x11_display; 4848 { 4849 int i; 4850 char *p, 4851 *s, 4852 *ns, 4853 *end, 4854 tmp[1024]; 4855 #define IMLEN_MAX 40 4856 char buf[IMLEN_MAX + 7]; 4857 XIM xim = NULL; 4858 XIMStyles *xim_styles; 4859 XIMStyle this_input_style = 0; 4860 Boolean found; 4861 XPoint over_spot; 4862 XVaNestedList preedit_list, status_list; 4863 4864 input_style = 0; 4865 status_area_enabled = FALSE; 4866 4867 if (xic != NULL) 4868 return FALSE; 4869 4870 if (gui.rsrc_input_method != NULL && *gui.rsrc_input_method != NUL) 4871 { 4872 strcpy(tmp, gui.rsrc_input_method); 4873 for (ns = s = tmp; ns != NULL && *s != NUL;) 4874 { 4875 s = (char *)skipwhite((char_u *)s); 4876 if (*s == NUL) 4877 break; 4878 if ((ns = end = strchr(s, ',')) == NULL) 4879 end = s + strlen(s); 4880 while (isspace(((char_u *)end)[-1])) 4881 end--; 4882 *end = NUL; 4883 4884 if (strlen(s) <= IMLEN_MAX) 4885 { 4886 strcpy(buf, "@im="); 4887 strcat(buf, s); 4888 if ((p = XSetLocaleModifiers(buf)) != NULL && *p != NUL 4889 && (xim = XOpenIM(x11_display, NULL, NULL, NULL)) 4890 != NULL) 4891 break; 4892 } 4893 4894 s = ns + 1; 4895 } 4896 } 4897 4898 if (xim == NULL && (p = XSetLocaleModifiers("")) != NULL && *p != NUL) 4899 xim = XOpenIM(x11_display, NULL, NULL, NULL); 4900 4901 /* This is supposed to be useful to obtain characters through 4902 * XmbLookupString() without really using a XIM. */ 4903 if (xim == NULL && (p = XSetLocaleModifiers("@im=none")) != NULL 4904 && *p != NUL) 4905 xim = XOpenIM(x11_display, NULL, NULL, NULL); 4906 4907 if (xim == NULL) 4908 { 4909 /* Only give this message when verbose is set, because too many people 4910 * got this message when they didn't want to use a XIM. */ 4911 if (p_verbose > 0) 4912 { 4913 verbose_enter(); 4914 EMSG(_("E286: Failed to open input method")); 4915 verbose_leave(); 4916 } 4917 return FALSE; 4918 } 4919 4920 #ifdef USE_X11R6_XIM 4921 { 4922 XIMCallback destroy_cb; 4923 4924 destroy_cb.callback = xim_destroy_cb; 4925 destroy_cb.client_data = NULL; 4926 if (XSetIMValues(xim, XNDestroyCallback, &destroy_cb, NULL)) 4927 EMSG(_("E287: Warning: Could not set destroy callback to IM")); 4928 } 4929 #endif 4930 4931 if (XGetIMValues(xim, XNQueryInputStyle, &xim_styles, NULL) || !xim_styles) 4932 { 4933 EMSG(_("E288: input method doesn't support any style")); 4934 XCloseIM(xim); 4935 return FALSE; 4936 } 4937 4938 found = False; 4939 strcpy(tmp, gui.rsrc_preedit_type_name); 4940 for (s = tmp; s && !found; ) 4941 { 4942 while (*s && isspace((unsigned char)*s)) 4943 s++; 4944 if (!*s) 4945 break; 4946 if ((ns = end = strchr(s, ',')) != 0) 4947 ns++; 4948 else 4949 end = s + strlen(s); 4950 while (isspace((unsigned char)*end)) 4951 end--; 4952 *end = '\0'; 4953 4954 if (!strcmp(s, "OverTheSpot")) 4955 this_input_style = (XIMPreeditPosition | XIMStatusArea); 4956 else if (!strcmp(s, "OffTheSpot")) 4957 this_input_style = (XIMPreeditArea | XIMStatusArea); 4958 else if (!strcmp(s, "Root")) 4959 this_input_style = (XIMPreeditNothing | XIMStatusNothing); 4960 4961 for (i = 0; (unsigned short)i < xim_styles->count_styles; i++) 4962 { 4963 if (this_input_style == xim_styles->supported_styles[i]) 4964 { 4965 found = True; 4966 break; 4967 } 4968 } 4969 if (!found) 4970 for (i = 0; (unsigned short)i < xim_styles->count_styles; i++) 4971 { 4972 if ((xim_styles->supported_styles[i] & this_input_style) 4973 == (this_input_style & ~XIMStatusArea)) 4974 { 4975 this_input_style &= ~XIMStatusArea; 4976 found = True; 4977 break; 4978 } 4979 } 4980 4981 s = ns; 4982 } 4983 XFree(xim_styles); 4984 4985 if (!found) 4986 { 4987 /* Only give this message when verbose is set, because too many people 4988 * got this message when they didn't want to use a XIM. */ 4989 if (p_verbose > 0) 4990 { 4991 verbose_enter(); 4992 EMSG(_("E289: input method doesn't support my preedit type")); 4993 verbose_leave(); 4994 } 4995 XCloseIM(xim); 4996 return FALSE; 4997 } 4998 4999 over_spot.x = TEXT_X(gui.col); 5000 over_spot.y = TEXT_Y(gui.row); 5001 input_style = this_input_style; 5002 5003 /* A crash was reported when trying to pass gui.norm_font as XNFontSet, 5004 * thus that has been removed. Hopefully the default works... */ 5005 #ifdef FEAT_XFONTSET 5006 if (gui.fontset != NOFONTSET) 5007 { 5008 preedit_list = XVaCreateNestedList(0, 5009 XNSpotLocation, &over_spot, 5010 XNForeground, (Pixel)gui.def_norm_pixel, 5011 XNBackground, (Pixel)gui.def_back_pixel, 5012 XNFontSet, (XFontSet)gui.fontset, 5013 NULL); 5014 status_list = XVaCreateNestedList(0, 5015 XNForeground, (Pixel)gui.def_norm_pixel, 5016 XNBackground, (Pixel)gui.def_back_pixel, 5017 XNFontSet, (XFontSet)gui.fontset, 5018 NULL); 5019 } 5020 else 5021 #endif 5022 { 5023 preedit_list = XVaCreateNestedList(0, 5024 XNSpotLocation, &over_spot, 5025 XNForeground, (Pixel)gui.def_norm_pixel, 5026 XNBackground, (Pixel)gui.def_back_pixel, 5027 NULL); 5028 status_list = XVaCreateNestedList(0, 5029 XNForeground, (Pixel)gui.def_norm_pixel, 5030 XNBackground, (Pixel)gui.def_back_pixel, 5031 NULL); 5032 } 5033 5034 xic = XCreateIC(xim, 5035 XNInputStyle, input_style, 5036 XNClientWindow, x11_window, 5037 XNFocusWindow, gui.wid, 5038 XNPreeditAttributes, preedit_list, 5039 XNStatusAttributes, status_list, 5040 NULL); 5041 XFree(status_list); 5042 XFree(preedit_list); 5043 if (xic != NULL) 5044 { 5045 if (input_style & XIMStatusArea) 5046 { 5047 xim_set_status_area(); 5048 status_area_enabled = TRUE; 5049 } 5050 else 5051 gui_set_shellsize(FALSE, FALSE); 5052 } 5053 else 5054 { 5055 EMSG(_(e_xim)); 5056 XCloseIM(xim); 5057 return FALSE; 5058 } 5059 5060 return TRUE; 5061 } 5062 5063 #endif /* FEAT_GUI_X11 */ 5064 5065 #if defined(FEAT_GUI_GTK) || defined(PROTO) 5066 5067 # ifdef FEAT_XFONTSET 5068 static char e_overthespot[] = N_("E290: over-the-spot style requires fontset"); 5069 # endif 5070 5071 # ifdef PROTO 5072 typedef int GdkIC; 5073 # endif 5074 5075 void 5076 xim_decide_input_style() 5077 { 5078 /* GDK_IM_STATUS_CALLBACKS was disabled, enabled it to allow Japanese 5079 * OverTheSpot. */ 5080 int supported_style = (int)GDK_IM_PREEDIT_NONE | 5081 (int)GDK_IM_PREEDIT_NOTHING | 5082 (int)GDK_IM_PREEDIT_POSITION | 5083 (int)GDK_IM_PREEDIT_CALLBACKS | 5084 (int)GDK_IM_STATUS_CALLBACKS | 5085 (int)GDK_IM_STATUS_AREA | 5086 (int)GDK_IM_STATUS_NONE | 5087 (int)GDK_IM_STATUS_NOTHING; 5088 5089 #ifdef XIM_DEBUG 5090 xim_log("xim_decide_input_style()\n"); 5091 #endif 5092 5093 if (!gdk_im_ready()) 5094 xim_input_style = 0; 5095 else 5096 { 5097 if (gtk_major_version > 1 5098 || (gtk_major_version == 1 5099 && (gtk_minor_version > 2 5100 || (gtk_minor_version == 2 && gtk_micro_version >= 3)))) 5101 use_status_area = TRUE; 5102 else 5103 { 5104 EMSG(_("E291: Your GTK+ is older than 1.2.3. Status area disabled")); 5105 use_status_area = FALSE; 5106 } 5107 #ifdef FEAT_XFONTSET 5108 if (gui.fontset == NOFONTSET || gui.fontset->type != GDK_FONT_FONTSET) 5109 #endif 5110 supported_style &= ~((int)GDK_IM_PREEDIT_POSITION 5111 | (int)GDK_IM_STATUS_AREA); 5112 if (!use_status_area) 5113 supported_style &= ~(int)GDK_IM_STATUS_AREA; 5114 xim_input_style = (int)gdk_im_decide_style((GdkIMStyle)supported_style); 5115 } 5116 } 5117 5118 /*ARGSUSED*/ 5119 static void 5120 preedit_start_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5121 { 5122 #ifdef XIM_DEBUG 5123 xim_log("xim_decide_input_style()\n"); 5124 #endif 5125 5126 draw_feedback = NULL; 5127 xim_can_preediting = TRUE; 5128 xim_has_preediting = TRUE; 5129 gui_update_cursor(TRUE, FALSE); 