src/ISpell.cpp

   1 /**
   2  * \file ISpell.cpp
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author unknown
   7  * \author Angus Leeming
   8  * \author John Levon
   9  *
  10  * Full author contact details are available in file CREDITS.
  11  */
  12
  13 #include <config.h>
  14
  15 #include "ISpell.h"
  16
  17 #include "BufferParams.h"
  18 #include "Encoding.h"
  19 #include "Language.h"
  20 #include "LyXRC.h"
  21 #include "WordLangTuple.h"
  22
  23 #include "support/debug.h"
  24 #include "support/gettext.h"
  25 #include "support/ForkedCalls.h"
  26 #include "support/lstrings.h"
  27 #include "support/unicode.h"
  28
  29 // HP-UX 11.x doesn't have this header
  30 #ifdef HAVE_SYS_SELECT_H
  31 # include <sys/select.h>
  32 #endif
  33 #ifdef HAVE_SYS_TIME_H
  34 # include <sys/time.h>
  35 #endif
  36
  37 using boost::shared_ptr;
  38
  39 using namespace std;
  40 using namespace lyx::support;
  41
  42 namespace lyx {
  43
  44 namespace {
  45
  46 class LaunchIspell : public ForkedProcess
  47 {
  48 public:
  49         ///
  50         LaunchIspell(BufferParams const & p, string const & l,
  51                      int * in, int * out, int * err)
  52                 : params(p), lang(l), pipein(in), pipeout(out), pipeerr(err) {}
  53         ///
  54         virtual shared_ptr<ForkedProcess> clone() const {
  55                 return shared_ptr<ForkedProcess>(new LaunchIspell(*this));
  56         }
  57         ///
  58         int start();
  59 private:
  60         ///
  61         virtual int generateChild();
  62
  63         ///
  64         BufferParams const & params;
  65         string const & lang;
  66         int * const pipein;
  67         int * const pipeout;
  68         int * const pipeerr;
  69 };
  70
  71
  72 int LaunchIspell::start()
  73 {
  74         command_ = lyxrc.isp_command;
  75         return run(DontWait);
  76 }
  77
  78
  79 int LaunchIspell::generateChild()
  80 {
  81         pid_t isp_pid = fork();
  82
  83         if (isp_pid != 0) {
  84                 // failed (-1) or parent process (>0)
  85                 return isp_pid;
  86         }
  87
  88         // child process
  89         dup2(pipein[0], STDIN_FILENO);
  90         dup2(pipeout[1], STDOUT_FILENO);
  91         dup2(pipeerr[1], STDERR_FILENO);
  92         close(pipein[0]);
  93         close(pipein[1]);
  94         close(pipeout[0]);
  95         close(pipeout[1]);
  96         close(pipeerr[0]);
  97         close(pipeerr[1]);
  98
  99         char * argv[14];
 100         int argc = 0;
 101
 102         char * tmp = new char[lyxrc.isp_command.length() + 1];
 103         lyxrc.isp_command.copy(tmp, lyxrc.isp_command.length());
 104         tmp[lyxrc.isp_command.length()] = '\0';
 105         argv[argc++] = tmp;
 106         tmp = new char[3];
 107         string("-a").copy(tmp, 2); tmp[2] = '\0'; // pipe mode
 108         argv[argc++] = tmp;
 109
 110         if (lang != "default") {
 111                 tmp = new char[3];
 112                 string("-d").copy(tmp, 2); tmp[2] = '\0'; // Dictionary file
 113                 argv[argc++] = tmp;
 114                 tmp = new char[lang.length() + 1];
 115                 lang.copy(tmp, lang.length()); tmp[lang.length()] = '\0';
 116                 argv[argc++] = tmp;
 117         }
 118
 119         if (lyxrc.isp_accept_compound) {
 120                 // Consider run-together words as legal compounds
 121                 tmp = new char[3];
 122                 string("-C").copy(tmp, 2); tmp[2] = '\0';
 123                 argv[argc++] = tmp;
 124         } else {
 125                 // Report run-together words with
 126                 // missing blanks as errors
 127                 tmp = new char[3];
 128                 string("-B").copy(tmp, 2); tmp[2] = '\0';
