1 /* vi:set ts=8 sts=4 sw=4: 2 * 3 * VIM - Vi IMproved by Bram Moolenaar 4 * Multibyte extensions partly by Sung-Hoon Baek 5 * 6 * Do ":help uganda" in Vim to read copying and usage conditions. 7 * Do ":help credits" in Vim to see a list of people who contributed. 8 * See README.txt for an overview of the Vim source code. 9 */ 10 /* 11 * mbyte.c: Code specifically for handling multi-byte characters. 12 * 13 * The encoding used in the core is set with 'encoding'. When 'encoding' is 14 * changed, the following four variables are set (for speed). 15 * Currently these types of character encodings are supported: 16 * 17 * "enc_dbcs" When non-zero it tells the type of double byte character 18 * encoding (Chinese, Korean, Japanese, etc.). 19 * The cell width on the display is equal to the number of 20 * bytes. (exception: DBCS_JPNU with first byte 0x8e) 21 * Recognizing the first or second byte is difficult, it 22 * requires checking a byte sequence from the start. 23 * "enc_utf8" When TRUE use Unicode characters in UTF-8 encoding. 24 * The cell width on the display needs to be determined from 25 * the character value. 26 * Recognizing bytes is easy: 0xxx.xxxx is a single-byte 27 * char, 10xx.xxxx is a trailing byte, 11xx.xxxx is a leading 28 * byte of a multi-byte character. 29 * To make things complicated, up to two composing characters 30 * are allowed. These are drawn on top of the first char. 31 * For most editing the sequence of bytes with composing 32 * characters included is considered to be one character. 33 * "enc_unicode" When 2 use 16-bit Unicode characters (or UTF-16). 34 * When 4 use 32-but Unicode characters. 35 * Internally characters are stored in UTF-8 encoding to 36 * avoid NUL bytes. Conversion happens when doing I/O. 37 * "enc_utf8" will also be TRUE. 38 * 39 * "has_mbyte" is set when "enc_dbcs" or "enc_utf8" is non-zero. 40 * 41 * If none of these is TRUE, 8-bit bytes are used for a character. The 42 * encoding isn't currently specified (TODO). 43 * 44 * 'encoding' specifies the encoding used in the core. This is in registers, 45 * text manipulation, buffers, etc. Conversion has to be done when characters 46 * in another encoding are received or send: 47 * 48 * clipboard 49 * ^ 50 * | (2) 51 * V 52 * +---------------+ 53 * (1) | | (3) 54 * keyboard ----->| core |-----> display 55 * | | 56 * +---------------+ 57 * ^ 58 * | (4) 59 * V 60 * file 61 * 62 * (1) Typed characters arrive in the current locale. Conversion is to be 63 * done when 'encoding' is different from 'termencoding'. 64 * (2) Text will be made available with the encoding specified with 65 * 'encoding'. If this is not sufficient, system-specific conversion 66 * might be required. 67 * (3) For the GUI the correct font must be selected, no conversion done. 68 * Otherwise, conversion is to be done when 'encoding' differs from 69 * 'termencoding'. (Different in the GTK+ 2 port -- 'termencoding' 70 * is always used for both input and output and must always be set to 71 * "utf-8". gui_mch_init() does this automatically.) 72 * (4) The encoding of the file is specified with 'fileencoding'. Conversion 73 * is to be done when it's different from 'encoding'. 74 * 75 * The viminfo file is a special case: Only text is converted, not file names. 76 * Vim scripts may contain an ":encoding" command. This has an effect for 77 * some commands, like ":menutrans" 78 */ 79 80 #include "vim.h" 81 82 #ifdef WIN32UNIX 83 # ifndef WIN32_LEAN_AND_MEAN 84 # define WIN32_LEAN_AND_MEAN 85 # endif 86 # include <windows.h> 87 # ifdef WIN32 88 # undef WIN32 /* Some windows.h define WIN32, we don't want that here. */ 89 # endif 90 #endif 91 92 #if (defined(WIN3264) || defined(WIN32UNIX)) && !defined(__MINGW32__) 93 # include <winnls.h> 94 #endif 95 96 #ifdef FEAT_GUI_X11 97 # include <X11/Intrinsic.h> 98 #endif 99 #ifdef X_LOCALE 100 #include <X11/Xlocale.h> 101 #endif 102 103 #if defined(FEAT_XIM) && defined(HAVE_GTK2) 104 # include <gdk/gdkkeysyms.h> 105 # ifdef WIN3264 106 # include <gdk/gdkwin32.h> 107 # else 108 # include <gdk/gdkx.h> 109 # endif 110 #endif 111 112 #ifdef HAVE_WCHAR_H 113 # include <wchar.h> 114 #endif 115 116 #if 0 117 /* This has been disabled, because several people reported problems with the 118 * wcwidth() and iswprint() library functions, esp. for Hebrew. */ 119 # ifdef __STDC_ISO_10646__ 120 # define USE_WCHAR_FUNCTIONS 121 # endif 122 #endif 123 124 #if defined(FEAT_MBYTE) || defined(PROTO) 125 126 static int enc_canon_search __ARGS((char_u *name)); 127 static int dbcs_char2len __ARGS((int c)); 128 static int dbcs_char2bytes __ARGS((int c, char_u *buf)); 129 static int dbcs_ptr2len_check __ARGS((char_u *p)); 130 static int dbcs_char2cells __ARGS((int c)); 131 static int dbcs_ptr2char __ARGS((char_u *p)); 132 133 /* Lookup table to quickly get the length in bytes of a UTF-8 character from 134 * the first byte of a UTF-8 string. Bytes which are illegal when used as the 135 * first byte have a one, because these will be used separately. */ 136 static char utf8len_tab[256] = 137 { 138 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 139 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 140 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 141 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 142 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /*bogus*/ 143 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /*bogus*/ 144 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, 145 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,4,4,4,4,4,4,4,4,5,5,5,5,6,6,1,1, 146 }; 147 148 /* 149 * XIM often causes trouble. Define XIM_DEBUG to get a log of XIM callbacks 150 * in the "xim.log" file. 151 */ 152 /* #define XIM_DEBUG */ 153 #ifdef XIM_DEBUG 154 static void 155 xim_log(char *s, ...) 156 { 157 va_list arglist; 158 static FILE *fd = NULL; 159 160 if (fd == (FILE *)-1) 161 return; 162 if (fd == NULL) 163 { 164 fd = fopen("xim.log", "w"); 165 if (fd == NULL) 166 { 167 EMSG("Cannot open xim.log"); 168 fd = (FILE *)-1; 169 return; 170 } 171 } 172 173 va_start(arglist, s); 174 vfprintf(fd, s, arglist); 175 va_end(arglist); 176 } 177 #endif 178 179 #endif 180 181 #if defined(FEAT_MBYTE) || defined(FEAT_POSTSCRIPT) || defined(PROTO) 182 /* 183 * Canonical encoding names and their properties. 184 * "iso-8859-n" is handled by enc_canonize() directly. 185 */ 186 static struct 187 { char *name; int prop; int codepage;} 188 enc_canon_table[] = 189 { 190 #define IDX_LATIN_1 0 191 {"latin1", ENC_8BIT + ENC_LATIN1, 1252}, 192 #define IDX_ISO_2 1 193 {"iso-8859-2", ENC_8BIT, 0}, 194 #define IDX_ISO_3 2 195 {"iso-8859-3", ENC_8BIT, 0}, 196 #define IDX_ISO_4 3 197 {"iso-8859-4", ENC_8BIT, 0}, 198 #define IDX_ISO_5 4 199 {"iso-8859-5", ENC_8BIT, 0}, 200 #define IDX_ISO_6 5 201 {"iso-8859-6", ENC_8BIT, 0}, 202 #define IDX_ISO_7 6 203 {"iso-8859-7", ENC_8BIT, 0}, 204 #define IDX_CP1255 7 205 {"cp1255", ENC_8BIT, 1255}, /* close to iso-8859-8 */ 206 #define IDX_ISO_8 8 207 {"iso-8859-8", ENC_8BIT, 0}, 208 #define IDX_ISO_9 9 209 {"iso-8859-9", ENC_8BIT, 0}, 210 #define IDX_ISO_10 10 211 {"iso-8859-10", ENC_8BIT, 0}, 212 #define IDX_ISO_11 11 213 {"iso-8859-11", ENC_8BIT, 0}, 214 #define IDX_ISO_13 12 215 {"iso-8859-13", ENC_8BIT, 0}, 216 #define IDX_ISO_14 13 217 {"iso-8859-14", ENC_8BIT, 0}, 218 #define IDX_ISO_15 14 219 {"iso-8859-15", ENC_8BIT + ENC_LATIN9, 0}, 220 #define IDX_KOI8_R 15 221 {"koi8-r", ENC_8BIT, 0}, 222 #define IDX_KOI8_U 16 223 {"koi8-u", ENC_8BIT, 0}, 224 #define IDX_UTF8 17 225 {"utf-8", ENC_UNICODE, 0}, 226 #define IDX_UCS2 18 227 {"ucs-2", ENC_UNICODE + ENC_ENDIAN_B + ENC_2BYTE, 0}, 228 #define IDX_UCS2LE 19 229 {"ucs-2le", ENC_UNICODE + ENC_ENDIAN_L + ENC_2BYTE, 0}, 230 #define IDX_UTF16 20 231 {"utf-16", ENC_UNICODE + ENC_ENDIAN_B + ENC_2WORD, 0}, 232 #define IDX_UTF16LE 21 233 {"utf-16le", ENC_UNICODE + ENC_ENDIAN_L + ENC_2WORD, 0}, 234 #define IDX_UCS4 22 235 {"ucs-4", ENC_UNICODE + ENC_ENDIAN_B + ENC_4BYTE, 0}, 236 #define IDX_UCS4LE 23 237 {"ucs-4le", ENC_UNICODE + ENC_ENDIAN_L + ENC_4BYTE, 0}, 238 #define IDX_DEBUG 24 239 {"debug", ENC_DBCS, DBCS_DEBUG}, 240 #define IDX_CP932 25 241 {"cp932", ENC_DBCS, DBCS_JPN}, 242 #define IDX_CP949 26 243 {"cp949", ENC_DBCS, DBCS_KOR}, 244 #define IDX_CP936 27 245 {"cp936", ENC_DBCS, DBCS_CHS}, 246 #define IDX_CP950 28 247 {"cp950", ENC_DBCS, DBCS_CHT}, 248 #define IDX_EUC_JP 29 249 {"euc-jp", ENC_DBCS, DBCS_JPNU}, 250 #define IDX_SJIS 30 251 {"sjis", ENC_DBCS, DBCS_JPN}, 252 #define IDX_EUC_KR 31 253 {"euc-kr", ENC_DBCS, DBCS_KORU}, 254 #define IDX_EUC_CN 32 255 {"euc-cn", ENC_DBCS, DBCS_CHSU}, 256 #define IDX_EUC_TW 33 257 {"euc-tw", ENC_DBCS, DBCS_CHTU}, 258 #define IDX_BIG5 34 259 {"big5", ENC_DBCS, DBCS_CHT}, 260 #define IDX_CP1251 35 261 {"cp1251", ENC_8BIT, 1251}, 262 #define IDX_MACROMAN 36 263 {"macroman", ENC_8BIT + ENC_MACROMAN, 0}, 264 #define IDX_COUNT 37 265 }; 266 267 /* 268 * Aliases for encoding names. 269 */ 270 static struct 271 { char *name; int canon;} 272 enc_alias_table[] = 273 { 274 {"ansi", IDX_LATIN_1}, 275 {"iso-8859-1", IDX_LATIN_1}, 276 {"latin2", IDX_ISO_2}, 277 {"latin3", IDX_ISO_3}, 278 {"latin4", IDX_ISO_4}, 279 {"cyrillic", IDX_ISO_5}, 280 {"arabic", IDX_ISO_6}, 281 {"greek", IDX_ISO_7}, 282 #ifdef WIN3264 283 {"hebrew", IDX_CP1255}, 284 #else 285 {"hebrew", IDX_ISO_8}, 286 #endif 287 {"latin5", IDX_ISO_9}, 288 {"turkish", IDX_ISO_9}, /* ? */ 289 {"latin6", IDX_ISO_10}, 290 {"nordic", IDX_ISO_10}, /* ? */ 291 {"thai", IDX_ISO_11}, /* ? */ 292 {"latin7", IDX_ISO_13}, 293 {"latin8", IDX_ISO_14}, 294 {"latin9", IDX_ISO_15}, 295 {"utf8", IDX_UTF8}, 296 {"unicode", IDX_UCS2}, 297 {"ucs2", IDX_UCS2}, 298 {"ucs2be", IDX_UCS2}, 299 {"ucs-2be", IDX_UCS2}, 300 {"ucs2le", IDX_UCS2LE}, 301 {"utf16", IDX_UTF16}, 302 {"utf16be", IDX_UTF16}, 303 {"utf-16be", IDX_UTF16}, 304 {"utf16le", IDX_UTF16LE}, 305 {"ucs4", IDX_UCS4}, 306 {"ucs4be", IDX_UCS4}, 307 {"ucs-4be", IDX_UCS4}, 308 {"ucs4le", IDX_UCS4LE}, 309 {"932", IDX_CP932}, 310 {"949", IDX_CP949}, 311 {"936", IDX_CP936}, 312 {"950", IDX_CP950}, 313 {"eucjp", IDX_EUC_JP}, 314 {"unix-jis", IDX_EUC_JP}, 315 {"ujis", IDX_EUC_JP}, 316 {"shift-jis", IDX_SJIS}, 317 {"euckr", IDX_EUC_KR}, 318 {"5601", IDX_EUC_KR}, /* Sun: KS C 5601 */ 319 {"euccn", IDX_EUC_CN}, 320 {"gb2312", IDX_EUC_CN}, 321 {"euctw", IDX_EUC_TW}, 322 #if defined(WIN3264) || defined(WIN32UNIX) || defined(MACOS) 323 {"japan", IDX_CP932}, 324 {"korea", IDX_CP949}, 325 {"prc", IDX_CP936}, 326 {"chinese", IDX_CP936}, 327 {"taiwan", IDX_CP950}, 328 {"big5", IDX_CP950}, 329 #else 330 {"japan", IDX_EUC_JP}, 331 {"korea", IDX_EUC_KR}, 332 {"prc", IDX_EUC_CN}, 333 {"chinese", IDX_EUC_CN}, 334 {"taiwan", IDX_EUC_TW}, 335 {"cp950", IDX_BIG5}, 336 {"950", IDX_BIG5}, 337 #endif 338 {"mac", IDX_MACROMAN}, 339 {NULL, 0} 340 }; 341 342 #ifndef CP_UTF8 343 # define CP_UTF8 65001 /* magic number from winnls.h */ 344 #endif 345 346 /* 347 * Find encoding "name" in the list of canonical encoding names. 348 * Returns -1 if not found. 349 */ 350 static int 351 enc_canon_search(name) 352 char_u *name; 353 { 354 int i; 355 356 for (i = 0; i < IDX_COUNT; ++i) 357 if (STRCMP(name, enc_canon_table[i].name) == 0) 358 return i; 359 return -1; 360 } 361 362 #endif 363 364 #if defined(FEAT_MBYTE) || defined(PROTO) 365 366 /* 367 * Find canonical encoding "name" in the list and return its properties. 368 * Returns 0 if not found. 369 */ 370 int 371 enc_canon_props(name) 372 char_u *name; 373 { 374 int i; 375 376 i = enc_canon_search(name); 377 if (i >= 0) 378 return enc_canon_table[i].prop; 379 #ifdef WIN3264 380 if (name[0] == 'c' && name[1] == 'p' && VIM_ISDIGIT(name[2])) 381 { 382 CPINFO cpinfo; 383 384 /* Get info on this codepage to find out what it is. */ 385 if (GetCPInfo(atoi(name + 2), &cpinfo) != 0) 386 { 387 if (cpinfo.MaxCharSize == 1) /* some single-byte encoding */ 388 return ENC_8BIT; 389 if (cpinfo.MaxCharSize == 2 390 && (cpinfo.LeadByte[0] != 0 || cpinfo.LeadByte[1] != 0)) 391 /* must be a DBCS encoding */ 392 return ENC_DBCS; 393 } 394 return 0; 395 } 396 #endif 397 if (STRNCMP(name, "2byte-", 6) == 0) 398 return ENC_DBCS; 399 if (STRNCMP(name, "8bit-", 5) == 0 || STRNCMP(name, "iso-8859-", 9) == 0) 400 return ENC_8BIT; 401 return 0; 402 } 403 404 /* 405 * Set up for using multi-byte characters. 406 * Called in three cases: 407 * - by main() to initialize (p_enc == NULL) 408 * - by set_init_1() after 'encoding' was set to its default. 409 * - by do_set() when 'encoding' has been set. 410 * p_enc must have been passed through enc_canonize() already. 411 * Sets the "enc_unicode", "enc_utf8", "enc_dbcs" and "has_mbyte" flags. 412 * Fills mb_bytelen_tab[] and returns NULL when there are no problems. 413 * When there is something wrong: Returns an error message and doesn't change 414 * anything. 415 */ 416 char_u * 417 mb_init() 418 { 419 int i; 420 int idx; 421 int n; 422 int enc_dbcs_new = 0; 423 #if defined(USE_ICONV) && !defined(WIN3264) && !defined(WIN32UNIX) \ 424 && !defined(MACOS) 425 # define LEN_FROM_CONV 426 vimconv_T vimconv; 427 char_u *p; 428 #endif 429 #ifdef WIN32 430 int prev_enc_utf8 = enc_utf8; 431 #endif 432 433 if (p_enc == NULL) 434 { 435 /* Just starting up: set the whole table to one's. */ 436 for (i = 0; i < 256; ++i) 437 mb_bytelen_tab[i] = 1; 438 input_conv.vc_type = CONV_NONE; 439 input_conv.vc_factor = 1; 440 output_conv.vc_type = CONV_NONE; 441 return NULL; 442 } 443 444 #ifdef WIN3264 445 if (p_enc[0] == 'c' && p_enc[1] == 'p' && VIM_ISDIGIT(p_enc[2])) 446 { 447 CPINFO cpinfo; 448 449 /* Get info on this codepage to find out what it is. */ 450 if (GetCPInfo(atoi(p_enc + 2), &cpinfo) != 0) 451 { 452 if (cpinfo.MaxCharSize == 1) 453 { 454 /* some single-byte encoding */ 455 enc_unicode = 0; 456 enc_utf8 = FALSE; 457 } 458 else if (cpinfo.MaxCharSize == 2 459 && (cpinfo.LeadByte[0] != 0 || cpinfo.LeadByte[1] != 0)) 460 { 461 /* must be a DBCS encoding, check below */ 462 enc_dbcs_new = atoi(p_enc + 2); 463 } 464 else 465 goto codepage_invalid; 466 } 467 else if (GetLastError() == ERROR_INVALID_PARAMETER) 468 { 469 codepage_invalid: 470 return (char_u *)N_("E543: Not a valid codepage"); 471 } 472 } 473 #endif 474 else if (STRNCMP(p_enc, "8bit-", 5) == 0 475 || STRNCMP(p_enc, "iso-8859-", 9) == 0) 476 { 477 /* Accept any "8bit-" or "iso-8859-" name. */ 478 enc_unicode = 0; 479 enc_utf8 = FALSE; 480 } 481 else if (STRNCMP(p_enc, "2byte-", 6) == 0) 482 { 483 #ifdef WIN3264 484 /* Windows: accept only valid codepage numbers, check below. */ 485 if (p_enc[6] != 'c' || p_enc[7] != 'p' 486 || (enc_dbcs_new = atoi(p_enc + 8)) == 0) 487 return e_invarg; 488 #else 489 /* Unix: accept any "2byte-" name, assume current locale. */ 490 enc_dbcs_new = DBCS_2BYTE; 491 #endif 492 } 493 else if ((idx = enc_canon_search(p_enc)) >= 0) 494 { 495 i = enc_canon_table[idx].prop; 496 if (i & ENC_UNICODE) 497 { 498 /* Unicode */ 499 enc_utf8 = TRUE; 500 if (i & (ENC_2BYTE | ENC_2WORD)) 501 enc_unicode = 2; 502 else if (i & ENC_4BYTE) 503 enc_unicode = 4; 504 else 505 enc_unicode = 0; 506 } 507 else if (i & ENC_DBCS) 508 { 509 /* 2byte, handle below */ 510 enc_dbcs_new = enc_canon_table[idx].codepage; 511 } 512 else 513 { 514 /* Must be 8-bit. */ 515 enc_unicode = 0; 516 enc_utf8 = FALSE; 517 } 518 } 519 else /* Don't know what encoding this is, reject it. */ 520 return e_invarg; 521 522 if (enc_dbcs_new != 0) 523 { 524 #ifdef WIN3264 525 /* Check if the DBCS code page is OK. */ 526 if (!IsValidCodePage(enc_dbcs_new)) 527 goto codepage_invalid; 528 #endif 529 enc_unicode = 0; 530 enc_utf8 = FALSE; 531 } 532 enc_dbcs = enc_dbcs_new; 533 has_mbyte = (enc_dbcs != 0 || enc_utf8); 534 535 #ifdef WIN3264 536 enc_codepage = encname2codepage(p_enc); 537 enc_latin9 = (STRCMP(p_enc, "iso-8859-15") == 0); 538 #endif 539 540 /* 541 * Set the function pointers. 542 */ 543 if (enc_utf8) 544 { 545 mb_ptr2len_check = utfc_ptr2len_check; 546 mb_char2len = utf_char2len; 547 mb_char2bytes = utf_char2bytes; 548 mb_ptr2cells = utf_ptr2cells; 549 mb_char2cells = utf_char2cells; 550 mb_off2cells = utf_off2cells; 551 mb_ptr2char = utf_ptr2char; 552 mb_head_off = utf_head_off; 553 } 554 else if (enc_dbcs != 0) 555 { 556 mb_ptr2len_check = dbcs_ptr2len_check; 557 mb_char2len = dbcs_char2len; 558 mb_char2bytes = dbcs_char2bytes; 559 mb_ptr2cells = dbcs_ptr2cells; 560 mb_char2cells = dbcs_char2cells; 561 mb_off2cells = dbcs_off2cells; 562 mb_ptr2char = dbcs_ptr2char; 563 mb_head_off = dbcs_head_off; 564 } 565 else 566 { 567 mb_ptr2len_check = latin_ptr2len_check; 568 mb_char2len = latin_char2len; 569 mb_char2bytes = latin_char2bytes; 570 mb_ptr2cells = latin_ptr2cells; 571 mb_char2cells = latin_char2cells; 572 mb_off2cells = latin_off2cells; 573 mb_ptr2char = latin_ptr2char; 574 mb_head_off = latin_head_off; 575 } 576 577 /* 578 * Fill the mb_bytelen_tab[] for MB_BYTE2LEN(). 579 */ 580 #ifdef LEN_FROM_CONV 581 /* When 'encoding' is different from the current locale mblen() won't 582 * work. Use conversion to "utf-8" instead. */ 583 vimconv.vc_type = CONV_NONE; 584 if (enc_dbcs) 585 { 586 p = enc_locale(); 587 if (p == NULL || STRCMP(p, p_enc) != 0) 588 { 589 convert_setup(&vimconv, p_enc, (char_u *)"utf-8"); 590 vimconv.vc_fail = TRUE; 591 } 592 vim_free(p); 593 } 594 #endif 595 596 for (i = 0; i < 256; ++i) 597 { 598 /* Our own function to reliably check the length of UTF-8 characters, 599 * independent of mblen(). */ 600 if (enc_utf8) 601 n = utf8len_tab[i]; 602 else if (enc_dbcs == 0) 603 n = 1; 604 else 605 { 606 #if defined(WIN3264) || defined(WIN32UNIX) 607 /* enc_dbcs is set by setting 'fileencoding'. It becomes a Windows 608 * CodePage identifier, which we can pass directly in to Windows 609 * API */ 610 n = IsDBCSLeadByteEx(enc_dbcs, (BYTE)i) ? 2 : 1; 611 #else 612 # ifdef MACOS 613 /* 614 * if mblen() is not available, character which MSB is turned on 615 * are treated as leading byte character. (note : This assumption 616 * is not always true.) 617 */ 618 n = (i & 0x80) ? 2 : 1; 619 # else 620 char buf[MB_MAXBYTES]; 621 # ifdef X_LOCALE 622 # ifndef mblen 623 # define mblen _Xmblen 624 # endif 625 # endif 626 if (i == NUL) /* just in case mblen() can't handle "" */ 627 n = 1; 628 else 629 { 630 buf[0] = i; 631 buf[1] = 0; 632 #ifdef LEN_FROM_CONV 633 if (vimconv.vc_type != CONV_NONE) 634 { 635 /* 636 * string_convert() should fail when converting the first 637 * byte of a double-byte character. 638 */ 639 p = string_convert(&vimconv, (char_u *)buf, NULL); 640 if (p != NULL) 641 { 642 vim_free(p); 643 n = 1; 644 } 645 else 646 n = 2; 647 } 648 else 649 #endif 650 { 651 /* 652 * mblen() should return -1 for invalid (means the leading 653 * multibyte) character. However there are some platforms 654 * where mblen() returns 0 for invalid character. 655 * Therefore, following condition includes 0. 656 */ 657 if (mblen(buf, (size_t)1) <= 0) 658 n = 2; 659 else 660 n = 1; 661 } 662 } 663 # endif 664 #endif 665 } 666 667 mb_bytelen_tab[i] = n; 668 } 669 670 #ifdef LEN_FROM_CONV 671 convert_setup(&vimconv, NULL, NULL); 672 #endif 673 674 /* The cell width depends on the type of multi-byte characters. */ 675 (void)init_chartab(); 676 677 /* When enc_utf8 is set or reset, (de)allocate ScreenLinesUC[] */ 678 screenalloc(FALSE); 679 680 /* When using Unicode, set default for 'fileencodings'. */ 681 if (enc_utf8 && !option_was_set((char_u *)"fencs")) 682 set_string_option_direct((char_u *)"fencs", -1, 683 (char_u *)"ucs-bom,utf-8,latin1", OPT_FREE); 684 #if defined(HAVE_BIND_TEXTDOMAIN_CODESET) && defined(FEAT_GETTEXT) 685 /* GNU gettext 0.10.37 supports this feature: set the codeset used for 686 * translated messages independently from the current locale. */ 687 (void)bind_textdomain_codeset(VIMPACKAGE, 688 enc_utf8 ? "utf-8" : (char *)p_enc); 689 #endif 690 691 #ifdef WIN32 692 /* When changing 'encoding' while starting up, then convert the command 693 * line arguments from the active codepage to 'encoding'. */ 694 if (starting != 0) 695 { 696 extern void fix_arg_enc(void); 697 698 fix_arg_enc(); 699 } 700 #endif 701 702 #ifdef FEAT_AUTOCMD 703 /* Fire an autocommand to let people do custom font setup. This must be 704 * after Vim has been setup for the new encoding. */ 705 apply_autocmds(EVENT_ENCODINGCHANGED, NULL, (char_u *)"", FALSE, curbuf); 706 #endif 707 708 #ifdef FEAT_GUI_KDE 709 if (gui.in_use) 710 gui_mch_update_codec(); 711 #endif 712 713 return NULL; 714 } 715 716 /* 717 * Return the size of the BOM for the current buffer: 718 * 0 - no BOM 719 * 2 - UCS-2 or UTF-16 BOM 720 * 4 - UCS-4 BOM 721 * 3 - UTF-8 BOM 722 */ 723 int 724 bomb_size() 725 { 726 int n = 0; 727 728 if (curbuf->b_p_bomb && !curbuf->b_p_bin) 729 { 730 if (*curbuf->b_p_fenc == NUL) 731 { 732 if (enc_utf8) 733 { 734 if (enc_unicode != 0) 735 n = enc_unicode; 736 else 737 n = 3; 738 } 739 } 740 else if (STRCMP(curbuf->b_p_fenc, "utf-8") == 0) 741 n = 3; 742 else if (STRNCMP(curbuf->b_p_fenc, "ucs-2", 5) == 0 743 || STRNCMP(curbuf->b_p_fenc, "utf-16", 6) == 0) 744 n = 2; 745 else if (STRNCMP(curbuf->b_p_fenc, "ucs-4", 5) == 0) 746 n = 4; 747 } 748 return n; 749 } 750 751 /* 752 * Get class of pointer: 753 * 0 for blank or NUL 754 * 1 for punctuation 755 * 2 for an (ASCII) word character 756 * >2 for other word characters 757 */ 758 int 759 mb_get_class(p) 760 char_u *p; 761 { 762 if (MB_BYTE2LEN(p[0]) == 1) 763 { 764 if (p[0] == NUL || vim_iswhite(p[0])) 765 return 0; 766 if (vim_iswordc(p[0])) 767 return 2; 768 return 1; 769 } 770 if (enc_dbcs != 0 && p[0] != NUL && p[1] != NUL) 771 return dbcs_class(p[0], p[1]); 772 if (enc_utf8) 773 return utf_class(utf_ptr2char(p)); 774 return 0; 775 } 776 777 /* 778 * Get class of a double-byte character. This always returns 3 or bigger. 