5130 if (showmode() > 0) 5131 { 5132 setcursor(); 5133 out_flush(); 5134 } 5135 } 5136 5137 static void 5138 xim_back_delete(int n) 5139 { 5140 char_u str[3]; 5141 5142 str[0] = CSI; 5143 str[1] = 'k'; 5144 str[2] = 'b'; 5145 while (n-- > 0) 5146 add_to_input_buf(str, 3); 5147 } 5148 5149 static GSList *key_press_event_queue = NULL; 5150 static gboolean processing_queued_event = FALSE; 5151 5152 /*ARGSUSED*/ 5153 static void 5154 preedit_draw_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5155 { 5156 XIMPreeditDrawCallbackStruct *draw_data; 5157 XIMText *text; 5158 char *src; 5159 GSList *event_queue; 5160 5161 #ifdef XIM_DEBUG 5162 xim_log("preedit_draw_cbproc()\n"); 5163 #endif 5164 5165 draw_data = (XIMPreeditDrawCallbackStruct *) call_data; 5166 text = (XIMText *) draw_data->text; 5167 5168 if ((text == NULL && draw_data->chg_length == preedit_buf_len) 5169 || preedit_buf_len == 0) 5170 { 5171 init_preedit_start_col(); 5172 vim_free(draw_feedback); 5173 draw_feedback = NULL; 5174 } 5175 if (draw_data->chg_length > 0) 5176 { 5177 int bs_cnt; 5178 5179 if (draw_data->chg_length > preedit_buf_len) 5180 bs_cnt = preedit_buf_len; 5181 else 5182 bs_cnt = draw_data->chg_length; 5183 xim_back_delete(bs_cnt); 5184 preedit_buf_len -= bs_cnt; 5185 } 5186 if (text != NULL) 5187 { 5188 int len; 5189 #ifdef FEAT_MBYTE 5190 char_u *buf = NULL; 5191 unsigned int nfeedback = 0; 5192 #endif 5193 char_u *ptr; 5194 5195 src = text->string.multi_byte; 5196 if (src != NULL && !text->encoding_is_wchar) 5197 { 5198 len = strlen(src); 5199 ptr = (char_u *)src; 5200 /* Avoid the enter for decision */ 5201 if (*ptr == '\n') 5202 return; 5203 5204 #ifdef FEAT_MBYTE 5205 if (input_conv.vc_type != CONV_NONE 5206 && (buf = string_convert(&input_conv, 5207 (char_u *)src, &len)) != NULL) 5208 { 5209 /* Converted from 'termencoding' to 'encoding'. */ 5210 add_to_input_buf_csi(buf, len); 5211 ptr = buf; 5212 } 5213 else 5214 #endif 5215 add_to_input_buf_csi((char_u *)src, len); 5216 /* Add count of character to preedit_buf_len */ 5217 while (*ptr != NUL) 5218 { 5219 #ifdef FEAT_MBYTE 5220 if (draw_data->text->feedback != NULL) 5221 { 5222 if (draw_feedback == NULL) 5223 draw_feedback = (char *)alloc(draw_data->chg_first 5224 + text->length); 5225 else 5226 draw_feedback = realloc(draw_feedback, 5227 draw_data->chg_first + text->length); 5228 if (draw_feedback != NULL) 5229 { 5230 draw_feedback[nfeedback + draw_data->chg_first] 5231 = draw_data->text->feedback[nfeedback]; 5232 nfeedback++; 5233 } 5234 } 5235 if (has_mbyte) 5236 ptr += mb_ptr2len_check(ptr); 5237 else 5238 #endif 5239 ptr++; 5240 preedit_buf_len++; 5241 } 5242 #ifdef FEAT_MBYTE 5243 vim_free(buf); 5244 #endif 5245 preedit_end_col = MAXCOL; 5246 } 5247 } 5248 if (text != NULL || draw_data->chg_length > 0) 5249 { 5250 event_queue = key_press_event_queue; 5251 processing_queued_event = TRUE; 5252 while (event_queue != NULL && processing_queued_event) 5253 { 5254 GdkEvent *ev = event_queue->data; 5255 5256 gboolean *ret; 5257 gtk_signal_emit_by_name((GtkObject*)gui.mainwin, "key_press_event", 5258 ev, &ret); 5259 gdk_event_free(ev); 5260 event_queue = event_queue->next; 5261 } 5262 processing_queued_event = FALSE; 5263 if (key_press_event_queue) 5264 { 5265 g_slist_free(key_press_event_queue); 5266 key_press_event_queue = NULL; 5267 } 5268 } 5269 if (gtk_main_level() > 0) 5270 gtk_main_quit(); 5271 } 5272 5273 /* 5274 * Retrieve the highlighting attributes at column col in the preedit string. 5275 * Return -1 if not in preediting mode or if col is out of range. 