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_GUI_GTK) && defined(FEAT_XIM) && defined(HAVE_GTK2) 104 # include <gdk/gdkkeysyms.h> 105 # ifdef WIN3264 106 # include <gdk/gdkwin32.h> 107 # else 108 # include <gdk/gdkx.h> 109 # endif 110 #endif 111 112 #ifdef HAVE_WCHAR_H 113 # include <wchar.h> 114 #endif 115 116 #if 0 117 /* This has been disabled, because several people reported problems with the 118 * wcwidth() and iswprint() library functions, esp. for Hebrew. */ 119 # ifdef __STDC_ISO_10646__ 120 # define USE_WCHAR_FUNCTIONS 121 # endif 122 #endif 123 124 #if defined(FEAT_MBYTE) || defined(PROTO) 125 126 static int enc_canon_search __ARGS((char_u *name)); 127 static int dbcs_char2len __ARGS((int c)); 128 static int dbcs_char2bytes __ARGS((int c, char_u *buf)); 129 static int dbcs_ptr2len __ARGS((char_u *p)); 130 static int dbcs_char2cells __ARGS((int c)); 131 static int dbcs_ptr2char __ARGS((char_u *p)); 132 133 /* Lookup table to quickly get the length in bytes of a UTF-8 character from 134 * the first byte of a UTF-8 string. Bytes which are illegal when used as the 135 * first byte have a one, because these will be used separately. */ 136 static char utf8len_tab[256] = 137 { 138 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 139 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 140 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 141 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 142 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /*bogus*/ 143 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /*bogus*/ 144 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, 145 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,4,4,4,4,4,4,4,4,5,5,5,5,6,6,1,1, 146 }; 147 148 /* 149 * XIM often causes trouble. Define XIM_DEBUG to get a log of XIM callbacks 150 * in the "xim.log" file. 151 */ 152 /* #define XIM_DEBUG */ 153 #ifdef XIM_DEBUG 154 static void 155 xim_log(char *s, ...) 156 { 157 va_list arglist; 158 static FILE *fd = NULL; 159 160 if (fd == (FILE *)-1) 161 return; 162 if (fd == NULL) 163 { 164 fd = mch_fopen("xim.log", "w"); 165 if (fd == NULL) 166 { 167 EMSG("Cannot open xim.log"); 168 fd = (FILE *)-1; 169 return; 170 } 171 } 172 173 va_start(arglist, s); 174 vfprintf(fd, s, arglist); 175 va_end(arglist); 176 } 177 #endif 178 179 #endif 180 181 #if defined(FEAT_MBYTE) || defined(FEAT_POSTSCRIPT) || defined(PROTO) 182 /* 183 * Canonical encoding names and their properties. 184 * "iso-8859-n" is handled by enc_canonize() directly. 185 */ 186 static struct 187 { char *name; int prop; int codepage;} 188 enc_canon_table[] = 189 { 190 #define IDX_LATIN_1 0 191 {"latin1", ENC_8BIT + ENC_LATIN1, 1252}, 192 #define IDX_ISO_2 1 193 {"iso-8859-2", ENC_8BIT, 0}, 194 #define IDX_ISO_3 2 195 {"iso-8859-3", ENC_8BIT, 0}, 196 #define IDX_ISO_4 3 197 {"iso-8859-4", ENC_8BIT, 0}, 198 #define IDX_ISO_5 4 199 {"iso-8859-5", ENC_8BIT, 0}, 200 #define IDX_ISO_6 5 201 {"iso-8859-6", ENC_8BIT, 0}, 202 #define IDX_ISO_7 6 203 {"iso-8859-7", ENC_8BIT, 0}, 204 #define IDX_ISO_8 7 205 {"iso-8859-8", ENC_8BIT, 0}, 206 #define IDX_ISO_9 8 207 {"iso-8859-9", ENC_8BIT, 0}, 208 #define IDX_ISO_10 9 209 {"iso-8859-10", ENC_8BIT, 0}, 210 #define IDX_ISO_11 10 211 {"iso-8859-11", ENC_8BIT, 0}, 212 #define IDX_ISO_13 11 213 {"iso-8859-13", ENC_8BIT, 0}, 214 #define IDX_ISO_14 12 215 {"iso-8859-14", ENC_8BIT, 0}, 216 #define IDX_ISO_15 13 217 {"iso-8859-15", ENC_8BIT + ENC_LATIN9, 0}, 218 #define IDX_KOI8_R 14 219 {"koi8-r", ENC_8BIT, 0}, 220 #define IDX_KOI8_U 15 221 {"koi8-u", ENC_8BIT, 0}, 222 #define IDX_UTF8 16 223 {"utf-8", ENC_UNICODE, 0}, 224 #define IDX_UCS2 17 225 {"ucs-2", ENC_UNICODE + ENC_ENDIAN_B + ENC_2BYTE, 0}, 226 #define IDX_UCS2LE 18 227 {"ucs-2le", ENC_UNICODE + ENC_ENDIAN_L + ENC_2BYTE, 0}, 228 #define IDX_UTF16 19 229 {"utf-16", ENC_UNICODE + ENC_ENDIAN_B + ENC_2WORD, 0}, 230 #define IDX_UTF16LE 20 231 {"utf-16le", ENC_UNICODE + ENC_ENDIAN_L + ENC_2WORD, 0}, 232 #define IDX_UCS4 21 233 {"ucs-4", ENC_UNICODE + ENC_ENDIAN_B + ENC_4BYTE, 0}, 234 #define IDX_UCS4LE 22 235 {"ucs-4le", ENC_UNICODE + ENC_ENDIAN_L + ENC_4BYTE, 0}, 236 237 /* For debugging DBCS encoding on Unix. */ 238 #define IDX_DEBUG 23 239 {"debug", ENC_DBCS, DBCS_DEBUG}, 240 #define IDX_EUC_JP 24 241 {"euc-jp", ENC_DBCS, DBCS_JPNU}, 242 #define IDX_SJIS 25 243 {"sjis", ENC_DBCS, DBCS_JPN}, 244 #define IDX_EUC_KR 26 245 {"euc-kr", ENC_DBCS, DBCS_KORU}, 246 #define IDX_EUC_CN 27 247 {"euc-cn", ENC_DBCS, DBCS_CHSU}, 248 #define IDX_EUC_TW 28 249 {"euc-tw", ENC_DBCS, DBCS_CHTU}, 250 #define IDX_BIG5 29 251 {"big5", ENC_DBCS, DBCS_CHT}, 252 253 /* MS-DOS and MS-Windows codepages are included here, so that they can be 254 * used on Unix too. Most of them are similar to ISO-8859 encodings, but 255 * not exactly the same. */ 256 #define IDX_CP437 30 257 {"cp437", ENC_8BIT, 437}, /* like iso-8859-1 */ 258 #define IDX_CP737 31 259 {"cp737", ENC_8BIT, 737}, /* like iso-8859-7 */ 260 #define IDX_CP775 32 261 {"cp775", ENC_8BIT, 775}, /* Baltic */ 262 #define IDX_CP850 33 263 {"cp850", ENC_8BIT, 850}, /* like iso-8859-4 */ 264 #define IDX_CP852 34 265 {"cp852", ENC_8BIT, 852}, /* like iso-8859-1 */ 266 #define IDX_CP855 35 267 {"cp855", ENC_8BIT, 855}, /* like iso-8859-2 */ 268 #define IDX_CP857 36 269 {"cp857", ENC_8BIT, 857}, /* like iso-8859-5 */ 270 #define IDX_CP860 37 271 {"cp860", ENC_8BIT, 860}, /* like iso-8859-9 */ 272 #define IDX_CP861 38 273 {"cp861", ENC_8BIT, 861}, /* like iso-8859-1 */ 274 #define IDX_CP862 39 275 {"cp862", ENC_8BIT, 862}, /* like iso-8859-1 */ 276 #define IDX_CP863 40 277 {"cp863", ENC_8BIT, 863}, /* like iso-8859-8 */ 278 #define IDX_CP865 41 279 {"cp865", ENC_8BIT, 865}, /* like iso-8859-1 */ 280 #define IDX_CP866 42 281 {"cp866", ENC_8BIT, 866}, /* like iso-8859-5 */ 282 #define IDX_CP869 43 283 {"cp869", ENC_8BIT, 869}, /* like iso-8859-7 */ 284 #define IDX_CP874 44 285 {"cp874", ENC_8BIT, 874}, /* Thai */ 286 #define IDX_CP932 45 287 {"cp932", ENC_DBCS, DBCS_JPN}, 288 #define IDX_CP936 46 289 {"cp936", ENC_DBCS, DBCS_CHS}, 290 #define IDX_CP949 47 291 {"cp949", ENC_DBCS, DBCS_KOR}, 292 #define IDX_CP950 48 293 {"cp950", ENC_DBCS, DBCS_CHT}, 294 #define IDX_CP1250 49 295 {"cp1250", ENC_8BIT, 1250}, /* Czech, Polish, etc. */ 296 #define IDX_CP1251 50 297 {"cp1251", ENC_8BIT, 1251}, /* Cyrillic */ 298 /* cp1252 is considered to be equal to latin1 */ 299 #define IDX_CP1253 51 300 {"cp1253", ENC_8BIT, 1253}, /* Greek */ 301 #define IDX_CP1254 52 302 {"cp1254", ENC_8BIT, 1254}, /* Turkish */ 303 #define IDX_CP1255 53 304 {"cp1255", ENC_8BIT, 1255}, /* Hebrew */ 305 #define IDX_CP1256 54 306 {"cp1256", ENC_8BIT, 1256}, /* Arabic */ 307 #define IDX_CP1257 55 308 {"cp1257", ENC_8BIT, 1257}, /* Baltic */ 309 #define IDX_CP1258 56 310 {"cp1258", ENC_8BIT, 1258}, /* Vietnamese */ 311 312 #define IDX_MACROMAN 57 313 {"macroman", ENC_8BIT + ENC_MACROMAN, 0}, /* Mac OS */ 314 #define IDX_COUNT 58 315 }; 316 317 /* 318 * Aliases for encoding names. 319 */ 320 static struct 321 { char *name; int canon;} 322 enc_alias_table[] = 323 { 324 {"ansi", IDX_LATIN_1}, 325 {"iso-8859-1", IDX_LATIN_1}, 326 {"latin2", IDX_ISO_2}, 327 {"latin3", IDX_ISO_3}, 328 {"latin4", IDX_ISO_4}, 329 {"cyrillic", IDX_ISO_5}, 330 {"arabic", IDX_ISO_6}, 331 {"greek", IDX_ISO_7}, 332 #ifdef WIN3264 333 {"hebrew", IDX_CP1255}, 334 #else 335 {"hebrew", IDX_ISO_8}, 336 #endif 337 {"latin5", IDX_ISO_9}, 338 {"turkish", IDX_ISO_9}, /* ? */ 339 {"latin6", IDX_ISO_10}, 340 {"nordic", IDX_ISO_10}, /* ? */ 341 {"thai", IDX_ISO_11}, /* ? */ 342 {"latin7", IDX_ISO_13}, 343 {"latin8", IDX_ISO_14}, 344 {"latin9", IDX_ISO_15}, 345 {"utf8", IDX_UTF8}, 346 {"unicode", IDX_UCS2}, 347 {"ucs2", IDX_UCS2}, 348 {"ucs2be", IDX_UCS2}, 349 {"ucs-2be", IDX_UCS2}, 350 {"ucs2le", IDX_UCS2LE}, 351 {"utf16", IDX_UTF16}, 352 {"utf16be", IDX_UTF16}, 353 {"utf-16be", IDX_UTF16}, 354 {"utf16le", IDX_UTF16LE}, 355 {"ucs4", IDX_UCS4}, 356 {"ucs4be", IDX_UCS4}, 357 {"ucs-4be", IDX_UCS4}, 358 {"ucs4le", IDX_UCS4LE}, 359 {"932", IDX_CP932}, 360 {"949", IDX_CP949}, 361 {"936", IDX_CP936}, 362 {"950", IDX_CP950}, 363 {"eucjp", IDX_EUC_JP}, 364 {"unix-jis", IDX_EUC_JP}, 365 {"ujis", IDX_EUC_JP}, 366 {"shift-jis", IDX_SJIS}, 367 {"euckr", IDX_EUC_KR}, 368 {"5601", IDX_EUC_KR}, /* Sun: KS C 5601 */ 369 {"euccn", IDX_EUC_CN}, 370 {"gb2312", IDX_EUC_CN}, 371 {"euctw", IDX_EUC_TW}, 372 #if defined(WIN3264) || defined(WIN32UNIX) || defined(MACOS) 373 {"japan", IDX_CP932}, 374 {"korea", IDX_CP949}, 375 {"prc", IDX_CP936}, 376 {"chinese", IDX_CP936}, 377 {"taiwan", IDX_CP950}, 378 {"big5", IDX_CP950}, 379 #else 380 {"japan", IDX_EUC_JP}, 381 {"korea", IDX_EUC_KR}, 382 {"prc", IDX_EUC_CN}, 383 {"chinese", IDX_EUC_CN}, 384 {"taiwan", IDX_EUC_TW}, 385 {"cp950", IDX_BIG5}, 386 {"950", IDX_BIG5}, 387 #endif 388 {"mac", IDX_MACROMAN}, 389 {NULL, 0} 390 }; 391 392 #ifndef CP_UTF8 393 # define CP_UTF8 65001 /* magic number from winnls.h */ 394 #endif 395 396 /* 397 * Find encoding "name" in the list of canonical encoding names. 398 * Returns -1 if not found. 399 */ 400 static int 401 enc_canon_search(name) 402 char_u *name; 403 { 404 int i; 405 406 for (i = 0; i < IDX_COUNT; ++i) 407 if (STRCMP(name, enc_canon_table[i].name) == 0) 408 return i; 409 return -1; 410 } 411 412 #endif 413 414 #if defined(FEAT_MBYTE) || defined(PROTO) 415 416 /* 417 * Find canonical encoding "name" in the list and return its properties. 418 * Returns 0 if not found. 419 */ 420 int 421 enc_canon_props(name) 422 char_u *name; 423 { 424 int i; 425 426 i = enc_canon_search(name); 427 if (i >= 0) 428 return enc_canon_table[i].prop; 429 #ifdef WIN3264 430 if (name[0] == 'c' && name[1] == 'p' && VIM_ISDIGIT(name[2])) 431 { 432 CPINFO cpinfo; 433 434 /* Get info on this codepage to find out what it is. */ 435 if (GetCPInfo(atoi(name + 2), &cpinfo) != 0) 436 { 437 if (cpinfo.MaxCharSize == 1) /* some single-byte encoding */ 438 return ENC_8BIT; 439 if (cpinfo.MaxCharSize == 2 440 && (cpinfo.LeadByte[0] != 0 || cpinfo.LeadByte[1] != 0)) 441 /* must be a DBCS encoding */ 442 return ENC_DBCS; 443 } 444 return 0; 445 } 446 #endif 447 if (STRNCMP(name, "2byte-", 6) == 0) 448 return ENC_DBCS; 449 if (STRNCMP(name, "8bit-", 5) == 0 || STRNCMP(name, "iso-8859-", 9) == 0) 450 return ENC_8BIT; 451 return 0; 452 } 453 454 /* 455 * Set up for using multi-byte characters. 456 * Called in three cases: 457 * - by main() to initialize (p_enc == NULL) 458 * - by set_init_1() after 'encoding' was set to its default. 459 * - by do_set() when 'encoding' has been set. 460 * p_enc must have been passed through enc_canonize() already. 461 * Sets the "enc_unicode", "enc_utf8", "enc_dbcs" and "has_mbyte" flags. 462 * Fills mb_bytelen_tab[] and returns NULL when there are no problems. 463 * When there is something wrong: Returns an error message and doesn't change 464 * anything. 465 */ 466 char_u * 467 mb_init() 468 { 469 int i; 470 int idx; 471 int n; 472 int enc_dbcs_new = 0; 473 #if defined(USE_ICONV) && !defined(WIN3264) && !defined(WIN32UNIX) \ 474 && !defined(MACOS) 475 # define LEN_FROM_CONV 476 vimconv_T vimconv; 477 char_u *p; 478 #endif 479 480 if (p_enc == NULL) 481 { 482 /* Just starting up: set the whole table to one's. */ 483 for (i = 0; i < 256; ++i) 484 mb_bytelen_tab[i] = 1; 485 input_conv.vc_type = CONV_NONE; 486 input_conv.vc_factor = 1; 487 output_conv.vc_type = CONV_NONE; 488 return NULL; 489 } 490 491 #ifdef WIN3264 492 if (p_enc[0] == 'c' && p_enc[1] == 'p' && VIM_ISDIGIT(p_enc[2])) 493 { 494 CPINFO cpinfo; 495 496 /* Get info on this codepage to find out what it is. */ 497 if (GetCPInfo(atoi(p_enc + 2), &cpinfo) != 0) 498 { 499 if (cpinfo.MaxCharSize == 1) 500 { 501 /* some single-byte encoding */ 502 enc_unicode = 0; 503 enc_utf8 = FALSE; 504 } 505 else if (cpinfo.MaxCharSize == 2 506 && (cpinfo.LeadByte[0] != 0 || cpinfo.LeadByte[1] != 0)) 507 { 508 /* must be a DBCS encoding, check below */ 509 enc_dbcs_new = atoi(p_enc + 2); 510 } 511 else 512 goto codepage_invalid; 513 } 514 else if (GetLastError() == ERROR_INVALID_PARAMETER) 515 { 516 codepage_invalid: 517 return (char_u *)N_("E543: Not a valid codepage"); 518 } 519 } 520 #endif 521 else if (STRNCMP(p_enc, "8bit-", 5) == 0 522 || STRNCMP(p_enc, "iso-8859-", 9) == 0) 523 { 524 /* Accept any "8bit-" or "iso-8859-" name. */ 525 enc_unicode = 0; 526 enc_utf8 = FALSE; 527 } 528 else if (STRNCMP(p_enc, "2byte-", 6) == 0) 529 { 530 #ifdef WIN3264 531 /* Windows: accept only valid codepage numbers, check below. */ 532 if (p_enc[6] != 'c' || p_enc[7] != 'p' 533 || (enc_dbcs_new = atoi(p_enc + 8)) == 0) 534 return e_invarg; 535 #else 536 /* Unix: accept any "2byte-" name, assume current locale. */ 537 enc_dbcs_new = DBCS_2BYTE; 538 #endif 539 } 540 else if ((idx = enc_canon_search(p_enc)) >= 0) 541 { 542 i = enc_canon_table[idx].prop; 543 if (i & ENC_UNICODE) 544 { 545 /* Unicode */ 546 enc_utf8 = TRUE; 547 if (i & (ENC_2BYTE | ENC_2WORD)) 548 enc_unicode = 2; 549 else if (i & ENC_4BYTE) 550 enc_unicode = 4; 551 else 552 enc_unicode = 0; 553 } 554 else if (i & ENC_DBCS) 555 { 556 /* 2byte, handle below */ 557 enc_dbcs_new = enc_canon_table[idx].codepage; 558 } 559 else 560 { 561 /* Must be 8-bit. */ 562 enc_unicode = 0; 563 enc_utf8 = FALSE; 564 } 565 } 566 else /* Don't know what encoding this is, reject it. */ 567 return e_invarg; 568 569 if (enc_dbcs_new != 0) 570 { 571 #ifdef WIN3264 572 /* Check if the DBCS code page is OK. */ 573 if (!IsValidCodePage(enc_dbcs_new)) 574 goto codepage_invalid; 575 #endif 576 enc_unicode = 0; 577 enc_utf8 = FALSE; 578 } 579 enc_dbcs = enc_dbcs_new; 580 has_mbyte = (enc_dbcs != 0 || enc_utf8); 581 582 #ifdef WIN3264 583 enc_codepage = encname2codepage(p_enc); 584 enc_latin9 = (STRCMP(p_enc, "iso-8859-15") == 0); 585 #endif 586 587 /* Detect an encoding that uses latin1 characters. */ 588 enc_latin1like = (enc_utf8 || STRCMP(p_enc, "latin1") == 0 589 || STRCMP(p_enc, "iso-8859-15") == 0); 590 591 /* 592 * Set the function pointers. 593 */ 594 if (enc_utf8) 595 { 596 mb_ptr2len = utfc_ptr2len; 597 mb_char2len = utf_char2len; 598 mb_char2bytes = utf_char2bytes; 599 mb_ptr2cells = utf_ptr2cells; 600 mb_char2cells = utf_char2cells; 601 mb_off2cells = utf_off2cells; 602 mb_ptr2char = utf_ptr2char; 603 mb_head_off = utf_head_off; 604 } 605 else if (enc_dbcs != 0) 606 { 607 mb_ptr2len = dbcs_ptr2len; 608 mb_char2len = dbcs_char2len; 609 mb_char2bytes = dbcs_char2bytes; 610 mb_ptr2cells = dbcs_ptr2cells; 611 mb_char2cells = dbcs_char2cells; 612 mb_off2cells = dbcs_off2cells; 613 mb_ptr2char = dbcs_ptr2char; 614 mb_head_off = dbcs_head_off; 615 } 616 else 617 { 618 mb_ptr2len = latin_ptr2len; 619 mb_char2len = latin_char2len; 620 mb_char2bytes = latin_char2bytes; 621 mb_ptr2cells = latin_ptr2cells; 622 mb_char2cells = latin_char2cells; 623 mb_off2cells = latin_off2cells; 624 mb_ptr2char = latin_ptr2char; 625 mb_head_off = latin_head_off; 626 } 627 628 /* 629 * Fill the mb_bytelen_tab[] for MB_BYTE2LEN(). 630 */ 631 #ifdef LEN_FROM_CONV 632 /* When 'encoding' is different from the current locale mblen() won't 633 * work. Use conversion to "utf-8" instead. */ 634 vimconv.vc_type = CONV_NONE; 635 if (enc_dbcs) 636 { 637 p = enc_locale(); 638 if (p == NULL || STRCMP(p, p_enc) != 0) 639 { 640 convert_setup(&vimconv, p_enc, (char_u *)"utf-8"); 641 vimconv.vc_fail = TRUE; 642 } 643 vim_free(p); 644 } 645 #endif 646 647 for (i = 0; i < 256; ++i) 648 { 649 /* Our own function to reliably check the length of UTF-8 characters, 650 * independent of mblen(). */ 651 if (enc_utf8) 652 n = utf8len_tab[i]; 653 else if (enc_dbcs == 0) 654 n = 1; 655 else 656 { 657 #if defined(WIN3264) || defined(WIN32UNIX) 658 /* enc_dbcs is set by setting 'fileencoding'. It becomes a Windows 659 * CodePage identifier, which we can pass directly in to Windows 660 * API */ 661 n = IsDBCSLeadByteEx(enc_dbcs, (BYTE)i) ? 2 : 1; 662 #else 663 # ifdef MACOS 664 /* 665 * if mblen() is not available, character which MSB is turned on 666 * are treated as leading byte character. (note : This assumption 667 * is not always true.) 668 */ 669 n = (i & 0x80) ? 2 : 1; 670 # else 671 char buf[MB_MAXBYTES]; 672 # ifdef X_LOCALE 673 # ifndef mblen 674 # define mblen _Xmblen 675 # endif 676 # endif 677 if (i == NUL) /* just in case mblen() can't handle "" */ 678 n = 1; 679 else 680 { 681 buf[0] = i; 682 buf[1] = 0; 683 #ifdef LEN_FROM_CONV 684 if (vimconv.vc_type != CONV_NONE) 685 { 686 /* 687 * string_convert() should fail when converting the first 688 * byte of a double-byte character. 689 */ 690 p = string_convert(&vimconv, (char_u *)buf, NULL); 691 if (p != NULL) 692 { 693 vim_free(p); 694 n = 1; 695 } 696 else 697 n = 2; 698 } 699 else 700 #endif 701 { 702 /* 703 * mblen() should return -1 for invalid (means the leading 704 * multibyte) character. However there are some platforms 705 * where mblen() returns 0 for invalid character. 706 * Therefore, following condition includes 0. 707 */ 708 (void)mblen(NULL, 0); /* First reset the state. */ 709 if (mblen(buf, (size_t)1) <= 0) 710 n = 2; 711 else 712 n = 1; 713 } 714 } 715 # endif 716 #endif 717 } 718 719 mb_bytelen_tab[i] = n; 720 } 721 722 #ifdef LEN_FROM_CONV 723 convert_setup(&vimconv, NULL, NULL); 724 #endif 725 726 /* The cell width depends on the type of multi-byte characters. */ 727 (void)init_chartab(); 728 729 /* When enc_utf8 is set or reset, (de)allocate ScreenLinesUC[] */ 730 screenalloc(FALSE); 731 732 /* When using Unicode, set default for 'fileencodings'. */ 733 if (enc_utf8 && !option_was_set((char_u *)"fencs")) 734 set_string_option_direct((char_u *)"fencs", -1, 735 (char_u *)"ucs-bom,utf-8,default,latin1", OPT_FREE, 0); 736 737 #if defined(HAVE_BIND_TEXTDOMAIN_CODESET) && defined(FEAT_GETTEXT) 738 /* GNU gettext 0.10.37 supports this feature: set the codeset used for 739 * translated messages independently from the current locale. */ 740 (void)bind_textdomain_codeset(VIMPACKAGE, 741 enc_utf8 ? "utf-8" : (char *)p_enc); 742 #endif 743 744 #ifdef WIN32 745 /* When changing 'encoding' while starting up, then convert the command 746 * line arguments from the active codepage to 'encoding'. */ 747 if (starting != 0) 748 fix_arg_enc(); 749 #endif 750 751 #ifdef FEAT_AUTOCMD 752 /* Fire an autocommand to let people do custom font setup. This must be 753 * after Vim has been setup for the new encoding. */ 754 apply_autocmds(EVENT_ENCODINGCHANGED, NULL, (char_u *)"", FALSE, curbuf); 755 #endif 756 757 #ifdef FEAT_SPELL 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() 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(p) 1065 char_u *p; 1066 { 1067 return MB_BYTE2LEN(*p); 1068 } 1069 1070 static int 1071 dbcs_ptr2len(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(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)(*pp); 1415 return c; 1416 } 1417 1418 /* 1419 * Get character at **pp and advance *pp to the next character. 