coding.c 252 KB
Newer Older
Karl Heuer's avatar
Karl Heuer committed
1
/* Coding system handler (conversion, detection, and etc).
2
   Copyright (C) 2001, 2002, 2003, 2004, 2005,
Glenn Morris's avatar
Glenn Morris committed
3
                 2006, 2007 Free Software Foundation, Inc.
Kenichi Handa's avatar
Kenichi Handa committed
4
   Copyright (C) 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004,
Glenn Morris's avatar
Glenn Morris committed
5
     2005, 2006, 2007
Kenichi Handa's avatar
Kenichi Handa committed
6 7
     National Institute of Advanced Industrial Science and Technology (AIST)
     Registration Number H14PRO021
Karl Heuer's avatar
Karl Heuer committed
8

Karl Heuer's avatar
Karl Heuer committed
9 10 11 12 13 14
This file is part of GNU Emacs.

GNU Emacs is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2, or (at your option)
any later version.
Karl Heuer's avatar
Karl Heuer committed
15

Karl Heuer's avatar
Karl Heuer committed
16 17 18 19
GNU Emacs is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.
Karl Heuer's avatar
Karl Heuer committed
20

Karl Heuer's avatar
Karl Heuer committed
21 22
You should have received a copy of the GNU General Public License
along with GNU Emacs; see the file COPYING.  If not, write to
Lute Kamstra's avatar
Lute Kamstra committed
23 24
the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
Boston, MA 02110-1301, USA.  */
Karl Heuer's avatar
Karl Heuer committed
25 26 27

/*** TABLE OF CONTENTS ***

28
  0. General comments
Karl Heuer's avatar
Karl Heuer committed
29
  1. Preamble
Kenichi Handa's avatar
Kenichi Handa committed
30
  2. Emacs' internal format (emacs-mule) handlers
Karl Heuer's avatar
Karl Heuer committed
31 32
  3. ISO2022 handlers
  4. Shift-JIS and BIG5 handlers
Kenichi Handa's avatar
Kenichi Handa committed
33 34 35 36 37
  5. CCL handlers
  6. End-of-line handlers
  7. C library functions
  8. Emacs Lisp library functions
  9. Post-amble
Karl Heuer's avatar
Karl Heuer committed
38 39 40

*/

41 42 43
/*** 0. General comments ***/


Dave Love's avatar
Dave Love committed
44
/*** GENERAL NOTE on CODING SYSTEMS ***
Karl Heuer's avatar
Karl Heuer committed
45

Dave Love's avatar
Dave Love committed
46
  A coding system is an encoding mechanism for one or more character
Karl Heuer's avatar
Karl Heuer committed
47 48
  sets.  Here's a list of coding systems which Emacs can handle.  When
  we say "decode", it means converting some other coding system to
Dave Love's avatar
Dave Love committed
49
  Emacs' internal format (emacs-mule), and when we say "encode",
Kenichi Handa's avatar
Kenichi Handa committed
50 51
  it means converting the coding system emacs-mule to some other
  coding system.
Karl Heuer's avatar
Karl Heuer committed
52

Kenichi Handa's avatar
Kenichi Handa committed
53
  0. Emacs' internal format (emacs-mule)
Karl Heuer's avatar
Karl Heuer committed
54

Dave Love's avatar
Dave Love committed
55
  Emacs itself holds a multi-lingual character in buffers and strings
Richard M. Stallman's avatar
Richard M. Stallman committed
56
  in a special format.  Details are described in section 2.
Karl Heuer's avatar
Karl Heuer committed
57 58 59 60

  1. ISO2022

  The most famous coding system for multiple character sets.  X's
Richard M. Stallman's avatar
Richard M. Stallman committed
61 62 63
  Compound Text, various EUCs (Extended Unix Code), and coding
  systems used in Internet communication such as ISO-2022-JP are
  all variants of ISO2022.  Details are described in section 3.
Karl Heuer's avatar
Karl Heuer committed
64 65

  2. SJIS (or Shift-JIS or MS-Kanji-Code)
66

Karl Heuer's avatar
Karl Heuer committed
67 68
  A coding system to encode character sets: ASCII, JISX0201, and
  JISX0208.  Widely used for PC's in Japan.  Details are described in
Richard M. Stallman's avatar
Richard M. Stallman committed
69
  section 4.
Karl Heuer's avatar
Karl Heuer committed
70 71 72

  3. BIG5

Dave Love's avatar
Dave Love committed
73 74
  A coding system to encode the character sets ASCII and Big5.  Widely
  used for Chinese (mainly in Taiwan and Hong Kong).  Details are
Richard M. Stallman's avatar
Richard M. Stallman committed
75 76 77
  described in section 4.  In this file, when we write "BIG5"
  (all uppercase), we mean the coding system, and when we write
  "Big5" (capitalized), we mean the character set.
Karl Heuer's avatar
Karl Heuer committed
78

79 80
  4. Raw text

Dave Love's avatar
Dave Love committed
81 82
  A coding system for text containing random 8-bit code.  Emacs does
  no code conversion on such text except for end-of-line format.
83 84

  5. Other
Karl Heuer's avatar
Karl Heuer committed
85

Dave Love's avatar
Dave Love committed
86 87
  If a user wants to read/write text encoded in a coding system not
  listed above, he can supply a decoder and an encoder for it as CCL
Karl Heuer's avatar
Karl Heuer committed
88 89 90
  (Code Conversion Language) programs.  Emacs executes the CCL program
  while reading/writing.