5276 */ 5277 int 5278 im_get_feedback_attr(int col) 5279 { 5280 if (draw_feedback != NULL && col < preedit_buf_len) 5281 { 5282 if (draw_feedback[col] & XIMReverse) 5283 return HL_INVERSE; 5284 else if (draw_feedback[col] & XIMUnderline) 5285 return HL_UNDERLINE; 5286 else 5287 return hl_attr(HLF_V); 5288 } 5289 5290 return -1; 5291 } 5292 5293 /*ARGSUSED*/ 5294 static void 5295 preedit_caret_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5296 { 5297 #ifdef XIM_DEBUG 5298 xim_log("preedit_caret_cbproc()\n"); 5299 #endif 5300 } 5301 5302 /*ARGSUSED*/ 5303 static void 5304 preedit_done_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5305 { 5306 #ifdef XIM_DEBUG 5307 xim_log("preedit_done_cbproc()\n"); 5308 #endif 5309 5310 vim_free(draw_feedback); 5311 draw_feedback = NULL; 5312 xim_can_preediting = FALSE; 5313 xim_has_preediting = FALSE; 5314 gui_update_cursor(TRUE, FALSE); 5315 if (showmode() > 0) 5316 { 5317 setcursor(); 5318 out_flush(); 5319 } 5320 } 5321 5322 void 5323 xim_reset(void) 5324 { 5325 char *text; 5326 5327 #ifdef XIM_DEBUG 5328 xim_log("xim_reset()\n"); 5329 #endif 5330 5331 if (xic != NULL) 5332 { 5333 text = XmbResetIC(((GdkICPrivate *)xic)->xic); 5334 if (text != NULL && !(xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS)) 5335 add_to_input_buf_csi((char_u *)text, strlen(text)); 5336 else 5337 preedit_buf_len = 0; 5338 if (text != NULL) 5339 XFree(text); 5340 } 5341 } 5342 5343 /*ARGSUSED*/ 5344 int 5345 xim_queue_key_press_event(GdkEventKey *event, int down) 5346 { 5347 #ifdef XIM_DEBUG 5348 xim_log("xim_queue_key_press_event()\n"); 5349 #endif 5350 5351 if (preedit_buf_len <= 0) 5352 return FALSE; 5353 if (processing_queued_event) 5354 processing_queued_event = FALSE; 5355 5356 key_press_event_queue = g_slist_append(key_press_event_queue, 5357 gdk_event_copy((GdkEvent *)event)); 5358 return TRUE; 5359 } 5360 5361 /*ARGSUSED*/ 5362 static void 5363 preedit_callback_setup(GdkIC *ic) 5364 { 5365 XIC xxic; 5366 XVaNestedList preedit_attr; 5367 XIMCallback preedit_start_cb; 5368 XIMCallback preedit_draw_cb; 5369 XIMCallback preedit_caret_cb; 5370 XIMCallback preedit_done_cb; 5371 5372 xxic = ((GdkICPrivate*)xic)->xic; 5373 preedit_start_cb.callback = (XIMProc)preedit_start_cbproc; 5374 preedit_draw_cb.callback = (XIMProc)preedit_draw_cbproc; 5375 preedit_caret_cb.callback = (XIMProc)preedit_caret_cbproc; 5376 preedit_done_cb.callback = (XIMProc)preedit_done_cbproc; 5377 preedit_attr 5378 = XVaCreateNestedList (0, 5379 XNPreeditStartCallback, &preedit_start_cb, 5380 XNPreeditDrawCallback, &preedit_draw_cb, 5381 XNPreeditCaretCallback, &preedit_caret_cb, 5382 XNPreeditDoneCallback, &preedit_done_cb, 5383 0); 5384 XSetICValues (xxic, XNPreeditAttributes, preedit_attr, 0); 5385 XFree(preedit_attr); 5386 } 5387 5388 /*ARGSUSED*/ 5389 static void 5390 reset_state_setup(GdkIC *ic) 5391 { 5392 #ifdef USE_X11R6_XIM 5393 /* don't change the input context when we call reset */ 5394 XSetICValues(((GdkICPrivate*)ic)->xic, XNResetState, XIMPreserveState, 0); 5395 #endif 5396 } 5397 5398 void 5399 xim_init(void) 5400 { 5401 #ifdef XIM_DEBUG 5402 xim_log("xim_init()\n"); 5403 #endif 5404 5405 xic = NULL; 5406 xic_attr = NULL; 5407 5408 if (!gdk_im_ready()) 5409 { 5410 if (p_verbose > 0) 5411 { 5412 verbose_enter(); 5413 EMSG(_("E292: Input Method Server is not running")); 5414 verbose_leave(); 5415 } 5416 return; 5417 } 5418 if ((xic_attr = gdk_ic_attr_new()) != NULL) 5419 { 5420 #ifdef FEAT_XFONTSET 5421 gint width, height; 5422 #endif 5423 int mask; 5424 GdkColormap *colormap; 5425 GdkICAttr *attr = xic_attr; 5426 int attrmask = (int)GDK_IC_ALL_REQ; 5427 GtkWidget *widget = gui.drawarea; 5428 5429 attr->style = (GdkIMStyle)xim_input_style; 5430 attr->client_window = gui.mainwin->window; 5431 5432 if ((colormap = gtk_widget_get_colormap(widget)) != 5433 gtk_widget_get_default_colormap()) 5434 { 5435 attrmask |= (int)GDK_IC_PREEDIT_COLORMAP; 5436 attr->preedit_colormap = colormap; 5437 } 5438 attrmask |= (int)GDK_IC_PREEDIT_FOREGROUND; 5439 attrmask |= (int)GDK_IC_PREEDIT_BACKGROUND; 5440 attr->preedit_foreground = widget->style->fg[GTK_STATE_NORMAL]; 5441 attr->preedit_background = widget->style->base[GTK_STATE_NORMAL]; 5442 5443 #ifdef FEAT_XFONTSET 5444 if ((xim_input_style & (int)GDK_IM_PREEDIT_MASK) 5445 == (int)GDK_IM_PREEDIT_POSITION) 5446 { 5447 if (gui.fontset == NOFONTSET 5448 || gui.fontset->type != GDK_FONT_FONTSET) 5449 { 5450 EMSG(_(e_overthespot)); 5451 } 5452 else 5453 { 5454 gdk_window_get_size(widget->window, &width, &height); 5455 5456 attrmask |= (int)GDK_IC_PREEDIT_POSITION_REQ; 5457 attr->spot_location.x = TEXT_X(0); 5458 attr->spot_location.y = TEXT_Y(0); 5459 attr->preedit_area.x = gui.border_offset; 5460 attr->preedit_area.y = gui.border_offset; 5461 attr->preedit_area.width = width - 2*gui.border_offset; 5462 attr->preedit_area.height = height - 2*gui.border_offset; 5463 attr->preedit_fontset = gui.fontset; 5464 } 5465 } 5466 5467 if ((xim_input_style & (int)GDK_IM_STATUS_MASK) 5468 == (int)GDK_IM_STATUS_AREA) 5469 { 5470 if (gui.fontset == NOFONTSET 5471 || gui.fontset->type != GDK_FONT_FONTSET) 5472 { 5473 EMSG(_(e_overthespot)); 5474 } 5475 else 5476 { 5477 gdk_window_get_size(gui.mainwin->window, &width, &height); 5478 attrmask |= (int)GDK_IC_STATUS_AREA_REQ; 5479 attr->status_area.x = 0; 5480 attr->status_area.y = height - gui.char_height - 1; 5481 attr->status_area.width = width; 5482 attr->status_area.height = gui.char_height; 5483 attr->status_fontset = gui.fontset; 5484 } 5485 } 5486 else if ((xim_input_style & (int)GDK_IM_STATUS_MASK) 5487 == (int)GDK_IM_STATUS_CALLBACKS) 5488 { 5489 /* FIXME */ 5490 } 5491 #endif 5492 5493 xic = gdk_ic_new(attr, (GdkICAttributesType)attrmask); 5494 5495 if (xic == NULL) 5496 EMSG(_(e_xim)); 5497 else 5498 { 5499 mask = (int)gdk_window_get_events(widget->window); 5500 mask |= (int)gdk_ic_get_events(xic); 5501 gdk_window_set_events(widget->window, (GdkEventMask)mask); 5502 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS) 5503 preedit_callback_setup(xic); 5504 reset_state_setup(xic); 5505 } 5506 } 5507 } 5508 5509 void 5510 im_shutdown(void) 5511 { 5512 #ifdef XIM_DEBUG 5513 xim_log("im_shutdown()\n"); 5514 #endif 5515 5516 if (xic != NULL) 5517 { 5518 gdk_im_end(); 5519 gdk_ic_destroy(xic); 5520 xic = NULL; 5521 } 5522 xim_is_active = FALSE; 5523 xim_can_preediting = FALSE; 5524 preedit_start_col = MAXCOL; 5525 xim_has_preediting = FALSE; 5526 } 5527 5528 #endif /* FEAT_GUI_GTK */ 5529 5530 int 5531 xim_get_status_area_height() 5532 { 5533 #ifdef FEAT_GUI_GTK 5534 if (xim_input_style & (int)GDK_IM_STATUS_AREA) 5535 return gui.char_height; 5536 #else 5537 # if defined FEAT_GUI_KDE 5538 /* always return zero? */ 5539 # else 5540 if (status_area_enabled) 5541 return gui.char_height; 5542 # endif 5543 #endif 5544 return 0; 5545 } 5546 5547 /* 5548 * Get IM status. When IM is on, return TRUE. Else return FALSE. 5549 * FIXME: This doesn't work correctly: Having focus doesn't always mean XIM is 5550 * active, when not having focus XIM may still be active (e.g., when using a 5551 * tear-off menu item). 5552 */ 5553 int 5554 im_get_status() 5555 { 5556 # ifdef FEAT_GUI_GTK 5557 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS) 5558 return xim_can_preediting; 5559 # endif 5560 # ifdef FEAT_GUI_KDE 5561 if (preedit_start_col != MAXCOL) 5562 return TRUE; 5563 # endif 5564 return xim_has_focus; 5565 } 5566 5567 # endif /* !HAVE_GTK2 */ 5568 5569 # if defined(FEAT_GUI_GTK) || defined(PROTO) 5570 int 5571 im_is_preediting() 5572 { 5573 return xim_has_preediting; 5574 } 5575 # endif 5576 #endif /* FEAT_XIM */ 5577 5578 #if defined(FEAT_MBYTE) || defined(PROTO) 5579 5580 /* 5581 * Setup "vcp" for conversion from "from" to "to". 