1420 * Note: composing characters are returned as separate characters. 1421 */ 1422 int 1423 mb_cptr2char_adv(pp) 1424 char_u **pp; 1425 { 1426 int c; 1427 1428 c = (*mb_ptr2char)(*pp); 1429 if (enc_utf8) 1430 *pp += utf_ptr2len(*pp); 1431 else 1432 *pp += (*mb_ptr2len)(*pp); 1433 return c; 1434 } 1435 1436 #if defined(FEAT_ARABIC) || defined(PROTO) 1437 /* 1438 * Check whether we are dealing with Arabic combining characters. 1439 * Note: these are NOT really composing characters! 1440 */ 1441 int 1442 arabic_combine(one, two) 1443 int one; /* first character */ 1444 int two; /* character just after "one" */ 1445 { 1446 if (one == a_LAM) 1447 return arabic_maycombine(two); 1448 return FALSE; 1449 } 1450 1451 /* 1452 * Check whether we are dealing with a character that could be regarded as an 1453 * Arabic combining character, need to check the character before this. 1454 */ 1455 int 1456 arabic_maycombine(two) 1457 int two; 1458 { 1459 if (p_arshape && !p_tbidi) 1460 return (two == a_ALEF_MADDA 1461 || two == a_ALEF_HAMZA_ABOVE 1462 || two == a_ALEF_HAMZA_BELOW 1463 || two == a_ALEF); 1464 return FALSE; 1465 } 1466 1467 /* 1468 * Check if the character pointed to by "p2" is a composing character when it 1469 * comes after "p1". For Arabic sometimes "ab" is replaced with "c", which 1470 * behaves like a composing character. 1471 */ 1472 int 1473 utf_composinglike(p1, p2) 1474 char_u *p1; 1475 char_u *p2; 1476 { 1477 int c2; 1478 1479 c2 = utf_ptr2char(p2); 1480 if (utf_iscomposing(c2)) 1481 return TRUE; 1482 if (!arabic_maycombine(c2)) 1483 return FALSE; 1484 return arabic_combine(utf_ptr2char(p1), c2); 1485 } 1486 #endif 1487 1488 /* 1489 * Convert a UTF-8 byte string to a wide chararacter. Also get up to MAX_MCO 1490 * composing characters. 1491 */ 1492 int 1493 utfc_ptr2char(p, pcc) 1494 char_u *p; 1495 int *pcc; /* return: composing chars, last one is 0 */ 1496 { 1497 int len; 1498 int c; 1499 int cc; 1500 int i = 0; 1501 1502 c = utf_ptr2char(p); 1503 len = utf_ptr2len(p); 1504 1505 /* Only accept a composing char when the first char isn't illegal. */ 1506 if ((len > 1 || *p < 0x80) 1507 && p[len] >= 0x80 1508 && UTF_COMPOSINGLIKE(p, p + len)) 1509 { 1510 cc = utf_ptr2char(p + len); 1511 for (;;) 1512 { 1513 pcc[i++] = cc; 1514 if (i == MAX_MCO) 1515 break; 1516 len += utf_ptr2len(p + len); 1517 if (p[len] < 0x80 || !utf_iscomposing(cc = utf_ptr2char(p + len))) 1518 break; 1519 } 1520 } 1521 1522 if (i < MAX_MCO) /* last composing char must be 0 */ 1523 pcc[i] = 0; 1524 1525 return c; 1526 } 1527 1528 /* 1529 * Convert a UTF-8 byte string to a wide chararacter. Also get up to MAX_MCO 1530 * composing characters. Use no more than p[maxlen]. 1531 */ 1532 int 1533 utfc_ptr2char_len(p, pcc, maxlen) 1534 char_u *p; 1535 int *pcc; /* return: composing chars, last one is 0 */ 1536 int maxlen; 1537 { 1538 int len; 1539 int c; 1540 int cc; 1541 int i = 0; 1542 1543 c = utf_ptr2char(p); 1544 len = utf_ptr2len_len(p, maxlen); 1545 /* Only accept a composing char when the first char isn't illegal. */ 1546 if ((len > 1 || *p < 0x80) 1547 && len < maxlen 1548 && p[len] >= 0x80 1549 && UTF_COMPOSINGLIKE(p, p + len)) 1550 { 1551 cc = utf_ptr2char(p + len); 1552 for (;;) 1553 { 1554 pcc[i++] = cc; 1555 if (i == MAX_MCO) 1556 break; 1557 len += utf_ptr2len_len(p + len, maxlen - len); 1558 if (len >= maxlen 1559 || p[len] < 0x80 1560 || !utf_iscomposing(cc = utf_ptr2char(p + len))) 1561 break; 1562 } 1563 } 1564 1565 if (i < MAX_MCO) /* last composing char must be 0 */ 1566 pcc[i] = 0; 1567 1568 return c; 1569 } 1570 1571 /* 1572 * Convert the character at screen position "off" to a sequence of bytes. 1573 * Includes the composing characters. 1574 * "buf" must at least have the length MB_MAXBYTES. 1575 * Returns the produced number of bytes. 1576 */ 1577 int 1578 utfc_char2bytes(off, buf) 1579 int off; 1580 char_u *buf; 1581 { 1582 int len; 1583 int i; 1584 1585 len = utf_char2bytes(ScreenLinesUC[off], buf); 1586 for (i = 0; i < Screen_mco; ++i) 1587 { 1588 if (ScreenLinesC[i][off] == 0) 1589 break; 1590 len += utf_char2bytes(ScreenLinesC[i][off], buf + len); 1591 } 1592 return len; 1593 } 1594 1595 /* 1596 * Get the length of a UTF-8 byte sequence, not including any following 1597 * composing characters. 1598 * Returns 0 for "". 1599 * Returns 1 for an illegal byte sequence. 1600 */ 1601 int 1602 utf_ptr2len(p) 1603 char_u *p; 1604 { 1605 int len; 1606 int i; 1607 1608 if (*p == NUL) 1609 return 0; 1610 len = utf8len_tab[*p]; 1611 for (i = 1; i < len; ++i) 1612 if ((p[i] & 0xc0) != 0x80) 1613 return 1; 1614 return len; 1615 } 1616 1617 /* 1618 * Return length of UTF-8 character, obtained from the first byte. 1619 * "b" must be between 0 and 255! 1620 */ 1621 int 1622 utf_byte2len(b) 1623 int b; 1624 { 1625 return utf8len_tab[b]; 1626 } 1627 1628 /* 1629 * Get the length of UTF-8 byte sequence "p[size]". Does not include any 1630 * following composing characters. 1631 * Returns 1 for "". 1632 * Returns 1 for an illegal byte sequence. 1633 * Returns number > "size" for an incomplete byte sequence. 1634 */ 1635 int 1636 utf_ptr2len_len(p, size) 1637 char_u *p; 1638 int size; 1639 { 1640 int len; 1641 int i; 1642 1643 if (*p == NUL) 1644 return 1; 1645 len = utf8len_tab[*p]; 1646 if (len > size) 1647 return len; /* incomplete byte sequence. */ 1648 for (i = 1; i < len; ++i) 1649 if ((p[i] & 0xc0) != 0x80) 1650 return 1; 1651 return len; 1652 } 1653 1654 /* 1655 * Return the number of bytes the UTF-8 encoding of the character at "p" takes. 1656 * This includes following composing characters. 1657 */ 1658 int 1659 utfc_ptr2len(p) 1660 char_u *p; 1661 { 1662 int len; 1663 int b0 = *p; 1664 #ifdef FEAT_ARABIC 1665 int prevlen; 1666 #endif 1667 1668 if (b0 == NUL) 1669 return 0; 1670 if (b0 < 0x80 && p[1] < 0x80) /* be quick for ASCII */ 1671 return 1; 1672 1673 /* Skip over first UTF-8 char, stopping at a NUL byte. */ 1674 len = utf_ptr2len(p); 1675 1676 /* Check for illegal byte. */ 1677 if (len == 1 && b0 >= 0x80) 1678 return 1; 1679 1680 /* 1681 * Check for composing characters. We can handle only the first two, but 1682 * skip all of them (otherwise the cursor would get stuck). 1683 */ 1684 #ifdef FEAT_ARABIC 1685 prevlen = 0; 1686 #endif 1687 for (;;) 1688 { 1689 if (p[len] < 0x80 || !UTF_COMPOSINGLIKE(p + prevlen, p + len)) 1690 return len; 1691 1692 /* Skip over composing char */ 1693 #ifdef FEAT_ARABIC 1694 prevlen = len; 1695 #endif 1696 len += utf_ptr2len(p + len); 1697 } 1698 } 1699 1700 /* 1701 * Return the number of bytes the UTF-8 encoding of the character at "p[size]" 1702 * takes. This includes following composing characters. 1703 * Returns 1 for an illegal char or an incomplete byte sequence. 1704 */ 1705 int 1706 utfc_ptr2len_len(p, size) 1707 char_u *p; 1708 int size; 1709 { 1710 int len; 1711 #ifdef FEAT_ARABIC 1712 int prevlen; 1713 #endif 1714 1715 if (*p == NUL) 1716 return 0; 1717 if (p[0] < 0x80 && (size == 1 || p[1] < 0x80)) /* be quick for ASCII */ 1718 return 1; 1719 1720 /* Skip over first UTF-8 char, stopping at a NUL byte. */ 1721 len = utf_ptr2len_len(p, size); 1722 1723 /* Check for illegal byte and incomplete byte sequence. */ 1724 if ((len == 1 && p[0] >= 0x80) || len > size) 1725 return 1; 1726 1727 /* 1728 * Check for composing characters. We can handle only the first two, but 1729 * skip all of them (otherwise the cursor would get stuck). 1730 */ 1731 #ifdef FEAT_ARABIC 1732 prevlen = 0; 1733 #endif 1734 while (len < size) 1735 { 1736 if (p[len] < 0x80 || !UTF_COMPOSINGLIKE(p + prevlen, p + len)) 1737 break; 1738 1739 /* Skip over composing char */ 1740 #ifdef FEAT_ARABIC 1741 prevlen = len; 1742 #endif 1743 len += utf_ptr2len_len(p + len, size - len); 1744 } 1745 return len; 1746 } 1747 1748 /* 1749 * Return the number of bytes the UTF-8 encoding of character "c" takes. 1750 * This does not include composing characters. 1751 */ 1752 int 1753 utf_char2len(c) 1754 int c; 1755 { 1756 if (c < 0x80) 1757 return 1; 1758 if (c < 0x800) 1759 return 2; 1760 if (c < 0x10000) 1761 return 3; 1762 if (c < 0x200000) 1763 return 4; 1764 if (c < 0x4000000) 1765 return 5; 1766 return 6; 1767 } 1768 1769 /* 1770 * Convert Unicode character "c" to UTF-8 string in "buf[]". 1771 * Returns the number of bytes. 1772 * This does not include composing characters. 1773 */ 1774 int 1775 utf_char2bytes(c, buf) 1776 int c; 1777 char_u *buf; 1778 { 1779 if (c < 0x80) /* 7 bits */ 1780 { 1781 buf[0] = c; 1782 return 1; 1783 } 1784 if (c < 0x800) /* 11 bits */ 1785 { 1786 buf[0] = 0xc0 + ((unsigned)c >> 6); 1787 buf[1] = 0x80 + (c & 0x3f); 1788 return 2; 1789 } 1790 if (c < 0x10000) /* 16 bits */ 1791 { 1792 buf[0] = 0xe0 + ((unsigned)c >> 12); 1793 buf[1] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1794 buf[2] = 0x80 + (c & 0x3f); 1795 return 3; 1796 } 1797 if (c < 0x200000) /* 21 bits */ 1798 { 1799 buf[0] = 0xf0 + ((unsigned)c >> 18); 1800 buf[1] = 0x80 + (((unsigned)c >> 12) & 0x3f); 1801 buf[2] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1802 buf[3] = 0x80 + (c & 0x3f); 1803 return 4; 1804 } 1805 if (c < 0x4000000) /* 26 bits */ 1806 { 1807 buf[0] = 0xf8 + ((unsigned)c >> 24); 1808 buf[1] = 0x80 + (((unsigned)c >> 18) & 0x3f); 1809 buf[2] = 0x80 + (((unsigned)c >> 12) & 0x3f); 1810 buf[3] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1811 buf[4] = 0x80 + (c & 0x3f); 1812 return 5; 1813 } 1814 /* 31 bits */ 1815 buf[0] = 0xfc + ((unsigned)c >> 30); 1816 buf[1] = 0x80 + (((unsigned)c >> 24) & 0x3f); 1817 buf[2] = 0x80 + (((unsigned)c >> 18) & 0x3f); 1818 buf[3] = 0x80 + (((unsigned)c >> 12) & 0x3f); 1819 buf[4] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1820 buf[5] = 0x80 + (c & 0x3f); 1821 return 6; 1822 } 1823 1824 /* 1825 * Return TRUE if "c" is a composing UTF-8 character. This means it will be 1826 * drawn on top of the preceding character. 1827 * Based on code from Markus Kuhn. 1828 */ 1829 int 1830 utf_iscomposing(c) 1831 int c; 1832 { 1833 /* sorted list of non-overlapping intervals */ 1834 static struct interval combining[] = 1835 { 1836 {0x0300, 0x034f}, {0x0360, 0x036f}, {0x0483, 0x0486}, {0x0488, 0x0489}, 1837 {0x0591, 0x05a1}, {0x05a3, 0x05b9}, {0x05bb, 0x05bd}, {0x05bf, 0x05bf}, 1838 {0x05c1, 0x05c2}, {0x05c4, 0x05c4}, {0x0610, 0x0615}, {0x064b, 0x0658}, 1839 {0x0670, 0x0670}, {0x06d6, 0x06dc}, {0x06de, 0x06e4}, {0x06e7, 0x06e8}, 1840 {0x06ea, 0x06ed}, {0x0711, 0x0711}, {0x0730, 0x074a}, {0x07a6, 0x07b0}, 1841 {0x0901, 0x0903}, {0x093c, 0x093c}, {0x093e, 0x094d}, {0x0951, 0x0954}, 1842 {0x0962, 0x0963}, {0x0981, 0x0983}, {0x09bc, 0x09bc}, {0x09be, 0x09c4}, 1843 {0x09c7, 0x09c8}, {0x09cb, 0x09cd}, {0x09d7, 0x09d7}, {0x09e2, 0x09e3}, 1844 {0x0a01, 0x0a03}, {0x0a3c, 0x0a3c}, {0x0a3e, 0x0a42}, {0x0a47, 0x0a48}, 1845 {0x0a4b, 0x0a4d}, {0x0a70, 0x0a71}, {0x0a81, 0x0a83}, {0x0abc, 0x0abc}, 1846 {0x0abe, 0x0ac5}, {0x0ac7, 0x0ac9}, {0x0acb, 0x0acd}, {0x0ae2, 0x0ae3}, 1847 {0x0b01, 0x0b03}, {0x0b3c, 0x0b3c}, {0x0b3e, 0x0b43}, {0x0b47, 0x0b48}, 1848 {0x0b4b, 0x0b4d}, {0x0b56, 0x0b57}, {0x0b82, 0x0b82}, {0x0bbe, 0x0bc2}, 1849 {0x0bc6, 0x0bc8}, {0x0bca, 0x0bcd}, {0x0bd7, 0x0bd7}, {0x0c01, 0x0c03}, 1850 {0x0c3e, 0x0c44}, {0x0c46, 0x0c48}, {0x0c4a, 0x0c4d}, {0x0c55, 0x0c56}, 1851 {0x0c82, 0x0c83}, {0x0cbc, 0x0cbc}, {0x0cbe, 0x0cc4}, {0x0cc6, 0x0cc8}, 1852 {0x0cca, 0x0ccd}, {0x0cd5, 0x0cd6}, {0x0d02, 0x0d03}, {0x0d3e, 0x0d43}, 1853 {0x0d46, 0x0d48}, {0x0d4a, 0x0d4d}, {0x0d57, 0x0d57}, {0x0d82, 0x0d83}, 1854 {0x0dca, 0x0dca}, {0x0dcf, 0x0dd4}, {0x0dd6, 0x0dd6}, {0x0dd8, 0x0ddf}, 1855 {0x0df2, 0x0df3}, {0x0e31, 0x0e31}, {0x0e34, 0x0e3a}, {0x0e47, 0x0e4e}, 1856 {0x0eb1, 0x0eb1}, {0x0eb4, 0x0eb9}, {0x0ebb, 0x0ebc}, {0x0ec8, 0x0ecd}, 1857 {0x0f18, 0x0f19}, {0x0f35, 0x0f35}, {0x0f37, 0x0f37}, {0x0f39, 0x0f39}, 1858 {0x0f3e, 0x0f3f}, {0x0f71, 0x0f84}, {0x0f86, 0x0f87}, {0x0f90, 0x0f97}, 1859 {0x0f99, 0x0fbc}, {0x0fc6, 0x0fc6}, {0x102c, 0x1032}, {0x1036, 0x1039}, 1860 {0x1056, 0x1059}, {0x1712, 0x1714}, {0x1732, 0x1734}, {0x1752, 0x1753}, 1861 {0x1772, 0x1773}, {0x17b6, 0x17d3}, {0x17dd, 0x17dd}, {0x180b, 0x180d}, 1862 {0x18a9, 0x18a9}, {0x1920, 0x192b}, {0x1930, 0x193b}, {0x20d0, 0x20ea}, 1863 {0x302a, 0x302f}, {0x3099, 0x309a}, {0xfb1e, 0xfb1e}, {0xfe00, 0xfe0f}, 1864 {0xfe20, 0xfe23}, 1865 }; 1866 1867 return intable(combining, sizeof(combining), c); 1868 } 1869 1870 /* 1871 * Return TRUE for characters that can be displayed in a normal way. 1872 * Only for characters of 0x100 and above! 1873 */ 1874 int 1875 utf_printable(c) 1876 int c; 1877 { 1878 #ifdef USE_WCHAR_FUNCTIONS 1879 /* 1880 * Assume the iswprint() library function works better than our own stuff. 1881 */ 1882 return iswprint(c); 1883 #else 1884 /* Sorted list of non-overlapping intervals. 1885 * 0xd800-0xdfff is reserved for UTF-16, actually illegal. */ 1886 static struct interval nonprint[] = 1887 { 1888 {0x070f, 0x070f}, {0x180b, 0x180e}, {0x200b, 0x200f}, {0x202a, 0x202e}, 1889 {0x206a, 0x206f}, {0xd800, 0xdfff}, {0xfeff, 0xfeff}, {0xfff9, 0xfffb}, 1890 {0xfffe, 0xffff} 1891 }; 1892 1893 return !intable(nonprint, sizeof(nonprint), c); 1894 #endif 1895 } 1896 1897 /* 1898 * Get class of a Unicode character. 1899 * 0: white space 1900 * 1: punctuation 1901 * 2 or bigger: some class of word character. 1902 */ 1903 int 1904 utf_class(c) 1905 int c; 1906 { 1907 /* sorted list of non-overlapping intervals */ 1908 static struct clinterval 1909 { 1910 unsigned short first; 1911 unsigned short last; 1912 unsigned short class; 1913 } classes[] = 1914 { 1915 {0x037e, 0x037e, 1}, /* Greek question mark */ 1916 {0x0387, 0x0387, 1}, /* Greek ano teleia */ 1917 {0x055a, 0x055f, 1}, /* Armenian punctuation */ 1918 {0x0589, 0x0589, 1}, /* Armenian full stop */ 1919 {0x05be, 0x05be, 1}, 1920 {0x05c0, 0x05c0, 1}, 1921 {0x05c3, 0x05c3, 1}, 1922 {0x05f3, 0x05f4, 1}, 1923 {0x060c, 0x060c, 1}, 1924 {0x061b, 0x061b, 1}, 1925 {0x061f, 0x061f, 1}, 1926 {0x066a, 0x066d, 1}, 1927 {0x06d4, 0x06d4, 1}, 1928 {0x0700, 0x070d, 1}, /* Syriac punctuation */ 1929 {0x0964, 0x0965, 1}, 1930 {0x0970, 0x0970, 1}, 1931 {0x0df4, 0x0df4, 1}, 1932 {0x0e4f, 0x0e4f, 1}, 1933 {0x0e5a, 0x0e5b, 1}, 1934 {0x0f04, 0x0f12, 1}, 1935 {0x0f3a, 0x0f3d, 1}, 1936 {0x0f85, 0x0f85, 1}, 1937 {0x104a, 0x104f, 1}, /* Myanmar punctuation */ 1938 {0x10fb, 0x10fb, 1}, /* Georgian punctuation */ 1939 {0x1361, 0x1368, 1}, /* Ethiopic punctuation */ 1940 {0x166d, 0x166e, 1}, /* Canadian Syl. punctuation */ 1941 {0x1680, 0x1680, 0}, 1942 {0x169b, 0x169c, 1}, 1943 {0x16eb, 0x16ed, 1}, 1944 {0x1735, 0x1736, 1}, 1945 {0x17d4, 0x17dc, 1}, /* Khmer punctuation */ 1946 {0x1800, 0x180a, 1}, /* Mongolian punctuation */ 1947 {0x2000, 0x200b, 0}, /* spaces */ 1948 {0x200c, 0x2027, 1}, /* punctuation and symbols */ 1949 {0x2028, 0x2029, 0}, 1950 {0x202a, 0x202e, 1}, /* punctuation and symbols */ 1951 {0x202f, 0x202f, 0}, 1952 {0x2030, 0x205e, 1}, /* punctuation and symbols */ 1953 {0x205f, 0x205f, 0}, 1954 {0x2060, 0x27ff, 1}, /* punctuation and symbols */ 1955 {0x2070, 0x207f, 0x2070}, /* superscript */ 1956 {0x2080, 0x208f, 0x2080}, /* subscript */ 1957 {0x2983, 0x2998, 1}, 1958 {0x29d8, 0x29db, 1}, 1959 {0x29fc, 0x29fd, 1}, 1960 {0x3000, 0x3000, 0}, /* ideographic space */ 1961 {0x3001, 0x3020, 1}, /* ideographic punctuation */ 1962 {0x3030, 0x3030, 1}, 1963 {0x303d, 0x303d, 1}, 1964 {0x3040, 0x309f, 0x3040}, /* Hiragana */ 1965 {0x30a0, 0x30ff, 0x30a0}, /* Katakana */ 1966 {0x3300, 0x9fff, 0x4e00}, /* CJK Ideographs */ 1967 {0xac00, 0xd7a3, 0xac00}, /* Hangul Syllables */ 1968 {0xf900, 0xfaff, 0x4e00}, /* CJK Ideographs */ 1969 {0xfd3e, 0xfd3f, 1}, 1970 {0xfe30, 0xfe6b, 1}, /* punctuation forms */ 1971 {0xff00, 0xff0f, 1}, /* half/fullwidth ASCII */ 1972 {0xff1a, 0xff20, 1}, /* half/fullwidth ASCII */ 1973 {0xff3b, 0xff40, 1}, /* half/fullwidth ASCII */ 1974 {0xff5b, 0xff65, 1}, /* half/fullwidth ASCII */ 1975 }; 1976 int bot = 0; 1977 int top = sizeof(classes) / sizeof(struct clinterval) - 1; 1978 int mid; 1979 1980 /* First quick check for Latin1 characters, use 'iskeyword'. */ 1981 if (c < 0x100) 1982 { 1983 if (c == ' ' || c == '\t' || c == NUL || c == 0xa0) 1984 return 0; /* blank */ 1985 if (vim_iswordc(c)) 1986 return 2; /* word character */ 1987 return 1; /* punctuation */ 1988 } 1989 1990 /* binary search in table */ 1991 while (top >= bot) 1992 { 1993 mid = (bot + top) / 2; 1994 if (classes[mid].last < c) 1995 bot = mid + 1; 1996 else if (classes[mid].first > c) 1997 top = mid - 1; 1998 else 1999 return (int)classes[mid].class; 2000 } 2001 2002 /* most other characters are "word" characters */ 2003 return 2; 2004 } 2005 2006 /* 2007 * Code for Unicode case-dependent operations. Based on notes in 2008 * http://www.unicode.org/Public/UNIDATA/CaseFolding.txt 2009 * This code uses simple case folding, not full case folding. 2010 */ 2011 2012 /* 2013 * The following table is built by foldExtract.pl < CaseFolding.txt . 2014 * It must be in numeric order, because we use binary search on it. 2015 * An entry such as {0x41,0x5a,1,32} means that UCS-4 characters in the range 2016 * from 0x41 to 0x5a inclusive, stepping by 1, are folded by adding 32. 2017 */ 2018 2019 typedef struct 2020 { 2021 int rangeStart; 2022 int rangeEnd; 2023 int step; 2024 int offset; 2025 } convertStruct; 2026 2027 static convertStruct foldCase[] = 2028 { 2029 {0x41,0x5a,1,32}, {0xc0,0xd6,1,32}, {0xd8,0xde,1,32}, 2030 {0x100,0x12e,2,1}, {0x130,0x130,-1,-199}, {0x132,0x136,2,1}, 2031 {0x139,0x147,2,1}, {0x14a,0x176,2,1}, {0x178,0x178,-1,-121}, 2032 {0x179,0x17d,2,1}, {0x181,0x181,-1,210}, {0x182,0x184,2,1}, 2033 {0x186,0x186,-1,206}, {0x187,0x187,-1,1}, {0x189,0x18a,1,205}, 2034 {0x18b,0x18b,-1,1}, {0x18e,0x18e,-1,79}, {0x18f,0x18f,-1,202}, 2035 {0x190,0x190,-1,203}, {0x191,0x191,-1,1}, {0x193,0x193,-1,205}, 2036 {0x194,0x194,-1,207}, {0x196,0x196,-1,211}, {0x197,0x197,-1,209}, 2037 {0x198,0x198,-1,1}, {0x19c,0x19c,-1,211}, {0x19d,0x19d,-1,213}, 2038 {0x19f,0x19f,-1,214}, {0x1a0,0x1a4,2,1}, {0x1a6,0x1a6,-1,218}, 2039 {0x1a7,0x1a7,-1,1}, {0x1a9,0x1a9,-1,218}, {0x1ac,0x1ac,-1,1}, 2040 {0x1ae,0x1ae,-1,218}, {0x1af,0x1af,-1,1}, {0x1b1,0x1b2,1,217}, 2041 {0x1b3,0x1b5,2,1}, {0x1b7,0x1b7,-1,219}, {0x1b8,0x1bc,4,1}, 2042 {0x1c4,0x1c4,-1,2}, {0x1c5,0x1c5,-1,1}, {0x1c7,0x1c7,-1,2}, 2043 {0x1c8,0x1c8,-1,1}, {0x1ca,0x1ca,-1,2}, {0x1cb,0x1db,2,1}, 2044 {0x1de,0x1ee,2,1}, {0x1f1,0x1f1,-1,2}, {0x1f2,0x1f4,2,1}, 2045 {0x1f6,0x1f6,-1,-97}, {0x1f7,0x1f7,-1,-56}, {0x1f8,0x21e,2,1}, 2046 {0x220,0x220,-1,-130}, {0x222,0x232,2,1}, {0x386,0x386,-1,38}, 2047 {0x388,0x38a,1,37}, {0x38c,0x38c,-1,64}, {0x38e,0x38f,1,63}, 2048 {0x391,0x3a1,1,32}, {0x3a3,0x3ab,1,32}, {0x3d8,0x3ee,2,1}, 2049 {0x3f4,0x3f4,-1,-60}, {0x3f7,0x3f7,-1,1}, {0x3f9,0x3f9,-1,-7}, 2050 {0x3fa,0x3fa,-1,1}, {0x400,0x40f,1,80}, {0x410,0x42f,1,32}, 2051 {0x460,0x480,2,1}, {0x48a,0x4be,2,1}, {0x4c1,0x4cd,2,1}, 2052 {0x4d0,0x4f4,2,1}, {0x4f8,0x500,8,1}, {0x502,0x50e,2,1}, 2053 {0x531,0x556,1,48}, {0x1e00,0x1e94,2,1}, {0x1ea0,0x1ef8,2,1}, 2054 {0x1f08,0x1f0f,1,-8}, {0x1f18,0x1f1d,1,-8}, {0x1f28,0x1f2f,1,-8}, 2055 {0x1f38,0x1f3f,1,-8}, {0x1f48,0x1f4d,1,-8}, {0x1f59,0x1f5f,2,-8}, 2056 {0x1f68,0x1f6f,1,-8}, {0x1f88,0x1f8f,1,-8}, {0x1f98,0x1f9f,1,-8}, 2057 {0x1fa8,0x1faf,1,-8}, {0x1fb8,0x1fb9,1,-8}, {0x1fba,0x1fbb,1,-74}, 2058 {0x1fbc,0x1fbc,-1,-9}, {0x1fc8,0x1fcb,1,-86}, {0x1fcc,0x1fcc,-1,-9}, 2059 {0x1fd8,0x1fd9,1,-8}, {0x1fda,0x1fdb,1,-100}, {0x1fe8,0x1fe9,1,-8}, 2060 {0x1fea,0x1feb,1,-112}, {0x1fec,0x1fec,-1,-7}, {0x1ff8,0x1ff9,1,-128}, 2061 {0x1ffa,0x1ffb,1,-126}, {0x1ffc,0x1ffc,-1,-9}, {0x2126,0x2126,-1,-7517}, 2062 {0x212a,0x212a,-1,-8383}, {0x212b,0x212b,-1,-8262}, 2063 {0x2160,0x216f,1,16}, {0x24b6,0x24cf,1,26}, {0xff21,0xff3a,1,32}, 2064 {0x10400,0x10427,1,40} 2065 }; 2066 2067 static int utf_convert(int a, convertStruct table[], int tableSize); 2068 2069 /* 2070 * Generic conversion function for case operations. 