91 92
  Emacs represents a coding system by a Lisp symbol that has a property
  `coding-system'.  But, before actually using the coding system, the
Karl Heuer's avatar
Karl Heuer committed
93
  information about it is set in a structure of type `struct
Richard M. Stallman's avatar
Richard M. Stallman committed
94
  coding_system' for rapid processing.  See section 6 for more details.
Karl Heuer's avatar
Karl Heuer committed
95 96 97 98 99

*/

/*** GENERAL NOTES on END-OF-LINE FORMAT ***

Dave Love's avatar
Dave Love committed
100 101
  How end-of-line of text is encoded depends on the operating system.
  For instance, Unix's format is just one byte of `line-feed' code,
Richard M. Stallman's avatar
Richard M. Stallman committed
102
  whereas DOS's format is two-byte sequence of `carriage-return' and
103 104
  `line-feed' codes.  MacOS's format is usually one byte of
  `carriage-return'.
Karl Heuer's avatar
Karl Heuer committed
105

Dave Love's avatar
Dave Love committed
106 107 108 109
  Since text character encoding and end-of-line encoding are
  independent, any coding system described above can have any
  end-of-line format.  So Emacs has information about end-of-line
  format in each coding-system.  See section 6 for more details.
Karl Heuer's avatar
Karl Heuer committed
110 111 112 113 114 115 116

*/

/*** GENERAL NOTES on `detect_coding_XXX ()' functions ***

  These functions check if a text between SRC and SRC_END is encoded
  in the coding system category XXX.  Each returns an integer value in
Dave Love's avatar
Dave Love committed
117
  which appropriate flag bits for the category XXX are set.  The flag
Karl Heuer's avatar
Karl Heuer committed
118
  bits are defined in macros CODING_CATEGORY_MASK_XXX.  Below is the
Dave Love's avatar
Dave Love committed
119
  template for these functions.  If MULTIBYTEP is nonzero, 8-bit codes
120
  of the range 0x80..0x9F are in multibyte form.  */
Karl Heuer's avatar
Karl Heuer committed
121 122
#if 0
int
123
detect_coding_emacs_mule (src, src_end, multibytep)
Karl Heuer's avatar
Karl Heuer committed
124
     unsigned char *src, *src_end;
125
     int multibytep;
Karl Heuer's avatar
Karl Heuer committed
126 127 128 129 130 131 132
{
  ...
}
#endif

/*** GENERAL NOTES on `decode_coding_XXX ()' functions ***

133 134 135 136
  These functions decode SRC_BYTES length of unibyte text at SOURCE
  encoded in CODING to Emacs' internal format.  The resulting
  multibyte text goes to a place pointed to by DESTINATION, the length
  of which should not exceed DST_BYTES.
137

Dave Love's avatar
Dave Love committed
138 139 140 141 142
  These functions set the information about original and decoded texts
  in the members `produced', `produced_char', `consumed', and
  `consumed_char' of the structure *CODING.  They also set the member
  `result' to one of CODING_FINISH_XXX indicating how the decoding
  finished.
143

Dave Love's avatar
Dave Love committed
144
  DST_BYTES zero means that the source area and destination area are
145
  overlapped, which means that we can produce a decoded text until it
Dave Love's avatar
Dave Love committed
146
  reaches the head of the not-yet-decoded source text.
147

Dave Love's avatar
Dave Love committed
148
  Below is a template for these functions.  */
Karl Heuer's avatar
Karl Heuer committed
149
#if 0
150
static void
151
decode_coding_XXX (coding, source, destination, src_bytes, dst_bytes)
Karl Heuer's avatar
Karl Heuer committed
152
     struct coding_system *coding;
153 154
     const unsigned char *source;
     unsigned char *destination;
Karl Heuer's avatar
Karl Heuer committed
155 156 157 158 159 160 161 162
     int src_bytes, dst_bytes;
{
  ...
}
#endif

/*** GENERAL NOTES on `encode_coding_XXX ()' functions ***

Dave Love's avatar
Dave Love committed
163
  These functions encode SRC_BYTES length text at SOURCE from Emacs'
164 165 166
  internal multibyte format to CODING.  The resulting unibyte text
  goes to a place pointed to by DESTINATION, the length of which
  should not exceed DST_BYTES.
167

Dave Love's avatar
Dave Love committed
168 169 170 171 172
  These functions set the information about original and encoded texts
  in the members `produced', `produced_char', `consumed', and
  `consumed_char' of the structure *CODING.  They also set the member
  `result' to one of CODING_FINISH_XXX indicating how the encoding
  finished.
173

Dave Love's avatar
Dave Love committed
174 175 176
  DST_BYTES zero means that the source area and destination area are
  overlapped, which means that we can produce encoded text until it
  reaches at the head of the not-yet-encoded source text.
177