779 * TODO: Should return 1 for punctuation. 780 */ 781 int 782 dbcs_class(lead, trail) 783 unsigned lead; 784 unsigned trail; 785 { 786 switch (enc_dbcs) 787 { 788 /* please add classfy routine for your language in here */ 789 790 case DBCS_JPNU: /* ? */ 791 case DBCS_JPN: 792 { 793 /* JIS code classification */ 794 unsigned char lb = lead; 795 unsigned char tb = trail; 796 797 /* convert process code to JIS */ 798 # if defined(WIN3264) || defined(WIN32UNIX) || defined(MACOS) 799 /* process code is SJIS */ 800 if (lb <= 0x9f) 801 lb = (lb - 0x81) * 2 + 0x21; 802 else 803 lb = (lb - 0xc1) * 2 + 0x21; 804 if (tb <= 0x7e) 805 tb -= 0x1f; 806 else if (tb <= 0x9e) 807 tb -= 0x20; 808 else 809 { 810 tb -= 0x7e; 811 lb += 1; 812 } 813 # else 814 /* 815 * XXX: Code page identification can not use with all 816 * system! So, some other encoding information 817 * will be needed. 818 * In japanese: SJIS,EUC,UNICODE,(JIS) 819 * Note that JIS-code system don't use as 820 * process code in most system because it uses 821 * escape sequences(JIS is context depend encoding). 822 */ 823 /* assume process code is JAPANESE-EUC */ 824 lb &= 0x7f; 825 tb &= 0x7f; 826 # endif 827 /* exceptions */ 828 switch (lb << 8 | tb) 829 { 830 case 0x2121: /* ZENKAKU space */ 831 return 0; 832 case 0x2122: /* KU-TEN (Japanese comma) */ 833 case 0x2123: /* TOU-TEN (Japanese period) */ 834 case 0x2124: /* ZENKAKU comma */ 835 case 0x2125: /* ZENKAKU period */ 836 return 1; 837 case 0x213c: /* prolongedsound handled as KATAKANA */ 838 return 13; 839 } 840 /* sieved by KU code */ 841 switch (lb) 842 { 843 case 0x21: 844 case 0x22: 845 /* special symbols */ 846 return 10; 847 case 0x23: 848 /* alpha-numeric */ 849 return 11; 850 case 0x24: 851 /* hiragana */ 852 return 12; 853 case 0x25: 854 /* katakana */ 855 return 13; 856 case 0x26: 857 /* greek */ 858 return 14; 859 case 0x27: 860 /* russian */ 861 return 15; 862 case 0x28: 863 /* lines */ 864 return 16; 865 default: 866 /* kanji */ 867 return 17; 868 } 869 } 870 871 case DBCS_KORU: /* ? */ 872 case DBCS_KOR: 873 { 874 /* KS code classification */ 875 unsigned char c1 = lead; 876 unsigned char c2 = trail; 877 878 /* 879 * 20 : Hangul 880 * 21 : Hanja 881 * 22 : Symbols 882 * 23 : Alpha-numeric/Roman Letter (Full width) 883 * 24 : Hangul Letter(Alphabet) 884 * 25 : Roman Numeral/Greek Letter 885 * 26 : Box Drawings 886 * 27 : Unit Symbols 887 * 28 : Circled/Parenthesized Letter 888 * 29 : Hirigana/Katakana 889 * 30 : Cyrillic Letter 890 */ 891 892 if (c1 >= 0xB0 && c1 <= 0xC8) 893 /* Hangul */ 894 return 20; 895 #if defined(WIN3264) || defined(WIN32UNIX) 896 else if (c1 <= 0xA0 || c2 <= 0xA0) 897 /* Extended Hangul Region : MS UHC(Unified Hangul Code) */ 898 /* c1: 0x81-0xA0 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xFE 899 * c1: 0xA1-0xC6 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xA0 900 */ 901 return 20; 902 #endif 903 904 else if (c1 >= 0xCA && c1 <= 0xFD) 905 /* Hanja */ 906 return 21; 907 else switch (c1) 908 { 909 case 0xA1: 910 case 0xA2: 911 /* Symbols */ 912 return 22; 913 case 0xA3: 914 /* Alpha-numeric */ 915 return 23; 916 case 0xA4: 917 /* Hangul Letter(Alphabet) */ 918 return 24; 919 case 0xA5: 920 /* Roman Numeral/Greek Letter */ 921 return 25; 922 case 0xA6: 923 /* Box Drawings */ 924 return 26; 925 case 0xA7: 926 /* Unit Symbols */ 927 return 27; 928 case 0xA8: 929 case 0xA9: 930 if (c2 <= 0xAF) 931 return 25; /* Roman Letter */ 932 else if (c2 >= 0xF6) 933 return 22; /* Symbols */ 934 else 935 /* Circled/Parenthesized Letter */ 936 return 28; 937 case 0xAA: 938 case 0xAB: 939 /* Hirigana/Katakana */ 940 return 29; 941 case 0xAC: 942 /* Cyrillic Letter */ 943 return 30; 944 } 945 } 946 default: 947 break; 948 } 949 return 3; 950 } 951 952 /* 953 * mb_char2len() function pointer. 954 * Return length in bytes of character "c". 955 * Returns 1 for a single-byte character. 956 */ 957 /* ARGSUSED */ 958 int 959 latin_char2len(c) 960 int c; 961 { 962 return 1; 963 } 964 965 static int 966 dbcs_char2len(c) 967 int c; 968 { 969 if (c >= 0x100) 970 return 2; 971 return 1; 972 } 973 974 /* 975 * mb_char2bytes() function pointer. 976 * Convert a character to its bytes. 977 * Returns the length in bytes. 978 */ 979 int 980 latin_char2bytes(c, buf) 981 int c; 982 char_u *buf; 983 { 984 buf[0] = c; 985 return 1; 986 } 987 988 static int 989 dbcs_char2bytes(c, buf) 990 int c; 991 char_u *buf; 992 { 993 if (c >= 0x100) 994 { 995 buf[0] = (unsigned)c >> 8; 996 buf[1] = c; 997 return 2; 998 } 999 buf[0] = c; 1000 return 1; 1001 } 1002 1003 /* 1004 * mb_ptr2len_check() function pointer. 1005 * Get byte length of character at "*p" but stop at a NUL. 1006 * For UTF-8 this includes following composing characters. 1007 * Returns 0 when *p is NUL. 1008 * 1009 */ 1010 int 1011 latin_ptr2len_check(p) 1012 char_u *p; 1013 { 1014 return MB_BYTE2LEN(*p); 1015 } 1016 1017 static int 1018 dbcs_ptr2len_check(p) 1019 char_u *p; 1020 { 1021 int len; 1022 1023 /* Check if second byte is not missing. */ 1024 len = MB_BYTE2LEN(*p); 1025 if (len == 2 && p[1] == NUL) 1026 len = 1; 1027 return len; 1028 } 1029 1030 struct interval 1031 { 1032 unsigned short first; 1033 unsigned short last; 1034 }; 1035 static int intable __ARGS((struct interval *table, size_t size, int c)); 1036 1037 /* 1038 * Return TRUE if "c" is in "table[size / sizeof(struct interval)]". 1039 */ 1040 static int 1041 intable(table, size, c) 1042 struct interval *table; 1043 size_t size; 1044 int c; 1045 { 1046 int mid, bot, top; 1047 1048 /* first quick check for Latin1 etc. characters */ 1049 if (c < table[0].first) 1050 return FALSE; 1051 1052 /* binary search in table */ 1053 bot = 0; 1054 top = size / sizeof(struct interval) - 1; 1055 while (top >= bot) 1056 { 1057 mid = (bot + top) / 2; 1058 if (table[mid].last < c) 1059 bot = mid + 1; 1060 else if (table[mid].first > c) 1061 top = mid - 1; 1062 else 1063 return TRUE; 1064 } 1065 return FALSE; 1066 } 1067 1068 /* 1069 * For UTF-8 character "c" return 2 for a double-width character, 1 for others. 1070 * Returns 4 or 6 for an unprintable character. 1071 * Is only correct for characters >= 0x80. 1072 * When p_ambw is "double", return 2 for a character with East Asian Width 1073 * class 'A'(mbiguous). 1074 */ 1075 int 1076 utf_char2cells(c) 1077 int c; 1078 { 1079 /* sorted list of non-overlapping intervals of East Asian Ambiguous 1080 * characters, generated with: 1081 * "uniset +WIDTH-A -cat=Me -cat=Mn -cat=Cf c" */ 1082 static struct interval ambiguous[] = { 1083 {0x00A1, 0x00A1}, {0x00A4, 0x00A4}, {0x00A7, 0x00A8}, 1084 {0x00AA, 0x00AA}, {0x00AE, 0x00AE}, {0x00B0, 0x00B4}, 1085 {0x00B6, 0x00BA}, {0x00BC, 0x00BF}, {0x00C6, 0x00C6}, 1086 {0x00D0, 0x00D0}, {0x00D7, 0x00D8}, {0x00DE, 0x00E1}, 1087 {0x00E6, 0x00E6}, {0x00E8, 0x00EA}, {0x00EC, 0x00ED}, 1088 {0x00F0, 0x00F0}, {0x00F2, 0x00F3}, {0x00F7, 0x00FA}, 1089 {0x00FC, 0x00FC}, {0x00FE, 0x00FE}, {0x0101, 0x0101}, 1090 {0x0111, 0x0111}, {0x0113, 0x0113}, {0x011B, 0x011B}, 1091 {0x0126, 0x0127}, {0x012B, 0x012B}, {0x0131, 0x0133}, 1092 {0x0138, 0x0138}, {0x013F, 0x0142}, {0x0144, 0x0144}, 1093 {0x0148, 0x014B}, {0x014D, 0x014D}, {0x0152, 0x0153}, 1094 {0x0166, 0x0167}, {0x016B, 0x016B}, {0x01CE, 0x01CE}, 1095 {0x01D0, 0x01D0}, {0x01D2, 0x01D2}, {0x01D4, 0x01D4}, 1096 {0x01D6, 0x01D6}, {0x01D8, 0x01D8}, {0x01DA, 0x01DA}, 1097 {0x01DC, 0x01DC}, {0x0251, 0x0251}, {0x0261, 0x0261}, 1098 {0x02C4, 0x02C4}, {0x02C7, 0x02C7}, {0x02C9, 0x02CB}, 1099 {0x02CD, 0x02CD}, {0x02D0, 0x02D0}, {0x02D8, 0x02DB}, 1100 {0x02DD, 0x02DD}, {0x02DF, 0x02DF}, {0x0391, 0x03A1}, 1101 {0x03A3, 0x03A9}, {0x03B1, 0x03C1}, {0x03C3, 0x03C9}, 1102 {0x0401, 0x0401}, {0x0410, 0x044F}, {0x0451, 0x0451}, 1103 {0x2010, 0x2010}, {0x2013, 0x2016}, {0x2018, 0x2019}, 1104 {0x201C, 0x201D}, {0x2020, 0x2022}, {0x2024, 0x2027}, 1105 {0x2030, 0x2030}, {0x2032, 0x2033}, {0x2035, 0x2035}, 1106 {0x203B, 0x203B}, {0x203E, 0x203E}, {0x2074, 0x2074}, 1107 {0x207F, 0x207F}, {0x2081, 0x2084}, {0x20AC, 0x20AC}, 1108 {0x2103, 0x2103}, {0x2105, 0x2105}, {0x2109, 0x2109}, 1109 {0x2113, 0x2113}, {0x2116, 0x2116}, {0x2121, 0x2122}, 1110 {0x2126, 0x2126}, {0x212B, 0x212B}, {0x2153, 0x2154}, 1111 {0x215B, 0x215E}, {0x2160, 0x216B}, {0x2170, 0x2179}, 1112 {0x2190, 0x2199}, {0x21B8, 0x21B9}, {0x21D2, 0x21D2}, 1113 {0x21D4, 0x21D4}, {0x21E7, 0x21E7}, {0x2200, 0x2200}, 1114 {0x2202, 0x2203}, {0x2207, 0x2208}, {0x220B, 0x220B}, 1115 {0x220F, 0x220F}, {0x2211, 0x2211}, {0x2215, 0x2215}, 1116 {0x221A, 0x221A}, {0x221D, 0x2220}, {0x2223, 0x2223}, 1117 {0x2225, 0x2225}, {0x2227, 0x222C}, {0x222E, 0x222E}, 1118 {0x2234, 0x2237}, {0x223C, 0x223D}, {0x2248, 0x2248}, 1119 {0x224C, 0x224C}, {0x2252, 0x2252}, {0x2260, 0x2261}, 1120 {0x2264, 0x2267}, {0x226A, 0x226B}, {0x226E, 0x226F}, 1121 {0x2282, 0x2283}, {0x2286, 0x2287}, {0x2295, 0x2295}, 1122 {0x2299, 0x2299}, {0x22A5, 0x22A5}, {0x22BF, 0x22BF}, 1123 {0x2312, 0x2312}, {0x2460, 0x24E9}, {0x24EB, 0x254B}, 1124 {0x2550, 0x2573}, {0x2580, 0x258F}, {0x2592, 0x2595}, 1125 {0x25A0, 0x25A1}, {0x25A3, 0x25A9}, {0x25B2, 0x25B3}, 1126 {0x25B6, 0x25B7}, {0x25BC, 0x25BD}, {0x25C0, 0x25C1}, 1127 {0x25C6, 0x25C8}, {0x25CB, 0x25CB}, {0x25CE, 0x25D1}, 1128 {0x25E2, 0x25E5}, {0x25EF, 0x25EF}, {0x2605, 0x2606}, 1129 {0x2609, 0x2609}, {0x260E, 0x260F}, {0x2614, 0x2615}, 1130 {0x261C, 0x261C}, {0x261E, 0x261E}, {0x2640, 0x2640}, 1131 {0x2642, 0x2642}, {0x2660, 0x2661}, {0x2663, 0x2665}, 1132 {0x2667, 0x266A}, {0x266C, 0x266D}, {0x266F, 0x266F}, 1133 {0x273D, 0x273D}, {0x2776, 0x277F}, {0xE000, 0xF8FF}, 1134 {0xFFFD, 0xFFFD}, /* {0xF0000, 0xFFFFD}, {0x100000, 0x10FFFD} */ 1135 }; 1136 1137 if (c >= 0x100) 1138 { 1139 #ifdef USE_WCHAR_FUNCTIONS 1140 /* 1141 * Assume the library function wcwidth() works better than our own 1142 * stuff. It should return 1 for ambiguous width chars! 1143 */ 1144 int n = wcwidth(c); 1145 1146 if (n < 0) 1147 return 6; /* unprintable, displays <xxxx> */ 1148 if (n > 1) 1149 return n; 1150 #else 1151 if (!utf_printable(c)) 1152 return 6; /* unprintable, displays <xxxx> */ 1153 if (c >= 0x1100 1154 && (c <= 0x115f /* Hangul Jamo */ 1155 || c == 0x2329 1156 || c == 0x232a 1157 || (c >= 0x2e80 && c <= 0xa4cf 1158 && c != 0x303f) /* CJK ... Yi */ 1159 || (c >= 0xac00 && c <= 0xd7a3) /* Hangul Syllables */ 1160 || (c >= 0xf900 && c <= 0xfaff) /* CJK Compatibility 1161 Ideographs */ 1162 || (c >= 0xfe30 && c <= 0xfe6f) /* CJK Compatibility Forms */ 1163 || (c >= 0xff00 && c <= 0xff60) /* Fullwidth Forms */ 1164 || (c >= 0xffe0 && c <= 0xffe6) 1165 || (c >= 0x20000 && c <= 0x2fffd) 1166 || (c >= 0x30000 && c <= 0x3fffd))) 1167 return 2; 1168 #endif 1169 } 1170 1171 /* Characters below 0x100 are influenced by 'isprint' option */ 1172 else if (c >= 0x80 && !vim_isprintc(c)) 1173 return 4; /* unprintable, displays <xx> */ 1174 1175 if (c >= 0x80 && *p_ambw == 'd' && intable(ambiguous, sizeof(ambiguous), c)) 1176 return 2; 1177 1178 return 1; 1179 } 1180 1181 /* 1182 * mb_ptr2cells() function pointer. 1183 * Return the number of display cells character at "*p" occupies. 1184 * This doesn't take care of unprintable characters, use ptr2cells() for that. 1185 */ 1186 /*ARGSUSED*/ 1187 int 1188 latin_ptr2cells(p) 1189 char_u *p; 1190 { 1191 return 1; 1192 } 1193 1194 int 1195 utf_ptr2cells(p) 1196 char_u *p; 1197 { 1198 int c; 1199 1200 /* Need to convert to a wide character. */ 1201 if (*p >= 0x80) 1202 { 1203 c = utf_ptr2char(p); 1204 /* An illegal byte is displayed as <xx>. */ 1205 if (utf_ptr2len_check(p) == 1 || c == NUL) 1206 return 4; 1207 /* If the char is ASCII it must be an overlong sequence. */ 1208 if (c < 0x80) 1209 return char2cells(c); 1210 return utf_char2cells(c); 1211 } 1212 return 1; 1213 } 1214 1215 int 1216 dbcs_ptr2cells(p) 1217 char_u *p; 1218 { 1219 /* Number of cells is equal to number of bytes, except for euc-jp when 1220 * the first byte is 0x8e. */ 1221 if (enc_dbcs == DBCS_JPNU && *p == 0x8e) 1222 return 1; 1223 return MB_BYTE2LEN(*p); 1224 } 1225 1226 /* 1227 * mb_char2cells() function pointer. 1228 * Return the number of display cells character "c" occupies. 1229 * Only takes care of multi-byte chars, not "^C" and such. 1230 */ 1231 /*ARGSUSED*/ 1232 int 1233 latin_char2cells(c) 1234 int c; 1235 { 1236 return 1; 1237 } 1238 1239 static int 1240 dbcs_char2cells(c) 1241 int c; 1242 { 1243 /* Number of cells is equal to number of bytes, except for euc-jp when 1244 * the first byte is 0x8e. */ 1245 if (enc_dbcs == DBCS_JPNU && ((unsigned)c >> 8) == 0x8e) 1246 return 1; 1247 /* use the first byte */ 1248 return MB_BYTE2LEN((unsigned)c >> 8); 1249 } 1250 1251 /* 1252 * mb_off2cells() function pointer. 1253 * Return number of display cells for char at ScreenLines[off]. 1254 * Caller must make sure "off" and "off + 1" are valid! 1255 */ 1256 /*ARGSUSED*/ 1257 int 1258 latin_off2cells(off) 1259 unsigned off; 1260 { 1261 return 1; 1262 } 1263 1264 int 1265 dbcs_off2cells(off) 1266 unsigned off; 1267 { 1268 /* Number of cells is equal to number of bytes, except for euc-jp when 1269 * the first byte is 0x8e. */ 1270 if (enc_dbcs == DBCS_JPNU && ScreenLines[off] == 0x8e) 1271 return 1; 1272 return MB_BYTE2LEN(ScreenLines[off]); 1273 } 1274 1275 int 1276 utf_off2cells(off) 1277 unsigned off; 1278 { 1279 return ScreenLines[off + 1] == 0 ? 2 : 1; 1280 } 1281 1282 /* 1283 * mb_ptr2char() function pointer. 1284 * Convert a byte sequence into a character. 1285 */ 1286 int 1287 latin_ptr2char(p) 1288 char_u *p; 1289 { 1290 return *p; 1291 } 1292 1293 static int 1294 dbcs_ptr2char(p) 1295 char_u *p; 1296 { 1297 if (MB_BYTE2LEN(*p) > 1 && p[1] != NUL) 1298 return (p[0] << 8) + p[1]; 1299 return *p; 1300 } 1301 1302 /* 1303 * Convert a UTF-8 byte sequence to a wide character. 1304 * If the sequence is illegal or truncated by a NUL the first byte is 1305 * returned. 1306 * Does not include composing characters, of course. 1307 */ 1308 int 1309 utf_ptr2char(p) 1310 char_u *p; 1311 { 1312 int len; 1313 1314 if (p[0] < 0x80) /* be quick for ASCII */ 1315 return p[0]; 1316 1317 len = utf8len_tab[p[0]]; 1318 if ((p[1] & 0xc0) == 0x80) 1319 { 1320 if (len == 2) 1321 return ((p[0] & 0x1f) << 6) + (p[1] & 0x3f); 1322 if ((p[2] & 0xc0) == 0x80) 1323 { 1324 if (len == 3) 1325 return ((p[0] & 0x0f) << 12) + ((p[1] & 0x3f) << 6) 1326 + (p[2] & 0x3f); 1327 if ((p[3] & 0xc0) == 0x80) 1328 { 1329 if (len == 4) 1330 return ((p[0] & 0x07) << 18) + ((p[1] & 0x3f) << 12) 1331 + ((p[2] & 0x3f) << 6) + (p[3] & 0x3f); 1332 if ((p[4] & 0xc0) == 0x80) 1333 { 1334 if (len == 5) 1335 return ((p[0] & 0x03) << 24) + ((p[1] & 0x3f) << 18) 1336 + ((p[2] & 0x3f) << 12) + ((p[3] & 0x3f) << 6) 1337 + (p[4] & 0x3f); 1338 if ((p[5] & 0xc0) == 0x80 && len == 6) 1339 return ((p[0] & 0x01) << 30) + ((p[1] & 0x3f) << 24) 1340 + ((p[2] & 0x3f) << 18) + ((p[3] & 0x3f) << 12) 1341 + ((p[4] & 0x3f) << 6) + (p[5] & 0x3f); 1342 } 1343 } 1344 } 1345 } 1346 /* Illegal value, just return the first byte */ 1347 return p[0]; 1348 } 1349 1350 /* 1351 * Get character at **pp and advance *pp to the next character. 1352 * Note: composing characters are skipped! 1353 */ 1354 int 1355 mb_ptr2char_adv(pp) 1356 char_u **pp; 1357 { 1358 int c; 1359 1360 c = (*mb_ptr2char)(*pp); 1361 *pp += (*mb_ptr2len_check)(*pp); 1362 return c; 1363 } 1364 1365 #if defined(FEAT_ARABIC) || defined(PROTO) 1366 /* 1367 * Check whether we are dealing with Arabic combining characters. 1368 * Note: these are NOT really composing characters! 1369 */ 1370 int 1371 arabic_combine(one, two) 1372 int one; /* first character */ 1373 int two; /* character just after "one" */ 1374 { 1375 if (one == a_LAM) 1376 return arabic_maycombine(two); 1377 return FALSE; 1378 } 1379 1380 /* 1381 * Check whether we are dealing with a character that could be regarded as an 1382 * Arabic combining character, need to check the character before this. 1383 */ 1384 int 1385 arabic_maycombine(two) 1386 int two; 1387 { 1388 if (p_arshape && !p_tbidi) 1389 return (two == a_ALEF_MADDA 1390 || two == a_ALEF_HAMZA_ABOVE 1391 || two == a_ALEF_HAMZA_BELOW 1392 || two == a_ALEF); 1393 return FALSE; 1394 } 1395 1396 /* 1397 * Check if the character pointed to by "p2" is a composing character when it 1398 * comes after "p1". For Arabic sometimes "ab" is replaced with "c", which 1399 * behaves like a composing character. 1400 */ 1401 int 1402 utf_composinglike(p1, p2) 1403 char_u *p1; 1404 char_u *p2; 1405 { 1406 int c2; 1407 1408 c2 = utf_ptr2char(p2); 1409 if (utf_iscomposing(c2)) 1410 return TRUE; 1411 if (!arabic_maycombine(c2)) 1412 return FALSE; 1413 return arabic_combine(utf_ptr2char(p1), c2); 1414 } 1415 #endif 1416 1417 /* 1418 * Convert a UTF-8 byte string to a wide chararacter. Also get up to two 1419 * composing characters. 1420 */ 1421 int 1422 utfc_ptr2char(p, p1, p2) 1423 char_u *p; 1424 int *p1; /* return: first composing char or 0 */ 1425 int *p2; /* return: second composing char or 0 */ 1426 { 1427 int len; 1428 int c; 1429 int cc; 1430 1431 c = utf_ptr2char(p); 1432 len = utf_ptr2len_check(p); 1433 /* Only accept a composing char when the first char isn't illegal. */ 1434 if ((len > 1 || *p < 0x80) 1435 && p[len] >= 0x80 1436 && UTF_COMPOSINGLIKE(p, p + len)) 1437 { 1438 *p1 = utf_ptr2char(p + len); 1439 len += utf_ptr2len_check(p + len); 1440 if (p[len] >= 0x80 && utf_iscomposing(cc = utf_ptr2char(p + len))) 1441 *p2 = cc; 1442 else 1443 *p2 = 0; 1444 } 1445 else 1446 { 1447 *p1 = 0; 1448 *p2 = 0; 1449 } 1450 return c; 1451 } 1452 1453 /* 1454 * Convert a UTF-8 byte string to a wide chararacter. Also get up to two 1455 * composing characters. Use no more than p[maxlen]. 1456 */ 1457 int 1458 utfc_ptr2char_len(p, p1, p2, maxlen) 1459 char_u *p; 1460 int *p1; /* return: first composing char or 0 */ 1461 int *p2; /* return: second composing char or 0 */ 1462 int maxlen; 1463 { 1464 int len; 1465 int c; 1466 int cc; 1467 1468 c = utf_ptr2char(p); 1469 len = utf_ptr2len_check_len(p, maxlen); 1470 /* Only accept a composing char when the first char isn't illegal. */ 1471 if ((len > 1 || *p < 0x80) 1472 && len < maxlen 1473 && p[len] >= 0x80 1474 && UTF_COMPOSINGLIKE(p, p + len)) 1475 { 1476 *p1 = utf_ptr2char(p + len); 1477 len += utf_ptr2len_check_len(p + len, maxlen - len); 1478 if (len < maxlen 1479 && p[len] >= 0x80 1480 && utf_iscomposing(cc = utf_ptr2char(p + len))) 1481 *p2 = cc; 1482 else 1483 *p2 = 0; 1484 } 1485 else 1486 { 1487 *p1 = 0; 1488 *p2 = 0; 1489 } 1490 return c; 1491 } 1492 1493 /* 1494 * Convert the character at screen position "off" to a sequence of bytes. 1495 * Includes the composing characters. 1496 * "buf" must at least have the length MB_MAXBYTES. 1497 * Returns the produced number of bytes. 1498 */ 1499 int 1500 utfc_char2bytes(off, buf) 1501 int off; 1502 char_u *buf; 1503 { 1504 int len; 1505 1506 len = utf_char2bytes(ScreenLinesUC[off], buf); 1507 if (ScreenLinesC1[off] != 0) 1508 { 1509 len += utf_char2bytes(ScreenLinesC1[off], buf + len); 1510 if (ScreenLinesC2[off] != 0) 1511 len += utf_char2bytes(ScreenLinesC2[off], buf + len); 1512 } 1513 return len; 1514 } 1515 1516 /* 1517 * Get the length of a UTF-8 byte sequence, not including any following 1518 * composing characters. 1519 * Returns 0 for "". 1520 * Returns 1 for an illegal byte sequence. 1521 */ 1522 int 1523 utf_ptr2len_check(p) 1524 char_u *p; 1525 { 1526 int len; 1527 int i; 1528 1529 if (*p == NUL) 1530 return 0; 1531 len = utf8len_tab[*p]; 1532 for (i = 1; i < len; ++i) 1533 if ((p[i] & 0xc0) != 0x80) 1534 return 1; 1535 return len; 1536 } 1537 1538 /* 1539 * Return length of UTF-8 character, obtained from the first byte. 1540 * "b" must be between 0 and 255! 