 129                 argv[argc++] = tmp;
 130         }
 131         if (lyxrc.isp_use_esc_chars) {
 132                 // Specify additional characters that
 133                 // can be part of a word
 134                 tmp = new char[3];
 135                 string("-w").copy(tmp, 2); tmp[2] = '\0';
 136                 argv[argc++] = tmp;
 137                 // Put the escape chars in ""s
 138                 string tms = '"' + lyxrc.isp_esc_chars + '"';
 139                 tmp = new char[tms.length() + 1];
 140                 tms.copy(tmp, tms.length()); tmp[tms.length()] = '\0';
 141                 argv[argc++] = tmp;
 142         }
 143         if (lyxrc.isp_use_pers_dict) {
 144                 // Specify an alternate personal dictionary
 145                 tmp = new char[3];
 146                 string("-p").copy(tmp, 2);
 147                 tmp[2]= '\0';
 148                 argv[argc++] = tmp;
 149                 tmp = new char[lyxrc.isp_pers_dict.length() + 1];
 150                 lyxrc.isp_pers_dict.copy(tmp, lyxrc.isp_pers_dict.length());
 151                 tmp[lyxrc.isp_pers_dict.length()] = '\0';
 152                 argv[argc++] = tmp;
 153         }
 154         if (lyxrc.isp_use_input_encoding &&
 155             params.inputenc != "default") {
 156                 string enc = (params.inputenc == "auto")
 157                         ? params.language->encoding()->latexName()
 158                         : params.inputenc;
 159                 size_t const n = enc.length();
 160                 tmp = new char[3];
 161                 string("-T").copy(tmp, 2);
 162                 tmp[2] = '\0';
 163                 argv[argc++] = tmp; // Input encoding
 164                 tmp = new char[n + 1];
 165                 enc.copy(tmp, n);
 166                 tmp[n] = '\0';
 167                 argv[argc++] = tmp;
 168         }
 169
 170         argv[argc++] = 0;
 171
 172         execvp(argv[0], const_cast<char * const *>(argv));
 173
 174         // free the memory used by string::copy in the
 175         // setup of argv
 176         for (int i = 0; i < argc - 1; ++i)
 177                 delete[] argv[i];
 178
 179         lyxerr << "LyX: Failed to start ispell!" << endl;
 180         _exit(0);
 181 }
 182
 183
 184 string const to_iconv_encoding(docstring const & s, string const & encoding)
 185 {
 186         if (lyxrc.isp_use_input_encoding) {
 187                 vector<char> const encoded =
 188                         ucs4_to_eightbit(s.data(), s.length(), encoding);
 189                 return string(encoded.begin(), encoded.end());
 190         }
 191         // FIXME UNICODE: we don't need to convert to UTF8, but probably to the locale encoding
 192         return to_utf8(s);
 193 }
 194
 195
 196 docstring const from_iconv_encoding(string const & s, string const & encoding)
 197 {
 198         if (lyxrc.isp_use_input_encoding) {
 199                 vector<char_type> const ucs4 =
 200                         eightbit_to_ucs4(s.data(), s.length(), encoding);
 201                 return docstring(ucs4.begin(), ucs4.end());
 202         }
 203         // FIXME UNICODE: s is not in UTF8, but probably the locale encoding
 204         return from_utf8(s);
 205 }
 206
 207 } // namespace anon
 208
 209
 210 ISpell::ISpell(BufferParams const & params, string const & lang)
 211         : in(0), out(0), inerr(0), str(0)
 212 {
 213         //LYXERR(Debug::GUI, "Created ispell");
 214
 215         encoding = params.encoding().iconvName();
 216
 217         // static due to the setvbuf. Ugly.
 218         static char o_buf[BUFSIZ];
 219
 220         // We need to throw an exception not do this
 221         pipein[0] = pipein[1] = pipeout[0] = pipeout[1]
 222                 = pipeerr[0] = pipeerr[1] = -1;
 223
 224         // This is what happens when goto gets banned.