Dave Love's avatar
Dave Love committed
178
  Below is a template for these functions.  */
Karl Heuer's avatar
Karl Heuer committed
179
#if 0
180
static void
181
encode_coding_XXX (coding, source, destination, src_bytes, dst_bytes)
Karl Heuer's avatar
Karl Heuer committed
182 183 184 185 186 187 188 189 190 191
     struct coding_system *coding;
     unsigned char *source, *destination;
     int src_bytes, dst_bytes;
{
  ...
}
#endif

/*** COMMONLY USED MACROS ***/

192 193 194 195 196 197 198
/* The following two macros ONE_MORE_BYTE and TWO_MORE_BYTES safely
   get one, two, and three bytes from the source text respectively.
   If there are not enough bytes in the source, they jump to
   `label_end_of_loop'.  The caller should set variables `coding',
   `src' and `src_end' to appropriate pointer in advance.  These
   macros are called from decoding routines `decode_coding_XXX', thus
   it is assumed that the source text is unibyte.  */
Karl Heuer's avatar
Karl Heuer committed
199

200 201 202 203 204 205 206 207
#define ONE_MORE_BYTE(c1)					\
  do {								\
    if (src >= src_end)						\
      {								\
	coding->result = CODING_FINISH_INSUFFICIENT_SRC;	\
	goto label_end_of_loop;					\
      }								\
    c1 = *src++;						\
Karl Heuer's avatar
Karl Heuer committed
208 209
  } while (0)

210 211 212 213 214 215 216 217 218
#define TWO_MORE_BYTES(c1, c2)					\
  do {								\
    if (src + 1 >= src_end)					\
      {								\
	coding->result = CODING_FINISH_INSUFFICIENT_SRC;	\
	goto label_end_of_loop;					\
      }								\
    c1 = *src++;						\
    c2 = *src++;						\
Karl Heuer's avatar
Karl Heuer committed
219 220 221
  } while (0)


222
/* Like ONE_MORE_BYTE, but 8-bit bytes of data at SRC are in multibyte
223 224
   form if MULTIBYTEP is nonzero.  In addition, if SRC is not less
   than SRC_END, return with RET.  */
225

226
#define ONE_MORE_BYTE_CHECK_MULTIBYTE(c1, multibytep, ret)	\
227 228 229 230
  do {								\
    if (src >= src_end)						\
      {								\
	coding->result = CODING_FINISH_INSUFFICIENT_SRC;	\
231
	return ret;						\
232 233 234 235 236 237
      }								\
    c1 = *src++;						\
    if (multibytep && c1 == LEADING_CODE_8_BIT_CONTROL)		\
      c1 = *src++ - 0x20;					\
  } while (0)

238 239 240 241 242 243 244 245 246
/* Set C to the next character at the source text pointed by `src'.
   If there are not enough characters in the source, jump to
   `label_end_of_loop'.  The caller should set variables `coding'
   `src', `src_end', and `translation_table' to appropriate pointers
   in advance.  This macro is used in encoding routines
   `encode_coding_XXX', thus it assumes that the source text is in
   multibyte form except for 8-bit characters.  8-bit characters are
   in multibyte form if coding->src_multibyte is nonzero, else they
   are represented by a single byte.  */
Karl Heuer's avatar
Karl Heuer committed
247

248 249 250 251 252 253 254 255 256 257 258 259 260 261 262
#define ONE_MORE_CHAR(c)					\
  do {								\
    int len = src_end - src;					\
    int bytes;							\
    if (len <= 0)						\
      {								\
	coding->result = CODING_FINISH_INSUFFICIENT_SRC;	\
	goto label_end_of_loop;					\
      }								\
    if (coding->src_multibyte					\
	|| UNIBYTE_STR_AS_MULTIBYTE_P (src, len, bytes))	\
      c = STRING_CHAR_AND_LENGTH (src, len, bytes);		\
    else							\
      c = *src, bytes = 1;					\
    if (!NILP (translation_table))				\
263
      c = translate_char (translation_table, c, -1, 0, 0);	\
264
    src += bytes;						\
Karl Heuer's avatar
Karl Heuer committed
265 266 267
  } while (0)


Dave Love's avatar
Dave Love committed
268
/* Produce a multibyte form of character C to `dst'.  Jump to
269 270
   `label_end_of_loop' if there's not enough space at `dst'.

Dave Love's avatar
Dave Love committed
271
   If we are now in the middle of a composition sequence, the decoded
272 273 274 275 276 277 278
   character may be ALTCHAR (for the current composition).  In that
   case, the character goes to coding->cmp_data->data instead of
   `dst'.

   This macro is used in decoding routines.  */

#define EMIT_CHAR(c)							\
Karl Heuer's avatar
Karl Heuer committed
279
  do {									\
280 281 282 283 284 285 286 287 288 289 290 291 292
    if (! COMPOSING_P (coding)						\
	|| coding->composing == COMPOSITION_RELATIVE			\
	|| coding->composing == COMPOSITION_WITH_RULE)			\
      {									\
	int bytes = CHAR_BYTES (c);					\
	if ((dst + bytes) > (dst_bytes ? dst_end : src))		\
	  {								\
	    coding->result = CODING_FINISH_INSUFFICIENT_DST;		\
	    goto label_end_of_loop;					\
	  }								\
	dst += CHAR_STRING (c, dst);					\
	coding->produced_char++;					\
      }									\
Kenichi Handa's avatar
Kenichi Handa committed
293
    									\
294 295 296 297 298 299 300
    if (COMPOSING_P (coding)						\
	&& coding->composing != COMPOSITION_RELATIVE)			\
      {									\
	CODING_ADD_COMPOSITION_COMPONENT (coding, c);			\
	coding->composition_rule_follows				\
	  = coding->composing != COMPOSITION_WITH_ALTCHARS;		\
      }									\
Karl Heuer's avatar
Karl Heuer committed
301 302 303
  } while (0)