2071 * Return the converted equivalent of "a", which is a UCS-4 character. Use 2072 * the given conversion "table". Uses binary search on "table". 2073 */ 2074 static int 2075 utf_convert(a, table, tableSize) 2076 int a; 2077 convertStruct table[]; 2078 int tableSize; 2079 { 2080 int start, mid, end; /* indices into table */ 2081 2082 start = 0; 2083 end = tableSize / sizeof(convertStruct); 2084 while (start < end) 2085 { 2086 /* need to search further */ 2087 mid = (end + start) /2; 2088 if (table[mid].rangeEnd < a) 2089 start = mid + 1; 2090 else 2091 end = mid; 2092 } 2093 if (table[start].rangeStart <= a && a <= table[start].rangeEnd 2094 && (a - table[start].rangeStart) % table[start].step == 0) 2095 return (a + table[start].offset); 2096 else 2097 return a; 2098 } 2099 2100 /* 2101 * Return the folded-case equivalent of "a", which is a UCS-4 character. Uses 2102 * simple case folding. 2103 */ 2104 int 2105 utf_fold(a) 2106 int a; 2107 { 2108 return utf_convert(a, foldCase, sizeof(foldCase)); 2109 } 2110 2111 /* 2112 * The following tables are built by upperLowerExtract.pl < UnicodeData.txt . 2113 * They must be in numeric order, because we use binary search on them. 2114 * An entry such as {0x41,0x5a,1,32} means that UCS-4 characters in the range 2115 * from 0x41 to 0x5a inclusive, stepping by 1, are switched to lower (for 2116 * example) by adding 32. 2117 */ 2118 static convertStruct toLower[] = 2119 { 2120 {0x41,0x5a,1,32}, {0xc0,0xd6,1,32}, {0xd8,0xde,1,32}, 2121 {0x100,0x12e,2,1}, {0x130,0x130,-1,-199}, {0x132,0x136,2,1}, 2122 {0x139,0x147,2,1}, {0x14a,0x176,2,1}, {0x178,0x178,-1,-121}, 2123 {0x179,0x17d,2,1}, {0x181,0x181,-1,210}, {0x182,0x184,2,1}, 2124 {0x186,0x186,-1,206}, {0x187,0x187,-1,1}, {0x189,0x18a,1,205}, 2125 {0x18b,0x18b,-1,1}, {0x18e,0x18e,-1,79}, {0x18f,0x18f,-1,202}, 2126 {0x190,0x190,-1,203}, {0x191,0x191,-1,1}, {0x193,0x193,-1,205}, 2127 {0x194,0x194,-1,207}, {0x196,0x196,-1,211}, {0x197,0x197,-1,209}, 2128 {0x198,0x198,-1,1}, {0x19c,0x19c,-1,211}, {0x19d,0x19d,-1,213}, 2129 {0x19f,0x19f,-1,214}, {0x1a0,0x1a4,2,1}, {0x1a6,0x1a6,-1,218}, 2130 {0x1a7,0x1a7,-1,1}, {0x1a9,0x1a9,-1,218}, {0x1ac,0x1ac,-1,1}, 2131 {0x1ae,0x1ae,-1,218}, {0x1af,0x1af,-1,1}, {0x1b1,0x1b2,1,217}, 2132 {0x1b3,0x1b5,2,1}, {0x1b7,0x1b7,-1,219}, {0x1b8,0x1bc,4,1}, 2133 {0x1c4,0x1ca,3,2}, {0x1cd,0x1db,2,1}, {0x1de,0x1ee,2,1}, 2134 {0x1f1,0x1f1,-1,2}, {0x1f4,0x1f4,-1,1}, {0x1f6,0x1f6,-1,-97}, 2135 {0x1f7,0x1f7,-1,-56}, {0x1f8,0x21e,2,1}, {0x220,0x220,-1,-130}, 2136 {0x222,0x232,2,1}, {0x386,0x386,-1,38}, {0x388,0x38a,1,37}, 2137 {0x38c,0x38c,-1,64}, {0x38e,0x38f,1,63}, {0x391,0x3a1,1,32}, 2138 {0x3a3,0x3ab,1,32}, {0x3d8,0x3ee,2,1}, {0x3f4,0x3f4,-1,-60}, 2139 {0x3f7,0x3f7,-1,1}, {0x3f9,0x3f9,-1,-7}, {0x3fa,0x3fa,-1,1}, 2140 {0x400,0x40f,1,80}, {0x410,0x42f,1,32}, {0x460,0x480,2,1}, 2141 {0x48a,0x4be,2,1}, {0x4c1,0x4cd,2,1}, {0x4d0,0x4f4,2,1}, 2142 {0x4f8,0x500,8,1}, {0x502,0x50e,2,1}, {0x531,0x556,1,48}, 2143 {0x1e00,0x1e94,2,1}, {0x1ea0,0x1ef8,2,1}, {0x1f08,0x1f0f,1,-8}, 2144 {0x1f18,0x1f1d,1,-8}, {0x1f28,0x1f2f,1,-8}, {0x1f38,0x1f3f,1,-8}, 2145 {0x1f48,0x1f4d,1,-8}, {0x1f59,0x1f5f,2,-8}, {0x1f68,0x1f6f,1,-8}, 2146 {0x1fb8,0x1fb9,1,-8}, {0x1fba,0x1fbb,1,-74}, {0x1fc8,0x1fcb,1,-86}, 2147 {0x1fd8,0x1fd9,1,-8}, {0x1fda,0x1fdb,1,-100}, {0x1fe8,0x1fe9,1,-8}, 2148 {0x1fea,0x1feb,1,-112}, {0x1fec,0x1fec,-1,-7}, {0x1ff8,0x1ff9,1,-128}, 2149 {0x1ffa,0x1ffb,1,-126}, {0x2126,0x2126,-1,-7517}, {0x212a,0x212a,-1,-8383}, 2150 {0x212b,0x212b,-1,-8262}, {0xff21,0xff3a,1,32}, {0x10400,0x10427,1,40} 2151 }; 2152 2153 static convertStruct toUpper[] = 2154 { 2155 {0x61,0x7a,1,-32}, {0xb5,0xb5,-1,743}, {0xe0,0xf6,1,-32}, 2156 {0xf8,0xfe,1,-32}, {0xff,0xff,-1,121}, {0x101,0x12f,2,-1}, 2157 {0x131,0x131,-1,-232}, {0x133,0x137,2,-1}, {0x13a,0x148,2,-1}, 2158 {0x14b,0x177,2,-1}, {0x17a,0x17e,2,-1}, {0x17f,0x17f,-1,-300}, 2159 {0x183,0x185,2,-1}, {0x188,0x18c,4,-1}, {0x192,0x192,-1,-1}, 2160 {0x195,0x195,-1,97}, {0x199,0x199,-1,-1}, {0x19e,0x19e,-1,130}, 2161 {0x1a1,0x1a5,2,-1}, {0x1a8,0x1ad,5,-1}, {0x1b0,0x1b4,4,-1}, 2162 {0x1b6,0x1b9,3,-1}, {0x1bd,0x1bd,-1,-1}, {0x1bf,0x1bf,-1,56}, 2163 {0x1c5,0x1c6,1,-1}, {0x1c8,0x1c9,1,-1}, {0x1cb,0x1cc,1,-1}, 2164 {0x1ce,0x1dc,2,-1}, {0x1dd,0x1dd,-1,-79}, {0x1df,0x1ef,2,-1}, 2165 {0x1f2,0x1f3,1,-1}, {0x1f5,0x1f9,4,-1}, {0x1fb,0x21f,2,-1}, 2166 {0x223,0x233,2,-1}, {0x253,0x253,-1,-210}, {0x254,0x254,-1,-206}, 2167 {0x256,0x257,1,-205}, {0x259,0x259,-1,-202}, {0x25b,0x25b,-1,-203}, 2168 {0x260,0x260,-1,-205}, {0x263,0x263,-1,-207}, {0x268,0x268,-1,-209}, 2169 {0x269,0x26f,6,-211}, {0x272,0x272,-1,-213}, {0x275,0x275,-1,-214}, 2170 {0x280,0x283,3,-218}, {0x288,0x288,-1,-218}, {0x28a,0x28b,1,-217}, 2171 {0x292,0x292,-1,-219}, {0x3ac,0x3ac,-1,-38}, {0x3ad,0x3af,1,-37}, 2172 {0x3b1,0x3c1,1,-32}, {0x3c2,0x3c2,-1,-31}, {0x3c3,0x3cb,1,-32}, 2173 {0x3cc,0x3cc,-1,-64}, {0x3cd,0x3ce,1,-63}, {0x3d0,0x3d0,-1,-62}, 2174 {0x3d1,0x3d1,-1,-57}, {0x3d5,0x3d5,-1,-47}, {0x3d6,0x3d6,-1,-54}, 2175 {0x3d9,0x3ef,2,-1}, {0x3f0,0x3f0,-1,-86}, {0x3f1,0x3f1,-1,-80}, 2176 {0x3f2,0x3f2,-1,7}, {0x3f5,0x3f5,-1,-96}, {0x3f8,0x3fb,3,-1}, 2177 {0x430,0x44f,1,-32}, {0x450,0x45f,1,-80}, {0x461,0x481,2,-1}, 2178 {0x48b,0x4bf,2,-1}, {0x4c2,0x4ce,2,-1}, {0x4d1,0x4f5,2,-1}, 2179 {0x4f9,0x501,8,-1}, {0x503,0x50f,2,-1}, {0x561,0x586,1,-48}, 2180 {0x1e01,0x1e95,2,-1}, {0x1e9b,0x1e9b,-1,-59}, {0x1ea1,0x1ef9,2,-1}, 2181 {0x1f00,0x1f07,1,8}, {0x1f10,0x1f15,1,8}, {0x1f20,0x1f27,1,8}, 2182 {0x1f30,0x1f37,1,8}, {0x1f40,0x1f45,1,8}, {0x1f51,0x1f57,2,8}, 2183 {0x1f60,0x1f67,1,8}, {0x1f70,0x1f71,1,74}, {0x1f72,0x1f75,1,86}, 2184 {0x1f76,0x1f77,1,100}, {0x1f78,0x1f79,1,128}, {0x1f7a,0x1f7b,1,112}, 2185 {0x1f7c,0x1f7d,1,126}, {0x1f80,0x1f87,1,8}, {0x1f90,0x1f97,1,8}, 2186 {0x1fa0,0x1fa7,1,8}, {0x1fb0,0x1fb1,1,8}, {0x1fb3,0x1fb3,-1,9}, 2187 {0x1fbe,0x1fbe,-1,-7205}, {0x1fc3,0x1fc3,-1,9}, {0x1fd0,0x1fd1,1,8}, 2188 {0x1fe0,0x1fe1,1,8}, {0x1fe5,0x1fe5,-1,7}, {0x1ff3,0x1ff3,-1,9}, 2189 {0xff41,0xff5a,1,-32}, {0x10428,0x1044f,1,-40} 2190 }; 2191 2192 /* 2193 * Return the upper-case equivalent of "a", which is a UCS-4 character. Use 2194 * simple case folding. 2195 */ 2196 int 2197 utf_toupper(a) 2198 int a; 2199 { 2200 /* If 'casemap' contains "keepascii" use ASCII style toupper(). */ 2201 if (a < 128 && (cmp_flags & CMP_KEEPASCII)) 2202 return TOUPPER_ASC(a); 2203 2204 #if defined(HAVE_TOWUPPER) && defined(__STDC__ISO_10646__) 2205 /* If towupper() is availble and handles Unicode, use it. */ 2206 if (!(cmp_flags & CMP_INTERNAL)) 2207 return towupper(a); 2208 #endif 2209 2210 /* For characters below 128 use locale sensitive toupper(). */ 2211 if (a < 128) 2212 return TOUPPER_LOC(a); 2213 2214 /* For any other characters use the above mapping table. */ 2215 return utf_convert(a, toUpper, sizeof(toUpper)); 2216 } 2217 2218 int 2219 utf_islower(a) 2220 int a; 2221 { 2222 return (utf_toupper(a) != a); 2223 } 2224 2225 /* 2226 * Return the lower-case equivalent of "a", which is a UCS-4 character. Use 2227 * simple case folding. 2228 */ 2229 int 2230 utf_tolower(a) 2231 int a; 2232 { 2233 /* If 'casemap' contains "keepascii" use ASCII style tolower(). */ 2234 if (a < 128 && (cmp_flags & CMP_KEEPASCII)) 2235 return TOLOWER_ASC(a); 2236 2237 #if defined(HAVE_TOWLOWER) && defined(__STDC__ISO_10646__) 2238 /* If towlower() is available and handles Unicode, use it. */ 2239 if (!(cmp_flags & CMP_INTERNAL)) 2240 return towlower(a); 2241 #endif 2242 2243 /* For characters below 128 use locale sensitive tolower(). */ 2244 if (a < 128) 2245 return TOLOWER_LOC(a); 2246 2247 /* For any other characters use the above mapping table. */ 2248 return utf_convert(a, toLower, sizeof(toLower)); 2249 } 2250 2251 int 2252 utf_isupper(a) 2253 int a; 2254 { 2255 return (utf_tolower(a) != a); 2256 } 2257 2258 /* 2259 * Version of strnicmp() that handles multi-byte characters. 2260 * Needed for Big5, Sjift-JIS and UTF-8 encoding. Other DBCS encodings can 2261 * probably use strnicmp(), because there are no ASCII characters in the 2262 * second byte. 2263 * Returns zero if s1 and s2 are equal (ignoring case), the difference between 2264 * two characters otherwise. 2265 */ 2266 int 2267 mb_strnicmp(s1, s2, nn) 2268 char_u *s1, *s2; 2269 size_t nn; 2270 { 2271 int i, j, l; 2272 int cdiff; 2273 int incomplete = FALSE; 2274 int n = nn; 2275 2276 for (i = 0; i < n; i += l) 2277 { 2278 if (s1[i] == NUL && s2[i] == NUL) /* both strings end */ 2279 return 0; 2280 if (enc_utf8) 2281 { 2282 l = utf_byte2len(s1[i]); 2283 if (l > n - i) 2284 { 2285 l = n - i; /* incomplete character */ 2286 incomplete = TRUE; 2287 } 2288 /* Check directly first, it's faster. */ 2289 for (j = 0; j < l; ++j) 2290 if (s1[i + j] != s2[i + j]) 2291 break; 2292 if (j < l) 2293 { 2294 /* If one of the two characters is incomplete return -1. */ 2295 if (incomplete || i + utf_byte2len(s2[i]) > n) 2296 return -1; 2297 cdiff = utf_fold(utf_ptr2char(s1 + i)) 2298 - utf_fold(utf_ptr2char(s2 + i)); 2299 if (cdiff != 0) 2300 return cdiff; 2301 } 2302 } 2303 else 2304 { 2305 l = (*mb_ptr2len)(s1 + i); 2306 if (l <= 1) 2307 { 2308 /* Single byte: first check normally, then with ignore case. */ 2309 if (s1[i] != s2[i]) 2310 { 2311 cdiff = TOLOWER_LOC(s1[i]) - TOLOWER_LOC(s2[i]); 2312 if (cdiff != 0) 2313 return cdiff; 2314 } 2315 } 2316 else 2317 { 2318 /* For non-Unicode multi-byte don't ignore case. */ 2319 if (l > n - i) 2320 l = n - i; 2321 cdiff = STRNCMP(s1 + i, s2 + i, l); 2322 if (cdiff != 0) 2323 return cdiff; 2324 } 2325 } 2326 } 2327 return 0; 2328 } 2329 2330 /* 2331 * "g8": show bytes of the UTF-8 char under the cursor. Doesn't matter what 2332 * 'encoding' has been set to. 2333 */ 2334 void 2335 show_utf8() 2336 { 2337 int len; 2338 int rlen = 0; 2339 char_u *line; 2340 int clen; 2341 int i; 2342 2343 /* Get the byte length of the char under the cursor, including composing 2344 * characters. */ 2345 line = ml_get_cursor(); 2346 len = utfc_ptr2len(line); 2347 if (len == 0) 2348 { 2349 MSG("NUL"); 2350 return; 2351 } 2352 2353 clen = 0; 2354 for (i = 0; i < len; ++i) 2355 { 2356 if (clen == 0) 2357 { 2358 /* start of (composing) character, get its length */ 2359 if (i > 0) 2360 { 2361 STRCPY(IObuff + rlen, "+ "); 2362 rlen += 2; 2363 } 2364 clen = utf_ptr2len(line + i); 2365 } 2366 sprintf((char *)IObuff + rlen, "%02x ", line[i]); 2367 --clen; 2368 rlen += STRLEN(IObuff + rlen); 2369 if (rlen > IOSIZE - 20) 2370 break; 2371 } 2372 2373 msg(IObuff); 2374 } 2375 2376 /* 2377 * mb_head_off() function pointer. 2378 * Return offset from "p" to the first byte of the character it points into. 2379 * Returns 0 when already at the first byte of a character. 2380 */ 2381 /*ARGSUSED*/ 2382 int 2383 latin_head_off(base, p) 2384 char_u *base; 2385 char_u *p; 2386 { 2387 return 0; 2388 } 2389 2390 int 2391 dbcs_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 if (p <= base || MB_BYTE2LEN(p[-1]) == 1) 2400 return 0; 2401 2402 /* This is slow: need to start at the base and go forward until the 2403 * byte we are looking for. Return 1 when we went past it, 0 otherwise. */ 2404 q = base; 2405 while (q < p) 2406 q += dbcs_ptr2len(q); 2407 return (q == p) ? 0 : 1; 2408 } 2409 2410 #if defined(FEAT_CLIPBOARD) || defined(FEAT_GUI) || defined(FEAT_RIGHTLEFT) \ 2411 || defined(PROTO) 2412 /* 2413 * Special version of dbcs_head_off() that works for ScreenLines[], where 2414 * single-width DBCS_JPNU characters are stored separately. 2415 */ 2416 int 2417 dbcs_screen_head_off(base, p) 2418 char_u *base; 2419 char_u *p; 2420 { 2421 char_u *q; 2422 2423 /* It can't be a trailing byte when not using DBCS, at the start of the 2424 * string or the previous byte can't start a double-byte. 2425 * For euc-jp an 0x8e byte in the previous cell always means we have a 2426 * lead byte in the current cell. */ 2427 if (p <= base 2428 || (enc_dbcs == DBCS_JPNU && p[-1] == 0x8e) 2429 || MB_BYTE2LEN(p[-1]) == 1) 2430 return 0; 2431 2432 /* This is slow: need to start at the base and go forward until the 2433 * byte we are looking for. Return 1 when we went past it, 0 otherwise. 2434 * For DBCS_JPNU look out for 0x8e, which means the second byte is not 2435 * stored as the next byte. */ 2436 q = base; 2437 while (q < p) 2438 { 2439 if (enc_dbcs == DBCS_JPNU && *q == 0x8e) 2440 ++q; 2441 else 2442 q += dbcs_ptr2len(q); 2443 } 2444 return (q == p) ? 0 : 1; 2445 } 2446 #endif 2447 2448 int 2449 utf_head_off(base, p) 2450 char_u *base; 2451 char_u *p; 2452 { 2453 char_u *q; 2454 char_u *s; 2455 int c; 2456 #ifdef FEAT_ARABIC 2457 char_u *j; 2458 #endif 2459 2460 if (*p < 0x80) /* be quick for ASCII */ 2461 return 0; 2462 2463 /* Skip backwards over trailing bytes: 10xx.xxxx 2464 * Skip backwards again if on a composing char. */ 2465 for (q = p; ; --q) 2466 { 2467 /* Move s to the last byte of this char. */ 2468 for (s = q; (s[1] & 0xc0) == 0x80; ++s) 2469 ; 2470 /* Move q to the first byte of this char. */ 2471 while (q > base && (*q & 0xc0) == 0x80) 2472 --q; 2473 /* Check for illegal sequence. Do allow an illegal byte after where we 2474 * started. */ 2475 if (utf8len_tab[*q] != (int)(s - q + 1) 2476 && utf8len_tab[*q] != (int)(p - q + 1)) 2477 return 0; 2478 2479 if (q <= base) 2480 break; 2481 2482 c = utf_ptr2char(q); 2483 if (utf_iscomposing(c)) 2484 continue; 2485 2486 #ifdef FEAT_ARABIC 2487 if (arabic_maycombine(c)) 2488 { 2489 /* Advance to get a sneak-peak at the next char */ 2490 j = q; 2491 --j; 2492 /* Move j to the first byte of this char. */ 2493 while (j > base && (*j & 0xc0) == 0x80) 2494 --j; 2495 if (arabic_combine(utf_ptr2char(j), c)) 2496 continue; 2497 } 2498 #endif 2499 break; 2500 } 2501 2502 return (int)(p - q); 2503 } 2504 2505 #if defined(FEAT_EVAL) || defined(PROTO) 2506 /* 2507 * Copy a character from "*fp" to "*tp" and advance the pointers. 2508 */ 2509 void 2510 mb_copy_char(fp, tp) 2511 char_u **fp; 2512 char_u **tp; 2513 { 2514 int l = (*mb_ptr2len)(*fp); 2515 2516 mch_memmove(*tp, *fp, (size_t)l); 2517 *tp += l; 2518 *fp += l; 2519 } 2520 #endif 2521 2522 /* 2523 * Return the offset from "p" to the first byte of a character. When "p" is 2524 * at the start of a character 0 is returned, otherwise the offset to the next 2525 * character. Can start anywhere in a stream of bytes. 2526 */ 2527 int 2528 mb_off_next(base, p) 2529 char_u *base; 2530 char_u *p; 2531 { 2532 int i; 2533 int j; 2534 2535 if (enc_utf8) 2536 { 2537 if (*p < 0x80) /* be quick for ASCII */ 2538 return 0; 2539 2540 /* Find the next character that isn't 10xx.xxxx */ 2541 for (i = 0; (p[i] & 0xc0) == 0x80; ++i) 2542 ; 2543 if (i > 0) 2544 { 2545 /* Check for illegal sequence. */ 2546 for (j = 0; p - j > base; ++j) 2547 if ((p[-j] & 0xc0) != 0x80) 2548 break; 2549 if (utf8len_tab[p[-j]] != i + j) 2550 return 0; 2551 } 2552 return i; 2553 } 2554 2555 /* Only need to check if we're on a trail byte, it doesn't matter if we 2556 * want the offset to the next or current character. */ 2557 return (*mb_head_off)(base, p); 2558 } 2559 2560 /* 2561 * Return the offset from "p" to the last byte of the character it points 2562 * into. Can start anywhere in a stream of bytes. 2563 */ 2564 int 2565 mb_tail_off(base, p) 2566 char_u *base; 2567 char_u *p; 2568 { 2569 int i; 2570 int j; 2571 2572 if (*p == NUL) 2573 return 0; 2574 2575 if (enc_utf8) 2576 { 2577 /* Find the last character that is 10xx.xxxx */ 2578 for (i = 0; (p[i + 1] & 0xc0) == 0x80; ++i) 2579 ; 2580 /* Check for illegal sequence. */ 2581 for (j = 0; p - j > base; ++j) 2582 if ((p[-j] & 0xc0) != 0x80) 2583 break; 2584 if (utf8len_tab[p[-j]] != i + j + 1) 2585 return 0; 2586 return i; 2587 } 2588 2589 /* It can't be the first byte if a double-byte when not using DBCS, at the 2590 * end of the string or the byte can't start a double-byte. */ 2591 if (enc_dbcs == 0 || p[1] == NUL || MB_BYTE2LEN(*p) == 1) 2592 return 0; 2593 2594 /* Return 1 when on the lead byte, 0 when on the tail byte. */ 2595 return 1 - dbcs_head_off(base, p); 2596 } 2597 2598 /* 2599 * Find the next illegal byte sequence. 2600 */ 2601 void 2602 utf_find_illegal() 2603 { 2604 pos_T pos = curwin->w_cursor; 2605 char_u *p; 2606 int len; 2607 vimconv_T vimconv; 2608 char_u *tofree = NULL; 2609 2610 vimconv.vc_type = CONV_NONE; 2611 if (enc_utf8 && (enc_canon_props(curbuf->b_p_fenc) & ENC_8BIT)) 2612 { 2613 /* 'encoding' is "utf-8" but we are editing a 8-bit encoded file, 2614 * possibly a utf-8 file with illegal bytes. Setup for conversion 2615 * from utf-8 to 'fileencoding'. */ 2616 convert_setup(&vimconv, p_enc, curbuf->b_p_fenc); 2617 } 2618 2619 #ifdef FEAT_VIRTUALEDIT 2620 curwin->w_cursor.coladd = 0; 2621 #endif 2622 for (;;) 2623 { 2624 p = ml_get_cursor(); 2625 if (vimconv.vc_type != CONV_NONE) 2626 { 2627 vim_free(tofree); 2628 tofree = string_convert(&vimconv, p, NULL); 2629 if (tofree == NULL) 2630 break; 2631 p = tofree; 2632 } 2633 2634 while (*p != NUL) 2635 { 2636 /* Illegal means that there are not enough trail bytes (checked by 2637 * utf_ptr2len()) or too many of them (overlong sequence). */ 2638 len = utf_ptr2len(p); 2639 if (*p >= 0x80 && (len == 1 2640 || utf_char2len(utf_ptr2char(p)) != len)) 2641 { 2642 if (vimconv.vc_type == CONV_NONE) 2643 curwin->w_cursor.col += p - ml_get_cursor(); 2644 else 2645 { 2646 int l; 2647 2648 len = p - tofree; 2649 for (p = ml_get_cursor(); *p != NUL && len-- > 0; p += l) 2650 { 2651 l = utf_ptr2len(p); 2652 curwin->w_cursor.col += l; 2653 } 2654 } 2655 goto theend; 2656 } 2657 p += len; 2658 } 2659 if (curwin->w_cursor.lnum == curbuf->b_ml.ml_line_count) 2660 break; 2661 ++curwin->w_cursor.lnum; 2662 curwin->w_cursor.col = 0; 2663 } 2664 2665 /* didn't find it: don't move and beep */ 2666 curwin->w_cursor = pos; 2667 beep_flush(); 2668 2669 theend: 2670 vim_free(tofree); 2671 convert_setup(&vimconv, NULL, NULL); 2672 } 2673 2674 #if defined(HAVE_GTK2) || defined(PROTO) 2675 /* 2676 * Return TRUE if string "s" is a valid utf-8 string. 