1541 */ 1542 int 1543 utf_byte2len(b) 1544 int b; 1545 { 1546 return utf8len_tab[b]; 1547 } 1548 1549 /* 1550 * Get the length of UTF-8 byte sequence "p[size]". Does not include any 1551 * following composing characters. 1552 * Returns 1 for "". 1553 * Returns 1 for an illegal byte sequence. 1554 * Returns number > "size" for an incomplete byte sequence. 1555 */ 1556 int 1557 utf_ptr2len_check_len(p, size) 1558 char_u *p; 1559 int size; 1560 { 1561 int len; 1562 int i; 1563 1564 if (*p == NUL) 1565 return 1; 1566 len = utf8len_tab[*p]; 1567 if (len > size) 1568 return len; /* incomplete byte sequence. */ 1569 for (i = 1; i < len; ++i) 1570 if ((p[i] & 0xc0) != 0x80) 1571 return 1; 1572 return len; 1573 } 1574 1575 /* 1576 * Return the number of bytes the UTF-8 encoding of the character at "p" takes. 1577 * This includes following composing characters. 1578 */ 1579 int 1580 utfc_ptr2len_check(p) 1581 char_u *p; 1582 { 1583 int len; 1584 #ifdef FEAT_ARABIC 1585 int prevlen; 1586 #endif 1587 1588 if (*p == NUL) 1589 return 0; 1590 if (p[0] < 0x80 && p[1] < 0x80) /* be quick for ASCII */ 1591 return 1; 1592 1593 /* Skip over first UTF-8 char, stopping at a NUL byte. */ 1594 len = utf_ptr2len_check(p); 1595 1596 /* Check for illegal byte. */ 1597 if (len == 1 && p[0] >= 0x80) 1598 return 1; 1599 1600 /* 1601 * Check for composing characters. We can handle only the first two, but 1602 * skip all of them (otherwise the cursor would get stuck). 1603 */ 1604 #ifdef FEAT_ARABIC 1605 prevlen = 0; 1606 #endif 1607 for (;;) 1608 { 1609 if (p[len] < 0x80 || !UTF_COMPOSINGLIKE(p + prevlen, p + len)) 1610 return len; 1611 1612 /* Skip over composing char */ 1613 #ifdef FEAT_ARABIC 1614 prevlen = len; 1615 #endif 1616 len += utf_ptr2len_check(p + len); 1617 } 1618 } 1619 1620 /* 1621 * Return the number of bytes the UTF-8 encoding of the character at "p[size]" 1622 * takes. This includes following composing characters. 1623 * Returns 1 for an illegal char or an incomplete byte sequence. 1624 */ 1625 int 1626 utfc_ptr2len_check_len(p, size) 1627 char_u *p; 1628 int size; 1629 { 1630 int len; 1631 #ifdef FEAT_ARABIC 1632 int prevlen; 1633 #endif 1634 1635 if (*p == NUL) 1636 return 0; 1637 if (p[0] < 0x80 && (size == 1 || p[1] < 0x80)) /* be quick for ASCII */ 1638 return 1; 1639 1640 /* Skip over first UTF-8 char, stopping at a NUL byte. */ 1641 len = utf_ptr2len_check_len(p, size); 1642 1643 /* Check for illegal byte and incomplete byte sequence. */ 1644 if ((len == 1 && p[0] >= 0x80) || len > size) 1645 return 1; 1646 1647 /* 1648 * Check for composing characters. We can handle only the first two, but 1649 * skip all of them (otherwise the cursor would get stuck). 1650 */ 1651 #ifdef FEAT_ARABIC 1652 prevlen = 0; 1653 #endif 1654 while (len < size) 1655 { 1656 if (p[len] < 0x80 || !UTF_COMPOSINGLIKE(p + prevlen, p + len)) 1657 break; 1658 1659 /* Skip over composing char */ 1660 #ifdef FEAT_ARABIC 1661 prevlen = len; 1662 #endif 1663 len += utf_ptr2len_check_len(p + len, size - len); 1664 } 1665 return len; 1666 } 1667 1668 /* 1669 * Return the number of bytes the UTF-8 encoding of character "c" takes. 1670 * This does not include composing characters. 1671 */ 1672 int 1673 utf_char2len(c) 1674 int c; 1675 { 1676 if (c < 0x80) 1677 return 1; 1678 if (c < 0x800) 1679 return 2; 1680 if (c < 0x10000) 1681 return 3; 1682 if (c < 0x200000) 1683 return 4; 1684 if (c < 0x4000000) 1685 return 5; 1686 return 6; 1687 } 1688 1689 /* 1690 * Convert Unicode character "c" to UTF-8 string in "buf[]". 1691 * Returns the number of bytes. 1692 * This does not include composing characters. 1693 */ 1694 int 1695 utf_char2bytes(c, buf) 1696 int c; 1697 char_u *buf; 1698 { 1699 if (c < 0x80) /* 7 bits */ 1700 { 1701 buf[0] = c; 1702 return 1; 1703 } 1704 if (c < 0x800) /* 11 bits */ 1705 { 1706 buf[0] = 0xc0 + ((unsigned)c >> 6); 1707 buf[1] = 0x80 + (c & 0x3f); 1708 return 2; 1709 } 1710 if (c < 0x10000) /* 16 bits */ 1711 { 1712 buf[0] = 0xe0 + ((unsigned)c >> 12); 1713 buf[1] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1714 buf[2] = 0x80 + (c & 0x3f); 1715 return 3; 1716 } 1717 if (c < 0x200000) /* 21 bits */ 1718 { 1719 buf[0] = 0xf0 + ((unsigned)c >> 18); 1720 buf[1] = 0x80 + (((unsigned)c >> 12) & 0x3f); 1721 buf[2] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1722 buf[3] = 0x80 + (c & 0x3f); 1723 return 4; 1724 } 1725 if (c < 0x4000000) /* 26 bits */ 1726 { 1727 buf[0] = 0xf8 + ((unsigned)c >> 24); 1728 buf[1] = 0x80 + (((unsigned)c >> 18) & 0x3f); 1729 buf[2] = 0x80 + (((unsigned)c >> 12) & 0x3f); 1730 buf[3] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1731 buf[4] = 0x80 + (c & 0x3f); 1732 return 5; 1733 } 1734 /* 31 bits */ 1735 buf[0] = 0xfc + ((unsigned)c >> 30); 1736 buf[1] = 0x80 + (((unsigned)c >> 24) & 0x3f); 1737 buf[2] = 0x80 + (((unsigned)c >> 18) & 0x3f); 1738 buf[3] = 0x80 + (((unsigned)c >> 12) & 0x3f); 1739 buf[4] = 0x80 + (((unsigned)c >> 6) & 0x3f); 1740 buf[5] = 0x80 + (c & 0x3f); 1741 return 6; 1742 } 1743 1744 /* 1745 * Return TRUE if "c" is a composing UTF-8 character. This means it will be 1746 * drawn on top of the preceding character. 1747 * Based on code from Markus Kuhn. 1748 */ 1749 int 1750 utf_iscomposing(c) 1751 int c; 1752 { 1753 /* sorted list of non-overlapping intervals */ 1754 static struct interval combining[] = 1755 { 1756 {0x0300, 0x034f}, {0x0360, 0x036f}, {0x0483, 0x0486}, {0x0488, 0x0489}, 1757 {0x0591, 0x05a1}, {0x05a3, 0x05b9}, {0x05bb, 0x05bd}, {0x05bf, 0x05bf}, 1758 {0x05c1, 0x05c2}, {0x05c4, 0x05c4}, {0x0610, 0x0615}, {0x064b, 0x0658}, 1759 {0x0670, 0x0670}, {0x06d6, 0x06dc}, {0x06de, 0x06e4}, {0x06e7, 0x06e8}, 1760 {0x06ea, 0x06ed}, {0x0711, 0x0711}, {0x0730, 0x074a}, {0x07a6, 0x07b0}, 1761 {0x0901, 0x0903}, {0x093c, 0x093c}, {0x093e, 0x094d}, {0x0951, 0x0954}, 1762 {0x0962, 0x0963}, {0x0981, 0x0983}, {0x09bc, 0x09bc}, {0x09be, 0x09c4}, 1763 {0x09c7, 0x09c8}, {0x09cb, 0x09cd}, {0x09d7, 0x09d7}, {0x09e2, 0x09e3}, 1764 {0x0a01, 0x0a03}, {0x0a3c, 0x0a3c}, {0x0a3e, 0x0a42}, {0x0a47, 0x0a48}, 1765 {0x0a4b, 0x0a4d}, {0x0a70, 0x0a71}, {0x0a81, 0x0a83}, {0x0abc, 0x0abc}, 1766 {0x0abe, 0x0ac5}, {0x0ac7, 0x0ac9}, {0x0acb, 0x0acd}, {0x0ae2, 0x0ae3}, 1767 {0x0b01, 0x0b03}, {0x0b3c, 0x0b3c}, {0x0b3e, 0x0b43}, {0x0b47, 0x0b48}, 1768 {0x0b4b, 0x0b4d}, {0x0b56, 0x0b57}, {0x0b82, 0x0b82}, {0x0bbe, 0x0bc2}, 1769 {0x0bc6, 0x0bc8}, {0x0bca, 0x0bcd}, {0x0bd7, 0x0bd7}, {0x0c01, 0x0c03}, 1770 {0x0c3e, 0x0c44}, {0x0c46, 0x0c48}, {0x0c4a, 0x0c4d}, {0x0c55, 0x0c56}, 1771 {0x0c82, 0x0c83}, {0x0cbc, 0x0cbc}, {0x0cbe, 0x0cc4}, {0x0cc6, 0x0cc8}, 1772 {0x0cca, 0x0ccd}, {0x0cd5, 0x0cd6}, {0x0d02, 0x0d03}, {0x0d3e, 0x0d43}, 1773 {0x0d46, 0x0d48}, {0x0d4a, 0x0d4d}, {0x0d57, 0x0d57}, {0x0d82, 0x0d83}, 1774 {0x0dca, 0x0dca}, {0x0dcf, 0x0dd4}, {0x0dd6, 0x0dd6}, {0x0dd8, 0x0ddf}, 1775 {0x0df2, 0x0df3}, {0x0e31, 0x0e31}, {0x0e34, 0x0e3a}, {0x0e47, 0x0e4e}, 1776 {0x0eb1, 0x0eb1}, {0x0eb4, 0x0eb9}, {0x0ebb, 0x0ebc}, {0x0ec8, 0x0ecd}, 1777 {0x0f18, 0x0f19}, {0x0f35, 0x0f35}, {0x0f37, 0x0f37}, {0x0f39, 0x0f39}, 1778 {0x0f3e, 0x0f3f}, {0x0f71, 0x0f84}, {0x0f86, 0x0f87}, {0x0f90, 0x0f97}, 1779 {0x0f99, 0x0fbc}, {0x0fc6, 0x0fc6}, {0x102c, 0x1032}, {0x1036, 0x1039}, 1780 {0x1056, 0x1059}, {0x1712, 0x1714}, {0x1732, 0x1734}, {0x1752, 0x1753}, 1781 {0x1772, 0x1773}, {0x17b6, 0x17d3}, {0x17dd, 0x17dd}, {0x180b, 0x180d}, 1782 {0x18a9, 0x18a9}, {0x1920, 0x192b}, {0x1930, 0x193b}, {0x20d0, 0x20ea}, 1783 {0x302a, 0x302f}, {0x3099, 0x309a}, {0xfb1e, 0xfb1e}, {0xfe00, 0xfe0f}, 1784 {0xfe20, 0xfe23}, 1785 }; 1786 1787 return intable(combining, sizeof(combining), c); 1788 } 1789 1790 /* 1791 * Return TRUE for characters that can be displayed in a normal way. 1792 * Only for characters of 0x100 and above! 1793 */ 1794 int 1795 utf_printable(c) 1796 int c; 1797 { 1798 #ifdef USE_WCHAR_FUNCTIONS 1799 /* 1800 * Assume the iswprint() library function works better than our own stuff. 1801 */ 1802 return iswprint(c); 1803 #else 1804 /* Sorted list of non-overlapping intervals. 1805 * 0xd800-0xdfff is reserved for UTF-16, actually illegal. */ 1806 static struct interval nonprint[] = 1807 { 1808 {0x070f, 0x070f}, {0x180b, 0x180e}, {0x200b, 0x200f}, {0x202a, 0x202e}, 1809 {0x206a, 0x206f}, {0xd800, 0xdfff}, {0xfeff, 0xfeff}, {0xfff9, 0xfffb}, 1810 {0xfffe, 0xffff} 1811 }; 1812 1813 return !intable(nonprint, sizeof(nonprint), c); 1814 #endif 1815 } 1816 1817 /* 1818 * Get class of a Unicode character. 1819 * 0: white space 1820 * 1: punctuation 1821 * 2 or bigger: some class of word character. 1822 */ 1823 int 1824 utf_class(c) 1825 int c; 1826 { 1827 /* sorted list of non-overlapping intervals */ 1828 static struct clinterval 1829 { 1830 unsigned short first; 1831 unsigned short last; 1832 unsigned short class; 1833 } classes[] = 1834 { 1835 {0x037e, 0x037e, 1}, /* Greek question mark */ 1836 {0x0387, 0x0387, 1}, /* Greek ano teleia */ 1837 {0x055a, 0x055f, 1}, /* Armenian punctuation */ 1838 {0x0589, 0x0589, 1}, /* Armenian full stop */ 1839 {0x05be, 0x05be, 1}, 1840 {0x05c0, 0x05c0, 1}, 1841 {0x05c3, 0x05c3, 1}, 1842 {0x05f3, 0x05f4, 1}, 1843 {0x060c, 0x060c, 1}, 1844 {0x061b, 0x061b, 1}, 1845 {0x061f, 0x061f, 1}, 1846 {0x066a, 0x066d, 1}, 1847 {0x06d4, 0x06d4, 1}, 1848 {0x0700, 0x070d, 1}, /* Syriac punctuation */ 1849 {0x0964, 0x0965, 1}, 1850 {0x0970, 0x0970, 1}, 1851 {0x0df4, 0x0df4, 1}, 1852 {0x0e4f, 0x0e4f, 1}, 1853 {0x0e5a, 0x0e5b, 1}, 1854 {0x0f04, 0x0f12, 1}, 1855 {0x0f3a, 0x0f3d, 1}, 1856 {0x0f85, 0x0f85, 1}, 1857 {0x104a, 0x104f, 1}, /* Myanmar punctuation */ 1858 {0x10fb, 0x10fb, 1}, /* Georgian punctuation */ 1859 {0x1361, 0x1368, 1}, /* Ethiopic punctuation */ 1860 {0x166d, 0x166e, 1}, /* Canadian Syl. punctuation */ 1861 {0x1680, 0x1680, 0}, 1862 {0x169b, 0x169c, 1}, 1863 {0x16eb, 0x16ed, 1}, 1864 {0x1735, 0x1736, 1}, 1865 {0x17d4, 0x17dc, 1}, /* Khmer punctuation */ 1866 {0x1800, 0x180a, 1}, /* Mongolian punctuation */ 1867 {0x2000, 0x200b, 0}, /* spaces */ 1868 {0x200c, 0x2027, 1}, /* punctuation and symbols */ 1869 {0x2028, 0x2029, 0}, 1870 {0x202a, 0x202e, 1}, /* punctuation and symbols */ 1871 {0x202f, 0x202f, 0}, 1872 {0x2030, 0x205e, 1}, /* punctuation and symbols */ 1873 {0x205f, 0x205f, 0}, 1874 {0x2060, 0x27ff, 1}, /* punctuation and symbols */ 1875 {0x2070, 0x207f, 0x2070}, /* superscript */ 1876 {0x2080, 0x208f, 0x2080}, /* subscript */ 1877 {0x2983, 0x2998, 1}, 1878 {0x29d8, 0x29db, 1}, 1879 {0x29fc, 0x29fd, 1}, 1880 {0x3000, 0x3000, 0}, /* ideographic space */ 1881 {0x3001, 0x3020, 1}, /* ideographic punctuation */ 1882 {0x3030, 0x3030, 1}, 1883 {0x303d, 0x303d, 1}, 1884 {0x3040, 0x309f, 0x3040}, /* Hiragana */ 1885 {0x30a0, 0x30ff, 0x30a0}, /* Katakana */ 1886 {0x3300, 0x9fff, 0x4e00}, /* CJK Ideographs */ 1887 {0xac00, 0xd7a3, 0xac00}, /* Hangul Syllables */ 1888 {0xf900, 0xfaff, 0x4e00}, /* CJK Ideographs */ 1889 {0xfd3e, 0xfd3f, 1}, 1890 {0xfe30, 0xfe6b, 1}, /* punctuation forms */ 1891 {0xff00, 0xff0f, 1}, /* half/fullwidth ASCII */ 1892 {0xff1a, 0xff20, 1}, /* half/fullwidth ASCII */ 1893 {0xff3b, 0xff40, 1}, /* half/fullwidth ASCII */ 1894 {0xff5b, 0xff65, 1}, /* half/fullwidth ASCII */ 1895 }; 1896 int bot = 0; 1897 int top = sizeof(classes) / sizeof(struct clinterval) - 1; 1898 int mid; 1899 1900 /* First quick check for Latin1 characters, use 'iskeyword'. */ 1901 if (c < 0x100) 1902 { 1903 if (c == ' ' || c == '\t' || c == NUL) 1904 return 0; /* blank */ 1905 if (vim_iswordc(c)) 1906 return 2; /* word character */ 1907 return 1; /* punctuation */ 1908 } 1909 1910 /* binary search in table */ 1911 while (top >= bot) 1912 { 1913 mid = (bot + top) / 2; 1914 if (classes[mid].last < c) 1915 bot = mid + 1; 1916 else if (classes[mid].first > c) 1917 top = mid - 1; 1918 else 1919 return (int)classes[mid].class; 1920 } 1921 1922 /* most other characters are "word" characters */ 1923 return 2; 1924 } 1925 1926 /* 1927 * Code for Unicode case-dependent operations. Based on notes in 1928 * http://www.unicode.org/Public/UNIDATA/CaseFolding.txt 1929 * This code uses simple case folding, not full case folding. 1930 */ 1931 1932 /* 1933 * The following table is built by foldExtract.pl < CaseFolding.txt . 1934 * It must be in numeric order, because we use binary search on it. 1935 * An entry such as {0x41,0x5a,1,32} means that UCS-4 characters in the range 1936 * from 0x41 to 0x5a inclusive, stepping by 1, are folded by adding 32. 1937 */ 1938 1939 typedef struct 1940 { 1941 int rangeStart; 1942 int rangeEnd; 1943 int step; 1944 int offset; 1945 } convertStruct; 1946 1947 convertStruct foldCase[] = 1948 { 1949 {0x41,0x5a,1,32}, {0xc0,0xd6,1,32}, {0xd8,0xde,1,32}, 1950 {0x100,0x12e,2,1}, {0x130,0x130,-1,-199}, {0x132,0x136,2,1}, 1951 {0x139,0x147,2,1}, {0x14a,0x176,2,1}, {0x178,0x178,-1,-121}, 1952 {0x179,0x17d,2,1}, {0x181,0x181,-1,210}, {0x182,0x184,2,1}, 1953 {0x186,0x186,-1,206}, {0x187,0x187,-1,1}, {0x189,0x18a,1,205}, 1954 {0x18b,0x18b,-1,1}, {0x18e,0x18e,-1,79}, {0x18f,0x18f,-1,202}, 1955 {0x190,0x190,-1,203}, {0x191,0x191,-1,1}, {0x193,0x193,-1,205}, 1956 {0x194,0x194,-1,207}, {0x196,0x196,-1,211}, {0x197,0x197,-1,209}, 1957 {0x198,0x198,-1,1}, {0x19c,0x19c,-1,211}, {0x19d,0x19d,-1,213}, 1958 {0x19f,0x19f,-1,214}, {0x1a0,0x1a4,2,1}, {0x1a6,0x1a6,-1,218}, 1959 {0x1a7,0x1a7,-1,1}, {0x1a9,0x1a9,-1,218}, {0x1ac,0x1ac,-1,1}, 1960 {0x1ae,0x1ae,-1,218}, {0x1af,0x1af,-1,1}, {0x1b1,0x1b2,1,217}, 1961 {0x1b3,0x1b5,2,1}, {0x1b7,0x1b7,-1,219}, {0x1b8,0x1bc,4,1}, 1962 {0x1c4,0x1c4,-1,2}, {0x1c5,0x1c5,-1,1}, {0x1c7,0x1c7,-1,2}, 1963 {0x1c8,0x1c8,-1,1}, {0x1ca,0x1ca,-1,2}, {0x1cb,0x1db,2,1}, 1964 {0x1de,0x1ee,2,1}, {0x1f1,0x1f1,-1,2}, {0x1f2,0x1f4,2,1}, 1965 {0x1f6,0x1f6,-1,-97}, {0x1f7,0x1f7,-1,-56}, {0x1f8,0x21e,2,1}, 1966 {0x220,0x220,-1,-130}, {0x222,0x232,2,1}, {0x386,0x386,-1,38}, 1967 {0x388,0x38a,1,37}, {0x38c,0x38c,-1,64}, {0x38e,0x38f,1,63}, 1968 {0x391,0x3a1,1,32}, {0x3a3,0x3ab,1,32}, {0x3d8,0x3ee,2,1}, 1969 {0x3f4,0x3f4,-1,-60}, {0x3f7,0x3f7,-1,1}, {0x3f9,0x3f9,-1,-7}, 1970 {0x3fa,0x3fa,-1,1}, {0x400,0x40f,1,80}, {0x410,0x42f,1,32}, 1971 {0x460,0x480,2,1}, {0x48a,0x4be,2,1}, {0x4c1,0x4cd,2,1}, 1972 {0x4d0,0x4f4,2,1}, {0x4f8,0x500,8,1}, {0x502,0x50e,2,1}, 1973 {0x531,0x556,1,48}, {0x1e00,0x1e94,2,1}, {0x1ea0,0x1ef8,2,1}, 1974 {0x1f08,0x1f0f,1,-8}, {0x1f18,0x1f1d,1,-8}, {0x1f28,0x1f2f,1,-8}, 1975 {0x1f38,0x1f3f,1,-8}, {0x1f48,0x1f4d,1,-8}, {0x1f59,0x1f5f,2,-8}, 1976 {0x1f68,0x1f6f,1,-8}, {0x1f88,0x1f8f,1,-8}, {0x1f98,0x1f9f,1,-8}, 1977 {0x1fa8,0x1faf,1,-8}, {0x1fb8,0x1fb9,1,-8}, {0x1fba,0x1fbb,1,-74}, 1978 {0x1fbc,0x1fbc,-1,-9}, {0x1fc8,0x1fcb,1,-86}, {0x1fcc,0x1fcc,-1,-9}, 1979 {0x1fd8,0x1fd9,1,-8}, {0x1fda,0x1fdb,1,-100}, {0x1fe8,0x1fe9,1,-8}, 1980 {0x1fea,0x1feb,1,-112}, {0x1fec,0x1fec,-1,-7}, {0x1ff8,0x1ff9,1,-128}, 1981 {0x1ffa,0x1ffb,1,-126}, {0x1ffc,0x1ffc,-1,-9}, {0x2126,0x2126,-1,-7517}, 1982 {0x212a,0x212a,-1,-8383}, {0x212b,0x212b,-1,-8262}, 1983 {0x2160,0x216f,1,16}, {0x24b6,0x24cf,1,26}, {0xff21,0xff3a,1,32}, 1984 {0x10400,0x10427,1,40} 1985 }; 1986 1987 static int utf_convert(int a, convertStruct table[], int tableSize); 1988 1989 /* 1990 * Generic conversion function for case operations. 1991 * Return the converted equivalent of "a", which is a UCS-4 character. Use 1992 * the given conversion "table". Uses binary search on "table". 1993 */ 1994 static int 1995 utf_convert(a, table, tableSize) 1996 int a; 1997 convertStruct table[]; 1998 int tableSize; 1999 { 2000 int start, mid, end; /* indices into table */ 2001 2002 start = 0; 2003 end = tableSize / sizeof(convertStruct); 2004 while (start < end) 2005 { 2006 /* need to search further */ 2007 mid = (end + start) /2; 2008 if (table[mid].rangeEnd < a) 2009 start = mid + 1; 2010 else 2011 end = mid; 2012 } 2013 if (table[start].rangeStart <= a && a <= table[start].rangeEnd 2014 && (a - table[start].rangeStart) % table[start].step == 0) 2015 return (a + table[start].offset); 2016 else 2017 return a; 2018 } 2019 2020 /* 2021 * Return the folded-case equivalent of "a", which is a UCS-4 character. Uses 2022 * simple case folding. 2023 */ 2024 int 2025 utf_fold(a) 2026 int a; 2027 { 2028 return utf_convert(a, foldCase, sizeof(foldCase)); 2029 } 2030 2031 /* 2032 * The following tables are built by upperLowerExtract.pl < UnicodeData.txt . 2033 * They must be in numeric order, because we use binary search on them. 2034 * An entry such as {0x41,0x5a,1,32} means that UCS-4 characters in the range 2035 * from 0x41 to 0x5a inclusive, stepping by 1, are switched to lower (for 2036 * example) by adding 32. 2037 */ 2038 convertStruct toLower[] = 2039 { 2040 {0x41,0x5a,1,32}, {0xc0,0xd6,1,32}, {0xd8,0xde,1,32}, 2041 {0x100,0x12e,2,1}, {0x130,0x130,-1,-199}, {0x132,0x136,2,1}, 2042 {0x139,0x147,2,1}, {0x14a,0x176,2,1}, {0x178,0x178,-1,-121}, 2043 {0x179,0x17d,2,1}, {0x181,0x181,-1,210}, {0x182,0x184,2,1}, 2044 {0x186,0x186,-1,206}, {0x187,0x187,-1,1}, {0x189,0x18a,1,205}, 2045 {0x18b,0x18b,-1,1}, {0x18e,0x18e,-1,79}, {0x18f,0x18f,-1,202}, 2046 {0x190,0x190,-1,203}, {0x191,0x191,-1,1}, {0x193,0x193,-1,205}, 2047 {0x194,0x194,-1,207}, {0x196,0x196,-1,211}, {0x197,0x197,-1,209}, 2048 {0x198,0x198,-1,1}, {0x19c,0x19c,-1,211}, {0x19d,0x19d,-1,213}, 2049 {0x19f,0x19f,-1,214}, {0x1a0,0x1a4,2,1}, {0x1a6,0x1a6,-1,218}, 2050 {0x1a7,0x1a7,-1,1}, {0x1a9,0x1a9,-1,218}, {0x1ac,0x1ac,-1,1}, 2051 {0x1ae,0x1ae,-1,218}, {0x1af,0x1af,-1,1}, {0x1b1,0x1b2,1,217}, 2052 {0x1b3,0x1b5,2,1}, {0x1b7,0x1b7,-1,219}, {0x1b8,0x1bc,4,1}, 2053 {0x1c4,0x1ca,3,2}, {0x1cd,0x1db,2,1}, {0x1de,0x1ee,2,1}, 2054 {0x1f1,0x1f1,-1,2}, {0x1f4,0x1f4,-1,1}, {0x1f6,0x1f6,-1,-97}, 2055 {0x1f7,0x1f7,-1,-56}, {0x1f8,0x21e,2,1}, {0x220,0x220,-1,-130}, 2056 {0x222,0x232,2,1}, {0x386,0x386,-1,38}, {0x388,0x38a,1,37}, 2057 {0x38c,0x38c,-1,64}, {0x38e,0x38f,1,63}, {0x391,0x3a1,1,32}, 2058 {0x3a3,0x3ab,1,32}, {0x3d8,0x3ee,2,1}, {0x3f4,0x3f4,-1,-60}, 2059 {0x3f7,0x3f7,-1,1}, {0x3f9,0x3f9,-1,-7}, {0x3fa,0x3fa,-1,1}, 2060 {0x400,0x40f,1,80}, {0x410,0x42f,1,32}, {0x460,0x480,2,1}, 2061 {0x48a,0x4be,2,1}, {0x4c1,0x4cd,2,1}, {0x4d0,0x4f4,2,1}, 2062 {0x4f8,0x500,8,1}, {0x502,0x50e,2,1}, {0x531,0x556,1,48}, 2063 {0x1e00,0x1e94,2,1}, {0x1ea0,0x1ef8,2,1}, {0x1f08,0x1f0f,1,-8}, 2064 {0x1f18,0x1f1d,1,-8}, {0x1f28,0x1f2f,1,-8}, {0x1f38,0x1f3f,1,-8}, 2065 {0x1f48,0x1f4d,1,-8}, {0x1f59,0x1f5f,2,-8}, {0x1f68,0x1f6f,1,-8}, 2066 {0x1fb8,0x1fb9,1,-8}, {0x1fba,0x1fbb,1,-74}, {0x1fc8,0x1fcb,1,-86}, 2067 {0x1fd8,0x1fd9,1,-8}, {0x1fda,0x1fdb,1,-100}, {0x1fe8,0x1fe9,1,-8}, 2068 {0x1fea,0x1feb,1,-112}, {0x1fec,0x1fec,-1,-7}, {0x1ff8,0x1ff9,1,-128}, 2069 {0x1ffa,0x1ffb,1,-126}, {0x2126,0x2126,-1,-7517}, {0x212a,0x212a,-1,-8383}, 2070 {0x212b,0x212b,-1,-8262}, {0xff21,0xff3a,1,32}, {0x10400,0x10427,1,40} 2071 }; 2072 2073 convertStruct toUpper[] = 2074 { 2075 {0x61,0x7a,1,-32}, {0xb5,0xb5,-1,743}, {0xe0,0xf6,1,-32}, 2076 {0xf8,0xfe,1,-32}, {0xff,0xff,-1,121}, {0x101,0x12f,2,-1}, 2077 {0x131,0x131,-1,-232}, {0x133,0x137,2,-1}, {0x13a,0x148,2,-1}, 2078 {0x14b,0x177,2,-1}, {0x17a,0x17e,2,-1}, {0x17f,0x17f,-1,-300}, 2079 {0x183,0x185,2,-1}, {0x188,0x18c,4,-1}, {0x192,0x192,-1,-1}, 2080 {0x195,0x195,-1,97}, {0x199,0x199,-1,-1}, {0x19e,0x19e,-1,130}, 2081 {0x1a1,0x1a5,2,-1}, {0x1a8,0x1ad,5,-1}, {0x1b0,0x1b4,4,-1}, 2082 {0x1b6,0x1b9,3,-1}, {0x1bd,0x1bd,-1,-1}, {0x1bf,0x1bf,-1,56}, 2083 {0x1c5,0x1c6,1,-1}, {0x1c8,0x1c9,1,-1}, {0x1cb,0x1cc,1,-1}, 2084 {0x1ce,0x1dc,2,-1}, {0x1dd,0x1dd,-1,-79}, {0x1df,0x1ef,2,-1}, 2085 {0x1f2,0x1f3,1,-1}, {0x1f5,0x1f9,4,-1}, {0x1fb,0x21f,2,-1}, 2086 {0x223,0x233,2,-1}, {0x253,0x253,-1,-210}, {0x254,0x254,-1,-206}, 2087 {0x256,0x257,1,-205}, {0x259,0x259,-1,-202}, {0x25b,0x25b,-1,-203}, 2088 {0x260,0x260,-1,-205}, {0x263,0x263,-1,-207}, {0x268,0x268,-1,-209}, 2089 {0x269,0x26f,6,-211}, {0x272,0x272,-1,-213}, {0x275,0x275,-1,-214}, 2090 {0x280,0x283,3,-218}, {0x288,0x288,-1,-218}, {0x28a,0x28b,1,-217}, 2091 {0x292,0x292,-1,-219}, {0x3ac,0x3ac,-1,-38}, {0x3ad,0x3af,1,-37}, 2092 {0x3b1,0x3c1,1,-32}, {0x3c2,0x3c2,-1,-31}, {0x3c3,0x3cb,1,-32}, 2093 {0x3cc,0x3cc,-1,-64}, {0x3cd,0x3ce,1,-63}, {0x3d0,0x3d0,-1,-62}, 2094 {0x3d1,0x3d1,-1,-57}, {0x3d5,0x3d5,-1,-47}, {0x3d6,0x3d6,-1,-54}, 2095 {0x3d9,0x3ef,2,-1}, {0x3f0,0x3f0,-1,-86}, {0x3f1,0x3f1,-1,-80}, 2096 {0x3f2,0x3f2,-1,7}, {0x3f5,0x3f5,-1,-96}, {0x3f8,0x3fb,3,-1}, 2097 {0x430,0x44f,1,-32}, {0x450,0x45f,1,-80}, {0x461,0x481,2,-1}, 2098 {0x48b,0x4bf,2,-1}, {0x4c2,0x4ce,2,-1}, {0x4d1,0x4f5,2,-1}, 2099 {0x4f9,0x501,8,-1}, {0x503,0x50f,2,-1}, {0x561,0x586,1,-48}, 2100 {0x1e01,0x1e95,2,-1}, {0x1e9b,0x1e9b,-1,-59}, {0x1ea1,0x1ef9,2,-1}, 2101 {0x1f00,0x1f07,1,8}, {0x1f10,0x1f15,1,8}, {0x1f20,0x1f27,1,8}, 2102 {0x1f30,0x1f37,1,8}, {0x1f40,0x1f45,1,8}, {0x1f51,0x1f57,2,8}, 2103 {0x1f60,0x1f67,1,8}, {0x1f70,0x1f71,1,74}, {0x1f72,0x1f75,1,86}, 2104 {0x1f76,0x1f77,1,100}, {0x1f78,0x1f79,1,128}, {0x1f7a,0x1f7b,1,112}, 2105 {0x1f7c,0x1f7d,1,126}, {0x1f80,0x1f87,1,8}, {0x1f90,0x1f97,1,8}, 2106 {0x1fa0,0x1fa7,1,8}, {0x1fb0,0x1fb1,1,8}, {0x1fb3,0x1fb3,-1,9}, 2107 {0x1fbe,0x1fbe,-1,-7205}, {0x1fc3,0x1fc3,-1,9}, {0x1fd0,0x1fd1,1,8}, 2108 {0x1fe0,0x1fe1,1,8}, {0x1fe5,0x1fe5,-1,7}, {0x1ff3,0x1ff3,-1,9}, 2109 {0xff41,0xff5a,1,-32}, {0x10428,0x1044f,1,-40} 2110 }; 2111 2112 /* 2113 * Return the upper-case equivalent of "a", which is a UCS-4 character. Use 2114 * simple case folding. 