304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332
#define EMIT_ONE_BYTE(c)					\
  do {								\
    if (dst >= (dst_bytes ? dst_end : src))			\
      {								\
	coding->result = CODING_FINISH_INSUFFICIENT_DST;	\
	goto label_end_of_loop;					\
      }								\
    *dst++ = c;							\
  } while (0)

#define EMIT_TWO_BYTES(c1, c2)					\
  do {								\
    if (dst + 2 > (dst_bytes ? dst_end : src))			\
      {								\
	coding->result = CODING_FINISH_INSUFFICIENT_DST;	\
	goto label_end_of_loop;					\
      }								\
    *dst++ = c1, *dst++ = c2;					\
  } while (0)

#define EMIT_BYTES(from, to)					\
  do {								\
    if (dst + (to - from) > (dst_bytes ? dst_end : src))	\
      {								\
	coding->result = CODING_FINISH_INSUFFICIENT_DST;	\
	goto label_end_of_loop;					\
      }								\
    while (from < to)						\
      *dst++ = *from++;						\
Karl Heuer's avatar
Karl Heuer committed
333 334 335 336 337
  } while (0)


/*** 1. Preamble ***/

338 339 340 341
#ifdef emacs
#include <config.h>
#endif

Karl Heuer's avatar
Karl Heuer committed
342 343 344 345 346 347 348
#include <stdio.h>

#ifdef emacs

#include "lisp.h"
#include "buffer.h"
#include "charset.h"
Kenichi Handa's avatar
Kenichi Handa committed
349
#include "composite.h"
Karl Heuer's avatar
Karl Heuer committed
350 351 352
#include "ccl.h"
#include "coding.h"
#include "window.h"
353
#include "intervals.h"
Karl Heuer's avatar
Karl Heuer committed
354 355 356 357 358 359 360 361 362 363

#else  /* not emacs */

#include "mulelib.h"

#endif /* not emacs */

Lisp_Object Qcoding_system, Qeol_type;
Lisp_Object Qbuffer_file_coding_system;
Lisp_Object Qpost_read_conversion, Qpre_write_conversion;
364
Lisp_Object Qno_conversion, Qundecided;
365
Lisp_Object Qcoding_system_history;
366
Lisp_Object Qsafe_chars;
Kenichi Handa's avatar
Kenichi Handa committed
367
Lisp_Object Qvalid_codes;
368
Lisp_Object Qascii_incompatible;
Karl Heuer's avatar
Karl Heuer committed
369 370

extern Lisp_Object Qinsert_file_contents, Qwrite_region;
Kim F. Storm's avatar
Kim F. Storm committed
371
Lisp_Object Qcall_process, Qcall_process_region;
Karl Heuer's avatar
Karl Heuer committed
372 373 374
Lisp_Object Qstart_process, Qopen_network_stream;
Lisp_Object Qtarget_idx;

375 376 377 378
/* If a symbol has this property, evaluate the value to define the
   symbol as a coding system.  */
Lisp_Object Qcoding_system_define_form;

379 380
Lisp_Object Vselect_safe_coding_system_function;

381 382
int coding_system_require_warning;

383 384 385
/* Mnemonic string for each format of end-of-line.  */
Lisp_Object eol_mnemonic_unix, eol_mnemonic_dos, eol_mnemonic_mac;
/* Mnemonic string to indicate format of end-of-line is not yet
Karl Heuer's avatar
Karl Heuer committed
386
   decided.  */
387
Lisp_Object eol_mnemonic_undecided;
Karl Heuer's avatar
Karl Heuer committed
388

389
/* Format of end-of-line decided by system.  This is CODING_EOL_LF on
390 391 392
   Unix, CODING_EOL_CRLF on DOS/Windows, and CODING_EOL_CR on Mac.
   This has an effect only for external encoding (i.e. for output to
   file and process), not for in-buffer or Lisp string encoding.  */
393 394
int system_eol_type;

Karl Heuer's avatar
Karl Heuer committed
395 396
#ifdef emacs

397 398 399 400 401 402 403 404 405 406
/* Information about which coding system is safe for which chars.
   The value has the form (GENERIC-LIST . NON-GENERIC-ALIST).

   GENERIC-LIST is a list of generic coding systems which can encode
   any characters.