 225
 226         if (pipe(pipein) == -1) {
 227                 error_ = _("Can't create pipe for spellchecker.");
 228                 return;
 229         }
 230
 231         if (pipe(pipeout) == -1) {
 232                 close(pipein[0]);
 233                 close(pipein[1]);
 234                 error_ = _("Can't create pipe for spellchecker.");
 235                 return;
 236         }
 237
 238         if (pipe(pipeerr) == -1) {
 239                 close(pipein[0]);
 240                 close(pipein[1]);
 241                 close(pipeout[0]);
 242                 close(pipeout[1]);
 243                 error_ = _("Can't create pipe for spellchecker.");
 244                 return;
 245         }
 246
 247         if ((out = fdopen(pipein[1], "w")) == 0) {
 248                 error_ = _("Can't open pipe for spellchecker.");
 249                 return;
 250         }
 251
 252         if ((in = fdopen(pipeout[0], "r")) == 0) {
 253                 error_ = _("Can't open pipe for spellchecker.");
 254                 return;
 255         }
 256
 257         if ((inerr = fdopen(pipeerr[0], "r")) == 0) {
 258                 error_ = _("Can't open pipe for spellchecker.");
 259                 return;
 260         }
 261
 262         setvbuf(out, o_buf, _IOLBF, BUFSIZ);
 263
 264         LaunchIspell * li = new LaunchIspell(params, lang, pipein, pipeout, pipeerr);
 265         child_.reset(li);
 266         if (li->start() == -1) {
 267                 error_ = _("Could not create an ispell process.\nYou may not have "
 268                                         "the right languages installed.");
 269                 child_.reset(0);
 270                 return;
 271         }
 272
 273         // Parent process: Read ispells identification message
 274
 275         bool err_read;
 276         bool error = select(err_read);
 277
 278         if (!error) {
 279                 if (!err_read) {
 280                         // Set terse mode (silently accept correct words)
 281                         fputs("!\n", out);
 282                         return;
 283                 }
 284
 285                 // must have read something from stderr
 286                 // FIXME UNICODE: buf is not in UTF8, but probably the locale encoding
 287                 error_ = from_utf8(buf);
 288         } else {
 289                 // select returned error
 290                 error_ = _("The ispell process returned an error.\nPerhaps "
 291                            "it has been configured wrongly ?");
 292         }
 293
 294         close(pipein[0]);
 295         close(pipein[1]);
 296         close(pipeout[0]);
 297         close(pipeout[1]);
 298         close(pipeerr[0]);
 299         close(pipeerr[1]);
 300         child_->kill();
 301         child_.reset(0);
 302 }
 303
 304
 305 ISpell::~ISpell()
 306 {
 307         //LYXERR(Debug::GUI, "Killing ispell");
 308
 309         if (in)
 310                 fclose(in);
 311
 312         if (inerr)
 313                 fclose(inerr);
 314
 315         if (out) {
 316                 fputs("#\n", out); // Save personal dictionary
 317
 318                 fflush(out);
 319                 fclose(out);
 320         }
 321
 322         close(pipein[0]);
 323         close(pipein[1]);
 324         close(pipeout[0]);
 325         close(pipeout[1]);
 326         close(pipeerr[0]);
 327         close(pipeerr[1]);
 328         delete [] str;
 329 }
 330
 331
 332 bool ISpell::select(bool & err_read)
 333 {
 334         fd_set infds;
 335         struct timeval tv;
 336         int retval = 0;
 337         FD_ZERO(&infds);
 338         FD_SET(pipeout[0], &infds);
 339         FD_SET(pipeerr[0], &infds);
 340         tv.tv_sec = 2;
 341         tv.tv_usec = 0;
 342
 343         retval = ::select(SELECT_TYPE_ARG1 (max(pipeout[0], pipeerr[0]) + 1),
 344                         SELECT_TYPE_ARG234 (&infds),
 345                         0,
 346                         0,
 347                         SELECT_TYPE_ARG5 (&tv));
 348
 349         // error
 350         if (retval <= 0)
 351                 return true;
 352
 353         if (FD_ISSET(pipeerr[0], &infds)) {
 354                 fgets(buf, BUFSIZ, inerr);
 355                 err_read = true;
 356                 return false;
 357         }
 358
 359         fgets(buf, BUFSIZ, in);
 360         err_read = false;
 361         return false;
 362 }
 363
 364
 365 docstring const ISpell::nextMiss()
 366 {
 367         // Well, somebody is a sick fuck.