2115 */ 2116 int 2117 utf_toupper(a) 2118 int a; 2119 { 2120 /* If 'casemap' contains "keepascii" use ASCII style toupper(). */ 2121 if (a < 128 && (cmp_flags & CMP_KEEPASCII)) 2122 return TOUPPER_ASC(a); 2123 2124 #if defined(HAVE_TOWUPPER) && defined(__STDC__ISO_10646__) 2125 /* If towupper() is availble and handles Unicode, use it. */ 2126 if (!(cmp_flags & CMP_INTERNAL)) 2127 return towupper(a); 2128 #endif 2129 2130 /* For characters below 128 use locale sensitive toupper(). */ 2131 if (a < 128) 2132 return TOUPPER_LOC(a); 2133 2134 /* For any other characters use the above mapping table. */ 2135 return utf_convert(a, toUpper, sizeof(toUpper)); 2136 } 2137 2138 int 2139 utf_islower(a) 2140 int a; 2141 { 2142 return (utf_toupper(a) != a); 2143 } 2144 2145 /* 2146 * Return the lower-case equivalent of "a", which is a UCS-4 character. Use 2147 * simple case folding. 2148 */ 2149 int 2150 utf_tolower(a) 2151 int a; 2152 { 2153 /* If 'casemap' contains "keepascii" use ASCII style tolower(). */ 2154 if (a < 128 && (cmp_flags & CMP_KEEPASCII)) 2155 return TOLOWER_ASC(a); 2156 2157 #if defined(HAVE_TOWLOWER) && defined(__STDC__ISO_10646__) 2158 /* If towlower() is availble and handles Unicode, use it. */ 2159 if (!(cmp_flags & CMP_INTERNAL)) 2160 return towlower(a); 2161 #endif 2162 2163 /* For characters below 128 use locale sensitive tolower(). */ 2164 if (a < 128) 2165 return TOLOWER_LOC(a); 2166 2167 /* For any other characters use the above mapping table. */ 2168 return utf_convert(a, toLower, sizeof(toLower)); 2169 } 2170 2171 int 2172 utf_isupper(a) 2173 int a; 2174 { 2175 return (utf_tolower(a) != a); 2176 } 2177 2178 /* 2179 * Version of strnicmp() that handles multi-byte characters. 2180 * Needed for Big5, Sjift-JIS and UTF-8 encoding. Other DBCS encodings can 2181 * probably use strnicmp(), because there are no ASCII characters in the 2182 * second byte. 2183 * Returns zero if s1 and s2 are equal (ignoring case), the difference between 2184 * two characters otherwise. 2185 */ 2186 int 2187 mb_strnicmp(s1, s2, n) 2188 char_u *s1, *s2; 2189 int n; 2190 { 2191 int i, j, l; 2192 int cdiff; 2193 2194 for (i = 0; i < n; i += l) 2195 { 2196 if (s1[i] == NUL && s2[i] == NUL) /* both strings end */ 2197 return 0; 2198 if (enc_utf8) 2199 { 2200 l = utf_byte2len(s1[i]); 2201 if (l > n - i) 2202 l = n - i; /* incomplete character */ 2203 /* Check directly first, it's faster. */ 2204 for (j = 0; j < l; ++j) 2205 if (s1[i + j] != s2[i + j]) 2206 break; 2207 if (j < l) 2208 { 2209 /* If one of the two characters is incomplete return -1. */ 2210 if (i + utf_byte2len(s1[i]) > n || i + utf_byte2len(s2[i]) > n) 2211 return -1; 2212 cdiff = utf_fold(utf_ptr2char(s1 + i)) 2213 - utf_fold(utf_ptr2char(s2 + i)); 2214 if (cdiff != 0) 2215 return cdiff; 2216 } 2217 } 2218 else 2219 { 2220 l = (*mb_ptr2len_check)(s1 + i); 2221 if (l <= 1) 2222 { 2223 /* Single byte: first check normally, then with ignore case. */ 2224 if (s1[i] != s2[i]) 2225 { 2226 cdiff = TOLOWER_LOC(s1[i]) - TOLOWER_LOC(s2[i]); 2227 if (cdiff != 0) 2228 return cdiff; 2229 } 2230 } 2231 else 2232 { 2233 /* For non-Unicode multi-byte don't ignore case. */ 2234 if (l > n - i) 2235 l = n - i; 2236 cdiff = STRNCMP(s1 + i, s2 + i, l); 2237 if (cdiff != 0) 2238 return cdiff; 2239 } 2240 } 2241 } 2242 return 0; 2243 } 2244 2245 /* 2246 * "g8": show bytes of the UTF-8 char under the cursor. Doesn't matter what 2247 * 'encoding' has been set to. 2248 */ 2249 void 2250 show_utf8() 2251 { 2252 int len; 2253 char_u *line; 2254 int clen; 2255 int i; 2256 2257 /* Get the byte length of the char under the cursor, including composing 2258 * characters. */ 2259 line = ml_get_cursor(); 2260 len = utfc_ptr2len_check(line); 2261 if (len == 0) 2262 { 2263 MSG("NUL"); 2264 return; 2265 } 2266 2267 IObuff[0] = NUL; 2268 clen = 0; 2269 for (i = 0; i < len; ++i) 2270 { 2271 if (clen == 0) 2272 { 2273 /* start of (composing) character, get its length */ 2274 if (i > 0) 2275 STRCAT(IObuff, "+ "); 2276 clen = utf_ptr2len_check(line + i); 2277 } 2278 sprintf((char *)IObuff + STRLEN(IObuff), "%02x ", line[i]); 2279 --clen; 2280 } 2281 2282 msg(IObuff); 2283 } 2284 2285 /* 2286 * mb_head_off() function pointer. 2287 * Return offset from "p" to the first byte of the character it points into. 2288 * Returns 0 when already at the first byte of a character. 2289 */ 2290 /*ARGSUSED*/ 2291 int 2292 latin_head_off(base, p) 2293 char_u *base; 2294 char_u *p; 2295 { 2296 return 0; 2297 } 2298 2299 int 2300 dbcs_head_off(base, p) 2301 char_u *base; 2302 char_u *p; 2303 { 2304 char_u *q; 2305 2306 /* It can't be a trailing byte when not using DBCS, at the start of the 2307 * string or the previous byte can't start a double-byte. */ 2308 if (p <= base || MB_BYTE2LEN(p[-1]) == 1) 2309 return 0; 2310 2311 /* This is slow: need to start at the base and go forward until the 2312 * byte we are looking for. Return 1 when we went past it, 0 otherwise. */ 2313 q = base; 2314 while (q < p) 2315 q += dbcs_ptr2len_check(q); 2316 return (q == p) ? 0 : 1; 2317 } 2318 2319 #if defined(FEAT_CLIPBOARD) || defined(FEAT_GUI) || defined(FEAT_RIGHTLEFT) \ 2320 || defined(PROTO) 2321 /* 2322 * Special version of dbcs_head_off() that works for ScreenLines[], where 2323 * single-width DBCS_JPNU characters are stored separately. 2324 */ 2325 int 2326 dbcs_screen_head_off(base, p) 2327 char_u *base; 2328 char_u *p; 2329 { 2330 char_u *q; 2331 2332 /* It can't be a trailing byte when not using DBCS, at the start of the 2333 * string or the previous byte can't start a double-byte. 2334 * For euc-jp an 0x8e byte in the previous cell always means we have a 2335 * lead byte in the current cell. */ 2336 if (p <= base 2337 || (enc_dbcs == DBCS_JPNU && p[-1] == 0x8e) 2338 || MB_BYTE2LEN(p[-1]) == 1) 2339 return 0; 2340 2341 /* This is slow: need to start at the base and go forward until the 2342 * byte we are looking for. Return 1 when we went past it, 0 otherwise. 2343 * For DBCS_JPNU look out for 0x8e, which means the second byte is not 2344 * stored as the next byte. */ 2345 q = base; 2346 while (q < p) 2347 { 2348 if (enc_dbcs == DBCS_JPNU && *q == 0x8e) 2349 ++q; 2350 else 2351 q += dbcs_ptr2len_check(q); 2352 } 2353 return (q == p) ? 0 : 1; 2354 } 2355 #endif 2356 2357 int 2358 utf_head_off(base, p) 2359 char_u *base; 2360 char_u *p; 2361 { 2362 char_u *q; 2363 char_u *s; 2364 int c; 2365 #ifdef FEAT_ARABIC 2366 char_u *j; 2367 #endif 2368 2369 if (*p < 0x80) /* be quick for ASCII */ 2370 return 0; 2371 2372 /* Skip backwards over trailing bytes: 10xx.xxxx 2373 * Skip backwards again if on a composing char. */ 2374 for (q = p; ; --q) 2375 { 2376 /* Move s to the last byte of this char. */ 2377 for (s = q; (s[1] & 0xc0) == 0x80; ++s) 2378 ; 2379 /* Move q to the first byte of this char. */ 2380 while (q > base && (*q & 0xc0) == 0x80) 2381 --q; 2382 /* Check for illegal sequence. Do allow an illegal byte after where we 2383 * started. */ 2384 if (utf8len_tab[*q] != (int)(s - q + 1) 2385 && utf8len_tab[*q] != (int)(p - q + 1)) 2386 return 0; 2387 2388 if (q <= base) 2389 break; 2390 2391 c = utf_ptr2char(q); 2392 if (utf_iscomposing(c)) 2393 continue; 2394 2395 #ifdef FEAT_ARABIC 2396 if (arabic_maycombine(c)) 2397 { 2398 /* Advance to get a sneak-peak at the next char */ 2399 j = q; 2400 --j; 2401 /* Move j to the first byte of this char. */ 2402 while (j > base && (*j & 0xc0) == 0x80) 2403 --j; 2404 if (arabic_combine(utf_ptr2char(j), c)) 2405 continue; 2406 } 2407 #endif 2408 break; 2409 } 2410 2411 return (int)(p - q); 2412 } 2413 2414 /* 2415 * Return the offset from "p" to the first byte of a character. When "p" is 2416 * at the start of a character 0 is returned, otherwise the offset to the next 2417 * character. Can start anywhere in a stream of bytes. 2418 */ 2419 int 2420 mb_off_next(base, p) 2421 char_u *base; 2422 char_u *p; 2423 { 2424 int i; 2425 int j; 2426 2427 if (enc_utf8) 2428 { 2429 if (*p < 0x80) /* be quick for ASCII */ 2430 return 0; 2431 2432 /* Find the next character that isn't 10xx.xxxx */ 2433 for (i = 0; (p[i] & 0xc0) == 0x80; ++i) 2434 ; 2435 if (i > 0) 2436 { 2437 /* Check for illegal sequence. */ 2438 for (j = 0; p - j > base; ++j) 2439 if ((p[-j] & 0xc0) != 0x80) 2440 break; 2441 if (utf8len_tab[p[-j]] != i + j) 2442 return 0; 2443 } 2444 return i; 2445 } 2446 2447 /* Only need to check if we're on a trail byte, it doesn't matter if we 2448 * want the offset to the next or current character. */ 2449 return (*mb_head_off)(base, p); 2450 } 2451 2452 /* 2453 * Return the offset from "p" to the last byte of the character it points 2454 * into. Can start anywhere in a stream of bytes. 2455 */ 2456 int 2457 mb_tail_off(base, p) 2458 char_u *base; 2459 char_u *p; 2460 { 2461 int i; 2462 int j; 2463 2464 if (*p == NUL) 2465 return 0; 2466 2467 if (enc_utf8) 2468 { 2469 /* Find the last character that is 10xx.xxxx */ 2470 for (i = 0; (p[i + 1] & 0xc0) == 0x80; ++i) 2471 ; 2472 /* Check for illegal sequence. */ 2473 for (j = 0; p - j > base; ++j) 2474 if ((p[-j] & 0xc0) != 0x80) 2475 break; 2476 if (utf8len_tab[p[-j]] != i + j + 1) 2477 return 0; 2478 return i; 2479 } 2480 2481 /* It can't be the first byte if a double-byte when not using DBCS, at the 2482 * end of the string or the byte can't start a double-byte. */ 2483 if (enc_dbcs == 0 || p[1] == NUL || MB_BYTE2LEN(*p) == 1) 2484 return 0; 2485 2486 /* Return 1 when on the lead byte, 0 when on the tail byte. */ 2487 return 1 - dbcs_head_off(base, p); 2488 } 2489 2490 #if defined(HAVE_GTK2) || defined(PROTO) 2491 /* 2492 * Return TRUE if string "s" is a valid utf-8 string. 2493 * When "end" is NULL stop at the first NUL. 2494 * When "end" is positive stop there. 2495 */ 2496 int 2497 utf_valid_string(s, end) 2498 char_u *s; 2499 char_u *end; 2500 { 2501 int l; 2502 char_u *p = s; 2503 2504 while (end == NULL ? *p != NUL : p < end) 2505 { 2506 if ((*p & 0xc0) == 0x80) 2507 return FALSE; /* invalid lead byte */ 2508 l = utf8len_tab[*p]; 2509 if (end != NULL && p + l > end) 2510 return FALSE; /* incomplete byte sequence */ 2511 ++p; 2512 while (--l > 0) 2513 if ((*p++ & 0xc0) != 0x80) 2514 return FALSE; /* invalid trail byte */ 2515 } 2516 return TRUE; 2517 } 2518 #endif 2519 2520 #if defined(FEAT_GUI) || defined(PROTO) 2521 /* 2522 * Special version of mb_tail_off() for use in ScreenLines[]. 2523 */ 2524 int 2525 dbcs_screen_tail_off(base, p) 2526 char_u *base; 2527 char_u *p; 2528 { 2529 /* It can't be the first byte if a double-byte when not using DBCS, at the 2530 * end of the string or the byte can't start a double-byte. 2531 * For euc-jp an 0x8e byte always means we have a lead byte in the current 2532 * cell. */ 2533 if (*p == NUL || p[1] == NUL 2534 || (enc_dbcs == DBCS_JPNU && *p == 0x8e) 2535 || MB_BYTE2LEN(*p) == 1) 2536 return 0; 2537 2538 /* Return 1 when on the lead byte, 0 when on the tail byte. */ 2539 return 1 - dbcs_screen_head_off(base, p); 2540 } 2541 #endif 2542 2543 /* 2544 * If the cursor moves on an trail byte, set the cursor on the lead byte. 2545 * Thus it moves left if necessary. 2546 * Return TRUE when the cursor was adjusted. 2547 */ 2548 void 2549 mb_adjust_cursor() 2550 { 2551 mb_adjustpos(&curwin->w_cursor); 2552 } 2553 2554 /* 2555 * Adjust position "*lp" to point to the first byte of a multi-byte character. 2556 * If it points to a tail byte it's moved backwards to the head byte. 2557 */ 2558 void 2559 mb_adjustpos(lp) 2560 pos_T *lp; 2561 { 2562 char_u *p; 2563 2564 if (lp->col > 0 2565 #ifdef FEAT_VIRTUALEDIT 2566 || lp->coladd > 1 2567 #endif 2568 ) 2569 { 2570 p = ml_get(lp->lnum); 2571 lp->col -= (*mb_head_off)(p, p + lp->col); 2572 #ifdef FEAT_VIRTUALEDIT 2573 /* Reset "coladd" when the cursor would be on the right half of a 2574 * double-wide character. */ 2575 if (lp->coladd == 1 2576 && p[lp->col] != TAB 2577 && vim_isprintc((*mb_ptr2char)(p + lp->col)) 2578 && ptr2cells(p + lp->col) > 1) 2579 lp->coladd = 0; 2580 #endif 2581 } 2582 } 2583 2584 /* 2585 * Return a pointer to the character before "*p", if there is one. 2586 */ 2587 char_u * 2588 mb_prevptr(line, p) 2589 char_u *line; /* start of the string */ 2590 char_u *p; 2591 { 2592 if (p > line) 2593 p = p - (*mb_head_off)(line, p - 1) - 1; 2594 return p; 2595 } 2596 2597 /* 2598 * Return the character length of "str". multi-byte characters counts as one. 2599 */ 2600 int 2601 mb_charlen(str) 2602 char_u *str; 2603 { 2604 int count; 2605 2606 if (str == NULL) 2607 return 0; 2608 2609 for (count = 0; *str != NUL; count++) 2610 str += (*mb_ptr2len_check)(str); 2611 2612 return count; 2613 } 2614 2615 /* 2616 * Try to un-escape a multi-byte character. 2617 * Used for the "to" and "from" part of a mapping. 2618 * Return the un-escaped string if it is a multi-byte character, and advance 2619 * "pp" to just after the bytes that formed it. 2620 * Return NULL if no multi-byte char was found. 2621 */ 2622 char_u * 2623 mb_unescape(pp) 2624 char_u **pp; 2625 { 2626 static char_u buf[MB_MAXBYTES + 1]; 2627 int n, m = 0; 2628 char_u *str = *pp; 2629 2630 /* Must translate K_SPECIAL KS_SPECIAL KE_FILLER to K_SPECIAL and CSI 2631 * KS_EXTRA KE_CSI to CSI. */ 2632 for (n = 0; str[n] != NUL && m <= MB_MAXBYTES; ++n) 2633 { 2634 if (str[n] == K_SPECIAL 2635 && str[n + 1] == KS_SPECIAL 2636 && str[n + 2] == KE_FILLER) 2637 { 2638 buf[m++] = K_SPECIAL; 2639 n += 2; 2640 } 2641 # ifdef FEAT_GUI 2642 else if (str[n] == CSI 2643 && str[n + 1] == KS_EXTRA 2644 && str[n + 2] == (int)KE_CSI) 2645 { 2646 buf[m++] = CSI; 2647 n += 2; 2648 } 2649 # endif 2650 else if (str[n] == K_SPECIAL 2651 # ifdef FEAT_GUI 2652 || str[n] == CSI 2653 # endif 2654 ) 2655 break; /* a special key can't be a multibyte char */ 2656 else 2657 buf[m++] = str[n]; 2658 buf[m] = NUL; 2659 2660 /* Return a multi-byte character if it's found. An illegal sequence 2661 * will result in a 1 here. */ 2662 if ((*mb_ptr2len_check)(buf) > 1) 2663 { 2664 *pp = str + n + 1; 2665 return buf; 2666 } 2667 } 2668 return NULL; 2669 } 2670 2671 /* 2672 * Return TRUE if the character at "row"/"col" on the screen is the left side 2673 * of a double-width character. 2674 * Caller must make sure "row" and "col" are not invalid! 2675 */ 2676 int 2677 mb_lefthalve(row, col) 2678 int row; 2679 int col; 2680 { 2681 #ifdef FEAT_HANGULIN 2682 if (composing_hangul) 2683 return TRUE; 2684 #endif 2685 if (enc_dbcs != 0) 2686 return dbcs_off2cells(LineOffset[row] + col) > 1; 2687 if (enc_utf8) 2688 return (col + 1 < Columns 2689 && ScreenLines[LineOffset[row] + col + 1] == 0); 2690 return FALSE; 2691 } 2692 2693 # if defined(FEAT_CLIPBOARD) || defined(FEAT_GUI) || defined(FEAT_RIGHTLEFT) \ 2694 || defined(PROTO) 2695 /* 2696 * Correct a position on the screen, if it's the right halve of a double-wide 2697 * char move it to the left halve. Returns the corrected column. 2698 */ 2699 int 2700 mb_fix_col(col, row) 2701 int col; 2702 int row; 2703 { 2704 col = check_col(col); 2705 row = check_row(row); 2706 if (has_mbyte && ScreenLines != NULL && col > 0 2707 && ((enc_dbcs 2708 && ScreenLines[LineOffset[row] + col] != NUL 2709 && dbcs_screen_head_off(ScreenLines + LineOffset[row], 2710 ScreenLines + LineOffset[row] + col)) 2711 || (enc_utf8 && ScreenLines[LineOffset[row] + col] == 0))) 2712 --col; 2713 return col; 2714 } 2715 # endif 2716 #endif 2717 2718 #if defined(FEAT_MBYTE) || defined(FEAT_POSTSCRIPT) || defined(PROTO) 2719 static int enc_alias_search __ARGS((char_u *name)); 2720 2721 /* 2722 * Skip the Vim specific head of a 'encoding' name. 2723 */ 2724 char_u * 2725 enc_skip(p) 2726 char_u *p; 2727 { 2728 if (STRNCMP(p, "2byte-", 6) == 0) 2729 return p + 6; 2730 if (STRNCMP(p, "8bit-", 5) == 0) 2731 return p + 5; 2732 return p; 2733 } 2734 2735 /* 2736 * Find the canonical name for encoding "enc". 2737 * When the name isn't recognized, returns "enc" itself, but with all lower 2738 * case characters and '_' replaced with '-'. 2739 * Returns an allocated string. NULL for out-of-memory. 2740 */ 2741 char_u * 2742 enc_canonize(enc) 2743 char_u *enc; 2744 { 2745 char_u *r; 2746 char_u *p, *s; 2747 int i; 2748 2749 /* copy "enc" to allocted memory, with room for two '-' */ 2750 r = alloc((unsigned)(STRLEN(enc) + 3)); 2751 if (r != NULL) 2752 { 2753 /* Make it all lower case and replace '_' with '-'. */ 2754 p = r; 2755 for (s = enc; *s != NUL; ++s) 2756 { 2757 if (*s == '_') 2758 *p++ = '-'; 2759 else 2760 *p++ = TOLOWER_ASC(*s); 2761 } 2762 *p = NUL; 2763 2764 /* Skip "2byte-" and "8bit-". */ 2765 p = enc_skip(r); 2766 2767 /* "iso8859" -> "iso-8859" */ 2768 if (STRNCMP(p, "iso8859", 7) == 0) 2769 { 2770 mch_memmove(p + 4, p + 3, STRLEN(p + 2)); 2771 p[3] = '-'; 2772 } 2773 2774 /* "iso-8859n" -> "iso-8859-n" */ 2775 if (STRNCMP(p, "iso-8859", 8) == 0 && p[8] != '-') 2776 { 2777 mch_memmove(p + 9, p + 8, STRLEN(p + 7)); 2778 p[8] = '-'; 2779 } 2780 2781 /* "latin-N" -> "latinN" */ 2782 if (STRNCMP(p, "latin-", 6) == 0) 2783 mch_memmove(p + 5, p + 6, STRLEN(p + 5)); 2784 2785 if (enc_canon_search(p) >= 0) 2786 { 2787 /* canonical name can be used unmodified */ 2788 if (p != r) 2789 mch_memmove(r, p, STRLEN(p) + 1); 2790 } 2791 else if ((i = enc_alias_search(p)) >= 0) 2792 { 2793 /* alias recognized, get canonical name */ 2794 vim_free(r); 2795 r = vim_strsave((char_u *)enc_canon_table[i].name); 2796 } 2797 } 2798 return r; 2799 } 2800 2801 /* 2802 * Search for an encoding alias of "name". 2803 * Returns -1 when not found. 2804 */ 2805 static int 2806 enc_alias_search(name) 2807 char_u *name; 2808 { 2809 int i; 2810 2811 for (i = 0; enc_alias_table[i].name != NULL; ++i) 2812 if (STRCMP(name, enc_alias_table[i].name) == 0) 2813 return enc_alias_table[i].canon; 2814 return -1; 2815 } 2816 #endif 2817 2818 #if defined(FEAT_MBYTE) || defined(PROTO) 2819 2820 #ifdef HAVE_LANGINFO_H 2821 # include <langinfo.h> 2822 #endif 2823 2824 /* 2825 * Get the canonicalized encoding of the current locale. 2826 * Returns an allocated string when successful, NULL when not. 2827 */ 2828 char_u * 2829 enc_locale() 2830 { 2831 #ifndef WIN3264 2832 char *s; 2833 char *p; 2834 int i; 2835 #endif 2836 char buf[50]; 2837 #ifdef WIN3264 2838 long acp = GetACP(); 2839 2840 if (acp == 1200) 2841 STRCPY(buf, "ucs-2le"); 2842 else if (acp == 1252) 2843 STRCPY(buf, "latin1"); 2844 else 2845 sprintf(buf, "cp%ld", acp); 2846 #else 2847 # ifdef HAVE_NL_LANGINFO_CODESET 2848 if ((s = nl_langinfo(CODESET)) == NULL || *s == NUL) 2849 # endif 2850 # if defined(HAVE_LOCALE_H) || defined(X_LOCALE) 2851 if ((s = setlocale(LC_CTYPE, NULL)) == NULL || *s == NUL) 2852 # endif 2853 if ((s = getenv("LC_ALL")) == NULL || *s == NUL) 2854 if ((s = getenv("LC_CTYPE")) == NULL || *s == NUL) 2855 s = getenv("LANG"); 2856 2857 if (s == NULL || *s == NUL) 2858 return FAIL; 2859 2860 /* The most generic locale format is: 2861 * language[_territory][.codeset][@modifier][+special][,[sponsor][_revision]] 2862 * If there is a '.' remove the part before it. 2863 * if there is something after the codeset, remove it. 2864 * Make the name lowercase and replace '_' with '-'. 2865 * Exception: "ja_JP.EUC" == "euc-jp", "zh_CN.EUC" = "euc-cn", 2866 * "ko_KR.EUC" == "euc-kr" 2867 */ 2868 if ((p = (char *)vim_strchr((char_u *)s, '.')) != NULL) 2869 { 2870 if (p > s + 2 && STRNICMP(p + 1, "EUC", 3) == 0 2871 && !isalnum((int)p[4]) && p[4] != '-' && p[-3] == '_') 2872 { 2873 /* copy "XY.EUC" to "euc-XY" to buf[10] */ 2874 STRCPY(buf + 10, "euc-"); 2875 buf[14] = p[-2]; 2876 buf[15] = p[-1]; 2877 buf[16] = 0; 2878 s = buf + 10; 2879 } 2880 else 2881 s = p + 1; 2882 } 2883 for (i = 0; s[i] != NUL && i < sizeof(buf) - 1; ++i) 2884 { 2885 if (s[i] == '_' || s[i] == '-') 2886 buf[i] = '-'; 2887 else if (isalnum((int)s[i])) 2888 buf[i] = TOLOWER_ASC(s[i]); 2889 else 2890 break; 2891 } 2892 buf[i] = NUL; 2893 #endif 2894 2895 return enc_canonize((char_u *)buf); 2896 } 2897 2898 #if defined(WIN3264) || defined(PROTO) 2899 /* 2900 * Convert an encoding name to an MS-Windows codepage. 