   NON-GENERIC-ALIST is an alist of non generic coding systems vs the
   corresponding char table that contains safe chars.  */
Lisp_Object Vcoding_system_safe_chars;

407 408 409
Lisp_Object Vcoding_system_list, Vcoding_system_alist;

Lisp_Object Qcoding_system_p, Qcoding_system_error;
Karl Heuer's avatar
Karl Heuer committed
410

411 412 413
/* Coding system emacs-mule and raw-text are for converting only
   end-of-line format.  */
Lisp_Object Qemacs_mule, Qraw_text;
414

Dave Love's avatar
Dave Love committed
415 416
Lisp_Object Qutf_8;

Karl Heuer's avatar
Karl Heuer committed
417 418 419 420 421 422 423 424 425
/* Coding-systems are handed between Emacs Lisp programs and C internal
   routines by the following three variables.  */
/* Coding-system for reading files and receiving data from process.  */
Lisp_Object Vcoding_system_for_read;
/* Coding-system for writing files and sending data to process.  */
Lisp_Object Vcoding_system_for_write;
/* Coding-system actually used in the latest I/O.  */
Lisp_Object Vlast_coding_system_used;

426
/* A vector of length 256 which contains information about special
Karl Heuer's avatar
Karl Heuer committed
427
   Latin codes (especially for dealing with Microsoft codes).  */
428
Lisp_Object Vlatin_extra_code_table;
429

430 431 432
/* Flag to inhibit code conversion of end-of-line format.  */
int inhibit_eol_conversion;

433 434 435
/* Flag to inhibit ISO2022 escape sequence detection.  */
int inhibit_iso_escape_detection;

436 437 438
/* Flag to make buffer-file-coding-system inherit from process-coding.  */
int inherit_process_coding_system;

439
/* Coding system to be used to encode text for terminal display.  */
Karl Heuer's avatar
Karl Heuer committed
440 441
struct coding_system terminal_coding;

442 443 444 445 446
/* Coding system to be used to encode text for terminal display when
   terminal coding system is nil.  */
struct coding_system safe_terminal_coding;

/* Coding system of what is sent from terminal keyboard.  */
Karl Heuer's avatar
Karl Heuer committed
447 448
struct coding_system keyboard_coding;

449 450 451
/* Default coding system to be used to write a file.  */
struct coding_system default_buffer_file_coding;

452 453 454
Lisp_Object Vfile_coding_system_alist;
Lisp_Object Vprocess_coding_system_alist;
Lisp_Object Vnetwork_coding_system_alist;
Karl Heuer's avatar
Karl Heuer committed
455

456 457
Lisp_Object Vlocale_coding_system;

Karl Heuer's avatar
Karl Heuer committed
458 459
#endif /* emacs */

460
Lisp_Object Qcoding_category, Qcoding_category_index;
Karl Heuer's avatar
Karl Heuer committed
461 462 463 464

/* List of symbols `coding-category-xxx' ordered by priority.  */
Lisp_Object Vcoding_category_list;

465 466
/* Table of coding categories (Lisp symbols).  */
Lisp_Object Vcoding_category_table;
Karl Heuer's avatar
Karl Heuer committed
467 468 469

/* Table of names of symbol for each coding-category.  */
char *coding_category_name[CODING_CATEGORY_IDX_MAX] = {
Kenichi Handa's avatar
Kenichi Handa committed
470
  "coding-category-emacs-mule",
Karl Heuer's avatar
Karl Heuer committed
471 472
  "coding-category-sjis",
  "coding-category-iso-7",
473
  "coding-category-iso-7-tight",
Karl Heuer's avatar
Karl Heuer committed
474 475
  "coding-category-iso-8-1",
  "coding-category-iso-8-2",
476 477
  "coding-category-iso-7-else",
  "coding-category-iso-8-else",
478
  "coding-category-ccl",
Karl Heuer's avatar
Karl Heuer committed
479
  "coding-category-big5",
480 481 482
  "coding-category-utf-8",
  "coding-category-utf-16-be",
  "coding-category-utf-16-le",
483
  "coding-category-raw-text",
484
  "coding-category-binary"
Karl Heuer's avatar
Karl Heuer committed
485 486
};

487
/* Table of pointers to coding systems corresponding to each coding
488 489 490
   categories.  */
struct coding_system *coding_system_table[CODING_CATEGORY_IDX_MAX];

491
/* Table of coding category masks.  Nth element is a mask for a coding
Dave Love's avatar
Dave Love committed
492
   category of which priority is Nth.  */
493 494 495
static
int coding_priorities[CODING_CATEGORY_IDX_MAX];

496 497
/* Flag to tell if we look up translation table on character code
   conversion.  */
498
Lisp_Object Venable_character_translation;
499 500 501 502
/* Standard translation table to look up on decoding (reading).  */
Lisp_Object Vstandard_translation_table_for_decode;
/* Standard translation table to look up on encoding (writing).  */
Lisp_Object Vstandard_translation_table_for_encode;
503

504 505 506 507
Lisp_Object Qtranslation_table;
Lisp_Object Qtranslation_table_id;
Lisp_Object Qtranslation_table_for_decode;
Lisp_Object Qtranslation_table_for_encode;
Karl Heuer's avatar
Karl Heuer committed
508 509 510 511

/* Alist of charsets vs revision number.  */
Lisp_Object Vcharset_revision_alist;

512 513 514
/* Default coding systems used for process I/O.  */
Lisp_Object Vdefault_process_coding_system;

515 516 517
/* Char table for translating Quail and self-inserting input.  */
Lisp_Object Vtranslation_table_for_input;

518 519 520 521 522 523
/* Global flag to tell that we can't call post-read-conversion and
   pre-write-conversion functions.  Usually the value is zero, but it
   is set to 1 temporarily while such functions are running.  This is
   to avoid infinite recursive call.  */
static int inhibit_pre_post_conversion;

