iconv/iconv_charmap.c

   1 /* Convert using charmaps and possibly iconv().
   2    Copyright (C) 2001-2024 Free Software Foundation, Inc.
   3    This file is part of the GNU C Library.
   4
   5    This program is free software; you can redistribute it and/or modify
   6    it under the terms of the GNU General Public License as published
   7    by the Free Software Foundation; version 2 of the License, or
   8    (at your option) any later version.
   9
  10    This program is distributed in the hope that it will be useful,
  11    but WITHOUT ANY WARRANTY; without even the implied warranty of
  12    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  13    GNU General Public License for more details.
  14
  15    You should have received a copy of the GNU General Public License
  16    along with this program; if not, see <https://www.gnu.org/licenses/>.  */
  17
  18 #include <assert.h>
  19 #include <errno.h>
  20 #include <error.h>
  21 #include <fcntl.h>
  22 #include <iconv.h>
  23 #include <libintl.h>
  24 #include <stdio.h>
  25 #include <stdlib.h>
  26 #include <unistd.h>
  27 #include <stdint.h>
  28 #include <sys/mman.h>
  29 #include <sys/stat.h>
  30
  31 #include "iconv_prog.h"
  32
  33
  34 /* Prototypes for a few program-wide used functions.  */
  35 #include <programs/xmalloc.h>
  36
  37
  38 struct convtable
  39 {
  40   int term[256 / 8];
  41   union
  42   {
  43     struct convtable *sub;
  44     struct charseq *out;
  45   } val[256];
  46 };
  47
  48
  49 static inline struct convtable *
  50 allocate_table (void)
  51 {
  52   return (struct convtable *) xcalloc (1, sizeof (struct convtable));
  53 }
  54
  55 static inline void
  56 free_table (struct convtable *tbl)
  57 {
  58   free (tbl);
  59 }
  60
  61
  62 static inline int
  63 is_term (struct convtable *tbl, unsigned int idx)
  64 {
  65   return tbl->term[idx / 8] & (1 << (idx % 8));
  66 }
  67
  68
  69 static inline void
  70 clear_term (struct convtable *tbl, unsigned int idx)
  71 {
  72   tbl->term[idx / 8] &= ~(1 << (idx % 8));
  73 }
  74
  75
  76 static inline void
  77 set_term (struct convtable *tbl, unsigned int idx)
  78 {
  79   tbl->term[idx / 8] |= 1 << (idx % 8);
  80 }
  81
  82
  83 /* Generate the conversion table.  */
  84 static struct convtable *use_from_charmap (struct charmap_t *from_charmap,
  85                                            const char *to_code);
  86 static struct convtable *use_to_charmap (const char *from_code,
  87                                          struct charmap_t *to_charmap);
  88 static struct convtable *use_both_charmaps (struct charmap_t *from_charmap,
  89                                             struct charmap_t *to_charmap);
  90
  91 /* Prototypes for the functions doing the actual work.  */
  92 static int process_block (struct convtable *tbl, char *addr, size_t len,
  93                           FILE *output);
  94 static int process_fd (struct convtable *tbl, int fd, FILE *output);
  95 static int process_file (struct convtable *tbl, FILE *input, FILE *output);
  96
  97
  98 int
  99 charmap_conversion (const char *from_code, struct charmap_t *from_charmap,
 100                     const char *to_code, struct charmap_t *to_charmap,
 101                     int argc, int remaining, char *argv[],
 102                     const char *output_file)
 103 {
 104   struct convtable *cvtbl;
 105   int status = EXIT_SUCCESS;
 106
 107   /* We have three different cases to handle:
 108
 109      - both, from_charmap and to_charmap, are available.  This means we
 110        can assume that the symbolic names match and use them to create
 111        the mapping.
 112
 113      - only from_charmap is available.  In this case we can only hope that
 114        the symbolic names used are of the <Uxxxx> form in which case we
 115        can use a UCS4->"to_code" iconv() conversion for the second step.
 116
 117      - only to_charmap is available.  This is similar, only that we would
 118        use iconv() for the "to_code"->UCS4 conversion.