5582 * The names must have been made canonical with enc_canonize(). 5583 * vcp->vc_type must have been initialized to CONV_NONE. 5584 * Note: cannot be used for conversion from/to ucs-2 and ucs-4 (will use utf-8 5585 * instead). 5586 * Afterwards invoke with "from" and "to" equal to NULL to cleanup. 5587 * Return FAIL when conversion is not supported, OK otherwise. 5588 */ 5589 int 5590 convert_setup(vcp, from, to) 5591 vimconv_T *vcp; 5592 char_u *from; 5593 char_u *to; 5594 { 5595 int from_prop; 5596 int to_prop; 5597 5598 /* Reset to no conversion. */ 5599 # ifdef USE_ICONV 5600 if (vcp->vc_type == CONV_ICONV && vcp->vc_fd != (iconv_t)-1) 5601 iconv_close(vcp->vc_fd); 5602 # endif 5603 vcp->vc_type = CONV_NONE; 5604 vcp->vc_factor = 1; 5605 vcp->vc_fail = FALSE; 5606 5607 /* No conversion when one of the names is empty or they are equal. */ 5608 if (from == NULL || *from == NUL || to == NULL || *to == NUL 5609 || STRCMP(from, to) == 0) 5610 return OK; 5611 5612 from_prop = enc_canon_props(from); 5613 to_prop = enc_canon_props(to); 5614 if ((from_prop & ENC_LATIN1) && (to_prop & ENC_UNICODE)) 5615 { 5616 /* Internal latin1 -> utf-8 conversion. */ 5617 vcp->vc_type = CONV_TO_UTF8; 5618 vcp->vc_factor = 2; /* up to twice as long */ 5619 } 5620 else if ((from_prop & ENC_LATIN9) && (to_prop & ENC_UNICODE)) 5621 { 5622 /* Internal latin9 -> utf-8 conversion. */ 5623 vcp->vc_type = CONV_9_TO_UTF8; 5624 vcp->vc_factor = 3; /* up to three as long (euro sign) */ 5625 } 5626 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_LATIN1)) 5627 { 5628 /* Internal utf-8 -> latin1 conversion. */ 5629 vcp->vc_type = CONV_TO_LATIN1; 5630 } 5631 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_LATIN9)) 5632 { 5633 /* Internal utf-8 -> latin9 conversion. */ 5634 vcp->vc_type = CONV_TO_LATIN9; 5635 } 5636 #ifdef WIN3264 5637 /* Win32-specific codepage <-> codepage conversion without iconv. */ 5638 else if (((from_prop & ENC_UNICODE) || encname2codepage(from) > 0) 5639 && ((to_prop & ENC_UNICODE) || encname2codepage(to) > 0)) 5640 { 5641 vcp->vc_type = CONV_CODEPAGE; 5642 vcp->vc_factor = 2; /* up to twice as long */ 5643 vcp->vc_cpfrom = (from_prop & ENC_UNICODE) ? 0 : encname2codepage(from); 5644 vcp->vc_cpto = (to_prop & ENC_UNICODE) ? 0 : encname2codepage(to); 5645 } 5646 #endif 5647 #ifdef MACOS_X 5648 else if ((from_prop & ENC_MACROMAN) && (to_prop & ENC_LATIN1)) 5649 { 5650 vcp->vc_type = CONV_MAC_LATIN1; 5651 } 5652 else if ((from_prop & ENC_MACROMAN) && (to_prop & ENC_UNICODE)) 5653 { 5654 vcp->vc_type = CONV_MAC_UTF8; 5655 vcp->vc_factor = 2; /* up to twice as long */ 5656 } 5657 else if ((from_prop & ENC_LATIN1) && (to_prop & ENC_MACROMAN)) 5658 { 5659 vcp->vc_type = CONV_LATIN1_MAC; 5660 } 5661 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_MACROMAN)) 5662 { 5663 vcp->vc_type = CONV_UTF8_MAC; 5664 } 5665 #endif 5666 # ifdef USE_ICONV 5667 else 5668 { 5669 /* Use iconv() for conversion. */ 5670 vcp->vc_fd = (iconv_t)my_iconv_open( 5671 (to_prop & ENC_UNICODE) ? (char_u *)"utf-8" : to, 5672 (from_prop & ENC_UNICODE) ? (char_u *)"utf-8" : from); 5673 if (vcp->vc_fd != (iconv_t)-1) 5674 { 5675 vcp->vc_type = CONV_ICONV; 5676 vcp->vc_factor = 4; /* could be longer too... */ 5677 } 5678 } 5679 # endif 5680 if (vcp->vc_type == CONV_NONE) 5681 return FAIL; 5682 5683 return OK; 5684 } 5685 5686 #if defined(FEAT_GUI) || defined(AMIGA) || defined(WIN3264) \ 5687 || defined(MSDOS) || defined(PROTO) 5688 /* 5689 * Do conversion on typed input characters in-place. 5690 * The input and output are not NUL terminated! 