524 525
Lisp_Object Qchar_coding_system;

526 527
/* Return `safe-chars' property of CODING_SYSTEM (symbol).  Don't check
   its validity.  */
528 529

Lisp_Object
530 531
coding_safe_chars (coding_system)
     Lisp_Object coding_system;
532 533
{
  Lisp_Object coding_spec, plist, safe_chars;
534

535
  coding_spec = Fget (coding_system, Qcoding_system);
536 537 538 539 540 541 542 543
  plist = XVECTOR (coding_spec)->contents[3];
  safe_chars = Fplist_get (XVECTOR (coding_spec)->contents[3], Qsafe_chars);
  return (CHAR_TABLE_P (safe_chars) ? safe_chars : Qt);
}

#define CODING_SAFE_CHAR_P(safe_chars, c) \
  (EQ (safe_chars, Qt) || !NILP (CHAR_TABLE_REF (safe_chars, c)))

Karl Heuer's avatar
Karl Heuer committed
544

Kenichi Handa's avatar
Kenichi Handa committed
545
/*** 2. Emacs internal format (emacs-mule) handlers ***/
Karl Heuer's avatar
Karl Heuer committed
546

547 548 549
/* Emacs' internal format for representation of multiple character
   sets is a kind of multi-byte encoding, i.e. characters are
   represented by variable-length sequences of one-byte codes.
550 551 552 553 554 555 556 557 558 559 560 561 562 563 564

   ASCII characters and control characters (e.g. `tab', `newline') are
   represented by one-byte sequences which are their ASCII codes, in
   the range 0x00 through 0x7F.

   8-bit characters of the range 0x80..0x9F are represented by
   two-byte sequences of LEADING_CODE_8_BIT_CONTROL and (their 8-bit
   code + 0x20).

   8-bit characters of the range 0xA0..0xFF are represented by
   one-byte sequences which are their 8-bit code.

   The other characters are represented by a sequence of `base
   leading-code', optional `extended leading-code', and one or two
   `position-code's.  The length of the sequence is determined by the
565
   base leading-code.  Leading-code takes the range 0x81 through 0x9D,
566 567 568
   whereas extended leading-code and position-code take the range 0xA0
   through 0xFF.  See `charset.h' for more details about leading-code
   and position-code.
Richard M. Stallman's avatar
Richard M. Stallman committed
569

Karl Heuer's avatar
Karl Heuer committed
570
   --- CODE RANGE of Emacs' internal format ---
571 572 573 574 575
   character set	range
   -------------	-----
   ascii		0x00..0x7F
   eight-bit-control	LEADING_CODE_8_BIT_CONTROL + 0xA0..0xBF
   eight-bit-graphic	0xA0..0xBF
576
   ELSE			0x81..0x9D + [0xA0..0xFF]+
Karl Heuer's avatar
Karl Heuer committed
577 578
   ---------------------------------------------

579 580 581 582 583 584 585 586
   As this is the internal character representation, the format is
   usually not used externally (i.e. in a file or in a data sent to a
   process).  But, it is possible to have a text externally in this
   format (i.e. by encoding by the coding system `emacs-mule').

   In that case, a sequence of one-byte codes has a slightly different
   form.

Dave Love's avatar
comment  
Dave Love committed
587
   Firstly, all characters in eight-bit-control are represented by
588 589 590 591 592 593 594 595
   one-byte sequences which are their 8-bit code.

   Next, character composition data are represented by the byte
   sequence of the form: 0x80 METHOD BYTES CHARS COMPONENT ...,
   where,
	METHOD is 0xF0 plus one of composition method (enum
	composition_method),

Dave Love's avatar
comment  
Dave Love committed
596
	BYTES is 0xA0 plus the byte length of these composition data,
597

Dave Love's avatar
comment  
Dave Love committed
598
	CHARS is 0xA0 plus the number of characters composed by these
599 600
	data,

Dave Love's avatar
Dave Love committed
601
	COMPONENTs are characters of multibyte form or composition
602 603 604 605 606 607 608 609 610 611 612 613 614 615
	rules encoded by two-byte of ASCII codes.

   In addition, for backward compatibility, the following formats are
   also recognized as composition data on decoding.

   0x80 MSEQ ...
   0x80 0xFF MSEQ RULE MSEQ RULE ... MSEQ

   Here,
	MSEQ is a multibyte form but in these special format:
	  ASCII: 0xA0 ASCII_CODE+0x80,
	  other: LEADING_CODE+0x20 FOLLOWING-BYTE ...,
	RULE is a one byte code of the range 0xA0..0xF0 that
	represents a composition rule.
Karl Heuer's avatar
Karl Heuer committed
616 617 618 619 620 621
  */

enum emacs_code_class_type emacs_code_class[256];

/* See the above "GENERAL NOTES on `detect_coding_XXX ()' functions".
   Check if a text is encoded in Emacs' internal format.  If it is,
622
   return CODING_CATEGORY_MASK_EMACS_MULE, else return 0.  */
Karl Heuer's avatar
Karl Heuer committed
623