2677 * When "end" is NULL stop at the first NUL. 2678 * When "end" is positive stop there. 2679 */ 2680 int 2681 utf_valid_string(s, end) 2682 char_u *s; 2683 char_u *end; 2684 { 2685 int l; 2686 char_u *p = s; 2687 2688 while (end == NULL ? *p != NUL : p < end) 2689 { 2690 if ((*p & 0xc0) == 0x80) 2691 return FALSE; /* invalid lead byte */ 2692 l = utf8len_tab[*p]; 2693 if (end != NULL && p + l > end) 2694 return FALSE; /* incomplete byte sequence */ 2695 ++p; 2696 while (--l > 0) 2697 if ((*p++ & 0xc0) != 0x80) 2698 return FALSE; /* invalid trail byte */ 2699 } 2700 return TRUE; 2701 } 2702 #endif 2703 2704 #if defined(FEAT_GUI) || defined(PROTO) 2705 /* 2706 * Special version of mb_tail_off() for use in ScreenLines[]. 2707 */ 2708 int 2709 dbcs_screen_tail_off(base, p) 2710 char_u *base; 2711 char_u *p; 2712 { 2713 /* It can't be the first byte if a double-byte when not using DBCS, at the 2714 * end of the string or the byte can't start a double-byte. 2715 * For euc-jp an 0x8e byte always means we have a lead byte in the current 2716 * cell. */ 2717 if (*p == NUL || p[1] == NUL 2718 || (enc_dbcs == DBCS_JPNU && *p == 0x8e) 2719 || MB_BYTE2LEN(*p) == 1) 2720 return 0; 2721 2722 /* Return 1 when on the lead byte, 0 when on the tail byte. */ 2723 return 1 - dbcs_screen_head_off(base, p); 2724 } 2725 #endif 2726 2727 /* 2728 * If the cursor moves on an trail byte, set the cursor on the lead byte. 2729 * Thus it moves left if necessary. 2730 * Return TRUE when the cursor was adjusted. 2731 */ 2732 void 2733 mb_adjust_cursor() 2734 { 2735 mb_adjustpos(&curwin->w_cursor); 2736 } 2737 2738 /* 2739 * Adjust position "*lp" to point to the first byte of a multi-byte character. 2740 * If it points to a tail byte it's moved backwards to the head byte. 2741 */ 2742 void 2743 mb_adjustpos(lp) 2744 pos_T *lp; 2745 { 2746 char_u *p; 2747 2748 if (lp->col > 0 2749 #ifdef FEAT_VIRTUALEDIT 2750 || lp->coladd > 1 2751 #endif 2752 ) 2753 { 2754 p = ml_get(lp->lnum); 2755 lp->col -= (*mb_head_off)(p, p + lp->col); 2756 #ifdef FEAT_VIRTUALEDIT 2757 /* Reset "coladd" when the cursor would be on the right half of a 2758 * double-wide character. */ 2759 if (lp->coladd == 1 2760 && p[lp->col] != TAB 2761 && vim_isprintc((*mb_ptr2char)(p + lp->col)) 2762 && ptr2cells(p + lp->col) > 1) 2763 lp->coladd = 0; 2764 #endif 2765 } 2766 } 2767 2768 /* 2769 * Return a pointer to the character before "*p", if there is one. 2770 */ 2771 char_u * 2772 mb_prevptr(line, p) 2773 char_u *line; /* start of the string */ 2774 char_u *p; 2775 { 2776 if (p > line) 2777 mb_ptr_back(line, p); 2778 return p; 2779 } 2780 2781 /* 2782 * Return the character length of "str". Each multi-byte character (with 2783 * following composing characters) counts as one. 2784 */ 2785 int 2786 mb_charlen(str) 2787 char_u *str; 2788 { 2789 char_u *p = str; 2790 int count; 2791 2792 if (p == NULL) 2793 return 0; 2794 2795 for (count = 0; *p != NUL; count++) 2796 p += (*mb_ptr2len)(p); 2797 2798 return count; 2799 } 2800 2801 #if defined(FEAT_SPELL) || defined(PROTO) 2802 /* 2803 * Like mb_charlen() but for a string with specified length. 2804 */ 2805 int 2806 mb_charlen_len(str, len) 2807 char_u *str; 2808 int len; 2809 { 2810 char_u *p = str; 2811 int count; 2812 2813 for (count = 0; *p != NUL && p < str + len; count++) 2814 p += (*mb_ptr2len)(p); 2815 2816 return count; 2817 } 2818 #endif 2819 2820 /* 2821 * Try to un-escape a multi-byte character. 2822 * Used for the "to" and "from" part of a mapping. 2823 * Return the un-escaped string if it is a multi-byte character, and advance 2824 * "pp" to just after the bytes that formed it. 2825 * Return NULL if no multi-byte char was found. 2826 */ 2827 char_u * 2828 mb_unescape(pp) 2829 char_u **pp; 2830 { 2831 static char_u buf[MB_MAXBYTES + 1]; 2832 int n, m = 0; 2833 char_u *str = *pp; 2834 2835 /* Must translate K_SPECIAL KS_SPECIAL KE_FILLER to K_SPECIAL and CSI 2836 * KS_EXTRA KE_CSI to CSI. */ 2837 for (n = 0; str[n] != NUL && m <= MB_MAXBYTES; ++n) 2838 { 2839 if (str[n] == K_SPECIAL 2840 && str[n + 1] == KS_SPECIAL 2841 && str[n + 2] == KE_FILLER) 2842 { 2843 buf[m++] = K_SPECIAL; 2844 n += 2; 2845 } 2846 # ifdef FEAT_GUI 2847 else if (str[n] == CSI 2848 && str[n + 1] == KS_EXTRA 2849 && str[n + 2] == (int)KE_CSI) 2850 { 2851 buf[m++] = CSI; 2852 n += 2; 2853 } 2854 # endif 2855 else if (str[n] == K_SPECIAL 2856 # ifdef FEAT_GUI 2857 || str[n] == CSI 2858 # endif 2859 ) 2860 break; /* a special key can't be a multibyte char */ 2861 else 2862 buf[m++] = str[n]; 2863 buf[m] = NUL; 2864 2865 /* Return a multi-byte character if it's found. An illegal sequence 2866 * will result in a 1 here. */ 2867 if ((*mb_ptr2len)(buf) > 1) 2868 { 2869 *pp = str + n + 1; 2870 return buf; 2871 } 2872 } 2873 return NULL; 2874 } 2875 2876 /* 2877 * Return TRUE if the character at "row"/"col" on the screen is the left side 2878 * of a double-width character. 2879 * Caller must make sure "row" and "col" are not invalid! 2880 */ 2881 int 2882 mb_lefthalve(row, col) 2883 int row; 2884 int col; 2885 { 2886 #ifdef FEAT_HANGULIN 2887 if (composing_hangul) 2888 return TRUE; 2889 #endif 2890 if (enc_dbcs != 0) 2891 return dbcs_off2cells(LineOffset[row] + col) > 1; 2892 if (enc_utf8) 2893 return (col + 1 < Columns 2894 && ScreenLines[LineOffset[row] + col + 1] == 0); 2895 return FALSE; 2896 } 2897 2898 # if defined(FEAT_CLIPBOARD) || defined(FEAT_GUI) || defined(FEAT_RIGHTLEFT) \ 2899 || defined(PROTO) 2900 /* 2901 * Correct a position on the screen, if it's the right halve of a double-wide 2902 * char move it to the left halve. Returns the corrected column. 2903 */ 2904 int 2905 mb_fix_col(col, row) 2906 int col; 2907 int row; 2908 { 2909 col = check_col(col); 2910 row = check_row(row); 2911 if (has_mbyte && ScreenLines != NULL && col > 0 2912 && ((enc_dbcs 2913 && ScreenLines[LineOffset[row] + col] != NUL 2914 && dbcs_screen_head_off(ScreenLines + LineOffset[row], 2915 ScreenLines + LineOffset[row] + col)) 2916 || (enc_utf8 && ScreenLines[LineOffset[row] + col] == 0))) 2917 --col; 2918 return col; 2919 } 2920 # endif 2921 #endif 2922 2923 #if defined(FEAT_MBYTE) || defined(FEAT_POSTSCRIPT) || defined(PROTO) 2924 static int enc_alias_search __ARGS((char_u *name)); 2925 2926 /* 2927 * Skip the Vim specific head of a 'encoding' name. 2928 */ 2929 char_u * 2930 enc_skip(p) 2931 char_u *p; 2932 { 2933 if (STRNCMP(p, "2byte-", 6) == 0) 2934 return p + 6; 2935 if (STRNCMP(p, "8bit-", 5) == 0) 2936 return p + 5; 2937 return p; 2938 } 2939 2940 /* 2941 * Find the canonical name for encoding "enc". 2942 * When the name isn't recognized, returns "enc" itself, but with all lower 2943 * case characters and '_' replaced with '-'. 2944 * Returns an allocated string. NULL for out-of-memory. 2945 */ 2946 char_u * 2947 enc_canonize(enc) 2948 char_u *enc; 2949 { 2950 char_u *r; 2951 char_u *p, *s; 2952 int i; 2953 2954 # ifdef FEAT_MBYTE 2955 if (STRCMP(enc, "default") == 0) 2956 { 2957 /* Use the default encoding as it's found by set_init_1(). */ 2958 r = get_encoding_default(); 2959 if (r == NULL) 2960 r = (char_u *)"latin1"; 2961 return vim_strsave(r); 2962 } 2963 # endif 2964 2965 /* copy "enc" to allocted memory, with room for two '-' */ 2966 r = alloc((unsigned)(STRLEN(enc) + 3)); 2967 if (r != NULL) 2968 { 2969 /* Make it all lower case and replace '_' with '-'. */ 2970 p = r; 2971 for (s = enc; *s != NUL; ++s) 2972 { 2973 if (*s == '_') 2974 *p++ = '-'; 2975 else 2976 *p++ = TOLOWER_ASC(*s); 2977 } 2978 *p = NUL; 2979 2980 /* Skip "2byte-" and "8bit-". */ 2981 p = enc_skip(r); 2982 2983 /* Change "microsoft-cp" to "cp". Used in some spell files. */ 2984 if (STRNCMP(p, "microsoft-cp", 12) == 0) 2985 mch_memmove(p, p + 10, STRLEN(p + 10) + 1); 2986 2987 /* "iso8859" -> "iso-8859" */ 2988 if (STRNCMP(p, "iso8859", 7) == 0) 2989 { 2990 mch_memmove(p + 4, p + 3, STRLEN(p + 2)); 2991 p[3] = '-'; 2992 } 2993 2994 /* "iso-8859n" -> "iso-8859-n" */ 2995 if (STRNCMP(p, "iso-8859", 8) == 0 && p[8] != '-') 2996 { 2997 mch_memmove(p + 9, p + 8, STRLEN(p + 7)); 2998 p[8] = '-'; 2999 } 3000 3001 /* "latin-N" -> "latinN" */ 3002 if (STRNCMP(p, "latin-", 6) == 0) 3003 mch_memmove(p + 5, p + 6, STRLEN(p + 5)); 3004 3005 if (enc_canon_search(p) >= 0) 3006 { 3007 /* canonical name can be used unmodified */ 3008 if (p != r) 3009 mch_memmove(r, p, STRLEN(p) + 1); 3010 } 3011 else if ((i = enc_alias_search(p)) >= 0) 3012 { 3013 /* alias recognized, get canonical name */ 3014 vim_free(r); 3015 r = vim_strsave((char_u *)enc_canon_table[i].name); 3016 } 3017 } 3018 return r; 3019 } 3020 3021 /* 3022 * Search for an encoding alias of "name". 3023 * Returns -1 when not found. 3024 */ 3025 static int 3026 enc_alias_search(name) 3027 char_u *name; 3028 { 3029 int i; 3030 3031 for (i = 0; enc_alias_table[i].name != NULL; ++i) 3032 if (STRCMP(name, enc_alias_table[i].name) == 0) 3033 return enc_alias_table[i].canon; 3034 return -1; 3035 } 3036 #endif 3037 3038 #if defined(FEAT_MBYTE) || defined(PROTO) 3039 3040 #ifdef HAVE_LANGINFO_H 3041 # include <langinfo.h> 3042 #endif 3043 3044 /* 3045 * Get the canonicalized encoding of the current locale. 3046 * Returns an allocated string when successful, NULL when not. 3047 */ 3048 char_u * 3049 enc_locale() 3050 { 3051 #ifndef WIN3264 3052 char *s; 3053 char *p; 3054 int i; 3055 #endif 3056 char buf[50]; 3057 #ifdef WIN3264 3058 long acp = GetACP(); 3059 3060 if (acp == 1200) 3061 STRCPY(buf, "ucs-2le"); 3062 else if (acp == 1252) /* cp1252 is used as latin1 */ 3063 STRCPY(buf, "latin1"); 3064 else 3065 sprintf(buf, "cp%ld", acp); 3066 #else 3067 # ifdef HAVE_NL_LANGINFO_CODESET 3068 if ((s = nl_langinfo(CODESET)) == NULL || *s == NUL) 3069 # endif 3070 # if defined(HAVE_LOCALE_H) || defined(X_LOCALE) 3071 if ((s = setlocale(LC_CTYPE, NULL)) == NULL || *s == NUL) 3072 # endif 3073 if ((s = getenv("LC_ALL")) == NULL || *s == NUL) 3074 if ((s = getenv("LC_CTYPE")) == NULL || *s == NUL) 3075 s = getenv("LANG"); 3076 3077 if (s == NULL || *s == NUL) 3078 return FAIL; 3079 3080 /* The most generic locale format is: 3081 * language[_territory][.codeset][@modifier][+special][,[sponsor][_revision]] 3082 * If there is a '.' remove the part before it. 3083 * if there is something after the codeset, remove it. 3084 * Make the name lowercase and replace '_' with '-'. 3085 * Exception: "ja_JP.EUC" == "euc-jp", "zh_CN.EUC" = "euc-cn", 3086 * "ko_KR.EUC" == "euc-kr" 3087 */ 3088 if ((p = (char *)vim_strchr((char_u *)s, '.')) != NULL) 3089 { 3090 if (p > s + 2 && STRNICMP(p + 1, "EUC", 3) == 0 3091 && !isalnum((int)p[4]) && p[4] != '-' && p[-3] == '_') 3092 { 3093 /* copy "XY.EUC" to "euc-XY" to buf[10] */ 3094 STRCPY(buf + 10, "euc-"); 3095 buf[14] = p[-2]; 3096 buf[15] = p[-1]; 3097 buf[16] = 0; 3098 s = buf + 10; 3099 } 3100 else 3101 s = p + 1; 3102 } 3103 for (i = 0; s[i] != NUL && i < sizeof(buf) - 1; ++i) 3104 { 3105 if (s[i] == '_' || s[i] == '-') 3106 buf[i] = '-'; 3107 else if (isalnum((int)s[i])) 3108 buf[i] = TOLOWER_ASC(s[i]); 3109 else 3110 break; 3111 } 3112 buf[i] = NUL; 3113 #endif 3114 3115 return enc_canonize((char_u *)buf); 3116 } 3117 3118 #if defined(WIN3264) || defined(PROTO) 3119 /* 3120 * Convert an encoding name to an MS-Windows codepage. 3121 * Returns zero if no codepage can be figured out. 3122 */ 3123 int 3124 encname2codepage(name) 3125 char_u *name; 3126 { 3127 int cp; 3128 char_u *p = name; 3129 int idx; 3130 3131 if (STRNCMP(p, "8bit-", 5) == 0) 3132 p += 5; 3133 else if (STRNCMP(p_enc, "2byte-", 6) == 0) 3134 p += 6; 3135 3136 if (p[0] == 'c' && p[1] == 'p') 3137 cp = atoi(p + 2); 3138 else if ((idx = enc_canon_search(p)) >= 0) 3139 cp = enc_canon_table[idx].codepage; 3140 else 3141 return 0; 3142 if (IsValidCodePage(cp)) 3143 return cp; 3144 return 0; 3145 } 3146 #endif 3147 3148 # if defined(USE_ICONV) || defined(PROTO) 3149 3150 static char_u *iconv_string __ARGS((vimconv_T *vcp, char_u *str, int slen, int *unconvlenp)); 3151 3152 /* 3153 * Call iconv_open() with a check if iconv() works properly (there are broken 3154 * versions). 3155 * Returns (void *)-1 if failed. 3156 * (should return iconv_t, but that causes problems with prototypes). 3157 */ 3158 void * 3159 my_iconv_open(to, from) 3160 char_u *to; 3161 char_u *from; 3162 { 3163 iconv_t fd; 3164 #define ICONV_TESTLEN 400 3165 char_u tobuf[ICONV_TESTLEN]; 3166 char *p; 3167 size_t tolen; 3168 static int iconv_ok = -1; 3169 3170 if (iconv_ok == FALSE) 3171 return (void *)-1; /* detected a broken iconv() previously */ 3172 3173 #ifdef DYNAMIC_ICONV 3174 /* Check if the iconv.dll can be found. */ 3175 if (!iconv_enabled(TRUE)) 3176 return (void *)-1; 3177 #endif 3178 3179 fd = iconv_open((char *)enc_skip(to), (char *)enc_skip(from)); 3180 3181 if (fd != (iconv_t)-1 && iconv_ok == -1) 3182 { 3183 /* 3184 * Do a dummy iconv() call to check if it actually works. There is a 3185 * version of iconv() on Linux that is broken. We can't ignore it, 3186 * because it's wide-spread. The symptoms are that after outputting 3187 * the initial shift state the "to" pointer is NULL and conversion 3188 * stops for no apparent reason after about 8160 characters. 3189 */ 3190 p = (char *)tobuf; 3191 tolen = ICONV_TESTLEN; 3192 (void)iconv(fd, NULL, NULL, &p, &tolen); 3193 if (p == NULL) 3194 { 3195 iconv_ok = FALSE; 3196 iconv_close(fd); 3197 fd = (iconv_t)-1; 3198 } 3199 else 3200 iconv_ok = TRUE; 3201 } 3202 3203 return (void *)fd; 3204 } 3205 3206 /* 3207 * Convert the string "str[slen]" with iconv(). 3208 * If "unconvlenp" is not NULL handle the string ending in an incomplete 3209 * sequence and set "*unconvlenp" to the length of it. 3210 * Returns the converted string in allocated memory. NULL for an error. 3211 */ 3212 static char_u * 3213 iconv_string(vcp, str, slen, unconvlenp) 3214 vimconv_T *vcp; 3215 char_u *str; 3216 int slen; 3217 int *unconvlenp; 3218 { 3219 const char *from; 3220 size_t fromlen; 3221 char *to; 3222 size_t tolen; 3223 size_t len = 0; 3224 size_t done = 0; 3225 char_u *result = NULL; 3226 char_u *p; 3227 int l; 3228 3229 from = (char *)str; 3230 fromlen = slen; 3231 for (;;) 3232 { 3233 if (len == 0 || ICONV_ERRNO == ICONV_E2BIG) 3234 { 3235 /* Allocate enough room for most conversions. When re-allocating 3236 * increase the buffer size. */ 3237 len = len + fromlen * 2 + 40; 3238 p = alloc((unsigned)len); 3239 if (p != NULL && done > 0) 3240 mch_memmove(p, result, done); 3241 vim_free(result); 3242 result = p; 3243 if (result == NULL) /* out of memory */ 3244 break; 3245 } 3246 3247 to = (char *)result + done; 3248 tolen = len - done - 2; 3249 /* Avoid a warning for systems with a wrong iconv() prototype by 3250 * casting the second argument to void *. */ 3251 if (iconv(vcp->vc_fd, (void *)&from, &fromlen, &to, &tolen) 3252 != (size_t)-1) 3253 { 3254 /* Finished, append a NUL. */ 3255 *to = NUL; 3256 break; 3257 } 3258 3259 /* Check both ICONV_EINVAL and EINVAL, because the dynamically loaded 3260 * iconv library may use one of them. */ 3261 if (!vcp->vc_fail && unconvlenp != NULL 3262 && (ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL)) 3263 { 3264 /* Handle an incomplete sequence at the end. */ 3265 *to = NUL; 3266 *unconvlenp = fromlen; 3267 break; 3268 } 3269 3270 /* Check both ICONV_EILSEQ and EILSEQ, because the dynamically loaded 3271 * iconv library may use one of them. */ 3272 else if (!vcp->vc_fail 3273 && (ICONV_ERRNO == ICONV_EILSEQ || ICONV_ERRNO == EILSEQ 3274 || ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL)) 3275 { 3276 /* Can't convert: insert a '?' and skip a character. This assumes 3277 * conversion from 'encoding' to something else. In other 3278 * situations we don't know what to skip anyway. */ 3279 *to++ = '?'; 3280 if ((*mb_ptr2cells)((char_u *)from) > 1) 3281 *to++ = '?'; 3282 if (enc_utf8) 3283 l = utfc_ptr2len_len((char_u *)from, fromlen); 3284 else 3285 { 3286 l = (*mb_ptr2len)((char_u *)from); 3287 if (l > (int)fromlen) 3288 l = fromlen; 3289 } 3290 from += l; 3291 fromlen -= l; 3292 } 3293 else if (ICONV_ERRNO != ICONV_E2BIG) 3294 { 3295 /* conversion failed */ 3296 vim_free(result); 3297 result = NULL; 3298 break; 3299 } 3300 /* Not enough room or skipping illegal sequence. */ 3301 done = to - (char *)result; 3302 } 3303 return result; 3304 } 3305 3306 # if defined(DYNAMIC_ICONV) || defined(PROTO) 3307 /* 3308 * Dynamically load the "iconv.dll" on Win32. 3309 */ 3310 3311 #ifndef DYNAMIC_ICONV /* just generating prototypes */ 3312 # define HINSTANCE int 3313 #endif 3314 static HINSTANCE hIconvDLL = 0; 3315 static HINSTANCE hMsvcrtDLL = 0; 3316 3317 # ifndef DYNAMIC_ICONV_DLL 3318 # define DYNAMIC_ICONV_DLL "iconv.dll" 3319 # define DYNAMIC_ICONV_DLL_ALT "libiconv.dll" 3320 # endif 3321 # ifndef DYNAMIC_MSVCRT_DLL 3322 # define DYNAMIC_MSVCRT_DLL "msvcrt.dll" 3323 # endif 3324 3325 /* 3326 * Try opening the iconv.dll and return TRUE if iconv() can be used. 3327 */ 3328 int 3329 iconv_enabled(verbose) 3330 int verbose; 3331 { 3332 if (hIconvDLL != 0 && hMsvcrtDLL != 0) 3333 return TRUE; 3334 hIconvDLL = LoadLibrary(DYNAMIC_ICONV_DLL); 3335 if (hIconvDLL == 0) /* sometimes it's called libiconv.dll */ 3336 hIconvDLL = LoadLibrary(DYNAMIC_ICONV_DLL_ALT); 3337 if (hIconvDLL != 0) 3338 hMsvcrtDLL = LoadLibrary(DYNAMIC_MSVCRT_DLL); 3339 if (hIconvDLL == 0 || hMsvcrtDLL == 0) 3340 { 3341 /* Only give the message when 'verbose' is set, otherwise it might be 3342 * done whenever a conversion is attempted. */ 3343 if (verbose && p_verbose > 0) 3344 { 3345 verbose_enter(); 3346 EMSG2(_(e_loadlib), 3347 hIconvDLL == 0 ? DYNAMIC_ICONV_DLL : DYNAMIC_MSVCRT_DLL); 3348 verbose_leave(); 3349 } 3350 iconv_end(); 3351 return FALSE; 3352 } 3353 3354 iconv = (void *)GetProcAddress(hIconvDLL, "libiconv"); 3355 iconv_open = (void *)GetProcAddress(hIconvDLL, "libiconv_open"); 3356 iconv_close = (void *)GetProcAddress(hIconvDLL, "libiconv_close"); 3357 iconvctl = (void *)GetProcAddress(hIconvDLL, "libiconvctl"); 3358 iconv_errno = (void *)GetProcAddress(hMsvcrtDLL, "_errno"); 3359 if (iconv == NULL || iconv_open == NULL || iconv_close == NULL 3360 || iconvctl == NULL || iconv_errno == NULL) 3361 { 3362 iconv_end(); 3363 if (verbose && p_verbose > 0) 3364 { 3365 verbose_enter(); 3366 EMSG2(_(e_loadfunc), "for libiconv"); 3367 verbose_leave(); 3368 } 3369 return FALSE; 3370 } 3371 return TRUE; 3372 } 3373 3374 void 3375 iconv_end() 3376 { 3377 /* Don't use iconv() when inputting or outputting characters. */ 3378 if (input_conv.vc_type == CONV_ICONV) 3379 convert_setup(&input_conv, NULL, NULL); 3380 if (output_conv.vc_type == CONV_ICONV) 3381 convert_setup(&output_conv, NULL, NULL); 3382 3383 if (hIconvDLL != 0) 3384 FreeLibrary(hIconvDLL); 3385 if (hMsvcrtDLL != 0) 3386 FreeLibrary(hMsvcrtDLL); 3387 hIconvDLL = 0; 3388 hMsvcrtDLL = 0; 3389 } 3390 # endif /* DYNAMIC_ICONV */ 3391 # endif /* USE_ICONV */ 3392 3393 #endif /* FEAT_MBYTE */ 3394 3395 #if defined(FEAT_XIM) || defined(PROTO) 3396 3397 # ifdef FEAT_GUI_GTK 3398 static int xim_has_preediting INIT(= FALSE); /* IM current status */ 3399 3400 /* 3401 * Set preedit_start_col to the current cursor position. 