5691 * Returns the length after conversion. 5692 */ 5693 int 5694 convert_input(ptr, len, maxlen) 5695 char_u *ptr; 5696 int len; 5697 int maxlen; 5698 { 5699 return convert_input_safe(ptr, len, maxlen, NULL, NULL); 5700 } 5701 #endif 5702 5703 /* 5704 * Like convert_input(), but when there is an incomplete byte sequence at the 5705 * end return that as an allocated string in "restp" and set "*restlenp" to 5706 * the length. If "restp" is NULL it is not used. 5707 */ 5708 int 5709 convert_input_safe(ptr, len, maxlen, restp, restlenp) 5710 char_u *ptr; 5711 int len; 5712 int maxlen; 5713 char_u **restp; 5714 int *restlenp; 5715 { 5716 char_u *d; 5717 int dlen = len; 5718 int unconvertlen = 0; 5719 5720 d = string_convert_ext(&input_conv, ptr, &dlen, 5721 restp == NULL ? NULL : &unconvertlen); 5722 if (d != NULL) 5723 { 5724 if (dlen <= maxlen) 5725 { 5726 if (unconvertlen > 0) 5727 { 5728 /* Move the unconverted characters to allocated memory. */ 5729 *restp = alloc(unconvertlen); 5730 if (*restp != NULL) 5731 mch_memmove(*restp, ptr + len - unconvertlen, unconvertlen); 5732 *restlenp = unconvertlen; 5733 } 5734 mch_memmove(ptr, d, dlen); 5735 } 5736 else 5737 /* result is too long, keep the unconverted text (the caller must 5738 * have done something wrong!) */ 5739 dlen = len; 5740 vim_free(d); 5741 } 5742 return dlen; 5743 } 5744 5745 #if defined(MACOS_X) 5746 /* This is in os_mac_conv.c. */ 5747 extern char_u *mac_string_convert __ARGS((char_u *ptr, int len, int *lenp, int fail_on_error, int from, int to, int *unconvlenp)); 5748 #endif 5749 5750 /* 5751 * Convert text "ptr[*lenp]" according to "vcp". 5752 * Returns the result in allocated memory and sets "*lenp". 5753 * When "lenp" is NULL, use NUL terminated strings. 5754 * Illegal chars are often changed to "?", unless vcp->vc_fail is set. 5755 * When something goes wrong, NULL is returned and "*lenp" is unchanged. 5756 */ 5757 char_u * 5758 string_convert(vcp, ptr, lenp) 5759 vimconv_T *vcp; 5760 char_u *ptr; 5761 int *lenp; 5762 { 5763 return string_convert_ext(vcp, ptr, lenp, NULL); 5764 } 5765 5766 /* 5767 * Like string_convert(), but when "unconvlenp" is not NULL and there are is 5768 * an incomplete sequence at the end it is not converted and "*unconvlenp" is 5769 * set to the number of remaining bytes. 5770 */ 5771 char_u * 5772 string_convert_ext(vcp, ptr, lenp, unconvlenp) 5773 vimconv_T *vcp; 5774 char_u *ptr; 5775 int *lenp; 5776 int *unconvlenp; 5777 { 5778 char_u *retval = NULL; 5779 char_u *d; 5780 int len; 5781 int i; 5782 int l; 5783 int c; 5784 5785 if (lenp == NULL) 5786 len = (int)STRLEN(ptr); 5787 else 5788 len = *lenp; 5789 if (len == 0) 5790 return vim_strsave((char_u *)""); 5791 5792 switch (vcp->vc_type) 5793 { 5794 case CONV_TO_UTF8: /* latin1 to utf-8 conversion */ 5795 retval = alloc(len * 2 + 1); 5796 if (retval == NULL) 5797 break; 5798 d = retval; 5799 for (i = 0; i < len; ++i) 5800 { 5801 c = ptr[i]; 5802 if (c < 0x80) 5803 *d++ = c; 5804 else 5805 { 5806 *d++ = 0xc0 + ((unsigned)c >> 6); 5807 *d++ = 0x80 + (c & 0x3f); 5808 } 5809 } 5810 *d = NUL; 5811 if (lenp != NULL) 5812 *lenp = (int)(d - retval); 5813 break; 5814 5815 case CONV_9_TO_UTF8: /* latin9 to utf-8 conversion */ 5816 retval = alloc(len * 3 + 1); 5817 if (retval == NULL) 5818 break; 5819 d = retval; 5820 for (i = 0; i < len; ++i) 5821 { 5822 c = ptr[i]; 5823 switch (c) 5824 { 5825 case 0xa4: c = 0x20ac; break; /* euro */ 5826 case 0xa6: c = 0x0160; break; /* S hat */ 5827 case 0xa8: c = 0x0161; break; /* S -hat */ 5828 case 0xb4: c = 0x017d; break; /* Z hat */ 5829 case 0xb8: c = 0x017e; break; /* Z -hat */ 5830 case 0xbc: c = 0x0152; break; /* OE */ 5831 case 0xbd: c = 0x0153; break; /* oe */ 5832 case 0xbe: c = 0x0178; break; /* Y */ 5833 } 5834 d += utf_char2bytes(c, d); 5835 } 5836 *d = NUL; 5837 if (lenp != NULL) 5838 *lenp = (int)(d - retval); 5839 break; 5840 5841 case CONV_TO_LATIN1: /* utf-8 to latin1 conversion */ 5842 case CONV_TO_LATIN9: /* utf-8 to latin9 conversion */ 5843 retval = alloc(len + 1); 5844 if (retval == NULL) 5845 break; 5846 d = retval; 5847 for (i = 0; i < len; ++i) 5848 { 5849 l = utf_ptr2len_check(ptr + i); 5850 if (l == 0) 5851 *d++ = NUL; 5852 else if (l == 1) 5853 { 5854 if (unconvlenp != NULL && utf8len_tab[ptr[i]] > len - i) 5855 { 5856 /* Incomplete sequence at the end. */ 5857 *unconvlenp = len - i; 5858 break; 5859 } 5860 *d++ = ptr[i]; 5861 } 5862 else 5863 { 5864 c = utf_ptr2char(ptr + i); 5865 if (vcp->vc_type == CONV_TO_LATIN9) 5866 switch (c) 5867 { 5868 case 0x20ac: c = 0xa4; break; /* euro */ 5869 case 0x0160: c = 0xa6; break; /* S hat */ 5870 case 0x0161: c = 0xa8; break; /* S -hat */ 5871 case 0x017d: c = 0xb4; break; /* Z hat */ 5872 case 0x017e: c = 0xb8; break; /* Z -hat */ 5873 case 0x0152: c = 0xbc; break; /* OE */ 5874 case 0x0153: c = 0xbd; break; /* oe */ 5875 case 0x0178: c = 0xbe; break; /* Y */ 5876 case 0xa4: 5877 case 0xa6: 5878 case 0xa8: 5879 case 0xb4: 5880 case 0xb8: 5881 case 0xbc: 5882 case 0xbd: 5883 case 0xbe: c = 0x100; break; /* not in latin9 */ 5884 } 5885 if (!utf_iscomposing(c)) /* skip composing chars */ 5886 { 5887 if (c < 0x100) 5888 *d++ = c; 5889 else if (vcp->vc_fail) 5890 { 5891 vim_free(retval); 5892 return NULL; 5893 } 5894 else 5895 { 5896 *d++ = 0xbf; 5897 if (utf_char2cells(c) > 1) 5898 *d++ = '?'; 5899 } 5900 } 5901 i += l - 1; 5902 } 5903 } 5904 *d = NUL; 5905 if (lenp != NULL) 5906 *lenp = (int)(d - retval); 5907 break; 5908 5909 # ifdef MACOS_X 5910 case CONV_MAC_LATIN1: 5911 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 5912 'm', 'l', unconvlenp); 5913 break; 5914 5915 case CONV_LATIN1_MAC: 5916 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 5917 'l', 'm', unconvlenp); 5918 break; 5919 5920 case CONV_MAC_UTF8: 5921 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 5922 'm', 'u', unconvlenp); 5923 break; 5924 5925 case CONV_UTF8_MAC: 5926 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 5927 'u', 'm', unconvlenp); 5928 break; 5929 # endif 5930 5931 # ifdef USE_ICONV 5932 case CONV_ICONV: /* conversion with output_conv.vc_fd */ 5933 retval = iconv_string(vcp, ptr, len, unconvlenp); 5934 if (retval != NULL && lenp != NULL) 5935 *lenp = (int)STRLEN(retval); 5936 break; 5937 # endif 5938 # ifdef WIN3264 5939 case CONV_CODEPAGE: /* codepage -> codepage */ 5940 { 5941 int retlen; 5942 int tmp_len; 5943 short_u *tmp; 5944 5945 /* 1. codepage/UTF-8 -> ucs-2. */ 5946 if (vcp->vc_cpfrom == 0) 5947 tmp_len = utf8_to_ucs2(ptr, len, NULL, NULL); 5948 else 5949 tmp_len = MultiByteToWideChar(vcp->vc_cpfrom, 0, 5950 ptr, len, 0, 0); 5951 tmp = (short_u *)alloc(sizeof(short_u) * tmp_len); 5952 if (tmp == NULL) 5953 break; 5954 if (vcp->vc_cpfrom == 0) 5955 utf8_to_ucs2(ptr, len, tmp, unconvlenp); 5956 else 5957 MultiByteToWideChar(vcp->vc_cpfrom, 0, ptr, len, tmp, tmp_len); 5958 5959 /* 2. ucs-2 -> codepage/UTF-8. */ 5960 if (vcp->vc_cpto == 0) 5961 retlen = ucs2_to_utf8(tmp, tmp_len, NULL); 5962 else 5963 retlen = WideCharToMultiByte(vcp->vc_cpto, 0, 5964 tmp, tmp_len, 0, 0, 0, 0); 5965 retval = alloc(retlen + 1); 5966 if (retval != NULL) 5967 { 5968 if (vcp->vc_cpto == 0) 5969 ucs2_to_utf8(tmp, tmp_len, retval); 5970 else 5971 WideCharToMultiByte(vcp->vc_cpto, 0, 5972 tmp, tmp_len, retval, retlen, 0, 0); 5973 retval[retlen] = NUL; 5974 if (lenp != NULL) 5975 *lenp = retlen; 5976 } 5977 vim_free(tmp); 5978 break; 5979 } 5980 # endif 5981 } 5982 5983 return retval; 5984 } 5985 #endif 5986