624 625
static int
detect_coding_emacs_mule (src, src_end, multibytep)
626
      unsigned char *src, *src_end;
627
      int multibytep;
Karl Heuer's avatar
Karl Heuer committed
628 629 630
{
  unsigned char c;
  int composing = 0;
631 632 633
  /* Dummy for ONE_MORE_BYTE.  */
  struct coding_system dummy_coding;
  struct coding_system *coding = &dummy_coding;
Karl Heuer's avatar
Karl Heuer committed
634

635
  while (1)
Karl Heuer's avatar
Karl Heuer committed
636
    {
637 638
      ONE_MORE_BYTE_CHECK_MULTIBYTE (c, multibytep,
				     CODING_CATEGORY_MASK_EMACS_MULE);
Karl Heuer's avatar
Karl Heuer committed
639 640 641 642
      if (composing)
	{
	  if (c < 0xA0)
	    composing = 0;
643 644
	  else if (c == 0xA0)
	    {
645
	      ONE_MORE_BYTE_CHECK_MULTIBYTE (c, multibytep, 0);
646 647
	      c &= 0x7F;
	    }
Karl Heuer's avatar
Karl Heuer committed
648 649 650 651
	  else
	    c -= 0x20;
	}

652
      if (c < 0x20)
Karl Heuer's avatar
Karl Heuer committed
653 654 655
	{
	  if (c == ISO_CODE_ESC || c == ISO_CODE_SI || c == ISO_CODE_SO)
	    return 0;
656 657 658 659 660 661 662 663 664 665
	}
      else if (c >= 0x80 && c < 0xA0)
	{
	  if (c == 0x80)
	    /* Old leading code for a composite character.  */
	    composing = 1;
	  else
	    {
	      unsigned char *src_base = src - 1;
	      int bytes;
Karl Heuer's avatar
Karl Heuer committed
666

667 668 669 670 671 672 673 674
	      if (!UNIBYTE_STR_AS_MULTIBYTE_P (src_base, src_end - src_base,
					       bytes))
		return 0;
	      src = src_base + bytes;
	    }
	}
    }
}
Karl Heuer's avatar
Karl Heuer committed
675 676


677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701
/* Record the starting position START and METHOD of one composition.  */

#define CODING_ADD_COMPOSITION_START(coding, start, method)	\
  do {								\
    struct composition_data *cmp_data = coding->cmp_data;	\
    int *data = cmp_data->data + cmp_data->used;		\
    coding->cmp_data_start = cmp_data->used;			\
    data[0] = -1;						\
    data[1] = cmp_data->char_offset + start;			\
    data[3] = (int) method;					\
    cmp_data->used += 4;					\
  } while (0)

/* Record the ending position END of the current composition.  */

#define CODING_ADD_COMPOSITION_END(coding, end)			\
  do {								\
    struct composition_data *cmp_data = coding->cmp_data;	\
    int *data = cmp_data->data + coding->cmp_data_start;	\
    data[0] = cmp_data->used - coding->cmp_data_start;		\
    data[2] = cmp_data->char_offset + end;			\
  } while (0)

/* Record one COMPONENT (alternate character or composition rule).  */

702 703 704 705 706 707 708 709 710 711
#define CODING_ADD_COMPOSITION_COMPONENT(coding, component)		\
  do {									\
    coding->cmp_data->data[coding->cmp_data->used++] = component;	\
    if (coding->cmp_data->used - coding->cmp_data_start			\
	== COMPOSITION_DATA_MAX_BUNCH_LENGTH)				\
      {									\
	CODING_ADD_COMPOSITION_END (coding, coding->produced_char);	\
	coding->composing = COMPOSITION_NO;				\
      }									\
  } while (0)
712 713 714


/* Get one byte from a data pointed by SRC and increment SRC.  If SRC
Dave Love's avatar
Dave Love committed
715
   is not less than SRC_END, return -1 without incrementing Src.  */
716 717 718 719 720 721 722 723 724 725 726 727

#define SAFE_ONE_MORE_BYTE() (src >= src_end ? -1 : *src++)


/* Decode a character represented as a component of composition
   sequence of Emacs 20 style at SRC.  Set C to that character, store
   its multibyte form sequence at P, and set P to the end of that
   sequence.  If no valid character is found, set C to -1.  */

#define DECODE_EMACS_MULE_COMPOSITION_CHAR(c, p)		\
  do {								\
    int bytes;							\
728
								\
729 730 731 732 733 734 735 736 737 738 739 740
    c = SAFE_ONE_MORE_BYTE ();					\
    if (c < 0)							\
      break;							\
    if (CHAR_HEAD_P (c))					\
      c = -1;							\
    else if (c == 0xA0)						\
      {								\
	c = SAFE_ONE_MORE_BYTE ();				\
	if (c < 0xA0)						\
	  c = -1;						\
	else							\
	  {							\
741
	    c -= 0x80;						\
742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758
	    *p++ = c;						\
	  }							\
      }								\
    else if (BASE_LEADING_CODE_P (c - 0x20))			\
      {								\
	unsigned char *p0 = p;					\
								\
	c -= 0x20;						\
	*p++ = c;						\
	bytes = BYTES_BY_CHAR_HEAD (c);				\
	while (--bytes)						\
	  {							\
	    c = SAFE_ONE_MORE_BYTE ();				\
	    if (c < 0)						\
	      break;						\
	    *p++ = c;						\
	  }							\
759 760 761 762
	if (UNIBYTE_STR_AS_MULTIBYTE_P (p0, p - p0, bytes)	\
	    || (coding->flags /* We are recovering a file.  */	\
		&& p0[0] == LEADING_CODE_8_BIT_CONTROL		\
		&& ! CHAR_HEAD_P (p0[1])))			\
763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808
	  c = STRING_CHAR (p0, bytes);				\
	else							\
	  c = -1;						\
      }								\
    else							\
      c = -1;							\
  } while (0)


