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