2901 * Returns zero if no codepage can be figured out. 2902 */ 2903 int 2904 encname2codepage(name) 2905 char_u *name; 2906 { 2907 int cp; 2908 char_u *p = name; 2909 int idx; 2910 2911 if (STRNCMP(p, "8bit-", 5) == 0) 2912 p += 5; 2913 else if (STRNCMP(p_enc, "2byte-", 6) == 0) 2914 p += 6; 2915 2916 if (p[0] == 'c' && p[1] == 'p') 2917 cp = atoi(p + 2); 2918 else if ((idx = enc_canon_search(p)) >= 0) 2919 cp = enc_canon_table[idx].codepage; 2920 else 2921 return 0; 2922 if (IsValidCodePage(cp)) 2923 return cp; 2924 return 0; 2925 } 2926 #endif 2927 2928 # if defined(USE_ICONV) || defined(PROTO) 2929 2930 static char_u *iconv_string __ARGS((vimconv_T *vcp, char_u *str, int slen, int *unconvlenp)); 2931 2932 /* 2933 * Call iconv_open() with a check if iconv() works properly (there are broken 2934 * versions). 2935 * Returns (void *)-1 if failed. 2936 * (should return iconv_t, but that causes problems with prototypes). 2937 */ 2938 void * 2939 my_iconv_open(to, from) 2940 char_u *to; 2941 char_u *from; 2942 { 2943 iconv_t fd; 2944 #define ICONV_TESTLEN 400 2945 char_u tobuf[ICONV_TESTLEN]; 2946 char *p; 2947 size_t tolen; 2948 static int iconv_ok = -1; 2949 2950 if (iconv_ok == FALSE) 2951 return (void *)-1; /* detected a broken iconv() previously */ 2952 2953 #ifdef DYNAMIC_ICONV 2954 /* Check if the iconv.dll can be found. */ 2955 if (!iconv_enabled(TRUE)) 2956 return (void *)-1; 2957 #endif 2958 2959 fd = iconv_open((char *)enc_skip(to), (char *)enc_skip(from)); 2960 2961 if (fd != (iconv_t)-1 && iconv_ok == -1) 2962 { 2963 /* 2964 * Do a dummy iconv() call to check if it actually works. There is a 2965 * version of iconv() on Linux that is broken. We can't ignore it, 2966 * because it's wide-spread. The symptoms are that after outputting 2967 * the initial shift state the "to" pointer is NULL and conversion 2968 * stops for no apparent reason after about 8160 characters. 2969 */ 2970 p = (char *)tobuf; 2971 tolen = ICONV_TESTLEN; 2972 (void)iconv(fd, NULL, NULL, &p, &tolen); 2973 if (p == NULL) 2974 { 2975 iconv_ok = FALSE; 2976 iconv_close(fd); 2977 fd = (iconv_t)-1; 2978 } 2979 else 2980 iconv_ok = TRUE; 2981 } 2982 2983 return (void *)fd; 2984 } 2985 2986 /* 2987 * Convert the string "str[slen]" with iconv(). 2988 * If "unconvlenp" is not NULL handle the string ending in an incomplete 2989 * sequence and set "*unconvlenp" to the length of it. 2990 * Returns the converted string in allocated memory. NULL for an error. 2991 */ 2992 static char_u * 2993 iconv_string(vcp, str, slen, unconvlenp) 2994 vimconv_T *vcp; 2995 char_u *str; 2996 int slen; 2997 int *unconvlenp; 2998 { 2999 const char *from; 3000 size_t fromlen; 3001 char *to; 3002 size_t tolen; 3003 size_t len = 0; 3004 size_t done = 0; 3005 char_u *result = NULL; 3006 char_u *p; 3007 int l; 3008 3009 from = (char *)str; 3010 fromlen = slen; 3011 for (;;) 3012 { 3013 if (len == 0 || ICONV_ERRNO == ICONV_E2BIG) 3014 { 3015 /* Allocate enough room for most conversions. When re-allocating 3016 * increase the buffer size. */ 3017 len = len + fromlen * 2 + 40; 3018 p = alloc((unsigned)len); 3019 if (p != NULL && done > 0) 3020 mch_memmove(p, result, done); 3021 vim_free(result); 3022 result = p; 3023 if (result == NULL) /* out of memory */ 3024 break; 3025 } 3026 3027 to = (char *)result + done; 3028 tolen = len - done - 2; 3029 /* Avoid a warning for systems with a wrong iconv() prototype by 3030 * casting the second argument to void *. */ 3031 if (iconv(vcp->vc_fd, (void *)&from, &fromlen, &to, &tolen) 3032 != (size_t)-1) 3033 { 3034 /* Finished, append a NUL. */ 3035 *to = NUL; 3036 break; 3037 } 3038 3039 /* Check both ICONV_EINVAL and EINVAL, because the dynamically loaded 3040 * iconv library may use one of them. */ 3041 if (!vcp->vc_fail && unconvlenp != NULL 3042 && (ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL)) 3043 { 3044 /* Handle an incomplete sequence at the end. */ 3045 *to = NUL; 3046 *unconvlenp = fromlen; 3047 break; 3048 } 3049 3050 /* Check both ICONV_EILSEQ and EILSEQ, because the dynamically loaded 3051 * iconv library may use one of them. */ 3052 else if (!vcp->vc_fail 3053 && (ICONV_ERRNO == ICONV_EILSEQ || ICONV_ERRNO == EILSEQ 3054 || ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL)) 3055 { 3056 /* Can't convert: insert a '?' and skip a character. This assumes 3057 * conversion from 'encoding' to something else. In other 3058 * situations we don't know what to skip anyway. */ 3059 *to++ = '?'; 3060 if ((*mb_ptr2cells)((char_u *)from) > 1) 3061 *to++ = '?'; 3062 l = (*mb_ptr2len_check)((char_u *)from); 3063 from += l; 3064 fromlen -= l; 3065 } 3066 else if (ICONV_ERRNO != ICONV_E2BIG) 3067 { 3068 /* conversion failed */ 3069 vim_free(result); 3070 result = NULL; 3071 break; 3072 } 3073 /* Not enough room or skipping illegal sequence. */ 3074 done = to - (char *)result; 3075 } 3076 return result; 3077 } 3078 3079 # if defined(DYNAMIC_ICONV) || defined(PROTO) 3080 /* 3081 * Dynamically load the "iconv.dll" on Win32. 3082 */ 3083 3084 #ifndef DYNAMIC_ICONV /* just generating prototypes */ 3085 # define HINSTANCE int 3086 #endif 3087 HINSTANCE hIconvDLL = 0; 3088 HINSTANCE hMsvcrtDLL = 0; 3089 3090 # ifndef DYNAMIC_ICONV_DLL 3091 # define DYNAMIC_ICONV_DLL "iconv.dll" 3092 # define DYNAMIC_ICONV_DLL_ALT "libiconv.dll" 3093 # endif 3094 # ifndef DYNAMIC_MSVCRT_DLL 3095 # define DYNAMIC_MSVCRT_DLL "msvcrt.dll" 3096 # endif 3097 3098 /* 3099 * Try opening the iconv.dll and return TRUE if iconv() can be used. 3100 */ 3101 int 3102 iconv_enabled(verbose) 3103 int verbose; 3104 { 3105 if (hIconvDLL != 0 && hMsvcrtDLL != 0) 3106 return TRUE; 3107 hIconvDLL = LoadLibrary(DYNAMIC_ICONV_DLL); 3108 if (hIconvDLL == 0) /* sometimes it's called libiconv.dll */ 3109 hIconvDLL = LoadLibrary(DYNAMIC_ICONV_DLL_ALT); 3110 if (hIconvDLL != 0) 3111 hMsvcrtDLL = LoadLibrary(DYNAMIC_MSVCRT_DLL); 3112 if (hIconvDLL == 0 || hMsvcrtDLL == 0) 3113 { 3114 /* Only give the message when 'verbose' is set, otherwise it might be 3115 * done whenever a conversion is attempted. */ 3116 if (verbose && p_verbose > 0) 3117 EMSG2(_(e_loadlib), 3118 hIconvDLL == 0 ? DYNAMIC_ICONV_DLL : DYNAMIC_MSVCRT_DLL); 3119 iconv_end(); 3120 return FALSE; 3121 } 3122 3123 *((FARPROC*)&iconv) = GetProcAddress(hIconvDLL, "libiconv"); 3124 *((FARPROC*)&iconv_open) = GetProcAddress(hIconvDLL, "libiconv_open"); 3125 *((FARPROC*)&iconv_close) = GetProcAddress(hIconvDLL, "libiconv_close"); 3126 *((FARPROC*)&iconvctl) = GetProcAddress(hIconvDLL, "libiconvctl"); 3127 *((FARPROC*)&iconv_errno) = GetProcAddress(hMsvcrtDLL, "_errno"); 3128 if (iconv == NULL || iconv_open == NULL || iconv_close == NULL 3129 || iconvctl == NULL || iconv_errno == NULL) 3130 { 3131 iconv_end(); 3132 if (verbose && p_verbose > 0) 3133 EMSG2(_(e_loadfunc), "for libiconv"); 3134 return FALSE; 3135 } 3136 return TRUE; 3137 } 3138 3139 void 3140 iconv_end() 3141 { 3142 /* Don't use iconv() when inputting or outputting characters. */ 3143 if (input_conv.vc_type == CONV_ICONV) 3144 convert_setup(&input_conv, NULL, NULL); 3145 if (output_conv.vc_type == CONV_ICONV) 3146 convert_setup(&output_conv, NULL, NULL); 3147 3148 if (hIconvDLL != 0) 3149 FreeLibrary(hIconvDLL); 3150 if (hMsvcrtDLL != 0) 3151 FreeLibrary(hMsvcrtDLL); 3152 hIconvDLL = 0; 3153 hMsvcrtDLL = 0; 3154 } 3155 # endif /* DYNAMIC_ICONV */ 3156 # endif /* USE_ICONV */ 3157 3158 #endif /* FEAT_MBYTE */ 3159 3160 #if defined(FEAT_XIM) || defined(PROTO) 3161 3162 # ifdef FEAT_GUI_GTK 3163 static int xim_has_preediting INIT(= FALSE); /* IM current status */ 3164 3165 /* 3166 * Set preedit_start_col to the current cursor position. 3167 */ 3168 static void 3169 init_preedit_start_col(void) 3170 { 3171 if (State & CMDLINE) 3172 preedit_start_col = cmdline_getvcol_cursor(); 3173 else if (curwin != NULL) 3174 getvcol(curwin, &curwin->w_cursor, &preedit_start_col, NULL, NULL); 3175 /* Prevent that preediting marks the buffer as changed. */ 3176 xim_changed_while_preediting = curbuf->b_changed; 3177 } 3178 # endif 3179 3180 # if defined(HAVE_GTK2) && !defined(PROTO) 3181 3182 static int im_is_active = FALSE; /* IM is enabled for current mode */ 3183 static int im_preedit_cursor = 0; /* cursor offset in characters */ 3184 static int im_preedit_trailing = 0; /* number of characters after cursor */ 3185 3186 static unsigned long im_commit_handler_id = 0; 3187 static unsigned int im_activatekey_keyval = GDK_VoidSymbol; 3188 static unsigned int im_activatekey_state = 0; 3189 3190 void 3191 im_set_active(int active) 3192 { 3193 int was_active; 3194 3195 was_active = !!im_is_active; 3196 im_is_active = (active && !p_imdisable); 3197 3198 if (im_is_active != was_active) 3199 xim_reset(); 3200 } 3201 3202 void 3203 xim_set_focus(int focus) 3204 { 3205 if (xic != NULL) 3206 { 3207 if (focus) 3208 gtk_im_context_focus_in(xic); 3209 else 3210 gtk_im_context_focus_out(xic); 3211 } 3212 } 3213 3214 void 3215 im_set_position(int row, int col) 3216 { 3217 if (xic != NULL) 3218 { 3219 GdkRectangle area; 3220 3221 area.x = FILL_X(col); 3222 area.y = FILL_Y(row); 3223 area.width = gui.char_width * (mb_lefthalve(row, col) ? 2 : 1); 3224 area.height = gui.char_height; 3225 3226 gtk_im_context_set_cursor_location(xic, &area); 3227 } 3228 } 3229 3230 # if 0 || defined(PROTO) /* apparently only used in gui_x11.c */ 3231 void 3232 xim_set_preedit(void) 3233 { 3234 im_set_position(gui.row, gui.col); 3235 } 3236 # endif 3237 3238 static void 3239 im_add_to_input(char_u *str, int len) 3240 { 3241 /* Convert from 'termencoding' (always "utf-8") to 'encoding' */ 3242 if (input_conv.vc_type != CONV_NONE) 3243 { 3244 str = string_convert(&input_conv, str, &len); 3245 g_return_if_fail(str != NULL); 3246 } 3247 3248 add_to_input_buf_csi(str, len); 3249 3250 if (input_conv.vc_type != CONV_NONE) 3251 vim_free(str); 3252 3253 if (p_mh) /* blank out the pointer if necessary */ 3254 gui_mch_mousehide(TRUE); 3255 } 3256 3257 static void 3258 im_delete_preedit(void) 3259 { 3260 char_u bskey[] = {CSI, 'k', 'b'}; 3261 char_u delkey[] = {CSI, 'k', 'D'}; 3262 3263 if (State & NORMAL) 3264 { 3265 im_preedit_cursor = 0; 3266 return; 3267 } 3268 for (; im_preedit_cursor > 0; --im_preedit_cursor) 3269 add_to_input_buf(bskey, (int)sizeof(bskey)); 3270 3271 for (; im_preedit_trailing > 0; --im_preedit_trailing) 3272 add_to_input_buf(delkey, (int)sizeof(delkey)); 3273 } 3274 3275 static void 3276 im_correct_cursor(int num_move_back) 3277 { 3278 char_u backkey[] = {CSI, 'k', 'l'}; 3279 3280 if (State & NORMAL) 3281 return; 3282 # ifdef FEAT_RIGHTLEFT 3283 if ((State & CMDLINE) == 0 && curwin != NULL && curwin->w_p_rl) 3284 backkey[2] = 'r'; 3285 # endif 3286 for (; num_move_back > 0; --num_move_back) 3287 add_to_input_buf(backkey, (int)sizeof(backkey)); 3288 } 3289 3290 static int xim_expected_char = NUL; 3291 static int xim_ignored_char = FALSE; 3292 3293 /* 3294 * Update the mode and cursor while in an IM callback. 3295 */ 3296 static void 3297 im_show_info(void) 3298 { 3299 int old_vgetc_busy; 3300 old_vgetc_busy = vgetc_busy; 3301 vgetc_busy = TRUE; 3302 showmode(); 3303 vgetc_busy = old_vgetc_busy; 3304 setcursor(); 3305 out_flush(); 3306 } 3307 3308 /* 3309 * Callback invoked when the user finished preediting. 3310 * Put the final string into the input buffer. 3311 */ 3312 /*ARGSUSED0*/ 3313 static void 3314 im_commit_cb(GtkIMContext *context, const gchar *str, gpointer data) 3315 { 3316 int slen = (int)STRLEN(str); 3317 int add_to_input = TRUE; 3318 int clen; 3319 int len = slen; 3320 int commit_with_preedit = TRUE; 3321 char_u *im_str, *p; 3322 3323 #ifdef XIM_DEBUG 3324 xim_log("im_commit_cb(): %s\n", str); 3325 #endif 3326 3327 /* The imhangul module doesn't reset the preedit string before 3328 * committing. Call im_delete_preedit() to work around that. */ 3329 im_delete_preedit(); 3330 3331 /* Indicate that preediting has finished. */ 3332 if (preedit_start_col == MAXCOL) 3333 { 3334 init_preedit_start_col(); 3335 commit_with_preedit = FALSE; 3336 } 3337 3338 /* The thing which setting "preedit_start_col" to MAXCOL means that 3339 * "preedit_start_col" will be set forcely when calling 3340 * preedit_changed_cb() next time. 3341 * "preedit_start_col" should not reset with MAXCOL on this part. Vim 3342 * is simulating the preediting by using add_to_input_str(). when 3343 * preedit begin immediately before committed, the typebuf is not 3344 * flushed to screen, then it can't get correct "preedit_start_col". 3345 * Thus, it should calculate the cells by adding cells of the committed 3346 * string. */ 3347 if (input_conv.vc_type != CONV_NONE) 3348 { 3349 im_str = string_convert(&input_conv, (char_u *)str, &len); 3350 g_return_if_fail(im_str != NULL); 3351 } 3352 else 3353 im_str = (char_u *)str; 3354 clen = 0; 3355 for (p = im_str; p < im_str + len; p += (*mb_ptr2len_check)(p)) 3356 clen += (*mb_ptr2cells)(p); 3357 if (input_conv.vc_type != CONV_NONE) 3358 vim_free(im_str); 3359 preedit_start_col += clen; 3360 3361 /* Is this a single character that matches a keypad key that's just 3362 * been pressed? If so, we don't want it to be entered as such - let 3363 * us carry on processing the raw keycode so that it may be used in 3364 * mappings as <kSomething>. */ 3365 if (xim_expected_char != NUL) 3366 { 3367 /* We're currently processing a keypad or other special key */ 3368 if (slen == 1 && str[0] == xim_expected_char) 3369 { 3370 /* It's a match - don't do it here */ 3371 xim_ignored_char = TRUE; 3372 add_to_input = FALSE; 3373 } 3374 else 3375 { 3376 /* Not a match */ 3377 xim_ignored_char = FALSE; 3378 } 3379 } 3380 3381 if (add_to_input) 3382 im_add_to_input((char_u *)str, slen); 3383 3384 /* Inserting chars while "im_is_active" is set does not cause a change of 3385 * buffer. When the chars are committed the buffer must be marked as 3386 * changed. */ 3387 if (!commit_with_preedit) 3388 preedit_start_col = MAXCOL; 3389 3390 /* This flag is used in changed() at next call. */ 3391 xim_changed_while_preediting = TRUE; 3392 3393 if (gtk_main_level() > 0) 3394 gtk_main_quit(); 3395 } 3396 3397 /* 3398 * Callback invoked after start to the preedit. 3399 */ 3400 /*ARGSUSED*/ 3401 static void 3402 im_preedit_start_cb(GtkIMContext *context, gpointer data) 3403 { 3404 #ifdef XIM_DEBUG 3405 xim_log("im_preedit_start_cb()\n"); 3406 #endif 3407 3408 im_is_active = TRUE; 3409 gui_update_cursor(TRUE, FALSE); 3410 } 3411 3412 /* 3413 * Callback invoked after end to the preedit. 3414 */ 3415 /*ARGSUSED*/ 3416 static void 3417 im_preedit_end_cb(GtkIMContext *context, gpointer data) 3418 { 3419 #ifdef XIM_DEBUG 3420 xim_log("im_preedit_end_cb()\n"); 3421 #endif 3422 im_delete_preedit(); 3423 3424 /* Indicate that preediting has finished */ 3425 preedit_start_col = MAXCOL; 3426 xim_has_preediting = FALSE; 3427 3428 im_is_active = FALSE; 3429 gui_update_cursor(TRUE, FALSE); 3430 im_show_info(); 3431 } 3432 3433 /* 3434 * Callback invoked after changes to the preedit string. If the preedit 3435 * string was empty before, remember the preedit start column so we know 3436 * where to apply feedback attributes. Delete the previous preedit string 3437 * if there was one, save the new preedit cursor offset, and put the new 3438 * string into the input buffer. 3439 * 3440 * TODO: The pragmatic "put into input buffer" approach used here has 3441 * several fundamental problems: 3442 * 3443 * - The characters in the preedit string are subject to remapping. 3444 * That's broken, only the finally committed string should be remapped. 3445 * 3446 * - There is a race condition involved: The retrieved value for the 3447 * current cursor position will be wrong if any unprocessed characters 3448 * are still queued in the input buffer. 3449 * 3450 * - Due to the lack of synchronization between the file buffer in memory 3451 * and any typed characters, it's practically impossible to implement the 3452 * "retrieve_surrounding" and "delete_surrounding" signals reliably. IM 3453 * modules for languages such as Thai are likely to rely on this feature 3454 * for proper operation. 3455 * 3456 * Conclusions: I think support for preediting needs to be moved to the 3457 * core parts of Vim. Ideally, until it has been committed, the preediting 3458 * string should only be displayed and not affect the buffer content at all. 3459 * The question how to deal with the synchronization issue still remains. 3460 * Circumventing the input buffer is probably not desirable. Anyway, I think 3461 * implementing "retrieve_surrounding" is the only hard problem. 3462 * 3463 * One way to solve all of this in a clean manner would be to queue all key 3464 * press/release events "as is" in the input buffer, and apply the IM filtering 3465 * at the receiving end of the queue. This, however, would have a rather large 3466 * impact on the code base. If there is an easy way to force processing of all 3467 * remaining input from within the "retrieve_surrounding" signal handler, this 3468 * might not be necessary. Gotta ask on vim-dev for opinions. 3469 */ 3470 /*ARGSUSED1*/ 3471 static void 3472 im_preedit_changed_cb(GtkIMContext *context, gpointer data) 3473 { 3474 char *preedit_string = NULL; 3475 int cursor_index = 0; 3476 int num_move_back = 0; 3477 char_u *str; 3478 char_u *p; 3479 int i; 3480 3481 gtk_im_context_get_preedit_string(context, 3482 &preedit_string, NULL, 3483 &cursor_index); 3484 3485 #ifdef XIM_DEBUG 3486 xim_log("im_preedit_changed_cb(): %s\n", preedit_string); 3487 #endif 3488 3489 g_return_if_fail(preedit_string != NULL); /* just in case */ 3490 3491 /* If preedit_start_col is MAXCOL set it to the current cursor position. */ 3492 if (preedit_start_col == MAXCOL && preedit_string[0] != '\0') 3493 { 3494 xim_has_preediting = TRUE; 3495 3496 /* Urgh, this breaks if the input buffer isn't empty now */ 3497 init_preedit_start_col(); 3498 } 3499 else if (cursor_index == 0 && preedit_string[0] == '\0') 3500 { 3501 if (preedit_start_col == MAXCOL) 3502 xim_has_preediting = FALSE; 3503 3504 /* If at the start position (after typing backspace) 3505 * preedit_start_col must be reset. */ 3506 preedit_start_col = MAXCOL; 3507 } 3508 3509 im_delete_preedit(); 3510 3511 /* 3512 * Compute the end of the preediting area: "preedit_end_col". 3513 * According to the documentation of gtk_im_context_get_preedit_string(), 3514 * the cursor_pos output argument returns the offset in bytes. This is 3515 * unfortunately not true -- real life shows the offset is in characters, 3516 * and the GTK+ source code agrees with me. Will file a bug later. 3517 */ 3518 if (preedit_start_col != MAXCOL) 3519 preedit_end_col = preedit_start_col; 3520 str = (char_u *)preedit_string; 3521 for (p = str, i = 0; *p != NUL; p += utf_byte2len(*p), ++i) 3522 { 3523 int is_composing; 3524 3525 is_composing = ((*p & 0x80) != 0 && utf_iscomposing(utf_ptr2char(p))); 3526 /* 3527 * These offsets are used as counters when generating <BS> and <Del> 3528 * to delete the preedit string. So don't count composing characters 3529 * unless 'delcombine' is enabled. 3530 */ 3531 if (!is_composing || p_deco) 3532 { 3533 if (i < cursor_index) 3534 ++im_preedit_cursor; 3535 else 3536 ++im_preedit_trailing; 3537 } 3538 if (!is_composing && i >= cursor_index) 3539 { 3540 /* This is essentially the same as im_preedit_trailing, except 3541 * composing characters are not counted even if p_deco is set. */ 3542 ++num_move_back; 3543 } 3544 if (preedit_start_col != MAXCOL) 3545 preedit_end_col += utf_ptr2cells(p); 3546 } 3547 3548 if (p > str) 3549 { 3550 im_add_to_input(str, (int)(p - str)); 3551 im_correct_cursor(num_move_back); 3552 } 3553 3554 g_free(preedit_string); 3555 3556 if (gtk_main_level() > 0) 3557 gtk_main_quit(); 3558 } 3559 3560 /* 3561 * Translate the Pango attributes at iter to Vim highlighting attributes. 3562 * Ignore attributes not supported by Vim highlighting. This shouldn't have 3563 * too much impact -- right now we handle even more attributes than necessary 3564 * for the IM modules I tested with. 3565 */ 3566 static int 3567 translate_pango_attributes(PangoAttrIterator *iter) 3568 { 3569 PangoAttribute *attr; 3570 int char_attr = HL_NORMAL; 3571 3572 attr = pango_attr_iterator_get(iter, PANGO_ATTR_UNDERLINE); 3573 if (attr != NULL && ((PangoAttrInt *)attr)->value 3574 != (int)PANGO_UNDERLINE_NONE) 3575 char_attr |= HL_UNDERLINE; 3576 3577 attr = pango_attr_iterator_get(iter, PANGO_ATTR_WEIGHT); 3578 if (attr != NULL && ((PangoAttrInt *)attr)->value >= (int)PANGO_WEIGHT_BOLD) 3579 char_attr |= HL_BOLD; 3580 3581 attr = pango_attr_iterator_get(iter, PANGO_ATTR_STYLE); 3582 if (attr != NULL && ((PangoAttrInt *)attr)->value 3583 != (int)PANGO_STYLE_NORMAL) 3584 char_attr |= HL_ITALIC; 3585 3586 attr = pango_attr_iterator_get(iter, PANGO_ATTR_BACKGROUND); 3587 if (attr != NULL) 3588 { 3589 const PangoColor *color = &((PangoAttrColor *)attr)->color; 3590 3591 /* Assume inverse if black background is requested */ 3592 if ((color->red | color->green | color->blue) == 0) 3593 char_attr |= HL_INVERSE; 3594 } 3595 3596 return char_attr; 3597 } 3598 3599 /* 3600 * Retrieve the highlighting attributes at column col in the preedit string. 