3402 */ 3403 static void 3404 init_preedit_start_col(void) 3405 { 3406 if (State & CMDLINE) 3407 preedit_start_col = cmdline_getvcol_cursor(); 3408 else if (curwin != NULL) 3409 getvcol(curwin, &curwin->w_cursor, &preedit_start_col, NULL, NULL); 3410 /* Prevent that preediting marks the buffer as changed. */ 3411 xim_changed_while_preediting = curbuf->b_changed; 3412 } 3413 # endif 3414 3415 # if defined(HAVE_GTK2) && !defined(PROTO) 3416 3417 static int im_is_active = FALSE; /* IM is enabled for current mode */ 3418 static int im_preedit_cursor = 0; /* cursor offset in characters */ 3419 static int im_preedit_trailing = 0; /* number of characters after cursor */ 3420 3421 static unsigned long im_commit_handler_id = 0; 3422 static unsigned int im_activatekey_keyval = GDK_VoidSymbol; 3423 static unsigned int im_activatekey_state = 0; 3424 3425 void 3426 im_set_active(int active) 3427 { 3428 int was_active; 3429 3430 was_active = !!im_is_active; 3431 im_is_active = (active && !p_imdisable); 3432 3433 if (im_is_active != was_active) 3434 xim_reset(); 3435 } 3436 3437 void 3438 xim_set_focus(int focus) 3439 { 3440 if (xic != NULL) 3441 { 3442 if (focus) 3443 gtk_im_context_focus_in(xic); 3444 else 3445 gtk_im_context_focus_out(xic); 3446 } 3447 } 3448 3449 void 3450 im_set_position(int row, int col) 3451 { 3452 if (xic != NULL) 3453 { 3454 GdkRectangle area; 3455 3456 area.x = FILL_X(col); 3457 area.y = FILL_Y(row); 3458 area.width = gui.char_width * (mb_lefthalve(row, col) ? 2 : 1); 3459 area.height = gui.char_height; 3460 3461 gtk_im_context_set_cursor_location(xic, &area); 3462 } 3463 } 3464 3465 # if 0 || defined(PROTO) /* apparently only used in gui_x11.c */ 3466 void 3467 xim_set_preedit(void) 3468 { 3469 im_set_position(gui.row, gui.col); 3470 } 3471 # endif 3472 3473 static void 3474 im_add_to_input(char_u *str, int len) 3475 { 3476 /* Convert from 'termencoding' (always "utf-8") to 'encoding' */ 3477 if (input_conv.vc_type != CONV_NONE) 3478 { 3479 str = string_convert(&input_conv, str, &len); 3480 g_return_if_fail(str != NULL); 3481 } 3482 3483 add_to_input_buf_csi(str, len); 3484 3485 if (input_conv.vc_type != CONV_NONE) 3486 vim_free(str); 3487 3488 if (p_mh) /* blank out the pointer if necessary */ 3489 gui_mch_mousehide(TRUE); 3490 } 3491 3492 static void 3493 im_delete_preedit(void) 3494 { 3495 char_u bskey[] = {CSI, 'k', 'b'}; 3496 char_u delkey[] = {CSI, 'k', 'D'}; 3497 3498 if (State & NORMAL) 3499 { 3500 im_preedit_cursor = 0; 3501 return; 3502 } 3503 for (; im_preedit_cursor > 0; --im_preedit_cursor) 3504 add_to_input_buf(bskey, (int)sizeof(bskey)); 3505 3506 for (; im_preedit_trailing > 0; --im_preedit_trailing) 3507 add_to_input_buf(delkey, (int)sizeof(delkey)); 3508 } 3509 3510 static void 3511 im_correct_cursor(int num_move_back) 3512 { 3513 char_u backkey[] = {CSI, 'k', 'l'}; 3514 3515 if (State & NORMAL) 3516 return; 3517 # ifdef FEAT_RIGHTLEFT 3518 if ((State & CMDLINE) == 0 && curwin != NULL && curwin->w_p_rl) 3519 backkey[2] = 'r'; 3520 # endif 3521 for (; num_move_back > 0; --num_move_back) 3522 add_to_input_buf(backkey, (int)sizeof(backkey)); 3523 } 3524 3525 static int xim_expected_char = NUL; 3526 static int xim_ignored_char = FALSE; 3527 3528 /* 3529 * Update the mode and cursor while in an IM callback. 3530 */ 3531 static void 3532 im_show_info(void) 3533 { 3534 int old_vgetc_busy; 3535 3536 old_vgetc_busy = vgetc_busy; 3537 vgetc_busy = TRUE; 3538 showmode(); 3539 vgetc_busy = old_vgetc_busy; 3540 setcursor(); 3541 out_flush(); 3542 } 3543 3544 /* 3545 * Callback invoked when the user finished preediting. 3546 * Put the final string into the input buffer. 3547 */ 3548 /*ARGSUSED0*/ 3549 static void 3550 im_commit_cb(GtkIMContext *context, const gchar *str, gpointer data) 3551 { 3552 int slen = (int)STRLEN(str); 3553 int add_to_input = TRUE; 3554 int clen; 3555 int len = slen; 3556 int commit_with_preedit = TRUE; 3557 char_u *im_str, *p; 3558 3559 #ifdef XIM_DEBUG 3560 xim_log("im_commit_cb(): %s\n", str); 3561 #endif 3562 3563 /* The imhangul module doesn't reset the preedit string before 3564 * committing. Call im_delete_preedit() to work around that. */ 3565 im_delete_preedit(); 3566 3567 /* Indicate that preediting has finished. */ 3568 if (preedit_start_col == MAXCOL) 3569 { 3570 init_preedit_start_col(); 3571 commit_with_preedit = FALSE; 3572 } 3573 3574 /* The thing which setting "preedit_start_col" to MAXCOL means that 3575 * "preedit_start_col" will be set forcely when calling 3576 * preedit_changed_cb() next time. 3577 * "preedit_start_col" should not reset with MAXCOL on this part. Vim 3578 * is simulating the preediting by using add_to_input_str(). when 3579 * preedit begin immediately before committed, the typebuf is not 3580 * flushed to screen, then it can't get correct "preedit_start_col". 3581 * Thus, it should calculate the cells by adding cells of the committed 3582 * string. */ 3583 if (input_conv.vc_type != CONV_NONE) 3584 { 3585 im_str = string_convert(&input_conv, (char_u *)str, &len); 3586 g_return_if_fail(im_str != NULL); 3587 } 3588 else 3589 im_str = (char_u *)str; 3590 clen = 0; 3591 for (p = im_str; p < im_str + len; p += (*mb_ptr2len)(p)) 3592 clen += (*mb_ptr2cells)(p); 3593 if (input_conv.vc_type != CONV_NONE) 3594 vim_free(im_str); 3595 preedit_start_col += clen; 3596 3597 /* Is this a single character that matches a keypad key that's just 3598 * been pressed? If so, we don't want it to be entered as such - let 3599 * us carry on processing the raw keycode so that it may be used in 3600 * mappings as <kSomething>. */ 3601 if (xim_expected_char != NUL) 3602 { 3603 /* We're currently processing a keypad or other special key */ 3604 if (slen == 1 && str[0] == xim_expected_char) 3605 { 3606 /* It's a match - don't do it here */ 3607 xim_ignored_char = TRUE; 3608 add_to_input = FALSE; 3609 } 3610 else 3611 { 3612 /* Not a match */ 3613 xim_ignored_char = FALSE; 3614 } 3615 } 3616 3617 if (add_to_input) 3618 im_add_to_input((char_u *)str, slen); 3619 3620 /* Inserting chars while "im_is_active" is set does not cause a change of 3621 * buffer. When the chars are committed the buffer must be marked as 3622 * changed. */ 3623 if (!commit_with_preedit) 3624 preedit_start_col = MAXCOL; 3625 3626 /* This flag is used in changed() at next call. */ 3627 xim_changed_while_preediting = TRUE; 3628 3629 if (gtk_main_level() > 0) 3630 gtk_main_quit(); 3631 } 3632 3633 /* 3634 * Callback invoked after start to the preedit. 3635 */ 3636 /*ARGSUSED*/ 3637 static void 3638 im_preedit_start_cb(GtkIMContext *context, gpointer data) 3639 { 3640 #ifdef XIM_DEBUG 3641 xim_log("im_preedit_start_cb()\n"); 3642 #endif 3643 3644 im_is_active = TRUE; 3645 gui_update_cursor(TRUE, FALSE); 3646 } 3647 3648 /* 3649 * Callback invoked after end to the preedit. 3650 */ 3651 /*ARGSUSED*/ 3652 static void 3653 im_preedit_end_cb(GtkIMContext *context, gpointer data) 3654 { 3655 #ifdef XIM_DEBUG 3656 xim_log("im_preedit_end_cb()\n"); 3657 #endif 3658 im_delete_preedit(); 3659 3660 /* Indicate that preediting has finished */ 3661 preedit_start_col = MAXCOL; 3662 xim_has_preediting = FALSE; 3663 3664 im_is_active = FALSE; 3665 gui_update_cursor(TRUE, FALSE); 3666 im_show_info(); 3667 } 3668 3669 /* 3670 * Callback invoked after changes to the preedit string. If the preedit 3671 * string was empty before, remember the preedit start column so we know 3672 * where to apply feedback attributes. Delete the previous preedit string 3673 * if there was one, save the new preedit cursor offset, and put the new 3674 * string into the input buffer. 3675 * 3676 * TODO: The pragmatic "put into input buffer" approach used here has 3677 * several fundamental problems: 3678 * 3679 * - The characters in the preedit string are subject to remapping. 3680 * That's broken, only the finally committed string should be remapped. 3681 * 3682 * - There is a race condition involved: The retrieved value for the 3683 * current cursor position will be wrong if any unprocessed characters 3684 * are still queued in the input buffer. 3685 * 3686 * - Due to the lack of synchronization between the file buffer in memory 3687 * and any typed characters, it's practically impossible to implement the 3688 * "retrieve_surrounding" and "delete_surrounding" signals reliably. IM 3689 * modules for languages such as Thai are likely to rely on this feature 3690 * for proper operation. 3691 * 3692 * Conclusions: I think support for preediting needs to be moved to the 3693 * core parts of Vim. Ideally, until it has been committed, the preediting 3694 * string should only be displayed and not affect the buffer content at all. 3695 * The question how to deal with the synchronization issue still remains. 3696 * Circumventing the input buffer is probably not desirable. Anyway, I think 3697 * implementing "retrieve_surrounding" is the only hard problem. 3698 * 3699 * One way to solve all of this in a clean manner would be to queue all key 3700 * press/release events "as is" in the input buffer, and apply the IM filtering 3701 * at the receiving end of the queue. This, however, would have a rather large 3702 * impact on the code base. If there is an easy way to force processing of all 3703 * remaining input from within the "retrieve_surrounding" signal handler, this 3704 * might not be necessary. Gotta ask on vim-dev for opinions. 3705 */ 3706 /*ARGSUSED1*/ 3707 static void 3708 im_preedit_changed_cb(GtkIMContext *context, gpointer data) 3709 { 3710 char *preedit_string = NULL; 3711 int cursor_index = 0; 3712 int num_move_back = 0; 3713 char_u *str; 3714 char_u *p; 3715 int i; 3716 3717 gtk_im_context_get_preedit_string(context, 3718 &preedit_string, NULL, 3719 &cursor_index); 3720 3721 #ifdef XIM_DEBUG 3722 xim_log("im_preedit_changed_cb(): %s\n", preedit_string); 3723 #endif 3724 3725 g_return_if_fail(preedit_string != NULL); /* just in case */ 3726 3727 /* If preedit_start_col is MAXCOL set it to the current cursor position. */ 3728 if (preedit_start_col == MAXCOL && preedit_string[0] != '\0') 3729 { 3730 xim_has_preediting = TRUE; 3731 3732 /* Urgh, this breaks if the input buffer isn't empty now */ 3733 init_preedit_start_col(); 3734 } 3735 else if (cursor_index == 0 && preedit_string[0] == '\0') 3736 { 3737 if (preedit_start_col == MAXCOL) 3738 xim_has_preediting = FALSE; 3739 3740 /* If at the start position (after typing backspace) 3741 * preedit_start_col must be reset. */ 3742 preedit_start_col = MAXCOL; 3743 } 3744 3745 im_delete_preedit(); 3746 3747 /* 3748 * Compute the end of the preediting area: "preedit_end_col". 3749 * According to the documentation of gtk_im_context_get_preedit_string(), 3750 * the cursor_pos output argument returns the offset in bytes. This is 3751 * unfortunately not true -- real life shows the offset is in characters, 3752 * and the GTK+ source code agrees with me. Will file a bug later. 3753 */ 3754 if (preedit_start_col != MAXCOL) 3755 preedit_end_col = preedit_start_col; 3756 str = (char_u *)preedit_string; 3757 for (p = str, i = 0; *p != NUL; p += utf_byte2len(*p), ++i) 3758 { 3759 int is_composing; 3760 3761 is_composing = ((*p & 0x80) != 0 && utf_iscomposing(utf_ptr2char(p))); 3762 /* 3763 * These offsets are used as counters when generating <BS> and <Del> 3764 * to delete the preedit string. So don't count composing characters 3765 * unless 'delcombine' is enabled. 3766 */ 3767 if (!is_composing || p_deco) 3768 { 3769 if (i < cursor_index) 3770 ++im_preedit_cursor; 3771 else 3772 ++im_preedit_trailing; 3773 } 3774 if (!is_composing && i >= cursor_index) 3775 { 3776 /* This is essentially the same as im_preedit_trailing, except 3777 * composing characters are not counted even if p_deco is set. */ 3778 ++num_move_back; 3779 } 3780 if (preedit_start_col != MAXCOL) 3781 preedit_end_col += utf_ptr2cells(p); 3782 } 3783 3784 if (p > str) 3785 { 3786 im_add_to_input(str, (int)(p - str)); 3787 im_correct_cursor(num_move_back); 3788 } 3789 3790 g_free(preedit_string); 3791 3792 if (gtk_main_level() > 0) 3793 gtk_main_quit(); 3794 } 3795 3796 /* 3797 * Translate the Pango attributes at iter to Vim highlighting attributes. 3798 * Ignore attributes not supported by Vim highlighting. This shouldn't have 3799 * too much impact -- right now we handle even more attributes than necessary 3800 * for the IM modules I tested with. 3801 */ 3802 static int 3803 translate_pango_attributes(PangoAttrIterator *iter) 3804 { 3805 PangoAttribute *attr; 3806 int char_attr = HL_NORMAL; 3807 3808 attr = pango_attr_iterator_get(iter, PANGO_ATTR_UNDERLINE); 3809 if (attr != NULL && ((PangoAttrInt *)attr)->value 3810 != (int)PANGO_UNDERLINE_NONE) 3811 char_attr |= HL_UNDERLINE; 3812 3813 attr = pango_attr_iterator_get(iter, PANGO_ATTR_WEIGHT); 3814 if (attr != NULL && ((PangoAttrInt *)attr)->value >= (int)PANGO_WEIGHT_BOLD) 3815 char_attr |= HL_BOLD; 3816 3817 attr = pango_attr_iterator_get(iter, PANGO_ATTR_STYLE); 3818 if (attr != NULL && ((PangoAttrInt *)attr)->value 3819 != (int)PANGO_STYLE_NORMAL) 3820 char_attr |= HL_ITALIC; 3821 3822 attr = pango_attr_iterator_get(iter, PANGO_ATTR_BACKGROUND); 3823 if (attr != NULL) 3824 { 3825 const PangoColor *color = &((PangoAttrColor *)attr)->color; 3826 3827 /* Assume inverse if black background is requested */ 3828 if ((color->red | color->green | color->blue) == 0) 3829 char_attr |= HL_INVERSE; 3830 } 3831 3832 return char_attr; 3833 } 3834 3835 /* 3836 * Retrieve the highlighting attributes at column col in the preedit string. 3837 * Return -1 if not in preediting mode or if col is out of range. 3838 */ 3839 int 3840 im_get_feedback_attr(int col) 3841 { 3842 char *preedit_string = NULL; 3843 PangoAttrList *attr_list = NULL; 3844 int char_attr = -1; 3845 3846 if (xic == NULL) 3847 return char_attr; 3848 3849 gtk_im_context_get_preedit_string(xic, &preedit_string, &attr_list, NULL); 3850 3851 if (preedit_string != NULL && attr_list != NULL) 3852 { 3853 int index; 3854 3855 /* Get the byte index as used by PangoAttrIterator */ 3856 for (index = 0; col > 0 && preedit_string[index] != '\0'; --col) 3857 index += utfc_ptr2len((char_u *)preedit_string + index); 3858 3859 if (preedit_string[index] != '\0') 3860 { 3861 PangoAttrIterator *iter; 3862 int start, end; 3863 3864 char_attr = HL_NORMAL; 3865 iter = pango_attr_list_get_iterator(attr_list); 3866 3867 /* Extract all relevant attributes from the list. */ 3868 do 3869 { 3870 pango_attr_iterator_range(iter, &start, &end); 3871 3872 if (index >= start && index < end) 3873 char_attr |= translate_pango_attributes(iter); 3874 } 3875 while (pango_attr_iterator_next(iter)); 3876 3877 pango_attr_iterator_destroy(iter); 3878 } 3879 } 3880 3881 if (attr_list != NULL) 3882 pango_attr_list_unref(attr_list); 3883 g_free(preedit_string); 3884 3885 return char_attr; 3886 } 3887 3888 void 3889 xim_init(void) 3890 { 3891 #ifdef XIM_DEBUG 3892 xim_log("xim_init()\n"); 3893 #endif 3894 3895 g_return_if_fail(gui.drawarea != NULL); 3896 g_return_if_fail(gui.drawarea->window != NULL); 3897 3898 xic = gtk_im_multicontext_new(); 3899 g_object_ref(xic); 3900 3901 im_commit_handler_id = g_signal_connect(G_OBJECT(xic), "commit", 3902 G_CALLBACK(&im_commit_cb), NULL); 3903 g_signal_connect(G_OBJECT(xic), "preedit_changed", 3904 G_CALLBACK(&im_preedit_changed_cb), NULL); 3905 g_signal_connect(G_OBJECT(xic), "preedit_start", 3906 G_CALLBACK(&im_preedit_start_cb), NULL); 3907 g_signal_connect(G_OBJECT(xic), "preedit_end", 3908 G_CALLBACK(&im_preedit_end_cb), NULL); 3909 3910 gtk_im_context_set_client_window(xic, gui.drawarea->window); 3911 } 3912 3913 void 3914 im_shutdown(void) 3915 { 3916 #ifdef XIM_DEBUG 3917 xim_log("im_shutdown()\n"); 3918 #endif 3919 3920 if (xic != NULL) 3921 { 3922 gtk_im_context_focus_out(xic); 3923 g_object_unref(xic); 3924 xic = NULL; 3925 } 3926 im_is_active = FALSE; 3927 im_commit_handler_id = 0; 3928 preedit_start_col = MAXCOL; 3929 xim_has_preediting = FALSE; 3930 } 3931 3932 /* 3933 * Convert the string argument to keyval and state for GdkEventKey. 3934 * If str is valid return TRUE, otherwise FALSE. 3935 * 3936 * See 'imactivatekey' for documentation of the format. 3937 */ 3938 static int 3939 im_string_to_keyval(const char *str, unsigned int *keyval, unsigned int *state) 3940 { 3941 const char *mods_end; 3942 unsigned tmp_keyval; 3943 unsigned tmp_state = 0; 3944 3945 mods_end = strrchr(str, '-'); 3946 mods_end = (mods_end != NULL) ? mods_end + 1 : str; 3947 3948 /* Parse modifier keys */ 3949 while (str < mods_end) 3950 switch (*str++) 3951 { 3952 case '-': break; 3953 case 'S': case 's': tmp_state |= (unsigned)GDK_SHIFT_MASK; break; 3954 case 'L': case 'l': tmp_state |= (unsigned)GDK_LOCK_MASK; break; 3955 case 'C': case 'c': tmp_state |= (unsigned)GDK_CONTROL_MASK;break; 3956 case '1': tmp_state |= (unsigned)GDK_MOD1_MASK; break; 3957 case '2': tmp_state |= (unsigned)GDK_MOD2_MASK; break; 3958 case '3': tmp_state |= (unsigned)GDK_MOD3_MASK; break; 3959 case '4': tmp_state |= (unsigned)GDK_MOD4_MASK; break; 3960 case '5': tmp_state |= (unsigned)GDK_MOD5_MASK; break; 3961 default: 3962 return FALSE; 3963 } 3964 3965 tmp_keyval = gdk_keyval_from_name(str); 3966 3967 if (tmp_keyval == 0 || tmp_keyval == GDK_VoidSymbol) 3968 return FALSE; 3969 3970 if (keyval != NULL) 3971 *keyval = tmp_keyval; 3972 if (state != NULL) 3973 *state = tmp_state; 3974 3975 return TRUE; 3976 } 3977 3978 /* 3979 * Return TRUE if p_imak is valid, otherwise FALSE. As a special case, an 3980 * empty string is also regarded as valid. 3981 * 3982 * Note: The numerical key value of p_imak is cached if it was valid; thus 3983 * boldly assuming im_xim_isvalid_imactivate() will always be called whenever 3984 * 'imak' changes. This is currently the case but not obvious -- should 3985 * probably rename the function for clarity. 