 119
 120        We first create a table which maps input bytes into output bytes.
 121        Once this is done we can handle all three of the cases above
 122        equally.  */
 123   if (from_charmap != NULL)
 124     {
 125       if (to_charmap == NULL)
 126         cvtbl = use_from_charmap (from_charmap, to_code);
 127       else
 128         cvtbl = use_both_charmaps (from_charmap, to_charmap);
 129     }
 130   else
 131     {
 132       assert (to_charmap != NULL);
 133       cvtbl = use_to_charmap (from_code, to_charmap);
 134     }
 135
 136   /* If we couldn't generate a table stop now.  */
 137   if (cvtbl == NULL)
 138     return EXIT_FAILURE;
 139
 140   /* Determine output file.  */
 141   FILE *output;
 142   if (output_file != NULL && strcmp (output_file, "-") != 0)
 143     {
 144       output = fopen (output_file, "w");
 145       if (output == NULL)
 146         error (EXIT_FAILURE, errno, _("cannot open output file"));
 147     }
 148   else
 149     output = stdout;
 150
 151   /* We can now start the conversion.  */
 152   if (remaining == argc)
 153     {
 154       if (process_file (cvtbl, stdin, output) != 0)
 155         status = EXIT_FAILURE;
 156     }
 157   else
 158     do
 159       {
 160         int fd;
 161
 162         if (verbose)
 163           printf ("%s:\n", argv[remaining]);
 164         if (strcmp (argv[remaining], "-") == 0)
 165           fd = 0;
 166         else
 167           {
 168             fd = open (argv[remaining], O_RDONLY);
 169
 170             if (fd == -1)
 171               {
 172                 error (0, errno, _("cannot open input file `%s'"),
 173                        argv[remaining]);
 174                 status = EXIT_FAILURE;
 175                 continue;
 176               }
 177           }
 178
 179 #ifdef _POSIX_MAPPED_FILES
 180         struct stat64 st;
 181         char *addr;
 182         /* We have possibilities for reading the input file.  First try
 183            to mmap() it since this will provide the fastest solution.  */
 184         if (fstat64 (fd, &st) == 0
 185             && ((addr = mmap (NULL, st.st_size, PROT_READ, MAP_PRIVATE,
 186                               fd, 0)) != MAP_FAILED))
 187           {
 188             /* Yes, we can use mmap().  The descriptor is not needed
 189                anymore.  */
 190             if (close (fd) != 0)
 191               error (EXIT_FAILURE, errno,
 192                      _("error while closing input `%s'"), argv[remaining]);
 193
 194             if (process_block (cvtbl, addr, st.st_size, output) < 0)
 195               {
 196                 /* Something went wrong.  */
 197                 status = EXIT_FAILURE;
 198
 199                 /* We don't need the input data anymore.  */
 200                 munmap ((void *) addr, st.st_size);
 201
 202                 /* We cannot go on with producing output since it might
 203                    lead to problem because the last output might leave
 204                    the output stream in an undefined state.  */
 205                 break;
 206               }
 207
 208             /* We don't need the input data anymore.  */
 209             munmap ((void *) addr, st.st_size);
 210           }
 211         else
 212 #endif  /* _POSIX_MAPPED_FILES */
 213           {
 214             /* Read the file in pieces.  */
 215             if (process_fd (cvtbl, fd, output) != 0)
 216               {
 217                 /* Something went wrong.  */
 218                 status = EXIT_FAILURE;
 219
 220                 /* We don't need the input file anymore.  */
 221                 close (fd);
 222
 223                 /* We cannot go on with producing output since it might
 224                    lead to problem because the last output might leave
 225                    the output stream in an undefined state.  */
 226                 break;
 227               }
 228
 229             /* Now close the file.  */
 230             close (fd);
 231           }
 232       }
 233     while (++remaining < argc);
 234
 235   /* All done.  */
 236   if (output != stdout)
 237     fclose (output);
 238   free_table (cvtbl);
 239   return status;
 240 }
 241
 242
 243 /* Add the IN->OUT mapping to TBL.  OUT is potentially stored in the table.
 244    IN is used only here, so it need not be kept live afterwards.  */
 245 static void
 246 add_bytes (struct convtable *tbl, const struct charseq *in, struct charseq *out)
 247 {
 248   int n = 0;
 249   unsigned int byte;
 250
 251   assert (in->nbytes > 0);
 252
 253   byte = ((unsigned char *) in->bytes)[n];
 254   while (n + 1 < in->nbytes)
 255     {
 256       if (is_term (tbl, byte) || tbl->val[byte].sub == NULL)
 257         {
 258           /* Note that we simply ignore a definition for a byte sequence
 259              which is also the prefix for a longer one.  */
 260           clear_term (tbl, byte);
 261           tbl->val[byte].sub =
 262             (struct convtable *) xcalloc (1, sizeof (struct convtable));
 263         }
 264
 265       tbl = tbl->val[byte].sub;
 266
 267       byte = ((unsigned char *) in->bytes)[++n];
 268     }
 269
 270   /* Only add the new sequence if there is none yet and the byte sequence
 271      is not part of an even longer one.  */
 272   if (! is_term (tbl, byte) && tbl->val[byte].sub == NULL)
 273     {
 274       set_term (tbl, byte);
 275       tbl->val[byte].out = out;
 276     }
 277 }
 278
 279 /* Try to convert SEQ from WCHAR_T format using CD.