3601 * Return -1 if not in preediting mode or if col is out of range. 3602 */ 3603 int 3604 im_get_feedback_attr(int col) 3605 { 3606 char *preedit_string = NULL; 3607 PangoAttrList *attr_list = NULL; 3608 int char_attr = -1; 3609 3610 if (xic == NULL) 3611 return char_attr; 3612 3613 gtk_im_context_get_preedit_string(xic, &preedit_string, &attr_list, NULL); 3614 3615 if (preedit_string != NULL && attr_list != NULL) 3616 { 3617 int index; 3618 3619 /* Get the byte index as used by PangoAttrIterator */ 3620 for (index = 0; col > 0 && preedit_string[index] != '\0'; --col) 3621 index += utfc_ptr2len_check((char_u *)preedit_string + index); 3622 3623 if (preedit_string[index] != '\0') 3624 { 3625 PangoAttrIterator *iter; 3626 int start, end; 3627 3628 char_attr = HL_NORMAL; 3629 iter = pango_attr_list_get_iterator(attr_list); 3630 3631 /* Extract all relevant attributes from the list. */ 3632 do 3633 { 3634 pango_attr_iterator_range(iter, &start, &end); 3635 3636 if (index >= start && index < end) 3637 char_attr |= translate_pango_attributes(iter); 3638 } 3639 while (pango_attr_iterator_next(iter)); 3640 3641 pango_attr_iterator_destroy(iter); 3642 } 3643 } 3644 3645 if (attr_list != NULL) 3646 pango_attr_list_unref(attr_list); 3647 g_free(preedit_string); 3648 3649 return char_attr; 3650 } 3651 3652 void 3653 xim_init(void) 3654 { 3655 #ifdef XIM_DEBUG 3656 xim_log("xim_init()\n"); 3657 #endif 3658 3659 g_return_if_fail(gui.drawarea != NULL); 3660 g_return_if_fail(gui.drawarea->window != NULL); 3661 3662 xic = gtk_im_multicontext_new(); 3663 g_object_ref(xic); 3664 3665 im_commit_handler_id = g_signal_connect(G_OBJECT(xic), "commit", 3666 G_CALLBACK(&im_commit_cb), NULL); 3667 g_signal_connect(G_OBJECT(xic), "preedit_changed", 3668 G_CALLBACK(&im_preedit_changed_cb), NULL); 3669 g_signal_connect(G_OBJECT(xic), "preedit_start", 3670 G_CALLBACK(&im_preedit_start_cb), NULL); 3671 g_signal_connect(G_OBJECT(xic), "preedit_end", 3672 G_CALLBACK(&im_preedit_end_cb), NULL); 3673 3674 gtk_im_context_set_client_window(xic, gui.drawarea->window); 3675 } 3676 3677 void 3678 im_shutdown(void) 3679 { 3680 #ifdef XIM_DEBUG 3681 xim_log("im_shutdown()\n"); 3682 #endif 3683 3684 if (xic != NULL) 3685 { 3686 gtk_im_context_focus_out(xic); 3687 g_object_unref(xic); 3688 xic = NULL; 3689 } 3690 im_is_active = FALSE; 3691 im_commit_handler_id = 0; 3692 preedit_start_col = MAXCOL; 3693 xim_has_preediting = FALSE; 3694 } 3695 3696 /* 3697 * Convert the string argument to keyval and state for GdkEventKey. 3698 * If str is valid return TRUE, otherwise FALSE. 3699 * 3700 * See 'imactivatekey' for documentation of the format. 3701 */ 3702 static int 3703 im_string_to_keyval(const char *str, unsigned int *keyval, unsigned int *state) 3704 { 3705 const char *mods_end; 3706 unsigned tmp_keyval; 3707 unsigned tmp_state = 0; 3708 3709 mods_end = strrchr(str, '-'); 3710 mods_end = (mods_end != NULL) ? mods_end + 1 : str; 3711 3712 /* Parse modifier keys */ 3713 while (str < mods_end) 3714 switch (*str++) 3715 { 3716 case '-': break; 3717 case 'S': case 's': tmp_state |= (unsigned)GDK_SHIFT_MASK; break; 3718 case 'L': case 'l': tmp_state |= (unsigned)GDK_LOCK_MASK; break; 3719 case 'C': case 'c': tmp_state |= (unsigned)GDK_CONTROL_MASK;break; 3720 case '1': tmp_state |= (unsigned)GDK_MOD1_MASK; break; 3721 case '2': tmp_state |= (unsigned)GDK_MOD2_MASK; break; 3722 case '3': tmp_state |= (unsigned)GDK_MOD3_MASK; break; 3723 case '4': tmp_state |= (unsigned)GDK_MOD4_MASK; break; 3724 case '5': tmp_state |= (unsigned)GDK_MOD5_MASK; break; 3725 default: 3726 return FALSE; 3727 } 3728 3729 tmp_keyval = gdk_keyval_from_name(str); 3730 3731 if (tmp_keyval == 0 || tmp_keyval == GDK_VoidSymbol) 3732 return FALSE; 3733 3734 if (keyval != NULL) 3735 *keyval = tmp_keyval; 3736 if (state != NULL) 3737 *state = tmp_state; 3738 3739 return TRUE; 3740 } 3741 3742 /* 3743 * Return TRUE if p_imak is valid, otherwise FALSE. As a special case, an 3744 * empty string is also regarded as valid. 3745 * 3746 * Note: The numerical key value of p_imak is cached if it was valid; thus 3747 * boldly assuming im_xim_isvalid_imactivate() will always be called whenever 3748 * 'imak' changes. This is currently the case but not obvious -- should 3749 * probably rename the function for clarity. 3750 */ 3751 int 3752 im_xim_isvalid_imactivate(void) 3753 { 3754 if (p_imak[0] == NUL) 3755 { 3756 im_activatekey_keyval = GDK_VoidSymbol; 3757 im_activatekey_state = 0; 3758 return TRUE; 3759 } 3760 3761 return im_string_to_keyval((const char *)p_imak, 3762 &im_activatekey_keyval, 3763 &im_activatekey_state); 3764 } 3765 3766 static void 3767 im_synthesize_keypress(unsigned int keyval, unsigned int state) 3768 { 3769 GdkEventKey *event; 3770 3771 # ifdef HAVE_GTK_MULTIHEAD 3772 event = (GdkEventKey *)gdk_event_new(GDK_KEY_PRESS); 3773 g_object_ref(gui.drawarea->window); /* unreffed by gdk_event_free() */ 3774 # else 3775 event = (GdkEventKey *)g_malloc0((gulong)sizeof(GdkEvent)); 3776 event->type = GDK_KEY_PRESS; 3777 # endif 3778 event->window = gui.drawarea->window; 3779 event->send_event = TRUE; 3780 event->time = GDK_CURRENT_TIME; 3781 event->state = state; 3782 event->keyval = keyval; 3783 event->hardware_keycode = /* needed for XIM */ 3784 XKeysymToKeycode(GDK_WINDOW_XDISPLAY(event->window), (KeySym)keyval); 3785 event->length = 0; 3786 event->string = NULL; 3787 3788 gtk_im_context_filter_keypress(xic, event); 3789 3790 /* For consistency, also send the corresponding release event. */ 3791 event->type = GDK_KEY_RELEASE; 3792 event->send_event = FALSE; 3793 gtk_im_context_filter_keypress(xic, event); 3794 3795 # ifdef HAVE_GTK_MULTIHEAD 3796 gdk_event_free((GdkEvent *)event); 3797 # else 3798 g_free(event); 3799 # endif 3800 } 3801 3802 void 3803 xim_reset(void) 3804 { 3805 if (xic != NULL) 3806 { 3807 /* 3808 * The third-party imhangul module (and maybe others too) ignores 3809 * gtk_im_context_reset() or at least doesn't reset the active state. 3810 * Thus sending imactivatekey would turn it off if it was on before, 3811 * which is clearly not what we want. Fortunately we can work around 3812 * that for imhangul by sending GDK_Escape, but I don't know if it 3813 * works with all IM modules that support an activation key :/ 3814 * 3815 * An alternative approach would be to destroy the IM context and 3816 * recreate it. But that means loading/unloading the IM module on 3817 * every mode switch, which causes a quite noticable delay even on 3818 * my rather fast box... 3819 * * 3820 * Moreover, there are some XIM which cannot respond to 3821 * im_synthesize_keypress(). we hope that they reset by 3822 * xim_shutdown(). 3823 */ 3824 if (im_activatekey_keyval != GDK_VoidSymbol && im_is_active) 3825 im_synthesize_keypress(GDK_Escape, 0U); 3826 3827 gtk_im_context_reset(xic); 3828 3829 /* 3830 * HACK for Ami: This sequence of function calls makes Ami handle 3831 * the IM reset gratiously, without breaking loads of other stuff. 3832 * It seems to force English mode as well, which is exactly what we 3833 * want because it makes the Ami status display work reliably. 3834 */ 3835 gtk_im_context_set_use_preedit(xic, FALSE); 3836 3837 if (p_imdisable) 3838 im_shutdown(); 3839 else 3840 { 3841 gtk_im_context_set_use_preedit(xic, TRUE); 3842 xim_set_focus(gui.in_focus); 3843 3844 if (im_activatekey_keyval != GDK_VoidSymbol) 3845 { 3846 if (im_is_active) 3847 { 3848 g_signal_handler_block(xic, im_commit_handler_id); 3849 im_synthesize_keypress(im_activatekey_keyval, 3850 im_activatekey_state); 3851 g_signal_handler_unblock(xic, im_commit_handler_id); 3852 } 3853 } 3854 else 3855 { 3856 im_shutdown(); 3857 xim_init(); 3858 xim_set_focus(gui.in_focus); 3859 } 3860 } 3861 } 3862 3863 preedit_start_col = MAXCOL; 3864 xim_has_preediting = FALSE; 3865 } 3866 3867 int 3868 xim_queue_key_press_event(GdkEventKey *event, int down) 3869 { 3870 if (down) 3871 { 3872 /* 3873 * Workaround GTK2 XIM 'feature' that always converts keypad keys to 3874 * chars., even when not part of an IM sequence (ref. feature of 3875 * gdk/gdkkeyuni.c). 3876 * Flag any keypad keys that might represent a single char. 3877 * If this (on its own - i.e., not part of an IM sequence) is 3878 * committed while we're processing one of these keys, we can ignore 3879 * that commit and go ahead & process it ourselves. That way we can 3880 * still distinguish keypad keys for use in mappings. 3881 */ 3882 switch (event->keyval) 3883 { 3884 case GDK_KP_Add: xim_expected_char = '+'; break; 3885 case GDK_KP_Subtract: xim_expected_char = '-'; break; 3886 case GDK_KP_Divide: xim_expected_char = '/'; break; 3887 case GDK_KP_Multiply: xim_expected_char = '*'; break; 3888 case GDK_KP_Decimal: xim_expected_char = '.'; break; 3889 case GDK_KP_Equal: xim_expected_char = '='; break; 3890 case GDK_KP_0: xim_expected_char = '0'; break; 3891 case GDK_KP_1: xim_expected_char = '1'; break; 3892 case GDK_KP_2: xim_expected_char = '2'; break; 3893 case GDK_KP_3: xim_expected_char = '3'; break; 3894 case GDK_KP_4: xim_expected_char = '4'; break; 3895 case GDK_KP_5: xim_expected_char = '5'; break; 3896 case GDK_KP_6: xim_expected_char = '6'; break; 3897 case GDK_KP_7: xim_expected_char = '7'; break; 3898 case GDK_KP_8: xim_expected_char = '8'; break; 3899 case GDK_KP_9: xim_expected_char = '9'; break; 3900 default: xim_expected_char = NUL; 3901 } 3902 xim_ignored_char = FALSE; 3903 } 3904 3905 /* 3906 * When typing fFtT, XIM may be activated. Thus it must pass 3907 * gtk_im_context_filter_keypress() in Normal mode. 3908 * And while doing :sh too. 3909 */ 3910 if (xic != NULL && !p_imdisable 3911 && (State & (INSERT | CMDLINE | NORMAL | EXTERNCMD)) != 0) 3912 { 3913 /* 3914 * Filter 'imactivatekey' and map it to CTRL-^. This way, Vim is 3915 * always aware of the current status of IM, and can even emulate 3916 * the activation key for modules that don't support one. 3917 */ 3918 if (event->keyval == im_activatekey_keyval 3919 && (event->state & im_activatekey_state) == im_activatekey_state) 3920 { 3921 unsigned int state_mask; 3922 3923 /* Require the state of the 3 most used modifiers to match exactly. 3924 * Otherwise e.g. <S-C-space> would be unusable for other purposes 3925 * if the IM activate key is <S-space>. */ 3926 state_mask = im_activatekey_state; 3927 state_mask |= ((int)GDK_SHIFT_MASK | (int)GDK_CONTROL_MASK 3928 | (int)GDK_MOD1_MASK); 3929 3930 if ((event->state & state_mask) != im_activatekey_state) 3931 return FALSE; 3932 3933 /* Don't send it a second time on GDK_KEY_RELEASE. */ 3934 if (event->type != GDK_KEY_PRESS) 3935 return TRUE; 3936 3937 if (map_to_exists_mode((char_u *)"", LANGMAP)) 3938 { 3939 im_set_active(FALSE); 3940 3941 /* ":lmap" mappings exists, toggle use of mappings. */ 3942 State ^= LANGMAP; 3943 if (State & LANGMAP) 3944 { 3945 curbuf->b_p_iminsert = B_IMODE_NONE; 3946 State &= ~LANGMAP; 3947 } 3948 else 3949 { 3950 curbuf->b_p_iminsert = B_IMODE_LMAP; 3951 State |= LANGMAP; 3952 } 3953 return TRUE; 3954 } 3955 3956 return gtk_im_context_filter_keypress(xic, event); 3957 } 3958 3959 /* Don't filter events through the IM context if IM isn't active 3960 * right now. Unlike with GTK+ 1.2 we cannot rely on the IM module 3961 * not doing anything before the activation key was sent. */ 3962 if (im_activatekey_keyval == GDK_VoidSymbol || im_is_active) 3963 { 3964 int imresult = gtk_im_context_filter_keypress(xic, event); 3965 3966 /* Some XIM send following sequence: 3967 * 1. preedited string. 3968 * 2. committed string. 3969 * 3. line changed key. 3970 * 4. preedited string. 3971 * 5. remove preedited string. 3972 * if 3, Vim can't move back the above line for 5. 3973 * thus, this part should not parse the key. */ 3974 if (!imresult && preedit_start_col != MAXCOL 3975 && event->keyval == GDK_Return) 3976 { 3977 im_synthesize_keypress(GDK_Return, 0U); 3978 return FALSE; 3979 } 3980 3981 /* If XIM tried to commit a keypad key as a single char., 3982 * ignore it so we can use the keypad key 'raw', for mappings. */ 3983 if (xim_expected_char != NUL && xim_ignored_char) 3984 /* We had a keypad key, and XIM tried to thieve it */ 3985 return FALSE; 3986 3987 /* Normal processing */ 3988 return imresult; 3989 } 3990 } 3991 3992 return FALSE; 3993 } 3994 3995 int 3996 im_get_status(void) 3997 { 3998 return im_is_active; 3999 } 4000 4001 # else /* !HAVE_GTK2 */ 4002 4003 static int xim_is_active = FALSE; /* XIM should be active in the current 4004 mode */ 4005 static int xim_has_focus = FALSE; /* XIM is really being used for Vim */ 4006 #ifdef FEAT_GUI_X11 4007 static XIMStyle input_style; 4008 static int status_area_enabled = TRUE; 4009 #endif 4010 4011 #ifdef FEAT_GUI_GTK 4012 # ifdef WIN3264 4013 # include <gdk/gdkwin32.h> 4014 # else 4015 # include <gdk/gdkx.h> 4016 # endif 4017 #else 4018 # ifdef PROTO 4019 /* Define a few things to be able to generate prototypes while not configured 4020 * for GTK. */ 4021 # define GSList int 4022 # define gboolean int 4023 typedef int GdkEvent; 4024 typedef int GdkEventKey; 4025 # define GdkIC int 4026 # endif 4027 #endif 4028 4029 #if defined(FEAT_GUI_GTK) || defined(PROTO) || defined(FEAT_GUI_KDE) 4030 static int preedit_buf_len = 0; 4031 static int xim_can_preediting INIT(= FALSE); /* XIM in showmode() */ 4032 static int xim_input_style; 4033 #ifndef FEAT_GUI_GTK 4034 # define gboolean int 4035 #endif 4036 static gboolean use_status_area = 0; 4037 4038 static int im_xim_str2keycode __ARGS((unsigned int *code, unsigned int *state)); 4039 static void im_xim_send_event_imactivate __ARGS((void)); 4040 4041 /* 4042 * Convert string to keycode and state for XKeyEvent. 4043 * When string is valid return OK, when invalid return FAIL. 4044 * 4045 * See 'imactivatekey' documentation for the format. 4046 */ 4047 static int 4048 im_xim_str2keycode(code, state) 4049 unsigned int *code; 4050 unsigned int *state; 4051 { 4052 int retval = OK; 4053 int len; 4054 unsigned keycode = 0, keystate = 0; 4055 Window window; 4056 Display *display; 4057 char_u *flag_end; 4058 char_u *str; 4059 4060 if (*p_imak != NUL) 4061 { 4062 len = STRLEN(p_imak); 4063 for (flag_end = p_imak + len - 1; 4064 flag_end > p_imak && *flag_end != '-'; --flag_end) 4065 ; 4066 4067 /* Parse modifier keys */ 4068 for (str = p_imak; str < flag_end; ++str) 4069 { 4070 switch (*str) 4071 { 4072 case 's': case 'S': 4073 keystate |= ShiftMask; 4074 break; 4075 case 'l': case 'L': 4076 keystate |= LockMask; 4077 break; 4078 case 'c': case 'C': 4079 keystate |= ControlMask; 4080 break; 4081 case '1': 4082 keystate |= Mod1Mask; 4083 break; 4084 case '2': 4085 keystate |= Mod2Mask; 4086 break; 4087 case '3': 4088 keystate |= Mod3Mask; 4089 break; 4090 case '4': 4091 keystate |= Mod4Mask; 4092 break; 4093 case '5': 4094 keystate |= Mod5Mask; 4095 break; 4096 case '-': 4097 break; 4098 default: 4099 retval = FAIL; 4100 } 4101 } 4102 if (*str == '-') 4103 ++str; 4104 4105 /* Get keycode from string. */ 4106 gui_get_x11_windis(&window, &display); 4107 if (display) 4108 keycode = XKeysymToKeycode(display, XStringToKeysym((char *)str)); 4109 if (keycode == 0) 4110 retval = FAIL; 4111 4112 if (code != NULL) 4113 *code = keycode; 4114 if (state != NULL) 4115 *state = keystate; 4116 } 4117 return retval; 4118 } 4119 4120 static void 4121 im_xim_send_event_imactivate() 4122 { 4123 /* Force turn on preedit state by symulate keypress event. 4124 * Keycode and state is specified by 'imactivatekey'. 4125 */ 4126 XKeyEvent ev; 4127 4128 gui_get_x11_windis(&ev.window, &ev.display); 4129 ev.root = RootWindow(ev.display, DefaultScreen(ev.display)); 4130 ev.subwindow = None; 4131 ev.time = CurrentTime; 4132 ev.x = 1; 4133 ev.y = 1; 4134 ev.x_root = 1; 4135 ev.y_root = 1; 4136 ev.same_screen = 1; 4137 ev.type = KeyPress; 4138 if (im_xim_str2keycode(&ev.keycode, &ev.state) == OK) 4139 XSendEvent(ev.display, ev.window, 1, KeyPressMask, (XEvent*)&ev); 4140 } 4141 4142 /* 4143 * Return TRUE if 'imactivatekey' has a valid value. 4144 */ 4145 int 4146 im_xim_isvalid_imactivate() 4147 { 4148 return im_xim_str2keycode(NULL, NULL) == OK; 4149 } 4150 #endif /* FEAT_GUI_GTK */ 4151 4152 /* 4153 * Switch using XIM on/off. This is used by the code that changes "State". 4154 */ 4155 void 4156 im_set_active(active) 4157 int active; 4158 { 4159 if (xic == NULL) 4160 return; 4161 4162 /* If 'imdisable' is set, XIM is never active. */ 4163 if (p_imdisable) 4164 active = FALSE; 4165 #if !defined (FEAT_GUI_GTK) && !defined (FEAT_GUI_KDE) 4166 else if (input_style & XIMPreeditPosition) 4167 /* There is a problem in switching XIM off when preediting is used, 4168 * and it is not clear how this can be solved. For now, keep XIM on 4169 * all the time, like it was done in Vim 5.8. */ 4170 active = TRUE; 4171 #endif 4172 4173 /* Remember the active state, it is needed when Vim gets keyboard focus. */ 4174 xim_is_active = active; 4175 4176 #ifdef FEAT_GUI_GTK 4177 /* When 'imactivatekey' has valid key-string, try to control XIM preedit 4178 * state. When 'imactivatekey' has no or invalid string, try old XIM 4179 * focus control. 4180 */ 4181 if (*p_imak != NUL) 4182 { 4183 /* BASIC STRATEGY: 4184 * Destroy old Input Context (XIC), and create new one. New XIC 4185 * would have a state of preedit that is off. When argument:active 4186 * is false, that's all. Else argument:active is true, send a key 4187 * event specified by 'imactivatekey' to activate XIM preedit state. 4188 */ 4189 4190 xim_is_active = TRUE; /* Disable old XIM focus control */ 4191 /* If we can monitor preedit state with preedit callback functions, 4192 * try least creation of new XIC. 4193 */ 4194 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS) 4195 { 4196 if (xim_can_preediting && !active) 4197 { 4198 /* Force turn off preedit state. With some IM 4199 * implementations, we cannot turn off preedit state by 4200 * symulate keypress event. It is why using such a method 4201 * that destroy old IC (input context), and create new one. 4202 * When create new IC, its preedit state is usually off. 4203 */ 4204 xim_reset(); 4205 xim_set_focus(FALSE); 4206 gdk_ic_destroy(xic); 4207 xim_init(); 4208 xim_can_preediting = FALSE; 4209 } 4210 else if (!xim_can_preediting && active) 4211 im_xim_send_event_imactivate(); 4212 } 4213 else 4214 { 4215 /* First, force destroy old IC, and create new one. It 4216 * symulates "turning off preedit state". 4217 */ 4218 xim_set_focus(FALSE); 4219 gdk_ic_destroy(xic); 4220 xim_init(); 4221 xim_can_preediting = FALSE; 4222 4223 /* 2nd, when requested to activate IM, symulate this by sending 4224 * the event. 4225 */ 4226 if (active) 4227 { 4228 im_xim_send_event_imactivate(); 4229 xim_can_preediting = TRUE; 4230 } 4231 } 4232 } 4233 else 4234 { 4235 # ifndef XIMPreeditUnKnown 4236 /* X11R5 doesn't have these, it looks safe enough to define here. */ 4237 typedef unsigned long XIMPreeditState; 4238 # define XIMPreeditUnKnown 0L 4239 # define XIMPreeditEnable 1L 4240 # define XIMPreeditDisable (1L<<1) 4241 # define XNPreeditState "preeditState" 4242 # endif 4243 XIMPreeditState preedit_state = XIMPreeditUnKnown; 4244 XVaNestedList preedit_attr; 4245 XIC pxic; 4246 4247 preedit_attr = XVaCreateNestedList(0, 4248 XNPreeditState, &preedit_state, 4249 NULL); 4250 pxic = ((GdkICPrivate *)xic)->xic; 4251 4252 if (!XGetICValues(pxic, XNPreeditAttributes, preedit_attr, NULL)) 4253 { 4254 XFree(preedit_attr); 4255 preedit_attr = XVaCreateNestedList(0, 4256 XNPreeditState, 4257 active ? XIMPreeditEnable : XIMPreeditDisable, 4258 NULL); 4259 XSetICValues(pxic, XNPreeditAttributes, preedit_attr, NULL); 4260 xim_can_preediting = active; 4261 xim_is_active = active; 4262 } 4263 XFree(preedit_attr); 4264 } 4265 if (xim_input_style & XIMPreeditCallbacks) 4266 { 4267 preedit_buf_len = 0; 4268 init_preedit_start_col(); 4269 } 4270 #else 4271 # if 0 4272 /* When had tested kinput2 + canna + Athena GUI version with 4273 * 'imactivatekey' is "s-space", im_xim_send_event_imactivate() did not 4274 * work correctly. It just inserted one space. I don't know why we 4275 * couldn't switch state of XIM preediting. This is reason why these 4276 * codes are commented out. 4277 */ 4278 /* First, force destroy old IC, and create new one. It symulates 4279 * "turning off preedit state". 