3986 */ 3987 int 3988 im_xim_isvalid_imactivate(void) 3989 { 3990 if (p_imak[0] == NUL) 3991 { 3992 im_activatekey_keyval = GDK_VoidSymbol; 3993 im_activatekey_state = 0; 3994 return TRUE; 3995 } 3996 3997 return im_string_to_keyval((const char *)p_imak, 3998 &im_activatekey_keyval, 3999 &im_activatekey_state); 4000 } 4001 4002 static void 4003 im_synthesize_keypress(unsigned int keyval, unsigned int state) 4004 { 4005 GdkEventKey *event; 4006 4007 # ifdef HAVE_GTK_MULTIHEAD 4008 event = (GdkEventKey *)gdk_event_new(GDK_KEY_PRESS); 4009 g_object_ref(gui.drawarea->window); /* unreffed by gdk_event_free() */ 4010 # else 4011 event = (GdkEventKey *)g_malloc0((gulong)sizeof(GdkEvent)); 4012 event->type = GDK_KEY_PRESS; 4013 # endif 4014 event->window = gui.drawarea->window; 4015 event->send_event = TRUE; 4016 event->time = GDK_CURRENT_TIME; 4017 event->state = state; 4018 event->keyval = keyval; 4019 event->hardware_keycode = /* needed for XIM */ 4020 XKeysymToKeycode(GDK_WINDOW_XDISPLAY(event->window), (KeySym)keyval); 4021 event->length = 0; 4022 event->string = NULL; 4023 4024 gtk_im_context_filter_keypress(xic, event); 4025 4026 /* For consistency, also send the corresponding release event. */ 4027 event->type = GDK_KEY_RELEASE; 4028 event->send_event = FALSE; 4029 gtk_im_context_filter_keypress(xic, event); 4030 4031 # ifdef HAVE_GTK_MULTIHEAD 4032 gdk_event_free((GdkEvent *)event); 4033 # else 4034 g_free(event); 4035 # endif 4036 } 4037 4038 void 4039 xim_reset(void) 4040 { 4041 if (xic != NULL) 4042 { 4043 /* 4044 * The third-party imhangul module (and maybe others too) ignores 4045 * gtk_im_context_reset() or at least doesn't reset the active state. 4046 * Thus sending imactivatekey would turn it off if it was on before, 4047 * which is clearly not what we want. Fortunately we can work around 4048 * that for imhangul by sending GDK_Escape, but I don't know if it 4049 * works with all IM modules that support an activation key :/ 4050 * 4051 * An alternative approach would be to destroy the IM context and 4052 * recreate it. But that means loading/unloading the IM module on 4053 * every mode switch, which causes a quite noticable delay even on 4054 * my rather fast box... 4055 * * 4056 * Moreover, there are some XIM which cannot respond to 4057 * im_synthesize_keypress(). we hope that they reset by 4058 * xim_shutdown(). 4059 */ 4060 if (im_activatekey_keyval != GDK_VoidSymbol && im_is_active) 4061 im_synthesize_keypress(GDK_Escape, 0U); 4062 4063 gtk_im_context_reset(xic); 4064 4065 /* 4066 * HACK for Ami: This sequence of function calls makes Ami handle 4067 * the IM reset gratiously, without breaking loads of other stuff. 4068 * It seems to force English mode as well, which is exactly what we 4069 * want because it makes the Ami status display work reliably. 4070 */ 4071 gtk_im_context_set_use_preedit(xic, FALSE); 4072 4073 if (p_imdisable) 4074 im_shutdown(); 4075 else 4076 { 4077 gtk_im_context_set_use_preedit(xic, TRUE); 4078 xim_set_focus(gui.in_focus); 4079 4080 if (im_activatekey_keyval != GDK_VoidSymbol) 4081 { 4082 if (im_is_active) 4083 { 4084 g_signal_handler_block(xic, im_commit_handler_id); 4085 im_synthesize_keypress(im_activatekey_keyval, 4086 im_activatekey_state); 4087 g_signal_handler_unblock(xic, im_commit_handler_id); 4088 } 4089 } 4090 else 4091 { 4092 im_shutdown(); 4093 xim_init(); 4094 xim_set_focus(gui.in_focus); 4095 } 4096 } 4097 } 4098 4099 preedit_start_col = MAXCOL; 4100 xim_has_preediting = FALSE; 4101 } 4102 4103 int 4104 xim_queue_key_press_event(GdkEventKey *event, int down) 4105 { 4106 if (down) 4107 { 4108 /* 4109 * Workaround GTK2 XIM 'feature' that always converts keypad keys to 4110 * chars., even when not part of an IM sequence (ref. feature of 4111 * gdk/gdkkeyuni.c). 4112 * Flag any keypad keys that might represent a single char. 4113 * If this (on its own - i.e., not part of an IM sequence) is 4114 * committed while we're processing one of these keys, we can ignore 4115 * that commit and go ahead & process it ourselves. That way we can 4116 * still distinguish keypad keys for use in mappings. 4117 */ 4118 switch (event->keyval) 4119 { 4120 case GDK_KP_Add: xim_expected_char = '+'; break; 4121 case GDK_KP_Subtract: xim_expected_char = '-'; break; 4122 case GDK_KP_Divide: xim_expected_char = '/'; break; 4123 case GDK_KP_Multiply: xim_expected_char = '*'; break; 4124 case GDK_KP_Decimal: xim_expected_char = '.'; break; 4125 case GDK_KP_Equal: xim_expected_char = '='; break; 4126 case GDK_KP_0: xim_expected_char = '0'; break; 4127 case GDK_KP_1: xim_expected_char = '1'; break; 4128 case GDK_KP_2: xim_expected_char = '2'; break; 4129 case GDK_KP_3: xim_expected_char = '3'; break; 4130 case GDK_KP_4: xim_expected_char = '4'; break; 4131 case GDK_KP_5: xim_expected_char = '5'; break; 4132 case GDK_KP_6: xim_expected_char = '6'; break; 4133 case GDK_KP_7: xim_expected_char = '7'; break; 4134 case GDK_KP_8: xim_expected_char = '8'; break; 4135 case GDK_KP_9: xim_expected_char = '9'; break; 4136 default: xim_expected_char = NUL; 4137 } 4138 xim_ignored_char = FALSE; 4139 } 4140 4141 /* 4142 * When typing fFtT, XIM may be activated. Thus it must pass 4143 * gtk_im_context_filter_keypress() in Normal mode. 4144 * And while doing :sh too. 4145 */ 4146 if (xic != NULL && !p_imdisable 4147 && (State & (INSERT | CMDLINE | NORMAL | EXTERNCMD)) != 0) 4148 { 4149 /* 4150 * Filter 'imactivatekey' and map it to CTRL-^. This way, Vim is 4151 * always aware of the current status of IM, and can even emulate 4152 * the activation key for modules that don't support one. 4153 */ 4154 if (event->keyval == im_activatekey_keyval 4155 && (event->state & im_activatekey_state) == im_activatekey_state) 4156 { 4157 unsigned int state_mask; 4158 4159 /* Require the state of the 3 most used modifiers to match exactly. 4160 * Otherwise e.g. <S-C-space> would be unusable for other purposes 4161 * if the IM activate key is <S-space>. */ 4162 state_mask = im_activatekey_state; 4163 state_mask |= ((int)GDK_SHIFT_MASK | (int)GDK_CONTROL_MASK 4164 | (int)GDK_MOD1_MASK); 4165 4166 if ((event->state & state_mask) != im_activatekey_state) 4167 return FALSE; 4168 4169 /* Don't send it a second time on GDK_KEY_RELEASE. */ 4170 if (event->type != GDK_KEY_PRESS) 4171 return TRUE; 4172 4173 if (map_to_exists_mode((char_u *)"", LANGMAP, FALSE)) 4174 { 4175 im_set_active(FALSE); 4176 4177 /* ":lmap" mappings exists, toggle use of mappings. */ 4178 State ^= LANGMAP; 4179 if (State & LANGMAP) 4180 { 4181 curbuf->b_p_iminsert = B_IMODE_NONE; 4182 State &= ~LANGMAP; 4183 } 4184 else 4185 { 4186 curbuf->b_p_iminsert = B_IMODE_LMAP; 4187 State |= LANGMAP; 4188 } 4189 return TRUE; 4190 } 4191 4192 return gtk_im_context_filter_keypress(xic, event); 4193 } 4194 4195 /* Don't filter events through the IM context if IM isn't active 4196 * right now. Unlike with GTK+ 1.2 we cannot rely on the IM module 4197 * not doing anything before the activation key was sent. */ 4198 if (im_activatekey_keyval == GDK_VoidSymbol || im_is_active) 4199 { 4200 int imresult = gtk_im_context_filter_keypress(xic, event); 4201 4202 /* Some XIM send following sequence: 4203 * 1. preedited string. 4204 * 2. committed string. 4205 * 3. line changed key. 4206 * 4. preedited string. 4207 * 5. remove preedited string. 4208 * if 3, Vim can't move back the above line for 5. 4209 * thus, this part should not parse the key. */ 4210 if (!imresult && preedit_start_col != MAXCOL 4211 && event->keyval == GDK_Return) 4212 { 4213 im_synthesize_keypress(GDK_Return, 0U); 4214 return FALSE; 4215 } 4216 4217 /* If XIM tried to commit a keypad key as a single char., 4218 * ignore it so we can use the keypad key 'raw', for mappings. */ 4219 if (xim_expected_char != NUL && xim_ignored_char) 4220 /* We had a keypad key, and XIM tried to thieve it */ 4221 return FALSE; 4222 4223 /* Normal processing */ 4224 return imresult; 4225 } 4226 } 4227 4228 return FALSE; 4229 } 4230 4231 int 4232 im_get_status(void) 4233 { 4234 return im_is_active; 4235 } 4236 4237 # else /* !HAVE_GTK2 */ 4238 4239 static int xim_is_active = FALSE; /* XIM should be active in the current 4240 mode */ 4241 static int xim_has_focus = FALSE; /* XIM is really being used for Vim */ 4242 #ifdef FEAT_GUI_X11 4243 static XIMStyle input_style; 4244 static int status_area_enabled = TRUE; 4245 #endif 4246 4247 #ifdef FEAT_GUI_GTK 4248 # ifdef WIN3264 4249 # include <gdk/gdkwin32.h> 4250 # else 4251 # include <gdk/gdkx.h> 4252 # endif 4253 #else 4254 # ifdef PROTO 4255 /* Define a few things to be able to generate prototypes while not configured 4256 * for GTK. */ 4257 # define GSList int 4258 # define gboolean int 4259 typedef int GdkEvent; 4260 typedef int GdkEventKey; 4261 # define GdkIC int 4262 # endif 4263 #endif 4264 4265 #if defined(FEAT_GUI_GTK) || defined(PROTO) 4266 static int preedit_buf_len = 0; 4267 static int xim_can_preediting INIT(= FALSE); /* XIM in showmode() */ 4268 static int xim_input_style; 4269 #ifndef FEAT_GUI_GTK 4270 # define gboolean int 4271 #endif 4272 static gboolean use_status_area = 0; 4273 4274 static int im_xim_str2keycode __ARGS((unsigned int *code, unsigned int *state)); 4275 static void im_xim_send_event_imactivate __ARGS((void)); 4276 4277 /* 4278 * Convert string to keycode and state for XKeyEvent. 4279 * When string is valid return OK, when invalid return FAIL. 4280 * 4281 * See 'imactivatekey' documentation for the format. 4282 */ 4283 static int 4284 im_xim_str2keycode(code, state) 4285 unsigned int *code; 4286 unsigned int *state; 4287 { 4288 int retval = OK; 4289 int len; 4290 unsigned keycode = 0, keystate = 0; 4291 Window window; 4292 Display *display; 4293 char_u *flag_end; 4294 char_u *str; 4295 4296 if (*p_imak != NUL) 4297 { 4298 len = STRLEN(p_imak); 4299 for (flag_end = p_imak + len - 1; 4300 flag_end > p_imak && *flag_end != '-'; --flag_end) 4301 ; 4302 4303 /* Parse modifier keys */ 4304 for (str = p_imak; str < flag_end; ++str) 4305 { 4306 switch (*str) 4307 { 4308 case 's': case 'S': 4309 keystate |= ShiftMask; 4310 break; 4311 case 'l': case 'L': 4312 keystate |= LockMask; 4313 break; 4314 case 'c': case 'C': 4315 keystate |= ControlMask; 4316 break; 4317 case '1': 4318 keystate |= Mod1Mask; 4319 break; 4320 case '2': 4321 keystate |= Mod2Mask; 4322 break; 4323 case '3': 4324 keystate |= Mod3Mask; 4325 break; 4326 case '4': 4327 keystate |= Mod4Mask; 4328 break; 4329 case '5': 4330 keystate |= Mod5Mask; 4331 break; 4332 case '-': 4333 break; 4334 default: 4335 retval = FAIL; 4336 } 4337 } 4338 if (*str == '-') 4339 ++str; 4340 4341 /* Get keycode from string. */ 4342 gui_get_x11_windis(&window, &display); 4343 if (display) 4344 keycode = XKeysymToKeycode(display, XStringToKeysym((char *)str)); 4345 if (keycode == 0) 4346 retval = FAIL; 4347 4348 if (code != NULL) 4349 *code = keycode; 4350 if (state != NULL) 4351 *state = keystate; 4352 } 4353 return retval; 4354 } 4355 4356 static void 4357 im_xim_send_event_imactivate() 4358 { 4359 /* Force turn on preedit state by symulate keypress event. 4360 * Keycode and state is specified by 'imactivatekey'. 4361 */ 4362 XKeyEvent ev; 4363 4364 gui_get_x11_windis(&ev.window, &ev.display); 4365 ev.root = RootWindow(ev.display, DefaultScreen(ev.display)); 4366 ev.subwindow = None; 4367 ev.time = CurrentTime; 4368 ev.x = 1; 4369 ev.y = 1; 4370 ev.x_root = 1; 4371 ev.y_root = 1; 4372 ev.same_screen = 1; 4373 ev.type = KeyPress; 4374 if (im_xim_str2keycode(&ev.keycode, &ev.state) == OK) 4375 XSendEvent(ev.display, ev.window, 1, KeyPressMask, (XEvent*)&ev); 4376 } 4377 4378 /* 4379 * Return TRUE if 'imactivatekey' has a valid value. 4380 */ 4381 int 4382 im_xim_isvalid_imactivate() 4383 { 4384 return im_xim_str2keycode(NULL, NULL) == OK; 4385 } 4386 #endif /* FEAT_GUI_GTK */ 4387 4388 /* 4389 * Switch using XIM on/off. This is used by the code that changes "State". 4390 */ 4391 void 4392 im_set_active(active) 4393 int active; 4394 { 4395 if (xic == NULL) 4396 return; 4397 4398 /* If 'imdisable' is set, XIM is never active. */ 4399 if (p_imdisable) 4400 active = FALSE; 4401 #if !defined (FEAT_GUI_GTK) 4402 else if (input_style & XIMPreeditPosition) 4403 /* There is a problem in switching XIM off when preediting is used, 4404 * and it is not clear how this can be solved. For now, keep XIM on 4405 * all the time, like it was done in Vim 5.8. */ 4406 active = TRUE; 4407 #endif 4408 4409 /* Remember the active state, it is needed when Vim gets keyboard focus. */ 4410 xim_is_active = active; 4411 4412 #ifdef FEAT_GUI_GTK 4413 /* When 'imactivatekey' has valid key-string, try to control XIM preedit 4414 * state. When 'imactivatekey' has no or invalid string, try old XIM 4415 * focus control. 4416 */ 4417 if (*p_imak != NUL) 4418 { 4419 /* BASIC STRATEGY: 4420 * Destroy old Input Context (XIC), and create new one. New XIC 4421 * would have a state of preedit that is off. When argument:active 4422 * is false, that's all. Else argument:active is true, send a key 4423 * event specified by 'imactivatekey' to activate XIM preedit state. 4424 */ 4425 4426 xim_is_active = TRUE; /* Disable old XIM focus control */ 4427 /* If we can monitor preedit state with preedit callback functions, 4428 * try least creation of new XIC. 4429 */ 4430 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS) 4431 { 4432 if (xim_can_preediting && !active) 4433 { 4434 /* Force turn off preedit state. With some IM 4435 * implementations, we cannot turn off preedit state by 4436 * symulate keypress event. It is why using such a method 4437 * that destroy old IC (input context), and create new one. 4438 * When create new IC, its preedit state is usually off. 4439 */ 4440 xim_reset(); 4441 xim_set_focus(FALSE); 4442 gdk_ic_destroy(xic); 4443 xim_init(); 4444 xim_can_preediting = FALSE; 4445 } 4446 else if (!xim_can_preediting && active) 4447 im_xim_send_event_imactivate(); 4448 } 4449 else 4450 { 4451 /* First, force destroy old IC, and create new one. It 4452 * symulates "turning off preedit state". 4453 */ 4454 xim_set_focus(FALSE); 4455 gdk_ic_destroy(xic); 4456 xim_init(); 4457 xim_can_preediting = FALSE; 4458 4459 /* 2nd, when requested to activate IM, symulate this by sending 4460 * the event. 4461 */ 4462 if (active) 4463 { 4464 im_xim_send_event_imactivate(); 4465 xim_can_preediting = TRUE; 4466 } 4467 } 4468 } 4469 else 4470 { 4471 # ifndef XIMPreeditUnKnown 4472 /* X11R5 doesn't have these, it looks safe enough to define here. */ 4473 typedef unsigned long XIMPreeditState; 4474 # define XIMPreeditUnKnown 0L 4475 # define XIMPreeditEnable 1L 4476 # define XIMPreeditDisable (1L<<1) 4477 # define XNPreeditState "preeditState" 4478 # endif 4479 XIMPreeditState preedit_state = XIMPreeditUnKnown; 4480 XVaNestedList preedit_attr; 4481 XIC pxic; 4482 4483 preedit_attr = XVaCreateNestedList(0, 4484 XNPreeditState, &preedit_state, 4485 NULL); 4486 pxic = ((GdkICPrivate *)xic)->xic; 4487 4488 if (!XGetICValues(pxic, XNPreeditAttributes, preedit_attr, NULL)) 4489 { 4490 XFree(preedit_attr); 4491 preedit_attr = XVaCreateNestedList(0, 4492 XNPreeditState, 4493 active ? XIMPreeditEnable : XIMPreeditDisable, 4494 NULL); 4495 XSetICValues(pxic, XNPreeditAttributes, preedit_attr, NULL); 4496 xim_can_preediting = active; 4497 xim_is_active = active; 4498 } 4499 XFree(preedit_attr); 4500 } 4501 if (xim_input_style & XIMPreeditCallbacks) 4502 { 4503 preedit_buf_len = 0; 4504 init_preedit_start_col(); 4505 } 4506 #else 4507 # if 0 4508 /* When had tested kinput2 + canna + Athena GUI version with 4509 * 'imactivatekey' is "s-space", im_xim_send_event_imactivate() did not 4510 * work correctly. It just inserted one space. I don't know why we 4511 * couldn't switch state of XIM preediting. This is reason why these 4512 * codes are commented out. 4513 */ 4514 /* First, force destroy old IC, and create new one. It symulates 4515 * "turning off preedit state". 4516 */ 4517 xim_set_focus(FALSE); 4518 XDestroyIC(xic); 4519 xic = NULL; 4520 xim_init(); 4521 4522 /* 2nd, when requested to activate IM, symulate this by sending the 4523 * event. 4524 */ 4525 if (active) 4526 im_xim_send_event_imactivate(); 4527 # endif 4528 #endif 4529 xim_set_preedit(); 4530 } 4531 4532 /* 4533 * Adjust using XIM for gaining or losing keyboard focus. Also called when 4534 * "xim_is_active" changes. 4535 */ 4536 void 4537 xim_set_focus(focus) 4538 int focus; 4539 { 4540 if (xic == NULL) 4541 return; 4542 4543 /* 4544 * XIM only gets focus when the Vim window has keyboard focus and XIM has 4545 * been set active for the current mode. 4546 */ 4547 if (focus && xim_is_active) 4548 { 4549 if (!xim_has_focus) 4550 { 4551 xim_has_focus = TRUE; 4552 #ifdef FEAT_GUI_GTK 4553 gdk_im_begin(xic, gui.drawarea->window); 4554 #else 4555 XSetICFocus(xic); 4556 #endif 4557 } 4558 } 4559 else 4560 { 4561 if (xim_has_focus) 4562 { 4563 xim_has_focus = FALSE; 4564 #ifdef FEAT_GUI_GTK 4565 gdk_im_end(); 4566 #else 4567 XUnsetICFocus(xic); 4568 #endif 4569 } 4570 } 4571 } 4572 4573 /*ARGSUSED*/ 4574 void 4575 im_set_position(row, col) 4576 int row; 4577 int col; 4578 { 4579 xim_set_preedit(); 4580 } 4581 4582 /* 4583 * Set the XIM to the current cursor position. 4584 */ 4585 void 4586 xim_set_preedit() 4587 { 4588 if (xic == NULL) 4589 return; 4590 4591 xim_set_focus(TRUE); 4592 4593 #ifdef FEAT_GUI_GTK 4594 if (gdk_im_ready()) 4595 { 4596 int attrmask; 4597 GdkICAttr *attr; 4598 4599 if (!xic_attr) 4600 return; 4601 4602 attr = xic_attr; 4603 attrmask = 0; 4604 4605 # ifdef FEAT_XFONTSET 4606 if ((xim_input_style & (int)GDK_IM_PREEDIT_POSITION) 4607 && gui.fontset != NOFONTSET 4608 && gui.fontset->type == GDK_FONT_FONTSET) 4609 { 4610 if (!xim_has_focus) 4611 { 4612 if (attr->spot_location.y >= 0) 4613 { 4614 attr->spot_location.x = 0; 4615 attr->spot_location.y = -100; 4616 attrmask |= (int)GDK_IC_SPOT_LOCATION; 4617 } 4618 } 4619 else 4620 { 4621 gint width, height; 4622 4623 if (attr->spot_location.x != TEXT_X(gui.col) 4624 || attr->spot_location.y != TEXT_Y(gui.row)) 4625 { 4626 attr->spot_location.x = TEXT_X(gui.col); 4627 attr->spot_location.y = TEXT_Y(gui.row); 4628 attrmask |= (int)GDK_IC_SPOT_LOCATION; 4629 } 4630 4631 gdk_window_get_size(gui.drawarea->window, &width, &height); 4632 width -= 2 * gui.border_offset; 4633 height -= 2 * gui.border_offset; 4634 if (xim_input_style & (int)GDK_IM_STATUS_AREA) 4635 height -= gui.char_height; 4636 if (attr->preedit_area.width != width 4637 || attr->preedit_area.height != height) 4638 { 4639 attr->preedit_area.x = gui.border_offset; 4640 attr->preedit_area.y = gui.border_offset; 4641 attr->preedit_area.width = width; 4642 attr->preedit_area.height = height; 4643 attrmask |= (int)GDK_IC_PREEDIT_AREA; 4644 } 4645 4646 if (attr->preedit_fontset != gui.current_font) 4647 { 4648 attr->preedit_fontset = gui.current_font; 4649 attrmask |= (int)GDK_IC_PREEDIT_FONTSET; 4650 } 4651 } 4652 } 4653 # endif /* FEAT_XFONTSET */ 4654 4655 if (xim_fg_color == INVALCOLOR) 4656 { 4657 xim_fg_color = gui.def_norm_pixel; 4658 xim_bg_color = gui.def_back_pixel; 4659 } 4660 if (attr->preedit_foreground.pixel != xim_fg_color) 4661 { 4662 attr->preedit_foreground.pixel = xim_fg_color; 4663 attrmask |= (int)GDK_IC_PREEDIT_FOREGROUND; 4664 } 4665 if (attr->preedit_background.pixel != xim_bg_color) 4666 { 4667 attr->preedit_background.pixel = xim_bg_color; 4668 attrmask |= (int)GDK_IC_PREEDIT_BACKGROUND; 4669 } 4670 4671 if (attrmask != 0) 4672 gdk_ic_set_attr(xic, attr, (GdkICAttributesType)attrmask); 4673 } 4674 #else /* FEAT_GUI_GTK */ 4675 { 4676 XVaNestedList attr_list; 4677 XRectangle spot_area; 4678 XPoint over_spot; 4679 int line_space; 4680 4681 if (!xim_has_focus) 4682 { 4683 /* hide XIM cursor */ 4684 over_spot.x = 0; 4685 over_spot.y = -100; /* arbitrary invisible position */ 4686 attr_list = (XVaNestedList) XVaCreateNestedList(0, 4687 XNSpotLocation, 4688 &over_spot, 4689 NULL); 4690 XSetICValues(xic, XNPreeditAttributes, attr_list, NULL); 4691 XFree(attr_list); 4692 return; 4693 } 4694 4695 if (input_style & XIMPreeditPosition) 4696 { 4697 if (xim_fg_color == INVALCOLOR) 4698 { 4699 xim_fg_color = gui.def_norm_pixel; 4700 xim_bg_color = gui.def_back_pixel; 4701 } 4702 over_spot.x = TEXT_X(gui.col); 4703 over_spot.y = TEXT_Y(gui.row); 4704 spot_area.x = 0; 4705 spot_area.y = 0; 4706 spot_area.height = gui.char_height * Rows; 4707 spot_area.width = gui.char_width * Columns; 4708 line_space = gui.char_height; 4709 attr_list = (XVaNestedList) XVaCreateNestedList(0, 4710 XNSpotLocation, &over_spot, 4711 XNForeground, (Pixel) xim_fg_color, 4712 XNBackground, (Pixel) xim_bg_color, 4713 XNArea, &spot_area, 4714 XNLineSpace, line_space, 4715 NULL); 4716 if (XSetICValues(xic, XNPreeditAttributes, attr_list, NULL)) 4717 EMSG(_("E284: Cannot set IC values")); 4718 XFree(attr_list); 4719 } 4720 } 4721 #endif /* FEAT_GUI_GTK */ 4722 } 4723 4724 /* 4725 * Set up the status area. 4726 * 4727 * This should use a separate Widget, but that seems not possible, because 4728 * preedit_area and status_area should be set to the same window as for the 4729 * text input. Unfortunately this means the status area pollutes the text 4730 * window... 