 280    Returns a malloc'd struct or NULL.  */
 281 static struct charseq *
 282 convert_charseq (iconv_t cd, const struct charseq *seq)
 283 {
 284   struct charseq *result = NULL;
 285
 286   if (seq->ucs4 != UNINITIALIZED_CHAR_VALUE)
 287     {
 288       /* There is a chance.  Try the iconv module.  */
 289       wchar_t inbuf[1] = { seq->ucs4 };
 290       unsigned char outbuf[64];
 291       char *inptr = (char *) inbuf;
 292       size_t inlen = sizeof (inbuf);
 293       char *outptr = (char *) outbuf;
 294       size_t outlen = sizeof (outbuf);
 295
 296       (void) iconv (cd, &inptr, &inlen, &outptr, &outlen);
 297
 298       if (outptr != (char *) outbuf)
 299         {
 300           /* We got some output.  Good, use it.  */
 301           outlen = sizeof (outbuf) - outlen;
 302           assert ((char *) outbuf + outlen == outptr);
 303
 304           result = xmalloc (sizeof (struct charseq) + outlen);
 305           result->name = seq->name;
 306           result->ucs4 = seq->ucs4;
 307           result->nbytes = outlen;
 308           memcpy (result->bytes, outbuf, outlen);
 309         }
 310
 311       /* Clear any possible state left behind.  */
 312       (void) iconv (cd, NULL, NULL, NULL, NULL);
 313     }
 314
 315   return result;
 316 }
 317
 318
 319 static struct convtable *
 320 use_from_charmap (struct charmap_t *from_charmap, const char *to_code)
 321 {
 322   /* We iterate over all entries in the from_charmap and for those which
 323      have a known UCS4 representation we use an iconv() call to determine
 324      the mapping to the to_code charset.  */
 325   struct convtable *rettbl;
 326   iconv_t cd;
 327   void *ptr = NULL;
 328   const void *key;
 329   size_t keylen;
 330   void *data;
 331
 332   cd = iconv_open (to_code, "WCHAR_T");
 333   if (cd == (iconv_t) -1)
 334     /* We cannot do anything.  */
 335     return NULL;
 336
 337   rettbl = allocate_table ();
 338
 339   while (iterate_table (&from_charmap->char_table, &ptr, &key, &keylen, &data)
 340          >= 0)
 341     {
 342       struct charseq *in = data;
 343       struct charseq *newp = convert_charseq (cd, in);
 344       if (newp != NULL)
 345         add_bytes (rettbl, in, newp);
 346     }
 347
 348   iconv_close (cd);
 349
 350   return rettbl;
 351 }
 352
 353
 354 static struct convtable *
 355 use_to_charmap (const char *from_code, struct charmap_t *to_charmap)
 356 {
 357   /* We iterate over all entries in the to_charmap and for those which
 358      have a known UCS4 representation we use an iconv() call to determine
 359      the mapping to the from_code charset.  */
 360   struct convtable *rettbl;
 361   iconv_t cd;
 362   void *ptr = NULL;
 363   const void *key;
 364   size_t keylen;
 365   void *data;
 366
 367   /* Note that the conversion we use here is the reverse direction.  Without
 368      exhaustive search we cannot figure out which input yields the UCS4
 369      character we are looking for.  Therefore we determine it the other
 370      way round.  */
 371   cd = iconv_open (from_code, "WCHAR_T");
 372   if (cd == (iconv_t) -1)
 373     /* We cannot do anything.  */
 374     return NULL;
 375
 376   rettbl = allocate_table ();
 377
 378   while (iterate_table (&to_charmap->char_table, &ptr, &key, &keylen, &data)
 379          >= 0)
 380     {
 381       struct charseq *out = data;
 382       struct charseq *newp = convert_charseq (cd, out);
 383       if (newp != NULL)
 384         {
 385           add_bytes (rettbl, newp, out);
 386           free (newp);
 387         }
 388     }
 389
 390   iconv_close (cd);
 391
 392   return rettbl;
 393 }
 394
 395
 396 static struct convtable *
 397 use_both_charmaps (struct charmap_t *from_charmap,
 398                    struct charmap_t *to_charmap)
 399 {
 400   /* In this case we iterate over all the entries in the from_charmap,
 401      determine the internal name, and find an appropriate entry in the
 402      to_charmap (if it exists).  */
 403   struct convtable *rettbl = allocate_table ();
 404   void *ptr = NULL;
 405   const void *key;
 406   size_t keylen;
 407   void *data;
 408
 409   while (iterate_table (&from_charmap->char_table, &ptr, &key, &keylen, &data)
 410          >= 0)
 411     {
 412       struct charseq *in = (struct charseq *) data;
 413       struct charseq *out = charmap_find_value (to_charmap, key, keylen);
 414
 415       if (out != NULL)