4280 */ 4281 xim_set_focus(FALSE); 4282 XDestroyIC(xic); 4283 xic = NULL; 4284 xim_init(); 4285 4286 /* 2nd, when requested to activate IM, symulate this by sending the 4287 * event. 4288 */ 4289 if (active) 4290 im_xim_send_event_imactivate(); 4291 # endif 4292 #endif 4293 xim_set_preedit(); 4294 } 4295 4296 /* 4297 * Adjust using XIM for gaining or losing keyboard focus. Also called when 4298 * "xim_is_active" changes. 4299 */ 4300 void 4301 xim_set_focus(focus) 4302 int focus; 4303 { 4304 if (xic == NULL) 4305 return; 4306 4307 /* 4308 * XIM only gets focus when the Vim window has keyboard focus and XIM has 4309 * been set active for the current mode. 4310 */ 4311 if (focus && xim_is_active) 4312 { 4313 if (!xim_has_focus) 4314 { 4315 xim_has_focus = TRUE; 4316 #ifdef FEAT_GUI_GTK 4317 gdk_im_begin(xic, gui.drawarea->window); 4318 #else 4319 XSetICFocus(xic); 4320 #endif 4321 } 4322 } 4323 else 4324 { 4325 if (xim_has_focus) 4326 { 4327 xim_has_focus = FALSE; 4328 #ifdef FEAT_GUI_GTK 4329 gdk_im_end(); 4330 #else 4331 XUnsetICFocus(xic); 4332 #endif 4333 } 4334 } 4335 } 4336 4337 /*ARGSUSED*/ 4338 void 4339 im_set_position(row, col) 4340 int row; 4341 int col; 4342 { 4343 xim_set_preedit(); 4344 } 4345 4346 /* 4347 * Set the XIM to the current cursor position. 4348 */ 4349 void 4350 xim_set_preedit() 4351 { 4352 if (xic == NULL) 4353 return; 4354 4355 xim_set_focus(TRUE); 4356 4357 #ifdef FEAT_GUI_GTK 4358 if (gdk_im_ready()) 4359 { 4360 int attrmask; 4361 GdkICAttr *attr; 4362 4363 if (!xic_attr) 4364 return; 4365 4366 attr = xic_attr; 4367 attrmask = 0; 4368 4369 # ifdef FEAT_XFONTSET 4370 if ((xim_input_style & (int)GDK_IM_PREEDIT_POSITION) 4371 && gui.fontset != NOFONTSET 4372 && gui.fontset->type == GDK_FONT_FONTSET) 4373 { 4374 if (!xim_has_focus) 4375 { 4376 if (attr->spot_location.y >= 0) 4377 { 4378 attr->spot_location.x = 0; 4379 attr->spot_location.y = -100; 4380 attrmask |= (int)GDK_IC_SPOT_LOCATION; 4381 } 4382 } 4383 else 4384 { 4385 gint width, height; 4386 4387 if (attr->spot_location.x != TEXT_X(gui.col) 4388 || attr->spot_location.y != TEXT_Y(gui.row)) 4389 { 4390 attr->spot_location.x = TEXT_X(gui.col); 4391 attr->spot_location.y = TEXT_Y(gui.row); 4392 attrmask |= (int)GDK_IC_SPOT_LOCATION; 4393 } 4394 4395 gdk_window_get_size(gui.drawarea->window, &width, &height); 4396 width -= 2 * gui.border_offset; 4397 height -= 2 * gui.border_offset; 4398 if (xim_input_style & (int)GDK_IM_STATUS_AREA) 4399 height -= gui.char_height; 4400 if (attr->preedit_area.width != width 4401 || attr->preedit_area.height != height) 4402 { 4403 attr->preedit_area.x = gui.border_offset; 4404 attr->preedit_area.y = gui.border_offset; 4405 attr->preedit_area.width = width; 4406 attr->preedit_area.height = height; 4407 attrmask |= (int)GDK_IC_PREEDIT_AREA; 4408 } 4409 4410 if (attr->preedit_fontset != gui.current_font) 4411 { 4412 attr->preedit_fontset = gui.current_font; 4413 attrmask |= (int)GDK_IC_PREEDIT_FONTSET; 4414 } 4415 } 4416 } 4417 # endif /* FEAT_XFONTSET */ 4418 4419 if (xim_fg_color == INVALCOLOR) 4420 { 4421 xim_fg_color = gui.def_norm_pixel; 4422 xim_bg_color = gui.def_back_pixel; 4423 } 4424 if (attr->preedit_foreground.pixel != xim_fg_color) 4425 { 4426 attr->preedit_foreground.pixel = xim_fg_color; 4427 attrmask |= (int)GDK_IC_PREEDIT_FOREGROUND; 4428 } 4429 if (attr->preedit_background.pixel != xim_bg_color) 4430 { 4431 attr->preedit_background.pixel = xim_bg_color; 4432 attrmask |= (int)GDK_IC_PREEDIT_BACKGROUND; 4433 } 4434 4435 if (attrmask != 0) 4436 gdk_ic_set_attr(xic, attr, (GdkICAttributesType)attrmask); 4437 } 4438 #else /* FEAT_GUI_GTK */ 4439 # ifdef FEAT_GUI_KDE 4440 # else 4441 { 4442 XVaNestedList attr_list; 4443 XRectangle spot_area; 4444 XPoint over_spot; 4445 int line_space; 4446 4447 if (!xim_has_focus) 4448 { 4449 /* hide XIM cursor */ 4450 over_spot.x = 0; 4451 over_spot.y = -100; /* arbitrary invisible position */ 4452 attr_list = (XVaNestedList) XVaCreateNestedList(0, 4453 XNSpotLocation, 4454 &over_spot, 4455 NULL); 4456 XSetICValues(xic, XNPreeditAttributes, attr_list, NULL); 4457 XFree(attr_list); 4458 return; 4459 } 4460 4461 if (input_style & XIMPreeditPosition) 4462 { 4463 if (xim_fg_color == INVALCOLOR) 4464 { 4465 xim_fg_color = gui.def_norm_pixel; 4466 xim_bg_color = gui.def_back_pixel; 4467 } 4468 over_spot.x = TEXT_X(gui.col); 4469 over_spot.y = TEXT_Y(gui.row); 4470 spot_area.x = 0; 4471 spot_area.y = 0; 4472 spot_area.height = gui.char_height * Rows; 4473 spot_area.width = gui.char_width * Columns; 4474 line_space = gui.char_height; 4475 attr_list = (XVaNestedList) XVaCreateNestedList(0, 4476 XNSpotLocation, &over_spot, 4477 XNForeground, (Pixel) xim_fg_color, 4478 XNBackground, (Pixel) xim_bg_color, 4479 XNArea, &spot_area, 4480 XNLineSpace, line_space, 4481 NULL); 4482 if (XSetICValues(xic, XNPreeditAttributes, attr_list, NULL)) 4483 EMSG(_("E284: Cannot set IC values")); 4484 XFree(attr_list); 4485 } 4486 } 4487 # endif /* FEAT_GUI_KDE */ 4488 #endif /* FEAT_GUI_GTK */ 4489 } 4490 4491 /* 4492 * Set up the status area. 4493 * 4494 * This should use a separate Widget, but that seems not possible, because 4495 * preedit_area and status_area should be set to the same window as for the 4496 * text input. Unfortunately this means the status area pollutes the text 4497 * window... 4498 */ 4499 void 4500 xim_set_status_area() 4501 { 4502 if (xic == NULL) 4503 return; 4504 4505 #ifdef FEAT_GUI_GTK 4506 # if defined(FEAT_XFONTSET) 4507 if (use_status_area) 4508 { 4509 GdkICAttr *attr; 4510 int style; 4511 gint width, height; 4512 GtkWidget *widget; 4513 int attrmask; 4514 4515 if (!xic_attr) 4516 return; 4517 4518 attr = xic_attr; 4519 attrmask = 0; 4520 style = (int)gdk_ic_get_style(xic); 4521 if ((style & (int)GDK_IM_STATUS_MASK) == (int)GDK_IM_STATUS_AREA) 4522 { 4523 if (gui.fontset != NOFONTSET 4524 && gui.fontset->type == GDK_FONT_FONTSET) 4525 { 4526 widget = gui.mainwin; 4527 gdk_window_get_size(widget->window, &width, &height); 4528 4529 attrmask |= (int)GDK_IC_STATUS_AREA; 4530 attr->status_area.x = 0; 4531 attr->status_area.y = height - gui.char_height - 1; 4532 attr->status_area.width = width; 4533 attr->status_area.height = gui.char_height; 4534 } 4535 } 4536 if (attrmask != 0) 4537 gdk_ic_set_attr(xic, attr, (GdkICAttributesType)attrmask); 4538 } 4539 # endif 4540 #else 4541 # ifdef FEAT_GUI_KDE 4542 # else 4543 { 4544 XVaNestedList preedit_list = 0, status_list = 0, list = 0; 4545 XRectangle pre_area, status_area; 4546 4547 if (input_style & XIMStatusArea) 4548 { 4549 if (input_style & XIMPreeditArea) 4550 { 4551 XRectangle *needed_rect; 4552 4553 /* to get status_area width */ 4554 status_list = XVaCreateNestedList(0, XNAreaNeeded, 4555 &needed_rect, NULL); 4556 XGetICValues(xic, XNStatusAttributes, status_list, NULL); 4557 XFree(status_list); 4558 4559 status_area.width = needed_rect->width; 4560 } 4561 else 4562 status_area.width = gui.char_width * Columns; 4563 4564 status_area.x = 0; 4565 status_area.y = gui.char_height * Rows + gui.border_offset; 4566 if (gui.which_scrollbars[SBAR_BOTTOM]) 4567 status_area.y += gui.scrollbar_height; 4568 #ifdef FEAT_MENU 4569 if (gui.menu_is_active) 4570 status_area.y += gui.menu_height; 4571 #endif 4572 status_area.height = gui.char_height; 4573 status_list = XVaCreateNestedList(0, XNArea, &status_area, NULL); 4574 } 4575 else 4576 { 4577 status_area.x = 0; 4578 status_area.y = gui.char_height * Rows + gui.border_offset; 4579 if (gui.which_scrollbars[SBAR_BOTTOM]) 4580 status_area.y += gui.scrollbar_height; 4581 #ifdef FEAT_MENU 4582 if (gui.menu_is_active) 4583 status_area.y += gui.menu_height; 4584 #endif 4585 status_area.width = 0; 4586 status_area.height = gui.char_height; 4587 } 4588 4589 if (input_style & XIMPreeditArea) /* off-the-spot */ 4590 { 4591 pre_area.x = status_area.x + status_area.width; 4592 pre_area.y = gui.char_height * Rows + gui.border_offset; 4593 pre_area.width = gui.char_width * Columns - pre_area.x; 4594 if (gui.which_scrollbars[SBAR_BOTTOM]) 4595 pre_area.y += gui.scrollbar_height; 4596 #ifdef FEAT_MENU 4597 if (gui.menu_is_active) 4598 pre_area.y += gui.menu_height; 4599 #endif 4600 pre_area.height = gui.char_height; 4601 preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL); 4602 } 4603 else if (input_style & XIMPreeditPosition) /* over-the-spot */ 4604 { 4605 pre_area.x = 0; 4606 pre_area.y = 0; 4607 pre_area.height = gui.char_height * Rows; 4608 pre_area.width = gui.char_width * Columns; 4609 preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL); 4610 } 4611 4612 if (preedit_list && status_list) 4613 list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list, 4614 XNStatusAttributes, status_list, NULL); 4615 else if (preedit_list) 4616 list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list, 4617 NULL); 4618 else if (status_list) 4619 list = XVaCreateNestedList(0, XNStatusAttributes, status_list, 4620 NULL); 4621 else 4622 list = NULL; 4623 4624 if (list) 4625 { 4626 XSetICValues(xic, XNVaNestedList, list, NULL); 4627 XFree(list); 4628 } 4629 if (status_list) 4630 XFree(status_list); 4631 if (preedit_list) 4632 XFree(preedit_list); 4633 } 4634 # endif /* FEAT_GUI_KDE */ 4635 #endif 4636 } 4637 4638 #if defined(FEAT_GUI_X11) || defined(FEAT_GUI_GTK) 4639 static char e_xim[] = N_("E285: Failed to create input context"); 4640 #endif 4641 4642 #if defined(FEAT_GUI_X11) || defined(PROTO) 4643 # if defined(XtSpecificationRelease) && XtSpecificationRelease >= 6 && !defined(sun) 4644 # define USE_X11R6_XIM 4645 # endif 4646 4647 static int xim_real_init __ARGS((Window x11_window, Display *x11_display)); 4648 4649 4650 #ifdef USE_X11R6_XIM 4651 static void xim_instantiate_cb __ARGS((Display *display, XPointer client_data, XPointer call_data)); 4652 static void xim_destroy_cb __ARGS((XIM im, XPointer client_data, XPointer call_data)); 4653 4654 /*ARGSUSED*/ 4655 static void 4656 xim_instantiate_cb(display, client_data, call_data) 4657 Display *display; 4658 XPointer client_data; 4659 XPointer call_data; 4660 { 4661 Window x11_window; 4662 Display *x11_display; 4663 4664 #ifdef XIM_DEBUG 4665 xim_log("xim_instantiate_cb()\n"); 4666 #endif 4667 4668 gui_get_x11_windis(&x11_window, &x11_display); 4669 if (display != x11_display) 4670 return; 4671 4672 xim_real_init(x11_window, x11_display); 4673 gui_set_shellsize(FALSE, FALSE); 4674 if (xic != NULL) 4675 XUnregisterIMInstantiateCallback(x11_display, NULL, NULL, NULL, 4676 xim_instantiate_cb, NULL); 4677 } 4678 4679 /*ARGSUSED*/ 4680 static void 4681 xim_destroy_cb(im, client_data, call_data) 4682 XIM im; 4683 XPointer client_data; 4684 XPointer call_data; 4685 { 4686 Window x11_window; 4687 Display *x11_display; 4688 4689 #ifdef XIM_DEBUG 4690 xim_log("xim_destroy_cb()\n"); 4691 #endif 4692 gui_get_x11_windis(&x11_window, &x11_display); 4693 4694 xic = NULL; 4695 status_area_enabled = FALSE; 4696 4697 gui_set_shellsize(FALSE, FALSE); 4698 4699 XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL, 4700 xim_instantiate_cb, NULL); 4701 } 4702 #endif 4703 4704 void 4705 xim_init() 4706 { 4707 Window x11_window; 4708 Display *x11_display; 4709 4710 #ifdef XIM_DEBUG 4711 xim_log("xim_init()\n"); 4712 #endif 4713 4714 gui_get_x11_windis(&x11_window, &x11_display); 4715 4716 xic = NULL; 4717 4718 if (xim_real_init(x11_window, x11_display)) 4719 return; 4720 4721 gui_set_shellsize(FALSE, FALSE); 4722 4723 #ifdef USE_X11R6_XIM 4724 XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL, 4725 xim_instantiate_cb, NULL); 4726 #endif 4727 } 4728 4729 static int 4730 xim_real_init(x11_window, x11_display) 4731 Window x11_window; 4732 Display *x11_display; 4733 { 4734 int i; 4735 char *p, 4736 *s, 4737 *ns, 4738 *end, 4739 tmp[1024]; 4740 #define IMLEN_MAX 40 4741 char buf[IMLEN_MAX + 7]; 4742 XIM xim = NULL; 4743 XIMStyles *xim_styles; 4744 XIMStyle this_input_style = 0; 4745 Boolean found; 4746 XPoint over_spot; 4747 XVaNestedList preedit_list, status_list; 4748 4749 input_style = 0; 4750 status_area_enabled = FALSE; 4751 4752 if (xic != NULL) 4753 return FALSE; 4754 4755 if (gui.rsrc_input_method != NULL && *gui.rsrc_input_method != NUL) 4756 { 4757 strcpy(tmp, gui.rsrc_input_method); 4758 for (ns = s = tmp; ns != NULL && *s != NUL;) 4759 { 4760 s = (char *)skipwhite((char_u *)s); 4761 if (*s == NUL) 4762 break; 4763 if ((ns = end = strchr(s, ',')) == NULL) 4764 end = s + strlen(s); 4765 while (isspace(((char_u *)end)[-1])) 4766 end--; 4767 *end = NUL; 4768 4769 if (strlen(s) <= IMLEN_MAX) 4770 { 4771 strcpy(buf, "@im="); 4772 strcat(buf, s); 4773 if ((p = XSetLocaleModifiers(buf)) != NULL && *p != NUL 4774 && (xim = XOpenIM(x11_display, NULL, NULL, NULL)) 4775 != NULL) 4776 break; 4777 } 4778 4779 s = ns + 1; 4780 } 4781 } 4782 4783 if (xim == NULL && (p = XSetLocaleModifiers("")) != NULL && *p != NUL) 4784 xim = XOpenIM(x11_display, NULL, NULL, NULL); 4785 4786 /* This is supposed to be useful to obtain characters through 4787 * XmbLookupString() without really using a XIM. */ 4788 if (xim == NULL && (p = XSetLocaleModifiers("@im=none")) != NULL 4789 && *p != NUL) 4790 xim = XOpenIM(x11_display, NULL, NULL, NULL); 4791 4792 if (xim == NULL) 4793 { 4794 /* Only give this message when verbose is set, because too many people 4795 * got this message when they didn't want to use a XIM. */ 4796 if (p_verbose > 0) 4797 EMSG(_("E286: Failed to open input method")); 4798 return FALSE; 4799 } 4800 4801 #ifdef USE_X11R6_XIM 4802 { 4803 XIMCallback destroy_cb; 4804 4805 destroy_cb.callback = xim_destroy_cb; 4806 destroy_cb.client_data = NULL; 4807 if (XSetIMValues(xim, XNDestroyCallback, &destroy_cb, NULL)) 4808 EMSG(_("E287: Warning: Could not set destroy callback to IM")); 4809 } 4810 #endif 4811 4812 if (XGetIMValues(xim, XNQueryInputStyle, &xim_styles, NULL) || !xim_styles) 4813 { 4814 EMSG(_("E288: input method doesn't support any style")); 4815 XCloseIM(xim); 4816 return FALSE; 4817 } 4818 4819 found = False; 4820 strcpy(tmp, gui.rsrc_preedit_type_name); 4821 for (s = tmp; s && !found; ) 4822 { 4823 while (*s && isspace((unsigned char)*s)) 4824 s++; 4825 if (!*s) 4826 break; 4827 if ((ns = end = strchr(s, ',')) != 0) 4828 ns++; 4829 else 4830 end = s + strlen(s); 4831 while (isspace((unsigned char)*end)) 4832 end--; 4833 *end = '\0'; 4834 4835 if (!strcmp(s, "OverTheSpot")) 4836 this_input_style = (XIMPreeditPosition | XIMStatusArea); 4837 else if (!strcmp(s, "OffTheSpot")) 4838 this_input_style = (XIMPreeditArea | XIMStatusArea); 4839 else if (!strcmp(s, "Root")) 4840 this_input_style = (XIMPreeditNothing | XIMStatusNothing); 4841 4842 for (i = 0; (unsigned short)i < xim_styles->count_styles; i++) 4843 { 4844 if (this_input_style == xim_styles->supported_styles[i]) 4845 { 4846 found = True; 4847 break; 4848 } 4849 } 4850 if (!found) 4851 for (i = 0; (unsigned short)i < xim_styles->count_styles; i++) 4852 { 4853 if ((xim_styles->supported_styles[i] & this_input_style) 4854 == (this_input_style & ~XIMStatusArea)) 4855 { 4856 this_input_style &= ~XIMStatusArea; 4857 found = True; 4858 break; 4859 } 4860 } 4861 4862 s = ns; 4863 } 4864 XFree(xim_styles); 4865 4866 if (!found) 4867 { 4868 /* Only give this message when verbose is set, because too many people 4869 * got this message when they didn't want to use a XIM. */ 4870 if (p_verbose > 0) 4871 EMSG(_("E289: input method doesn't support my preedit type")); 4872 XCloseIM(xim); 4873 return FALSE; 4874 } 4875 4876 over_spot.x = TEXT_X(gui.col); 4877 over_spot.y = TEXT_Y(gui.row); 4878 input_style = this_input_style; 4879 4880 /* A crash was reported when trying to pass gui.norm_font as XNFontSet, 4881 * thus that has been removed. Hopefully the default works... */ 4882 #ifdef FEAT_XFONTSET 4883 if (gui.fontset != NOFONTSET) 4884 { 4885 preedit_list = XVaCreateNestedList(0, 4886 XNSpotLocation, &over_spot, 4887 XNForeground, (Pixel)gui.def_norm_pixel, 4888 XNBackground, (Pixel)gui.def_back_pixel, 4889 XNFontSet, (XFontSet)gui.fontset, 4890 NULL); 4891 status_list = XVaCreateNestedList(0, 4892 XNForeground, (Pixel)gui.def_norm_pixel, 4893 XNBackground, (Pixel)gui.def_back_pixel, 4894 XNFontSet, (XFontSet)gui.fontset, 4895 NULL); 4896 } 4897 else 4898 #endif 4899 { 4900 preedit_list = XVaCreateNestedList(0, 4901 XNSpotLocation, &over_spot, 4902 XNForeground, (Pixel)gui.def_norm_pixel, 4903 XNBackground, (Pixel)gui.def_back_pixel, 4904 NULL); 4905 status_list = XVaCreateNestedList(0, 4906 XNForeground, (Pixel)gui.def_norm_pixel, 4907 XNBackground, (Pixel)gui.def_back_pixel, 4908 NULL); 4909 } 4910 4911 xic = XCreateIC(xim, 4912 XNInputStyle, input_style, 4913 XNClientWindow, x11_window, 4914 XNFocusWindow, gui.wid, 4915 XNPreeditAttributes, preedit_list, 4916 XNStatusAttributes, status_list, 4917 NULL); 4918 XFree(status_list); 4919 XFree(preedit_list); 4920 if (xic != NULL) 4921 { 4922 if (input_style & XIMStatusArea) 4923 { 4924 xim_set_status_area(); 4925 status_area_enabled = TRUE; 4926 } 4927 else 4928 gui_set_shellsize(FALSE, FALSE); 4929 } 4930 else 4931 { 4932 EMSG(_(e_xim)); 4933 XCloseIM(xim); 4934 return FALSE; 4935 } 4936 4937 return TRUE; 4938 } 4939 4940 #endif /* FEAT_GUI_X11 */ 4941 4942 #if defined(FEAT_GUI_GTK) || defined(PROTO) 4943 4944 # ifdef FEAT_XFONTSET 4945 static char e_overthespot[] = N_("E290: over-the-spot style requires fontset"); 4946 # endif 4947 4948 # ifdef PROTO 4949 typedef int GdkIC; 4950 # endif 4951 4952 void 4953 xim_decide_input_style() 4954 { 4955 /* GDK_IM_STATUS_CALLBACKS was disabled, enabled it to allow Japanese 4956 * OverTheSpot. */ 4957 int supported_style = (int)GDK_IM_PREEDIT_NONE | 4958 (int)GDK_IM_PREEDIT_NOTHING | 4959 (int)GDK_IM_PREEDIT_POSITION | 4960 (int)GDK_IM_PREEDIT_CALLBACKS | 4961 (int)GDK_IM_STATUS_CALLBACKS | 4962 (int)GDK_IM_STATUS_AREA | 4963 (int)GDK_IM_STATUS_NONE | 4964 (int)GDK_IM_STATUS_NOTHING; 4965 4966 #ifdef XIM_DEBUG 4967 xim_log("xim_decide_input_style()\n"); 4968 #endif 4969 4970 if (!gdk_im_ready()) 4971 xim_input_style = 0; 4972 else 4973 { 4974 if (gtk_major_version > 1 4975 || (gtk_major_version == 1 4976 && (gtk_minor_version > 2 4977 || (gtk_minor_version == 2 && gtk_micro_version >= 3)))) 4978 use_status_area = TRUE; 4979 else 4980 { 4981 EMSG(_("E291: Your GTK+ is older than 1.2.3. Status area disabled")); 4982 use_status_area = FALSE; 4983 } 4984 #ifdef FEAT_XFONTSET 4985 if (gui.fontset == NOFONTSET || gui.fontset->type != GDK_FONT_FONTSET) 4986 #endif 4987 supported_style &= ~((int)GDK_IM_PREEDIT_POSITION 4988 | (int)GDK_IM_STATUS_AREA); 4989 if (!use_status_area) 4990 supported_style &= ~(int)GDK_IM_STATUS_AREA; 4991 xim_input_style = (int)gdk_im_decide_style((GdkIMStyle)supported_style); 4992 } 4993 } 4994 4995 /*ARGSUSED*/ 4996 static void 4997 preedit_start_cbproc(XIC xic, XPointer client_data, XPointer call_data) 4998 { 4999 #ifdef XIM_DEBUG 5000 xim_log("xim_decide_input_style()\n"); 5001 #endif 5002 5003 draw_feedback = NULL; 5004 xim_can_preediting = TRUE; 5005 xim_has_preediting = TRUE; 5006 gui_update_cursor(TRUE, FALSE); 5007 if (showmode() > 0) 5008 { 5009 setcursor(); 5010 out_flush(); 5011 } 5012 } 5013 5014 static void 5015 xim_back_delete(int n) 5016 { 5017 char_u str[3]; 5018 5019 str[0] = CSI; 5020 str[1] = 'k'; 5021 str[2] = 'b'; 5022 while (n-- > 0) 5023 add_to_input_buf(str, 3); 5024 } 5025 5026 static GSList *key_press_event_queue = NULL; 5027 static gboolean processing_queued_event = FALSE; 5028 5029 /*ARGSUSED*/ 5030 static void 5031 preedit_draw_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5032 { 5033 XIMPreeditDrawCallbackStruct *draw_data; 5034 XIMText *text; 5035 char *src; 5036 GSList *event_queue; 5037 5038 #ifdef XIM_DEBUG 5039 xim_log("preedit_draw_cbproc()\n"); 5040 #endif 5041 5042 draw_data = (XIMPreeditDrawCallbackStruct *) call_data; 5043 text = (XIMText *) draw_data->text; 5044 5045 if ((text == NULL && draw_data->chg_length == preedit_buf_len) 5046 || preedit_buf_len == 0) 5047 { 5048 init_preedit_start_col(); 5049 vim_free(draw_feedback); 5050 draw_feedback = NULL; 5051 } 5052 if (draw_data->chg_length > 0) 5053 { 5054 int bs_cnt; 5055 5056 if (draw_data->chg_length > preedit_buf_len) 5057 bs_cnt = preedit_buf_len; 5058 else 5059 bs_cnt = draw_data->chg_length; 5060 xim_back_delete(bs_cnt); 5061 preedit_buf_len -= bs_cnt; 5062 } 5063 if (text != NULL) 5064 { 5065 int len; 5066 #ifdef FEAT_MBYTE 5067 char_u *buf = NULL; 5068 unsigned int nfeedback = 0; 5069 #endif 5070 char_u *ptr; 5071 5072 src = text->string.multi_byte; 5073 if (src != NULL && !text->encoding_is_wchar) 5074 { 5075 len = strlen(src); 5076 ptr = (char_u *)src; 5077 /* Avoid the enter for decision */ 5078 if (*ptr == '\n') 5079 return; 5080 5081 #ifdef FEAT_MBYTE 5082 if (input_conv.vc_type != CONV_NONE 5083 && (buf = string_convert(&input_conv, 5084 (char_u *)src, &len)) != NULL) 5085 { 5086 /* Converted from 'termencoding' to 'encoding'. */ 5087 add_to_input_buf_csi(buf, len); 5088 ptr = buf; 5089 } 5090 else 5091 #endif 5092 add_to_input_buf_csi((char_u *)src, len); 5093 /* Add count of character to preedit_buf_len */ 5094 while (*ptr != NUL) 5095 { 5096 #ifdef FEAT_MBYTE 5097 if (draw_data->text->feedback != NULL) 5098 { 5099 if (draw_feedback == NULL) 5100 draw_feedback = (char *)alloc(draw_data->chg_first 5101 + text->length); 5102 else 5103 draw_feedback = realloc(draw_feedback, 5104 draw_data->chg_first + text->length); 5105 if (draw_feedback != NULL) 5106 { 5107 draw_feedback[nfeedback + draw_data->chg_first] 5108 = draw_data->text->feedback[nfeedback]; 5109 nfeedback++; 5110 } 5111 } 5112 if (has_mbyte) 5113 ptr += mb_ptr2len_check(ptr); 5114 else 5115 #endif 5116 ptr++; 5117 preedit_buf_len++; 5118 } 5119 #ifdef FEAT_MBYTE 5120 vim_free(buf); 5121 #endif 5122 preedit_end_col = MAXCOL; 5123 } 5124 } 5125 if (text != NULL || draw_data->chg_length > 0) 5126 { 5127 event_queue = key_press_event_queue; 5128 processing_queued_event = TRUE; 5129 while (event_queue != NULL && processing_queued_event) 5130 { 5131 GdkEvent *ev = event_queue->data; 5132 5133 gboolean *ret; 5134 gtk_signal_emit_by_name((GtkObject*)gui.mainwin, "key_press_event", 5135 ev, &ret); 5136 gdk_event_free(ev); 5137 event_queue = event_queue->next; 5138 } 5139 processing_queued_event = FALSE; 5140 if (key_press_event_queue) 5141 { 5142 g_slist_free(key_press_event_queue); 5143 key_press_event_queue = NULL; 5144 } 5145 } 5146 if (gtk_main_level() > 0) 5147 gtk_main_quit(); 5148 } 5149 5150 /* 5151 * Retrieve the highlighting attributes at column col in the preedit string. 