4731 */ 4732 void 4733 xim_set_status_area() 4734 { 4735 if (xic == NULL) 4736 return; 4737 4738 #ifdef FEAT_GUI_GTK 4739 # if defined(FEAT_XFONTSET) 4740 if (use_status_area) 4741 { 4742 GdkICAttr *attr; 4743 int style; 4744 gint width, height; 4745 GtkWidget *widget; 4746 int attrmask; 4747 4748 if (!xic_attr) 4749 return; 4750 4751 attr = xic_attr; 4752 attrmask = 0; 4753 style = (int)gdk_ic_get_style(xic); 4754 if ((style & (int)GDK_IM_STATUS_MASK) == (int)GDK_IM_STATUS_AREA) 4755 { 4756 if (gui.fontset != NOFONTSET 4757 && gui.fontset->type == GDK_FONT_FONTSET) 4758 { 4759 widget = gui.mainwin; 4760 gdk_window_get_size(widget->window, &width, &height); 4761 4762 attrmask |= (int)GDK_IC_STATUS_AREA; 4763 attr->status_area.x = 0; 4764 attr->status_area.y = height - gui.char_height - 1; 4765 attr->status_area.width = width; 4766 attr->status_area.height = gui.char_height; 4767 } 4768 } 4769 if (attrmask != 0) 4770 gdk_ic_set_attr(xic, attr, (GdkICAttributesType)attrmask); 4771 } 4772 # endif 4773 #else 4774 { 4775 XVaNestedList preedit_list = 0, status_list = 0, list = 0; 4776 XRectangle pre_area, status_area; 4777 4778 if (input_style & XIMStatusArea) 4779 { 4780 if (input_style & XIMPreeditArea) 4781 { 4782 XRectangle *needed_rect; 4783 4784 /* to get status_area width */ 4785 status_list = XVaCreateNestedList(0, XNAreaNeeded, 4786 &needed_rect, NULL); 4787 XGetICValues(xic, XNStatusAttributes, status_list, NULL); 4788 XFree(status_list); 4789 4790 status_area.width = needed_rect->width; 4791 } 4792 else 4793 status_area.width = gui.char_width * Columns; 4794 4795 status_area.x = 0; 4796 status_area.y = gui.char_height * Rows + gui.border_offset; 4797 if (gui.which_scrollbars[SBAR_BOTTOM]) 4798 status_area.y += gui.scrollbar_height; 4799 #ifdef FEAT_MENU 4800 if (gui.menu_is_active) 4801 status_area.y += gui.menu_height; 4802 #endif 4803 status_area.height = gui.char_height; 4804 status_list = XVaCreateNestedList(0, XNArea, &status_area, NULL); 4805 } 4806 else 4807 { 4808 status_area.x = 0; 4809 status_area.y = gui.char_height * Rows + gui.border_offset; 4810 if (gui.which_scrollbars[SBAR_BOTTOM]) 4811 status_area.y += gui.scrollbar_height; 4812 #ifdef FEAT_MENU 4813 if (gui.menu_is_active) 4814 status_area.y += gui.menu_height; 4815 #endif 4816 status_area.width = 0; 4817 status_area.height = gui.char_height; 4818 } 4819 4820 if (input_style & XIMPreeditArea) /* off-the-spot */ 4821 { 4822 pre_area.x = status_area.x + status_area.width; 4823 pre_area.y = gui.char_height * Rows + gui.border_offset; 4824 pre_area.width = gui.char_width * Columns - pre_area.x; 4825 if (gui.which_scrollbars[SBAR_BOTTOM]) 4826 pre_area.y += gui.scrollbar_height; 4827 #ifdef FEAT_MENU 4828 if (gui.menu_is_active) 4829 pre_area.y += gui.menu_height; 4830 #endif 4831 pre_area.height = gui.char_height; 4832 preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL); 4833 } 4834 else if (input_style & XIMPreeditPosition) /* over-the-spot */ 4835 { 4836 pre_area.x = 0; 4837 pre_area.y = 0; 4838 pre_area.height = gui.char_height * Rows; 4839 pre_area.width = gui.char_width * Columns; 4840 preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL); 4841 } 4842 4843 if (preedit_list && status_list) 4844 list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list, 4845 XNStatusAttributes, status_list, NULL); 4846 else if (preedit_list) 4847 list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list, 4848 NULL); 4849 else if (status_list) 4850 list = XVaCreateNestedList(0, XNStatusAttributes, status_list, 4851 NULL); 4852 else 4853 list = NULL; 4854 4855 if (list) 4856 { 4857 XSetICValues(xic, XNVaNestedList, list, NULL); 4858 XFree(list); 4859 } 4860 if (status_list) 4861 XFree(status_list); 4862 if (preedit_list) 4863 XFree(preedit_list); 4864 } 4865 #endif 4866 } 4867 4868 #if defined(FEAT_GUI_X11) || defined(FEAT_GUI_GTK) 4869 static char e_xim[] = N_("E285: Failed to create input context"); 4870 #endif 4871 4872 #if defined(FEAT_GUI_X11) || defined(PROTO) 4873 # if defined(XtSpecificationRelease) && XtSpecificationRelease >= 6 && !defined(sun) 4874 # define USE_X11R6_XIM 4875 # endif 4876 4877 static int xim_real_init __ARGS((Window x11_window, Display *x11_display)); 4878 4879 4880 #ifdef USE_X11R6_XIM 4881 static void xim_instantiate_cb __ARGS((Display *display, XPointer client_data, XPointer call_data)); 4882 static void xim_destroy_cb __ARGS((XIM im, XPointer client_data, XPointer call_data)); 4883 4884 /*ARGSUSED*/ 4885 static void 4886 xim_instantiate_cb(display, client_data, call_data) 4887 Display *display; 4888 XPointer client_data; 4889 XPointer call_data; 4890 { 4891 Window x11_window; 4892 Display *x11_display; 4893 4894 #ifdef XIM_DEBUG 4895 xim_log("xim_instantiate_cb()\n"); 4896 #endif 4897 4898 gui_get_x11_windis(&x11_window, &x11_display); 4899 if (display != x11_display) 4900 return; 4901 4902 xim_real_init(x11_window, x11_display); 4903 gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH); 4904 if (xic != NULL) 4905 XUnregisterIMInstantiateCallback(x11_display, NULL, NULL, NULL, 4906 xim_instantiate_cb, NULL); 4907 } 4908 4909 /*ARGSUSED*/ 4910 static void 4911 xim_destroy_cb(im, client_data, call_data) 4912 XIM im; 4913 XPointer client_data; 4914 XPointer call_data; 4915 { 4916 Window x11_window; 4917 Display *x11_display; 4918 4919 #ifdef XIM_DEBUG 4920 xim_log("xim_destroy_cb()\n"); 4921 #endif 4922 gui_get_x11_windis(&x11_window, &x11_display); 4923 4924 xic = NULL; 4925 status_area_enabled = FALSE; 4926 4927 gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH); 4928 4929 XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL, 4930 xim_instantiate_cb, NULL); 4931 } 4932 #endif 4933 4934 void 4935 xim_init() 4936 { 4937 Window x11_window; 4938 Display *x11_display; 4939 4940 #ifdef XIM_DEBUG 4941 xim_log("xim_init()\n"); 4942 #endif 4943 4944 gui_get_x11_windis(&x11_window, &x11_display); 4945 4946 xic = NULL; 4947 4948 if (xim_real_init(x11_window, x11_display)) 4949 return; 4950 4951 gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH); 4952 4953 #ifdef USE_X11R6_XIM 4954 XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL, 4955 xim_instantiate_cb, NULL); 4956 #endif 4957 } 4958 4959 static int 4960 xim_real_init(x11_window, x11_display) 4961 Window x11_window; 4962 Display *x11_display; 4963 { 4964 int i; 4965 char *p, 4966 *s, 4967 *ns, 4968 *end, 4969 tmp[1024]; 4970 #define IMLEN_MAX 40 4971 char buf[IMLEN_MAX + 7]; 4972 XIM xim = NULL; 4973 XIMStyles *xim_styles; 4974 XIMStyle this_input_style = 0; 4975 Boolean found; 4976 XPoint over_spot; 4977 XVaNestedList preedit_list, status_list; 4978 4979 input_style = 0; 4980 status_area_enabled = FALSE; 4981 4982 if (xic != NULL) 4983 return FALSE; 4984 4985 if (gui.rsrc_input_method != NULL && *gui.rsrc_input_method != NUL) 4986 { 4987 strcpy(tmp, gui.rsrc_input_method); 4988 for (ns = s = tmp; ns != NULL && *s != NUL;) 4989 { 4990 s = (char *)skipwhite((char_u *)s); 4991 if (*s == NUL) 4992 break; 4993 if ((ns = end = strchr(s, ',')) == NULL) 4994 end = s + strlen(s); 4995 while (isspace(((char_u *)end)[-1])) 4996 end--; 4997 *end = NUL; 4998 4999 if (strlen(s) <= IMLEN_MAX) 5000 { 5001 strcpy(buf, "@im="); 5002 strcat(buf, s); 5003 if ((p = XSetLocaleModifiers(buf)) != NULL && *p != NUL 5004 && (xim = XOpenIM(x11_display, NULL, NULL, NULL)) 5005 != NULL) 5006 break; 5007 } 5008 5009 s = ns + 1; 5010 } 5011 } 5012 5013 if (xim == NULL && (p = XSetLocaleModifiers("")) != NULL && *p != NUL) 5014 xim = XOpenIM(x11_display, NULL, NULL, NULL); 5015 5016 /* This is supposed to be useful to obtain characters through 5017 * XmbLookupString() without really using a XIM. */ 5018 if (xim == NULL && (p = XSetLocaleModifiers("@im=none")) != NULL 5019 && *p != NUL) 5020 xim = XOpenIM(x11_display, NULL, NULL, NULL); 5021 5022 if (xim == NULL) 5023 { 5024 /* Only give this message when verbose is set, because too many people 5025 * got this message when they didn't want to use a XIM. */ 5026 if (p_verbose > 0) 5027 { 5028 verbose_enter(); 5029 EMSG(_("E286: Failed to open input method")); 5030 verbose_leave(); 5031 } 5032 return FALSE; 5033 } 5034 5035 #ifdef USE_X11R6_XIM 5036 { 5037 XIMCallback destroy_cb; 5038 5039 destroy_cb.callback = xim_destroy_cb; 5040 destroy_cb.client_data = NULL; 5041 if (XSetIMValues(xim, XNDestroyCallback, &destroy_cb, NULL)) 5042 EMSG(_("E287: Warning: Could not set destroy callback to IM")); 5043 } 5044 #endif 5045 5046 if (XGetIMValues(xim, XNQueryInputStyle, &xim_styles, NULL) || !xim_styles) 5047 { 5048 EMSG(_("E288: input method doesn't support any style")); 5049 XCloseIM(xim); 5050 return FALSE; 5051 } 5052 5053 found = False; 5054 strcpy(tmp, gui.rsrc_preedit_type_name); 5055 for (s = tmp; s && !found; ) 5056 { 5057 while (*s && isspace((unsigned char)*s)) 5058 s++; 5059 if (!*s) 5060 break; 5061 if ((ns = end = strchr(s, ',')) != 0) 5062 ns++; 5063 else 5064 end = s + strlen(s); 5065 while (isspace((unsigned char)*end)) 5066 end--; 5067 *end = '\0'; 5068 5069 if (!strcmp(s, "OverTheSpot")) 5070 this_input_style = (XIMPreeditPosition | XIMStatusArea); 5071 else if (!strcmp(s, "OffTheSpot")) 5072 this_input_style = (XIMPreeditArea | XIMStatusArea); 5073 else if (!strcmp(s, "Root")) 5074 this_input_style = (XIMPreeditNothing | XIMStatusNothing); 5075 5076 for (i = 0; (unsigned short)i < xim_styles->count_styles; i++) 5077 { 5078 if (this_input_style == xim_styles->supported_styles[i]) 5079 { 5080 found = True; 5081 break; 5082 } 5083 } 5084 if (!found) 5085 for (i = 0; (unsigned short)i < xim_styles->count_styles; i++) 5086 { 5087 if ((xim_styles->supported_styles[i] & this_input_style) 5088 == (this_input_style & ~XIMStatusArea)) 5089 { 5090 this_input_style &= ~XIMStatusArea; 5091 found = True; 5092 break; 5093 } 5094 } 5095 5096 s = ns; 5097 } 5098 XFree(xim_styles); 5099 5100 if (!found) 5101 { 5102 /* Only give this message when verbose is set, because too many people 5103 * got this message when they didn't want to use a XIM. */ 5104 if (p_verbose > 0) 5105 { 5106 verbose_enter(); 5107 EMSG(_("E289: input method doesn't support my preedit type")); 5108 verbose_leave(); 5109 } 5110 XCloseIM(xim); 5111 return FALSE; 5112 } 5113 5114 over_spot.x = TEXT_X(gui.col); 5115 over_spot.y = TEXT_Y(gui.row); 5116 input_style = this_input_style; 5117 5118 /* A crash was reported when trying to pass gui.norm_font as XNFontSet, 5119 * thus that has been removed. Hopefully the default works... */ 5120 #ifdef FEAT_XFONTSET 5121 if (gui.fontset != NOFONTSET) 5122 { 5123 preedit_list = XVaCreateNestedList(0, 5124 XNSpotLocation, &over_spot, 5125 XNForeground, (Pixel)gui.def_norm_pixel, 5126 XNBackground, (Pixel)gui.def_back_pixel, 5127 XNFontSet, (XFontSet)gui.fontset, 5128 NULL); 5129 status_list = XVaCreateNestedList(0, 5130 XNForeground, (Pixel)gui.def_norm_pixel, 5131 XNBackground, (Pixel)gui.def_back_pixel, 5132 XNFontSet, (XFontSet)gui.fontset, 5133 NULL); 5134 } 5135 else 5136 #endif 5137 { 5138 preedit_list = XVaCreateNestedList(0, 5139 XNSpotLocation, &over_spot, 5140 XNForeground, (Pixel)gui.def_norm_pixel, 5141 XNBackground, (Pixel)gui.def_back_pixel, 5142 NULL); 5143 status_list = XVaCreateNestedList(0, 5144 XNForeground, (Pixel)gui.def_norm_pixel, 5145 XNBackground, (Pixel)gui.def_back_pixel, 5146 NULL); 5147 } 5148 5149 xic = XCreateIC(xim, 5150 XNInputStyle, input_style, 5151 XNClientWindow, x11_window, 5152 XNFocusWindow, gui.wid, 5153 XNPreeditAttributes, preedit_list, 5154 XNStatusAttributes, status_list, 5155 NULL); 5156 XFree(status_list); 5157 XFree(preedit_list); 5158 if (xic != NULL) 5159 { 5160 if (input_style & XIMStatusArea) 5161 { 5162 xim_set_status_area(); 5163 status_area_enabled = TRUE; 5164 } 5165 else 5166 gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH); 5167 } 5168 else 5169 { 5170 EMSG(_(e_xim)); 5171 XCloseIM(xim); 5172 return FALSE; 5173 } 5174 5175 return TRUE; 5176 } 5177 5178 #endif /* FEAT_GUI_X11 */ 5179 5180 #if defined(FEAT_GUI_GTK) || defined(PROTO) 5181 5182 # ifdef FEAT_XFONTSET 5183 static char e_overthespot[] = N_("E290: over-the-spot style requires fontset"); 5184 # endif 5185 5186 # ifdef PROTO 5187 typedef int GdkIC; 5188 # endif 5189 5190 void 5191 xim_decide_input_style() 5192 { 5193 /* GDK_IM_STATUS_CALLBACKS was disabled, enabled it to allow Japanese 5194 * OverTheSpot. */ 5195 int supported_style = (int)GDK_IM_PREEDIT_NONE | 5196 (int)GDK_IM_PREEDIT_NOTHING | 5197 (int)GDK_IM_PREEDIT_POSITION | 5198 (int)GDK_IM_PREEDIT_CALLBACKS | 5199 (int)GDK_IM_STATUS_CALLBACKS | 5200 (int)GDK_IM_STATUS_AREA | 5201 (int)GDK_IM_STATUS_NONE | 5202 (int)GDK_IM_STATUS_NOTHING; 5203 5204 #ifdef XIM_DEBUG 5205 xim_log("xim_decide_input_style()\n"); 5206 #endif 5207 5208 if (!gdk_im_ready()) 5209 xim_input_style = 0; 5210 else 5211 { 5212 if (gtk_major_version > 1 5213 || (gtk_major_version == 1 5214 && (gtk_minor_version > 2 5215 || (gtk_minor_version == 2 && gtk_micro_version >= 3)))) 5216 use_status_area = TRUE; 5217 else 5218 { 5219 EMSG(_("E291: Your GTK+ is older than 1.2.3. Status area disabled")); 5220 use_status_area = FALSE; 5221 } 5222 #ifdef FEAT_XFONTSET 5223 if (gui.fontset == NOFONTSET || gui.fontset->type != GDK_FONT_FONTSET) 5224 #endif 5225 supported_style &= ~((int)GDK_IM_PREEDIT_POSITION 5226 | (int)GDK_IM_STATUS_AREA); 5227 if (!use_status_area) 5228 supported_style &= ~(int)GDK_IM_STATUS_AREA; 5229 xim_input_style = (int)gdk_im_decide_style((GdkIMStyle)supported_style); 5230 } 5231 } 5232 5233 /*ARGSUSED*/ 5234 static void 5235 preedit_start_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5236 { 5237 #ifdef XIM_DEBUG 5238 xim_log("xim_decide_input_style()\n"); 5239 #endif 5240 5241 draw_feedback = NULL; 5242 xim_can_preediting = TRUE; 5243 xim_has_preediting = TRUE; 5244 gui_update_cursor(TRUE, FALSE); 5245 if (showmode() > 0) 5246 { 5247 setcursor(); 5248 out_flush(); 5249 } 5250 } 5251 5252 static void 5253 xim_back_delete(int n) 5254 { 5255 char_u str[3]; 5256 5257 str[0] = CSI; 5258 str[1] = 'k'; 5259 str[2] = 'b'; 5260 while (n-- > 0) 5261 add_to_input_buf(str, 3); 5262 } 5263 5264 static GSList *key_press_event_queue = NULL; 5265 static gboolean processing_queued_event = FALSE; 5266 5267 /*ARGSUSED*/ 5268 static void 5269 preedit_draw_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5270 { 5271 XIMPreeditDrawCallbackStruct *draw_data; 5272 XIMText *text; 5273 char *src; 5274 GSList *event_queue; 5275 5276 #ifdef XIM_DEBUG 5277 xim_log("preedit_draw_cbproc()\n"); 5278 #endif 5279 5280 draw_data = (XIMPreeditDrawCallbackStruct *) call_data; 5281 text = (XIMText *) draw_data->text; 5282 5283 if ((text == NULL && draw_data->chg_length == preedit_buf_len) 5284 || preedit_buf_len == 0) 5285 { 5286 init_preedit_start_col(); 5287 vim_free(draw_feedback); 5288 draw_feedback = NULL; 5289 } 5290 if (draw_data->chg_length > 0) 5291 { 5292 int bs_cnt; 5293 5294 if (draw_data->chg_length > preedit_buf_len) 5295 bs_cnt = preedit_buf_len; 5296 else 5297 bs_cnt = draw_data->chg_length; 5298 xim_back_delete(bs_cnt); 5299 preedit_buf_len -= bs_cnt; 5300 } 5301 if (text != NULL) 5302 { 5303 int len; 5304 #ifdef FEAT_MBYTE 5305 char_u *buf = NULL; 5306 unsigned int nfeedback = 0; 5307 #endif 5308 char_u *ptr; 5309 5310 src = text->string.multi_byte; 5311 if (src != NULL && !text->encoding_is_wchar) 5312 { 5313 len = strlen(src); 5314 ptr = (char_u *)src; 5315 /* Avoid the enter for decision */ 5316 if (*ptr == '\n') 5317 return; 5318 5319 #ifdef FEAT_MBYTE 5320 if (input_conv.vc_type != CONV_NONE 5321 && (buf = string_convert(&input_conv, 5322 (char_u *)src, &len)) != NULL) 5323 { 5324 /* Converted from 'termencoding' to 'encoding'. */ 5325 add_to_input_buf_csi(buf, len); 5326 ptr = buf; 5327 } 5328 else 5329 #endif 5330 add_to_input_buf_csi((char_u *)src, len); 5331 /* Add count of character to preedit_buf_len */ 5332 while (*ptr != NUL) 5333 { 5334 #ifdef FEAT_MBYTE 5335 if (draw_data->text->feedback != NULL) 5336 { 5337 if (draw_feedback == NULL) 5338 draw_feedback = (char *)alloc(draw_data->chg_first 5339 + text->length); 5340 else 5341 draw_feedback = realloc(draw_feedback, 5342 draw_data->chg_first + text->length); 5343 if (draw_feedback != NULL) 5344 { 5345 draw_feedback[nfeedback + draw_data->chg_first] 5346 = draw_data->text->feedback[nfeedback]; 5347 nfeedback++; 5348 } 5349 } 5350 if (has_mbyte) 5351 ptr += (*mb_ptr2len)(ptr); 5352 else 5353 #endif 5354 ptr++; 5355 preedit_buf_len++; 5356 } 5357 #ifdef FEAT_MBYTE 5358 vim_free(buf); 5359 #endif 5360 preedit_end_col = MAXCOL; 5361 } 5362 } 5363 if (text != NULL || draw_data->chg_length > 0) 5364 { 5365 event_queue = key_press_event_queue; 5366 processing_queued_event = TRUE; 5367 while (event_queue != NULL && processing_queued_event) 5368 { 5369 GdkEvent *ev = event_queue->data; 5370 5371 gboolean *ret; 5372 gtk_signal_emit_by_name((GtkObject*)gui.mainwin, "key_press_event", 5373 ev, &ret); 5374 gdk_event_free(ev); 5375 event_queue = event_queue->next; 5376 } 5377 processing_queued_event = FALSE; 5378 if (key_press_event_queue) 5379 { 5380 g_slist_free(key_press_event_queue); 5381 key_press_event_queue = NULL; 5382 } 5383 } 5384 if (gtk_main_level() > 0) 5385 gtk_main_quit(); 5386 } 5387 5388 /* 5389 * Retrieve the highlighting attributes at column col in the preedit string. 