 416         add_bytes (rettbl, in, out);
 417     }
 418
 419   return rettbl;
 420 }
 421
 422
 423 static int
 424 process_block (struct convtable *tbl, char *addr, size_t len, FILE *output)
 425 {
 426   size_t n = 0;
 427
 428   while (n < len)
 429     {
 430       struct convtable *cur = tbl;
 431       unsigned char *curp = (unsigned char *) addr;
 432       unsigned int byte = *curp;
 433       int cnt;
 434       struct charseq *out;
 435
 436       while (! is_term (cur, byte))
 437         if (cur->val[byte].sub == NULL)
 438           {
 439             /* This is an invalid sequence.  Skip the first byte if we are
 440                ignoring errors.  Otherwise punt.  */
 441             if (! omit_invalid)
 442               {
 443                 error (0, 0, _("illegal input sequence at position %zd"), n);
 444                 return -1;
 445               }
 446
 447             n -= curp - (unsigned char *) addr;
 448
 449             byte = *(curp = (unsigned char *) ++addr);
 450             if (++n >= len)
 451               /* All converted.  */
 452               return 0;
 453
 454             cur = tbl;
 455           }
 456         else
 457           {
 458             cur = cur->val[byte].sub;
 459
 460             if (++n >= len)
 461               {
 462                 error (0, 0, _("\
 463 incomplete character or shift sequence at end of buffer"));
 464                 return -1;
 465               }
 466
 467             byte = *++curp;
 468           }
 469
 470       /* We found a final byte.  Write the output bytes.  */
 471       out = cur->val[byte].out;
 472       for (cnt = 0; cnt < out->nbytes; ++cnt)
 473         fputc_unlocked (out->bytes[cnt], output);
 474
 475       addr = (char *) curp + 1;
 476       ++n;
 477     }
 478
 479   return 0;
 480 }
 481
 482
 483 static int
 484 process_fd (struct convtable *tbl, int fd, FILE *output)
 485 {
 486   /* We have a problem with reading from a descriptor since we must not
 487      provide the iconv() function an incomplete character or shift
 488      sequence at the end of the buffer.  Since we have to deal with
 489      arbitrary encodings we must read the whole text in a buffer and
 490      process it in one step.  */
 491   static char *inbuf = NULL;
 492   static size_t maxlen = 0;
 493   char *inptr = inbuf;
 494   size_t actlen = 0;
 495
 496   while (actlen < maxlen)
 497     {
 498       ssize_t n = read (fd, inptr, maxlen - actlen);
 499
 500       if (n == 0)
 501         /* No more text to read.  */
 502         break;
 503
 504       if (n == -1)
 505         {
 506           /* Error while reading.  */
 507           error (0, errno, _("error while reading the input"));
 508           return -1;
 509         }
 510
 511       inptr += n;
 512       actlen += n;
 513     }
 514
 515   if (actlen == maxlen)
 516     while (1)
 517       {
 518         ssize_t n;
 519         char *new_inbuf;
 520
 521         /* Increase the buffer.  */
 522         new_inbuf = (char *) realloc (inbuf, maxlen + 32768);
 523         if (new_inbuf == NULL)
 524           {
 525             error (0, errno, _("unable to allocate buffer for input"));
 526             return -1;
 527           }
 528         inbuf = new_inbuf;
 529         maxlen += 32768;
 530         inptr = inbuf + actlen;
 531
 532         do
 533           {
 534             n = read (fd, inptr, maxlen - actlen);
 535
 536             if (n == 0)
 537               /* No more text to read.  */
 538               break;
 539
 540             if (n == -1)
 541               {
 542                 /* Error while reading.  */
 543                 error (0, errno, _("error while reading the input"));
 544                 return -1;
 545               }
 546
 547             inptr += n;
 548             actlen += n;
 549           }
 550         while (actlen < maxlen);
 551
 552         if (n == 0)
 553           /* Break again so we leave both loops.  */
 554           break;
 555       }
 556
 557   /* Now we have all the input in the buffer.  Process it in one run.  */
 558   return process_block (tbl, inbuf, actlen, output);
 559 }
 560
 561
 562 static int
 563 process_file (struct convtable *tbl, FILE *input, FILE *output)
 564 {
 565   /* This should be safe since we use this function only for `stdin' and
 566      we haven't read anything so far.  */
 567   return process_fd (tbl, fileno (input), output);
 568 }