 368
 369         if (str == 0 || *(e+1) == '\0')
 370                 return docstring();
 371         char * b = e + 2;
 372         e = strpbrk(b, ",\n");
 373         *e = '\0';
 374         if (b)
 375                 return from_iconv_encoding(b, encoding);
 376         return docstring();
 377 }
 378
 379
 380 bool ISpell::alive()
 381 {
 382         return child_.get() && child_->running();
 383 }
 384
 385
 386 enum ISpell::Result ISpell::check(WordLangTuple const & word)
 387 {
 388         // FIXME Please rewrite to use string.
 389
 390         Result res;
 391
 392         string const encoded = to_iconv_encoding(word.word(), encoding);
 393         if (encoded.empty()) {
 394                 error_ = bformat(
 395                         _("Could not check word `%1$s' because it could not be converted to encoding `%2$s'."),
 396                         word.word(), from_ascii(encoding));
 397                 return UNKNOWN_WORD;
 398         }
 399         ::fputs(encoded.c_str(), out);
 400         ::fputc('\n', out);
 401
 402         bool err_read;
 403         bool error = select(err_read);
 404
 405         if (error) {
 406                 error_ = _("Could not communicate with the ispell spellchecker process.");
 407                 return UNKNOWN_WORD;
 408         }
 409
 410         if (err_read) {
 411                 // FIXME UNICODE: buf is not in UTF8, but probably the locale encoding
 412                 error_ = from_utf8(buf);
 413                 return UNKNOWN_WORD;
 414         }
 415
 416         // I think we have to check if ispell is still alive here because
 417         // the signal-handler could have disabled blocking on the fd
 418         if (!alive())
 419                 return UNKNOWN_WORD;
 420
 421         switch (*buf) {
 422         case '*':
 423                 res = OK;
 424                 break;
 425         case '+':
 426                 res = ROOT;
 427                 break;
 428         case '-':
 429                 res = COMPOUND_WORD;
 430                 break;
 431         case '\n':
 432                 res = IGNORED_WORD;
 433                 break;
 434         case '#': // Not found, no near misses and guesses
 435                 res = UNKNOWN_WORD;
 436                 break;
 437         case '?': // Not found, and no near misses, but guesses (guesses are ignored)
 438         case '&': // Not found, but we have near misses
 439         {
 440                 res = SUGGESTED_WORDS;
 441                 char * p = strpbrk(buf, ":");
 442                 str = new char[strlen(p) + 1];
 443                 e   = str;
 444                 strcpy(str, p);
 445                 break;
 446         }
 447         default: // This shouldn't happen, but you know Murphy
 448                 res = UNKNOWN_WORD;
 449         }
 450
 451         *buf = 0;
 452         if (res != IGNORED_WORD) {
 453                 /* wait for ispell to finish */
 454                 while (*buf!= '\n')
 455                         fgets(buf, 255, in);
 456         }
 457         return res;
 458 }
 459
 460
 461 void ISpell::insert(WordLangTuple const & word)
 462 {
 463         string const encoded = to_iconv_encoding(word.word(), encoding);
 464         if (encoded.empty()) {
 465                 error_ = bformat(
 466                         _("Could not insert word `%1$s' because it could not be converted to encoding `%2$s'."),
 467                         word.word(), from_ascii(encoding));
 468                 return;
 469         }
 470         ::fputc('*', out); // Insert word in personal dictionary
 471         ::fputs(encoded.c_str(), out);
 472         ::fputc('\n', out);
 473 }
 474
 475
 476 void ISpell::accept(WordLangTuple const & word)
 477 {
 478         string const encoded = to_iconv_encoding(word.word(), encoding);
 479         if (encoded.empty()) {
 480                 error_ = bformat(
 481                         _("Could not accept word `%1$s' because it could not be converted to encoding `%2$s'."),
 482                         word.word(), from_ascii(encoding));
 483                 return;
 484         }
 485         ::fputc('@', out); // Accept in this session
 486         ::fputs(encoded.c_str(), out);
 487         ::fputc('\n', out);
 488 }
 489
 490
 491 docstring const ISpell::error()
 492 {
 493         return error_;
 494 }
 495
 496
 497 } // namespace lyx