5390 * Return -1 if not in preediting mode or if col is out of range. 5391 */ 5392 int 5393 im_get_feedback_attr(int col) 5394 { 5395 if (draw_feedback != NULL && col < preedit_buf_len) 5396 { 5397 if (draw_feedback[col] & XIMReverse) 5398 return HL_INVERSE; 5399 else if (draw_feedback[col] & XIMUnderline) 5400 return HL_UNDERLINE; 5401 else 5402 return hl_attr(HLF_V); 5403 } 5404 5405 return -1; 5406 } 5407 5408 /*ARGSUSED*/ 5409 static void 5410 preedit_caret_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5411 { 5412 #ifdef XIM_DEBUG 5413 xim_log("preedit_caret_cbproc()\n"); 5414 #endif 5415 } 5416 5417 /*ARGSUSED*/ 5418 static void 5419 preedit_done_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5420 { 5421 #ifdef XIM_DEBUG 5422 xim_log("preedit_done_cbproc()\n"); 5423 #endif 5424 5425 vim_free(draw_feedback); 5426 draw_feedback = NULL; 5427 xim_can_preediting = FALSE; 5428 xim_has_preediting = FALSE; 5429 gui_update_cursor(TRUE, FALSE); 5430 if (showmode() > 0) 5431 { 5432 setcursor(); 5433 out_flush(); 5434 } 5435 } 5436 5437 void 5438 xim_reset(void) 5439 { 5440 char *text; 5441 5442 #ifdef XIM_DEBUG 5443 xim_log("xim_reset()\n"); 5444 #endif 5445 5446 if (xic != NULL) 5447 { 5448 text = XmbResetIC(((GdkICPrivate *)xic)->xic); 5449 if (text != NULL && !(xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS)) 5450 add_to_input_buf_csi((char_u *)text, strlen(text)); 5451 else 5452 preedit_buf_len = 0; 5453 if (text != NULL) 5454 XFree(text); 5455 } 5456 } 5457 5458 /*ARGSUSED*/ 5459 int 5460 xim_queue_key_press_event(GdkEventKey *event, int down) 5461 { 5462 #ifdef XIM_DEBUG 5463 xim_log("xim_queue_key_press_event()\n"); 5464 #endif 5465 5466 if (preedit_buf_len <= 0) 5467 return FALSE; 5468 if (processing_queued_event) 5469 processing_queued_event = FALSE; 5470 5471 key_press_event_queue = g_slist_append(key_press_event_queue, 5472 gdk_event_copy((GdkEvent *)event)); 5473 return TRUE; 5474 } 5475 5476 /*ARGSUSED*/ 5477 static void 5478 preedit_callback_setup(GdkIC *ic) 5479 { 5480 XIC xxic; 5481 XVaNestedList preedit_attr; 5482 XIMCallback preedit_start_cb; 5483 XIMCallback preedit_draw_cb; 5484 XIMCallback preedit_caret_cb; 5485 XIMCallback preedit_done_cb; 5486 5487 xxic = ((GdkICPrivate*)xic)->xic; 5488 preedit_start_cb.callback = (XIMProc)preedit_start_cbproc; 5489 preedit_draw_cb.callback = (XIMProc)preedit_draw_cbproc; 5490 preedit_caret_cb.callback = (XIMProc)preedit_caret_cbproc; 5491 preedit_done_cb.callback = (XIMProc)preedit_done_cbproc; 5492 preedit_attr 5493 = XVaCreateNestedList (0, 5494 XNPreeditStartCallback, &preedit_start_cb, 5495 XNPreeditDrawCallback, &preedit_draw_cb, 5496 XNPreeditCaretCallback, &preedit_caret_cb, 5497 XNPreeditDoneCallback, &preedit_done_cb, 5498 0); 5499 XSetICValues (xxic, XNPreeditAttributes, preedit_attr, 0); 5500 XFree(preedit_attr); 5501 } 5502 5503 /*ARGSUSED*/ 5504 static void 5505 reset_state_setup(GdkIC *ic) 5506 { 5507 #ifdef USE_X11R6_XIM 5508 /* don't change the input context when we call reset */ 5509 XSetICValues(((GdkICPrivate*)ic)->xic, XNResetState, XIMPreserveState, 0); 5510 #endif 5511 } 5512 5513 void 5514 xim_init(void) 5515 { 5516 #ifdef XIM_DEBUG 5517 xim_log("xim_init()\n"); 5518 #endif 5519 5520 xic = NULL; 5521 xic_attr = NULL; 5522 5523 if (!gdk_im_ready()) 5524 { 5525 if (p_verbose > 0) 5526 { 5527 verbose_enter(); 5528 EMSG(_("E292: Input Method Server is not running")); 5529 verbose_leave(); 5530 } 5531 return; 5532 } 5533 if ((xic_attr = gdk_ic_attr_new()) != NULL) 5534 { 5535 #ifdef FEAT_XFONTSET 5536 gint width, height; 5537 #endif 5538 int mask; 5539 GdkColormap *colormap; 5540 GdkICAttr *attr = xic_attr; 5541 int attrmask = (int)GDK_IC_ALL_REQ; 5542 GtkWidget *widget = gui.drawarea; 5543 5544 attr->style = (GdkIMStyle)xim_input_style; 5545 attr->client_window = gui.mainwin->window; 5546 5547 if ((colormap = gtk_widget_get_colormap(widget)) != 5548 gtk_widget_get_default_colormap()) 5549 { 5550 attrmask |= (int)GDK_IC_PREEDIT_COLORMAP; 5551 attr->preedit_colormap = colormap; 5552 } 5553 attrmask |= (int)GDK_IC_PREEDIT_FOREGROUND; 5554 attrmask |= (int)GDK_IC_PREEDIT_BACKGROUND; 5555 attr->preedit_foreground = widget->style->fg[GTK_STATE_NORMAL]; 5556 attr->preedit_background = widget->style->base[GTK_STATE_NORMAL]; 5557 5558 #ifdef FEAT_XFONTSET 5559 if ((xim_input_style & (int)GDK_IM_PREEDIT_MASK) 5560 == (int)GDK_IM_PREEDIT_POSITION) 5561 { 5562 if (gui.fontset == NOFONTSET 5563 || gui.fontset->type != GDK_FONT_FONTSET) 5564 { 5565 EMSG(_(e_overthespot)); 5566 } 5567 else 5568 { 5569 gdk_window_get_size(widget->window, &width, &height); 5570 5571 attrmask |= (int)GDK_IC_PREEDIT_POSITION_REQ; 5572 attr->spot_location.x = TEXT_X(0); 5573 attr->spot_location.y = TEXT_Y(0); 5574 attr->preedit_area.x = gui.border_offset; 5575 attr->preedit_area.y = gui.border_offset; 5576 attr->preedit_area.width = width - 2*gui.border_offset; 5577 attr->preedit_area.height = height - 2*gui.border_offset; 5578 attr->preedit_fontset = gui.fontset; 5579 } 5580 } 5581 5582 if ((xim_input_style & (int)GDK_IM_STATUS_MASK) 5583 == (int)GDK_IM_STATUS_AREA) 5584 { 5585 if (gui.fontset == NOFONTSET 5586 || gui.fontset->type != GDK_FONT_FONTSET) 5587 { 5588 EMSG(_(e_overthespot)); 5589 } 5590 else 5591 { 5592 gdk_window_get_size(gui.mainwin->window, &width, &height); 5593 attrmask |= (int)GDK_IC_STATUS_AREA_REQ; 5594 attr->status_area.x = 0; 5595 attr->status_area.y = height - gui.char_height - 1; 5596 attr->status_area.width = width; 5597 attr->status_area.height = gui.char_height; 5598 attr->status_fontset = gui.fontset; 5599 } 5600 } 5601 else if ((xim_input_style & (int)GDK_IM_STATUS_MASK) 5602 == (int)GDK_IM_STATUS_CALLBACKS) 5603 { 5604 /* FIXME */ 5605 } 5606 #endif 5607 5608 xic = gdk_ic_new(attr, (GdkICAttributesType)attrmask); 5609 5610 if (xic == NULL) 5611 EMSG(_(e_xim)); 5612 else 5613 { 5614 mask = (int)gdk_window_get_events(widget->window); 5615 mask |= (int)gdk_ic_get_events(xic); 5616 gdk_window_set_events(widget->window, (GdkEventMask)mask); 5617 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS) 5618 preedit_callback_setup(xic); 5619 reset_state_setup(xic); 5620 } 5621 } 5622 } 5623 5624 void 5625 im_shutdown(void) 5626 { 5627 #ifdef XIM_DEBUG 5628 xim_log("im_shutdown()\n"); 5629 #endif 5630 5631 if (xic != NULL) 5632 { 5633 gdk_im_end(); 5634 gdk_ic_destroy(xic); 5635 xic = NULL; 5636 } 5637 xim_is_active = FALSE; 5638 xim_can_preediting = FALSE; 5639 preedit_start_col = MAXCOL; 5640 xim_has_preediting = FALSE; 5641 } 5642 5643 #endif /* FEAT_GUI_GTK */ 5644 5645 int 5646 xim_get_status_area_height() 5647 { 5648 #ifdef FEAT_GUI_GTK 5649 if (xim_input_style & (int)GDK_IM_STATUS_AREA) 5650 return gui.char_height; 5651 #else 5652 if (status_area_enabled) 5653 return gui.char_height; 5654 #endif 5655 return 0; 5656 } 5657 5658 /* 5659 * Get IM status. When IM is on, return TRUE. Else return FALSE. 5660 * FIXME: This doesn't work correctly: Having focus doesn't always mean XIM is 5661 * active, when not having focus XIM may still be active (e.g., when using a 5662 * tear-off menu item). 5663 */ 5664 int 5665 im_get_status() 5666 { 5667 # ifdef FEAT_GUI_GTK 5668 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS) 5669 return xim_can_preediting; 5670 # endif 5671 return xim_has_focus; 5672 } 5673 5674 # endif /* !HAVE_GTK2 */ 5675 5676 # if defined(FEAT_GUI_GTK) || defined(PROTO) 5677 int 5678 im_is_preediting() 5679 { 5680 return xim_has_preediting; 5681 } 5682 # endif 5683 #endif /* FEAT_XIM */ 5684 5685 #if defined(FEAT_MBYTE) || defined(PROTO) 5686 5687 /* 5688 * Setup "vcp" for conversion from "from" to "to". 5689 * The names must have been made canonical with enc_canonize(). 5690 * vcp->vc_type must have been initialized to CONV_NONE. 5691 * Note: cannot be used for conversion from/to ucs-2 and ucs-4 (will use utf-8 5692 * instead). 5693 * Afterwards invoke with "from" and "to" equal to NULL to cleanup. 5694 * Return FAIL when conversion is not supported, OK otherwise. 5695 */ 5696 int 5697 convert_setup(vcp, from, to) 5698 vimconv_T *vcp; 5699 char_u *from; 5700 char_u *to; 5701 { 5702 int from_prop; 5703 int to_prop; 5704 5705 /* Reset to no conversion. */ 5706 # ifdef USE_ICONV 5707 if (vcp->vc_type == CONV_ICONV && vcp->vc_fd != (iconv_t)-1) 5708 iconv_close(vcp->vc_fd); 5709 # endif 5710 vcp->vc_type = CONV_NONE; 5711 vcp->vc_factor = 1; 5712 vcp->vc_fail = FALSE; 5713 5714 /* No conversion when one of the names is empty or they are equal. */ 5715 if (from == NULL || *from == NUL || to == NULL || *to == NUL 5716 || STRCMP(from, to) == 0) 5717 return OK; 5718 5719 from_prop = enc_canon_props(from); 5720 to_prop = enc_canon_props(to); 5721 if ((from_prop & ENC_LATIN1) && (to_prop & ENC_UNICODE)) 5722 { 5723 /* Internal latin1 -> utf-8 conversion. */ 5724 vcp->vc_type = CONV_TO_UTF8; 5725 vcp->vc_factor = 2; /* up to twice as long */ 5726 } 5727 else if ((from_prop & ENC_LATIN9) && (to_prop & ENC_UNICODE)) 5728 { 5729 /* Internal latin9 -> utf-8 conversion. */ 5730 vcp->vc_type = CONV_9_TO_UTF8; 5731 vcp->vc_factor = 3; /* up to three as long (euro sign) */ 5732 } 5733 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_LATIN1)) 5734 { 5735 /* Internal utf-8 -> latin1 conversion. */ 5736 vcp->vc_type = CONV_TO_LATIN1; 5737 } 5738 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_LATIN9)) 5739 { 5740 /* Internal utf-8 -> latin9 conversion. */ 5741 vcp->vc_type = CONV_TO_LATIN9; 5742 } 5743 #ifdef WIN3264 5744 /* Win32-specific codepage <-> codepage conversion without iconv. */ 5745 else if (((from_prop & ENC_UNICODE) || encname2codepage(from) > 0) 5746 && ((to_prop & ENC_UNICODE) || encname2codepage(to) > 0)) 5747 { 5748 vcp->vc_type = CONV_CODEPAGE; 5749 vcp->vc_factor = 2; /* up to twice as long */ 5750 vcp->vc_cpfrom = (from_prop & ENC_UNICODE) ? 0 : encname2codepage(from); 5751 vcp->vc_cpto = (to_prop & ENC_UNICODE) ? 0 : encname2codepage(to); 5752 } 5753 #endif 5754 #ifdef MACOS_X 5755 else if ((from_prop & ENC_MACROMAN) && (to_prop & ENC_LATIN1)) 5756 { 5757 vcp->vc_type = CONV_MAC_LATIN1; 5758 } 5759 else if ((from_prop & ENC_MACROMAN) && (to_prop & ENC_UNICODE)) 5760 { 5761 vcp->vc_type = CONV_MAC_UTF8; 5762 vcp->vc_factor = 2; /* up to twice as long */ 5763 } 5764 else if ((from_prop & ENC_LATIN1) && (to_prop & ENC_MACROMAN)) 5765 { 5766 vcp->vc_type = CONV_LATIN1_MAC; 5767 } 5768 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_MACROMAN)) 5769 { 5770 vcp->vc_type = CONV_UTF8_MAC; 5771 } 5772 #endif 5773 # ifdef USE_ICONV 5774 else 5775 { 5776 /* Use iconv() for conversion. */ 5777 vcp->vc_fd = (iconv_t)my_iconv_open( 5778 (to_prop & ENC_UNICODE) ? (char_u *)"utf-8" : to, 5779 (from_prop & ENC_UNICODE) ? (char_u *)"utf-8" : from); 5780 if (vcp->vc_fd != (iconv_t)-1) 5781 { 5782 vcp->vc_type = CONV_ICONV; 5783 vcp->vc_factor = 4; /* could be longer too... */ 5784 } 5785 } 5786 # endif 5787 if (vcp->vc_type == CONV_NONE) 5788 return FAIL; 5789 5790 return OK; 5791 } 5792 5793 #if defined(FEAT_GUI) || defined(AMIGA) || defined(WIN3264) \ 5794 || defined(MSDOS) || defined(PROTO) 5795 /* 5796 * Do conversion on typed input characters in-place. 5797 * The input and output are not NUL terminated! 5798 * Returns the length after conversion. 5799 */ 5800 int 5801 convert_input(ptr, len, maxlen) 5802 char_u *ptr; 5803 int len; 5804 int maxlen; 5805 { 5806 return convert_input_safe(ptr, len, maxlen, NULL, NULL); 5807 } 5808 #endif 5809 5810 /* 5811 * Like convert_input(), but when there is an incomplete byte sequence at the 5812 * end return that as an allocated string in "restp" and set "*restlenp" to 5813 * the length. If "restp" is NULL it is not used. 5814 */ 5815 int 5816 convert_input_safe(ptr, len, maxlen, restp, restlenp) 5817 char_u *ptr; 5818 int len; 5819 int maxlen; 5820 char_u **restp; 5821 int *restlenp; 5822 { 5823 char_u *d; 5824 int dlen = len; 5825 int unconvertlen = 0; 5826 5827 d = string_convert_ext(&input_conv, ptr, &dlen, 5828 restp == NULL ? NULL : &unconvertlen); 5829 if (d != NULL) 5830 { 5831 if (dlen <= maxlen) 5832 { 5833 if (unconvertlen > 0) 5834 { 5835 /* Move the unconverted characters to allocated memory. */ 5836 *restp = alloc(unconvertlen); 5837 if (*restp != NULL) 5838 mch_memmove(*restp, ptr + len - unconvertlen, unconvertlen); 5839 *restlenp = unconvertlen; 5840 } 5841 mch_memmove(ptr, d, dlen); 5842 } 5843 else 5844 /* result is too long, keep the unconverted text (the caller must 5845 * have done something wrong!) */ 5846 dlen = len; 5847 vim_free(d); 5848 } 5849 return dlen; 5850 } 5851 5852 /* 5853 * Convert text "ptr[*lenp]" according to "vcp". 5854 * Returns the result in allocated memory and sets "*lenp". 5855 * When "lenp" is NULL, use NUL terminated strings. 5856 * Illegal chars are often changed to "?", unless vcp->vc_fail is set. 5857 * When something goes wrong, NULL is returned and "*lenp" is unchanged. 5858 */ 5859 char_u * 5860 string_convert(vcp, ptr, lenp) 5861 vimconv_T *vcp; 5862 char_u *ptr; 5863 int *lenp; 5864 { 5865 return string_convert_ext(vcp, ptr, lenp, NULL); 5866 } 5867 5868 /* 5869 * Like string_convert(), but when "unconvlenp" is not NULL and there are is 5870 * an incomplete sequence at the end it is not converted and "*unconvlenp" is 5871 * set to the number of remaining bytes. 5872 */ 5873 char_u * 5874 string_convert_ext(vcp, ptr, lenp, unconvlenp) 5875 vimconv_T *vcp; 5876 char_u *ptr; 5877 int *lenp; 5878 int *unconvlenp; 5879 { 5880 char_u *retval = NULL; 5881 char_u *d; 5882 int len; 5883 int i; 5884 int l; 5885 int c; 5886 5887 if (lenp == NULL) 5888 len = (int)STRLEN(ptr); 5889 else 5890 len = *lenp; 5891 if (len == 0) 5892 return vim_strsave((char_u *)""); 5893 5894 switch (vcp->vc_type) 5895 { 5896 case CONV_TO_UTF8: /* latin1 to utf-8 conversion */ 5897 retval = alloc(len * 2 + 1); 5898 if (retval == NULL) 5899 break; 5900 d = retval; 5901 for (i = 0; i < len; ++i) 5902 { 5903 c = ptr[i]; 5904 if (c < 0x80) 5905 *d++ = c; 5906 else 5907 { 5908 *d++ = 0xc0 + ((unsigned)c >> 6); 5909 *d++ = 0x80 + (c & 0x3f); 5910 } 5911 } 5912 *d = NUL; 5913 if (lenp != NULL) 5914 *lenp = (int)(d - retval); 5915 break; 5916 5917 case CONV_9_TO_UTF8: /* latin9 to utf-8 conversion */ 5918 retval = alloc(len * 3 + 1); 5919 if (retval == NULL) 5920 break; 5921 d = retval; 5922 for (i = 0; i < len; ++i) 5923 { 5924 c = ptr[i]; 5925 switch (c) 5926 { 5927 case 0xa4: c = 0x20ac; break; /* euro */ 5928 case 0xa6: c = 0x0160; break; /* S hat */ 5929 case 0xa8: c = 0x0161; break; /* S -hat */ 5930 case 0xb4: c = 0x017d; break; /* Z hat */ 5931 case 0xb8: c = 0x017e; break; /* Z -hat */ 5932 case 0xbc: c = 0x0152; break; /* OE */ 5933 case 0xbd: c = 0x0153; break; /* oe */ 5934 case 0xbe: c = 0x0178; break; /* Y */ 5935 } 5936 d += utf_char2bytes(c, d); 5937 } 5938 *d = NUL; 5939 if (lenp != NULL) 5940 *lenp = (int)(d - retval); 5941 break; 5942 5943 case CONV_TO_LATIN1: /* utf-8 to latin1 conversion */ 5944 case CONV_TO_LATIN9: /* utf-8 to latin9 conversion */ 5945 retval = alloc(len + 1); 5946 if (retval == NULL) 5947 break; 5948 d = retval; 5949 for (i = 0; i < len; ++i) 5950 { 5951 l = utf_ptr2len(ptr + i); 5952 if (l == 0) 5953 *d++ = NUL; 5954 else if (l == 1) 5955 { 5956 if (unconvlenp != NULL && utf8len_tab[ptr[i]] > len - i) 5957 { 5958 /* Incomplete sequence at the end. */ 5959 *unconvlenp = len - i; 5960 break; 5961 } 5962 *d++ = ptr[i]; 5963 } 5964 else 5965 { 5966 c = utf_ptr2char(ptr + i); 5967 if (vcp->vc_type == CONV_TO_LATIN9) 5968 switch (c) 5969 { 5970 case 0x20ac: c = 0xa4; break; /* euro */ 5971 case 0x0160: c = 0xa6; break; /* S hat */ 5972 case 0x0161: c = 0xa8; break; /* S -hat */ 5973 case 0x017d: c = 0xb4; break; /* Z hat */ 5974 case 0x017e: c = 0xb8; break; /* Z -hat */ 5975 case 0x0152: c = 0xbc; break; /* OE */ 5976 case 0x0153: c = 0xbd; break; /* oe */ 5977 case 0x0178: c = 0xbe; break; /* Y */ 5978 case 0xa4: 5979 case 0xa6: 5980 case 0xa8: 5981 case 0xb4: 5982 case 0xb8: 5983 case 0xbc: 5984 case 0xbd: 5985 case 0xbe: c = 0x100; break; /* not in latin9 */ 5986 } 5987 if (!utf_iscomposing(c)) /* skip composing chars */ 5988 { 5989 if (c < 0x100) 5990 *d++ = c; 5991 else if (vcp->vc_fail) 5992 { 5993 vim_free(retval); 5994 return NULL; 5995 } 5996 else 5997 { 5998 *d++ = 0xbf; 5999 if (utf_char2cells(c) > 1) 6000 *d++ = '?'; 6001 } 6002 } 6003 i += l - 1; 6004 } 6005 } 6006 *d = NUL; 6007 if (lenp != NULL) 6008 *lenp = (int)(d - retval); 6009 break; 6010 6011 # ifdef MACOS_CONVERT 6012 case CONV_MAC_LATIN1: 6013 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 6014 'm', 'l', unconvlenp); 6015 break; 6016 6017 case CONV_LATIN1_MAC: 6018 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 6019 'l', 'm', unconvlenp); 6020 break; 6021 6022 case CONV_MAC_UTF8: 6023 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 6024 'm', 'u', unconvlenp); 6025 break; 6026 6027 case CONV_UTF8_MAC: 6028 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 6029 'u', 'm', unconvlenp); 6030 break; 6031 # endif 6032 6033 # ifdef USE_ICONV 6034 case CONV_ICONV: /* conversion with output_conv.vc_fd */ 6035 retval = iconv_string(vcp, ptr, len, unconvlenp); 6036 if (retval != NULL && lenp != NULL) 6037 *lenp = (int)STRLEN(retval); 6038 break; 6039 # endif 6040 # ifdef WIN3264 6041 case CONV_CODEPAGE: /* codepage -> codepage */ 6042 { 6043 int retlen; 6044 int tmp_len; 6045 short_u *tmp; 6046 6047 /* 1. codepage/UTF-8 -> ucs-2. */ 6048 if (vcp->vc_cpfrom == 0) 6049 tmp_len = utf8_to_ucs2(ptr, len, NULL, NULL); 6050 else 6051 tmp_len = MultiByteToWideChar(vcp->vc_cpfrom, 0, 6052 ptr, len, 0, 0); 6053 tmp = (short_u *)alloc(sizeof(short_u) * tmp_len); 6054 if (tmp == NULL) 6055 break; 6056 if (vcp->vc_cpfrom == 0) 6057 utf8_to_ucs2(ptr, len, tmp, unconvlenp); 6058 else 6059 MultiByteToWideChar(vcp->vc_cpfrom, 0, ptr, len, tmp, tmp_len); 6060 6061 /* 2. ucs-2 -> codepage/UTF-8. */ 6062 if (vcp->vc_cpto == 0) 6063 retlen = ucs2_to_utf8(tmp, tmp_len, NULL); 6064 else 6065 retlen = WideCharToMultiByte(vcp->vc_cpto, 0, 6066 tmp, tmp_len, 0, 0, 0, 0); 6067 retval = alloc(retlen + 1); 6068 if (retval != NULL) 6069 { 6070 if (vcp->vc_cpto == 0) 6071 ucs2_to_utf8(tmp, tmp_len, retval); 6072 else 6073 WideCharToMultiByte(vcp->vc_cpto, 0, 6074 tmp, tmp_len, retval, retlen, 0, 0); 6075 retval[retlen] = NUL; 6076 if (lenp != NULL) 6077 *lenp = retlen; 6078 } 6079 vim_free(tmp); 6080 break; 6081 } 6082 # endif 6083 } 6084 6085 return retval; 6086 } 6087 #endif 6088