5152 * Return -1 if not in preediting mode or if col is out of range. 5153 */ 5154 int 5155 im_get_feedback_attr(int col) 5156 { 5157 if (draw_feedback != NULL && col < preedit_buf_len) 5158 { 5159 if (draw_feedback[col] & XIMReverse) 5160 return HL_INVERSE; 5161 else if (draw_feedback[col] & XIMUnderline) 5162 return HL_UNDERLINE; 5163 else 5164 return hl_attr(HLF_V); 5165 } 5166 5167 return -1; 5168 } 5169 5170 /*ARGSUSED*/ 5171 static void 5172 preedit_caret_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5173 { 5174 #ifdef XIM_DEBUG 5175 xim_log("preedit_caret_cbproc()\n"); 5176 #endif 5177 } 5178 5179 /*ARGSUSED*/ 5180 static void 5181 preedit_done_cbproc(XIC xic, XPointer client_data, XPointer call_data) 5182 { 5183 #ifdef XIM_DEBUG 5184 xim_log("preedit_done_cbproc()\n"); 5185 #endif 5186 5187 vim_free(draw_feedback); 5188 draw_feedback = NULL; 5189 xim_can_preediting = FALSE; 5190 xim_has_preediting = FALSE; 5191 gui_update_cursor(TRUE, FALSE); 5192 if (showmode() > 0) 5193 { 5194 setcursor(); 5195 out_flush(); 5196 } 5197 } 5198 5199 void 5200 xim_reset(void) 5201 { 5202 char *text; 5203 5204 #ifdef XIM_DEBUG 5205 xim_log("xim_reset()\n"); 5206 #endif 5207 5208 if (xic != NULL) 5209 { 5210 text = XmbResetIC(((GdkICPrivate *)xic)->xic); 5211 if (text != NULL && !(xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS)) 5212 add_to_input_buf_csi((char_u *)text, strlen(text)); 5213 else 5214 preedit_buf_len = 0; 5215 if (text != NULL) 5216 XFree(text); 5217 } 5218 } 5219 5220 /*ARGSUSED*/ 5221 int 5222 xim_queue_key_press_event(GdkEventKey *event, int down) 5223 { 5224 #ifdef XIM_DEBUG 5225 xim_log("xim_queue_key_press_event()\n"); 5226 #endif 5227 5228 if (preedit_buf_len <= 0) 5229 return FALSE; 5230 if (processing_queued_event) 5231 processing_queued_event = FALSE; 5232 5233 key_press_event_queue = g_slist_append(key_press_event_queue, 5234 gdk_event_copy((GdkEvent *)event)); 5235 return TRUE; 5236 } 5237 5238 /*ARGSUSED*/ 5239 static void 5240 preedit_callback_setup(GdkIC *ic) 5241 { 5242 XIC xxic; 5243 XVaNestedList preedit_attr; 5244 XIMCallback preedit_start_cb; 5245 XIMCallback preedit_draw_cb; 5246 XIMCallback preedit_caret_cb; 5247 XIMCallback preedit_done_cb; 5248 5249 xxic = ((GdkICPrivate*)xic)->xic; 5250 preedit_start_cb.callback = (XIMProc)preedit_start_cbproc; 5251 preedit_draw_cb.callback = (XIMProc)preedit_draw_cbproc; 5252 preedit_caret_cb.callback = (XIMProc)preedit_caret_cbproc; 5253 preedit_done_cb.callback = (XIMProc)preedit_done_cbproc; 5254 preedit_attr 5255 = XVaCreateNestedList (0, 5256 XNPreeditStartCallback, &preedit_start_cb, 5257 XNPreeditDrawCallback, &preedit_draw_cb, 5258 XNPreeditCaretCallback, &preedit_caret_cb, 5259 XNPreeditDoneCallback, &preedit_done_cb, 5260 0); 5261 XSetICValues (xxic, XNPreeditAttributes, preedit_attr, 0); 5262 XFree(preedit_attr); 5263 } 5264 5265 /*ARGSUSED*/ 5266 static void 5267 reset_state_setup(GdkIC *ic) 5268 { 5269 #ifdef USE_X11R6_XIM 5270 /* don't change the input context when we call reset */ 5271 XSetICValues(((GdkICPrivate*)ic)->xic, XNResetState, XIMPreserveState, 0); 5272 #endif 5273 } 5274 5275 void 5276 xim_init(void) 5277 { 5278 #ifdef XIM_DEBUG 5279 xim_log("xim_init()\n"); 5280 #endif 5281 5282 xic = NULL; 5283 xic_attr = NULL; 5284 5285 if (!gdk_im_ready()) 5286 { 5287 if (p_verbose > 0) 5288 EMSG(_("E292: Input Method Server is not running")); 5289 return; 5290 } 5291 if ((xic_attr = gdk_ic_attr_new()) != NULL) 5292 { 5293 #ifdef FEAT_XFONTSET 5294 gint width, height; 5295 #endif 5296 int mask; 5297 GdkColormap *colormap; 5298 GdkICAttr *attr = xic_attr; 5299 int attrmask = (int)GDK_IC_ALL_REQ; 5300 GtkWidget *widget = gui.drawarea; 5301 5302 attr->style = (GdkIMStyle)xim_input_style; 5303 attr->client_window = gui.mainwin->window; 5304 5305 if ((colormap = gtk_widget_get_colormap(widget)) != 5306 gtk_widget_get_default_colormap()) 5307 { 5308 attrmask |= (int)GDK_IC_PREEDIT_COLORMAP; 5309 attr->preedit_colormap = colormap; 5310 } 5311 attrmask |= (int)GDK_IC_PREEDIT_FOREGROUND; 5312 attrmask |= (int)GDK_IC_PREEDIT_BACKGROUND; 5313 attr->preedit_foreground = widget->style->fg[GTK_STATE_NORMAL]; 5314 attr->preedit_background = widget->style->base[GTK_STATE_NORMAL]; 5315 5316 #ifdef FEAT_XFONTSET 5317 if ((xim_input_style & (int)GDK_IM_PREEDIT_MASK) 5318 == (int)GDK_IM_PREEDIT_POSITION) 5319 { 5320 if (gui.fontset == NOFONTSET 5321 || gui.fontset->type != GDK_FONT_FONTSET) 5322 { 5323 EMSG(_(e_overthespot)); 5324 } 5325 else 5326 { 5327 gdk_window_get_size(widget->window, &width, &height); 5328 5329 attrmask |= (int)GDK_IC_PREEDIT_POSITION_REQ; 5330 attr->spot_location.x = TEXT_X(0); 5331 attr->spot_location.y = TEXT_Y(0); 5332 attr->preedit_area.x = gui.border_offset; 5333 attr->preedit_area.y = gui.border_offset; 5334 attr->preedit_area.width = width - 2*gui.border_offset; 5335 attr->preedit_area.height = height - 2*gui.border_offset; 5336 attr->preedit_fontset = gui.fontset; 5337 } 5338 } 5339 5340 if ((xim_input_style & (int)GDK_IM_STATUS_MASK) 5341 == (int)GDK_IM_STATUS_AREA) 5342 { 5343 if (gui.fontset == NOFONTSET 5344 || gui.fontset->type != GDK_FONT_FONTSET) 5345 { 5346 EMSG(_(e_overthespot)); 5347 } 5348 else 5349 { 5350 gdk_window_get_size(gui.mainwin->window, &width, &height); 5351 attrmask |= (int)GDK_IC_STATUS_AREA_REQ; 5352 attr->status_area.x = 0; 5353 attr->status_area.y = height - gui.char_height - 1; 5354 attr->status_area.width = width; 5355 attr->status_area.height = gui.char_height; 5356 attr->status_fontset = gui.fontset; 5357 } 5358 } 5359 else if ((xim_input_style & (int)GDK_IM_STATUS_MASK) 5360 == (int)GDK_IM_STATUS_CALLBACKS) 5361 { 5362 /* FIXME */ 5363 } 5364 #endif 5365 5366 xic = gdk_ic_new(attr, (GdkICAttributesType)attrmask); 5367 5368 if (xic == NULL) 5369 EMSG(_(e_xim)); 5370 else 5371 { 5372 mask = (int)gdk_window_get_events(widget->window); 5373 mask |= (int)gdk_ic_get_events(xic); 5374 gdk_window_set_events(widget->window, (GdkEventMask)mask); 5375 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS) 5376 preedit_callback_setup(xic); 5377 reset_state_setup(xic); 5378 } 5379 } 5380 } 5381 5382 void 5383 im_shutdown(void) 5384 { 5385 #ifdef XIM_DEBUG 5386 xim_log("im_shutdown()\n"); 5387 #endif 5388 5389 if (xic != NULL) 5390 { 5391 gdk_im_end(); 5392 gdk_ic_destroy(xic); 5393 xic = NULL; 5394 } 5395 xim_is_active = FALSE; 5396 xim_can_preediting = FALSE; 5397 preedit_start_col = MAXCOL; 5398 xim_has_preediting = FALSE; 5399 } 5400 5401 #endif /* FEAT_GUI_GTK */ 5402 5403 int 5404 xim_get_status_area_height() 5405 { 5406 #ifdef FEAT_GUI_GTK 5407 if (xim_input_style & (int)GDK_IM_STATUS_AREA) 5408 return gui.char_height; 5409 #elif defined FEAT_GUI_KDE 5410 #warning FIXME 5411 #else 5412 if (status_area_enabled) 5413 return gui.char_height; 5414 #endif 5415 return 0; 5416 } 5417 5418 /* 5419 * Get IM status. When IM is on, return TRUE. Else return FALSE. 5420 * FIXME: This doesn't work correctly: Having focus doesn't always mean XIM is 5421 * active, when not having focus XIM may still be active (e.g., when using a 5422 * tear-off menu item). 5423 */ 5424 int 5425 im_get_status() 5426 { 5427 # ifdef FEAT_GUI_GTK 5428 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS) 5429 return xim_can_preediting; 5430 # endif 5431 return xim_has_focus; 5432 } 5433 5434 # endif /* !HAVE_GTK2 */ 5435 5436 # if defined(FEAT_GUI_GTK) || defined(PROTO) 5437 int 5438 im_is_preediting() 5439 { 5440 return xim_has_preediting; 5441 } 5442 # endif 5443 #endif /* FEAT_XIM */ 5444 5445 #if defined(FEAT_MBYTE) || defined(PROTO) 5446 5447 /* 5448 * Setup "vcp" for conversion from "from" to "to". 5449 * The names must have been made canonical with enc_canonize(). 5450 * vcp->vc_type must have been initialized to CONV_NONE. 5451 * Note: cannot be used for conversion from/to ucs-2 and ucs-4 (will use utf-8 5452 * instead). 5453 * Afterwards invoke with "from" and "to" equal to NULL to cleanup. 5454 * Return FAIL when conversion is not supported, OK otherwise. 5455 */ 5456 int 5457 convert_setup(vcp, from, to) 5458 vimconv_T *vcp; 5459 char_u *from; 5460 char_u *to; 5461 { 5462 int from_prop; 5463 int to_prop; 5464 5465 /* Reset to no conversion. */ 5466 # ifdef USE_ICONV 5467 if (vcp->vc_type == CONV_ICONV && vcp->vc_fd != (iconv_t)-1) 5468 iconv_close(vcp->vc_fd); 5469 # endif 5470 vcp->vc_type = CONV_NONE; 5471 vcp->vc_factor = 1; 5472 vcp->vc_fail = FALSE; 5473 5474 /* No conversion when one of the names is empty or they are equal. */ 5475 if (from == NULL || *from == NUL || to == NULL || *to == NUL 5476 || STRCMP(from, to) == 0) 5477 return OK; 5478 5479 from_prop = enc_canon_props(from); 5480 to_prop = enc_canon_props(to); 5481 if ((from_prop & ENC_LATIN1) && (to_prop & ENC_UNICODE)) 5482 { 5483 /* Internal latin1 -> utf-8 conversion. */ 5484 vcp->vc_type = CONV_TO_UTF8; 5485 vcp->vc_factor = 2; /* up to twice as long */ 5486 } 5487 else if ((from_prop & ENC_LATIN9) && (to_prop & ENC_UNICODE)) 5488 { 5489 /* Internal latin9 -> utf-8 conversion. */ 5490 vcp->vc_type = CONV_9_TO_UTF8; 5491 vcp->vc_factor = 3; /* up to three as long (euro sign) */ 5492 } 5493 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_LATIN1)) 5494 { 5495 /* Internal utf-8 -> latin1 conversion. */ 5496 vcp->vc_type = CONV_TO_LATIN1; 5497 } 5498 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_LATIN9)) 5499 { 5500 /* Internal utf-8 -> latin9 conversion. */ 5501 vcp->vc_type = CONV_TO_LATIN9; 5502 } 5503 #ifdef WIN3264 5504 /* Win32-specific codepage <-> codepage conversion without iconv. */ 5505 else if (((from_prop & ENC_UNICODE) || encname2codepage(from) > 0) 5506 && ((to_prop & ENC_UNICODE) || encname2codepage(to) > 0)) 5507 { 5508 vcp->vc_type = CONV_CODEPAGE; 5509 vcp->vc_factor = 2; /* up to twice as long */ 5510 vcp->vc_cpfrom = (from_prop & ENC_UNICODE) ? 0 : encname2codepage(from); 5511 vcp->vc_cpto = (to_prop & ENC_UNICODE) ? 0 : encname2codepage(to); 5512 } 5513 #endif 5514 #ifdef MACOS_X 5515 else if ((from_prop & ENC_MACROMAN) && (to_prop & ENC_LATIN1)) 5516 { 5517 vcp->vc_type = CONV_MAC_LATIN1; 5518 } 5519 else if ((from_prop & ENC_MACROMAN) && (to_prop & ENC_UNICODE)) 5520 { 5521 vcp->vc_type = CONV_MAC_UTF8; 5522 vcp->vc_factor = 2; /* up to twice as long */ 5523 } 5524 else if ((from_prop & ENC_LATIN1) && (to_prop & ENC_MACROMAN)) 5525 { 5526 vcp->vc_type = CONV_LATIN1_MAC; 5527 } 5528 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_MACROMAN)) 5529 { 5530 vcp->vc_type = CONV_UTF8_MAC; 5531 } 5532 #endif 5533 # ifdef USE_ICONV 5534 else 5535 { 5536 /* Use iconv() for conversion. */ 5537 vcp->vc_fd = (iconv_t)my_iconv_open( 5538 (to_prop & ENC_UNICODE) ? (char_u *)"utf-8" : to, 5539 (from_prop & ENC_UNICODE) ? (char_u *)"utf-8" : from); 5540 if (vcp->vc_fd != (iconv_t)-1) 5541 { 5542 vcp->vc_type = CONV_ICONV; 5543 vcp->vc_factor = 4; /* could be longer too... */ 5544 } 5545 } 5546 # endif 5547 if (vcp->vc_type == CONV_NONE) 5548 return FAIL; 5549 return OK; 5550 } 5551 5552 #if defined(FEAT_GUI) || defined(AMIGA) || defined(WIN3264) \ 5553 || defined(MSDOS) || defined(PROTO) 5554 /* 5555 * Do conversion on typed input characters in-place. 5556 * The input and output are not NUL terminated! 5557 * Returns the length after conversion. 5558 */ 5559 int 5560 convert_input(ptr, len, maxlen) 5561 char_u *ptr; 5562 int len; 5563 int maxlen; 5564 { 5565 return convert_input_safe(ptr, len, maxlen, NULL, NULL); 5566 } 5567 #endif 5568 5569 /* 5570 * Like convert_input(), but when there is an incomplete byte sequence at the 5571 * end return that as an allocated string in "restp" and set "*restlenp" to 5572 * the length. If "restp" is NULL it is not used. 5573 */ 5574 int 5575 convert_input_safe(ptr, len, maxlen, restp, restlenp) 5576 char_u *ptr; 5577 int len; 5578 int maxlen; 5579 char_u **restp; 5580 int *restlenp; 5581 { 5582 char_u *d; 5583 int dlen = len; 5584 int unconvertlen = 0; 5585 5586 d = string_convert_ext(&input_conv, ptr, &dlen, 5587 restp == NULL ? NULL : &unconvertlen); 5588 if (d != NULL) 5589 { 5590 if (dlen <= maxlen) 5591 { 5592 if (unconvertlen > 0) 5593 { 5594 /* Move the unconverted characters to allocated memory. */ 5595 *restp = alloc(unconvertlen); 5596 if (*restp != NULL) 5597 mch_memmove(*restp, ptr + len - unconvertlen, unconvertlen); 5598 *restlenp = unconvertlen; 5599 } 5600 mch_memmove(ptr, d, dlen); 5601 } 5602 else 5603 /* result is too long, keep the unconverted text (the caller must 5604 * have done something wrong!) */ 5605 dlen = len; 5606 vim_free(d); 5607 } 5608 return dlen; 5609 } 5610 5611 #if defined(MACOS_X) 5612 /* This is in os_mac_conv.c. */ 5613 extern char_u *mac_string_convert __ARGS((char_u *ptr, int len, int *lenp, int fail_on_error, int from, int to, int *unconvlenp)); 5614 #endif 5615 5616 /* 5617 * Convert text "ptr[*lenp]" according to "vcp". 5618 * Returns the result in allocated memory and sets "*lenp". 5619 * When "lenp" is NULL, use NUL terminated strings. 5620 * Illegal chars are often changed to "?", unless vcp->vc_fail is set. 5621 * When something goes wrong, NULL is returned and "*lenp" is unchanged. 5622 */ 5623 char_u * 5624 string_convert(vcp, ptr, lenp) 5625 vimconv_T *vcp; 5626 char_u *ptr; 5627 int *lenp; 5628 { 5629 return string_convert_ext(vcp, ptr, lenp, NULL); 5630 } 5631 5632 /* 5633 * Like string_convert(), but when "unconvlenp" is not NULL and there are is 5634 * an incomplete sequence at the end it is not converted and "*unconvlenp" is 5635 * set to the number of remaining bytes. 5636 */ 5637 char_u * 5638 string_convert_ext(vcp, ptr, lenp, unconvlenp) 5639 vimconv_T *vcp; 5640 char_u *ptr; 5641 int *lenp; 5642 int *unconvlenp; 5643 { 5644 char_u *retval = NULL; 5645 char_u *d; 5646 int len; 5647 int i; 5648 int l; 5649 int c; 5650 5651 if (lenp == NULL) 5652 len = (int)STRLEN(ptr); 5653 else 5654 len = *lenp; 5655 if (len == 0) 5656 return vim_strsave((char_u *)""); 5657 5658 switch (vcp->vc_type) 5659 { 5660 case CONV_TO_UTF8: /* latin1 to utf-8 conversion */ 5661 retval = alloc(len * 2 + 1); 5662 if (retval == NULL) 5663 break; 5664 d = retval; 5665 for (i = 0; i < len; ++i) 5666 { 5667 c = ptr[i]; 5668 if (c < 0x80) 5669 *d++ = c; 5670 else 5671 { 5672 *d++ = 0xc0 + ((unsigned)c >> 6); 5673 *d++ = 0x80 + (c & 0x3f); 5674 } 5675 } 5676 *d = NUL; 5677 if (lenp != NULL) 5678 *lenp = (int)(d - retval); 5679 break; 5680 5681 case CONV_9_TO_UTF8: /* latin9 to utf-8 conversion */ 5682 retval = alloc(len * 3 + 1); 5683 if (retval == NULL) 5684 break; 5685 d = retval; 5686 for (i = 0; i < len; ++i) 5687 { 5688 c = ptr[i]; 5689 switch (c) 5690 { 5691 case 0xa4: c = 0x20ac; break; /* euro */ 5692 case 0xa6: c = 0x0160; break; /* S hat */ 5693 case 0xa8: c = 0x0161; break; /* S -hat */ 5694 case 0xb4: c = 0x017d; break; /* Z hat */ 5695 case 0xb8: c = 0x017e; break; /* Z -hat */ 5696 case 0xbc: c = 0x0152; break; /* OE */ 5697 case 0xbd: c = 0x0153; break; /* oe */ 5698 case 0xbe: c = 0x0178; break; /* Y */ 5699 } 5700 d += utf_char2bytes(c, d); 5701 } 5702 *d = NUL; 5703 if (lenp != NULL) 5704 *lenp = (int)(d - retval); 5705 break; 5706 5707 case CONV_TO_LATIN1: /* utf-8 to latin1 conversion */ 5708 case CONV_TO_LATIN9: /* utf-8 to latin9 conversion */ 5709 retval = alloc(len + 1); 5710 if (retval == NULL) 5711 break; 5712 d = retval; 5713 for (i = 0; i < len; ++i) 5714 { 5715 l = utf_ptr2len_check(ptr + i); 5716 if (l == 0) 5717 *d++ = NUL; 5718 else if (l == 1) 5719 { 5720 if (unconvlenp != NULL && utf8len_tab[ptr[i]] > len - i) 5721 { 5722 /* Incomplete sequence at the end. */ 5723 *unconvlenp = len - i; 5724 break; 5725 } 5726 *d++ = ptr[i]; 5727 } 5728 else 5729 { 5730 c = utf_ptr2char(ptr + i); 5731 if (vcp->vc_type == CONV_TO_LATIN9) 5732 switch (c) 5733 { 5734 case 0x20ac: c = 0xa4; break; /* euro */ 5735 case 0x0160: c = 0xa6; break; /* S hat */ 5736 case 0x0161: c = 0xa8; break; /* S -hat */ 5737 case 0x017d: c = 0xb4; break; /* Z hat */ 5738 case 0x017e: c = 0xb8; break; /* Z -hat */ 5739 case 0x0152: c = 0xbc; break; /* OE */ 5740 case 0x0153: c = 0xbd; break; /* oe */ 5741 case 0x0178: c = 0xbe; break; /* Y */ 5742 case 0xa4: 5743 case 0xa6: 5744 case 0xa8: 5745 case 0xb4: 5746 case 0xb8: 5747 case 0xbc: 5748 case 0xbd: 5749 case 0xbe: c = 0x100; break; /* not in latin9 */ 5750 } 5751 if (!utf_iscomposing(c)) /* skip composing chars */ 5752 { 5753 if (c < 0x100) 5754 *d++ = c; 5755 else if (vcp->vc_fail) 5756 { 5757 vim_free(retval); 5758 return NULL; 5759 } 5760 else 5761 { 5762 *d++ = 0xbf; 5763 if (utf_char2cells(c) > 1) 5764 *d++ = '?'; 5765 } 5766 } 5767 i += l - 1; 5768 } 5769 } 5770 *d = NUL; 5771 if (lenp != NULL) 5772 *lenp = (int)(d - retval); 5773 break; 5774 5775 # ifdef MACOS_X 5776 case CONV_MAC_LATIN1: 5777 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 5778 'm', 'l', unconvlenp); 5779 break; 5780 5781 case CONV_LATIN1_MAC: 5782 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 5783 'l', 'm', unconvlenp); 5784 break; 5785 5786 case CONV_MAC_UTF8: 5787 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 5788 'm', 'u', unconvlenp); 5789 break; 5790 5791 case CONV_UTF8_MAC: 5792 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail, 5793 'u', 'm', unconvlenp); 5794 break; 5795 # endif 5796 5797 # ifdef USE_ICONV 5798 case CONV_ICONV: /* conversion with output_conv.vc_fd */ 5799 retval = iconv_string(vcp, ptr, len, unconvlenp); 5800 if (retval != NULL && lenp != NULL) 5801 *lenp = (int)STRLEN(retval); 5802 break; 5803 # endif 5804 # ifdef WIN3264 5805 case CONV_CODEPAGE: /* codepage -> codepage */ 5806 { 5807 int retlen; 5808 int tmp_len; 5809 short_u *tmp; 5810 5811 /* 1. codepage/UTF-8 -> ucs-2. */ 5812 if (vcp->vc_cpfrom == 0) 5813 tmp_len = utf8_to_ucs2(ptr, len, NULL, NULL); 5814 else 5815 tmp_len = MultiByteToWideChar(vcp->vc_cpfrom, 0, 5816 ptr, len, 0, 0); 5817 tmp = (short_u *)alloc(sizeof(short_u) * tmp_len); 5818 if (tmp == NULL) 5819 break; 5820 if (vcp->vc_cpfrom == 0) 5821 utf8_to_ucs2(ptr, len, tmp, unconvlenp); 5822 else 5823 MultiByteToWideChar(vcp->vc_cpfrom, 0, ptr, len, tmp, tmp_len); 5824 5825 /* 2. ucs-2 -> codepage/UTF-8. */ 5826 if (vcp->vc_cpto == 0) 5827 retlen = ucs2_to_utf8(tmp, tmp_len, NULL); 5828 else 5829 retlen = WideCharToMultiByte(vcp->vc_cpto, 0, 5830 tmp, tmp_len, 0, 0, 0, 0); 5831 retval = alloc(retlen + 1); 5832 if (retval != NULL) 5833 { 5834 if (vcp->vc_cpto == 0) 5835 ucs2_to_utf8(tmp, tmp_len, retval); 5836 else 5837 WideCharToMultiByte(vcp->vc_cpto, 0, 5838 tmp, tmp_len, retval, retlen, 0, 0); 5839 retval[retlen] = NUL; 5840 if (lenp != NULL) 5841 *lenp = retlen; 5842 } 5843 vim_free(tmp); 5844 break; 5845 } 5846 # endif 5847 } 5848 5849 return retval; 5850 } 5851 #endif 5852