/* Decode a composition rule represented as a component of composition
   sequence of Emacs 20 style at SRC.  Set C to the rule.  If not
   valid rule is found, set C to -1.  */

#define DECODE_EMACS_MULE_COMPOSITION_RULE(c)		\
  do {							\
    c = SAFE_ONE_MORE_BYTE ();				\
    c -= 0xA0;						\
    if (c < 0 || c >= 81)				\
      c = -1;						\
    else						\
      {							\
	gref = c / 9, nref = c % 9;			\
	c = COMPOSITION_ENCODE_RULE (gref, nref);	\
      }							\
  } while (0)


/* Decode composition sequence encoded by `emacs-mule' at the source
   pointed by SRC.  SRC_END is the end of source.  Store information
   of the composition in CODING->cmp_data.

   For backward compatibility, decode also a composition sequence of
   Emacs 20 style.  In that case, the composition sequence contains
   characters that should be extracted into a buffer or string.  Store
   those characters at *DESTINATION in multibyte form.

   If we encounter an invalid byte sequence, return 0.
   If we encounter an insufficient source or destination, or
   insufficient space in CODING->cmp_data, return 1.
   Otherwise, return consumed bytes in the source.

*/
static INLINE int
decode_composition_emacs_mule (coding, src, src_end,
			       destination, dst_end, dst_bytes)
     struct coding_system *coding;
809 810
     const unsigned char *src, *src_end;
     unsigned char **destination, *dst_end;
811 812 813 814
     int dst_bytes;
{
  unsigned char *dst = *destination;
  int method, data_len, nchars;
815
  const unsigned char *src_base = src++;
Dave Love's avatar
Dave Love committed
816
  /* Store components of composition.  */
817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851
  int component[COMPOSITION_DATA_MAX_BUNCH_LENGTH];
  int ncomponent;
  /* Store multibyte form of characters to be composed.  This is for
     Emacs 20 style composition sequence.  */
  unsigned char buf[MAX_COMPOSITION_COMPONENTS * MAX_MULTIBYTE_LENGTH];
  unsigned char *bufp = buf;
  int c, i, gref, nref;

  if (coding->cmp_data->used + COMPOSITION_DATA_MAX_BUNCH_LENGTH
      >= COMPOSITION_DATA_SIZE)
    {
      coding->result = CODING_FINISH_INSUFFICIENT_CMP;
      return -1;
    }

  ONE_MORE_BYTE (c);
  if (c - 0xF0 >= COMPOSITION_RELATIVE
	   && c - 0xF0 <= COMPOSITION_WITH_RULE_ALTCHARS)
    {
      int with_rule;

      method = c - 0xF0;
      with_rule = (method == COMPOSITION_WITH_RULE
		   || method == COMPOSITION_WITH_RULE_ALTCHARS);
      ONE_MORE_BYTE (c);
      data_len = c - 0xA0;
      if (data_len < 4
	  || src_base + data_len > src_end)
	return 0;
      ONE_MORE_BYTE (c);
      nchars = c - 0xA0;
      if (c < 1)
	return 0;
      for (ncomponent = 0; src < src_base + data_len; ncomponent++)
	{
852 853 854 855
	  /* If it is longer than this, it can't be valid.  */
	  if (ncomponent >= COMPOSITION_DATA_MAX_BUNCH_LENGTH)
	    return 0;

856 857 858 859 860 861 862 863 864 865 866
	  if (ncomponent % 2 && with_rule)
	    {
	      ONE_MORE_BYTE (gref);
	      gref -= 32;
	      ONE_MORE_BYTE (nref);
	      nref -= 32;
	      c = COMPOSITION_ENCODE_RULE (gref, nref);
	    }
	  else
	    {
	      int bytes;
867 868 869 870
	      if (UNIBYTE_STR_AS_MULTIBYTE_P (src, src_end - src, bytes)
		  || (coding->flags /* We are recovering a file.  */
		      && src[0] == LEADING_CODE_8_BIT_CONTROL
		      && ! CHAR_HEAD_P (src[1])))
871 872 873 874 875 876 877 878
		c = STRING_CHAR (src, bytes);
	      else
		c = *src, bytes = 1;
	      src += bytes;
	    }
	  component[ncomponent] = c;
	}
    }
879
  else if (c >= 0x80)
880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930
    {
      /* This may be an old Emacs 20 style format.  See the comment at
	 the section 2 of this file.  */
      while (src < src_end && !CHAR_HEAD_P (*src)) src++;
      if (src == src_end
	  && !(coding->mode & CODING_MODE_LAST_BLOCK))
	goto label_end_of_loop;

      src_end = src;
      src = src_base + 1;
      if (c < 0xC0)
	{
	  method = COMPOSITION_RELATIVE;
	  for (ncomponent = 0; ncomponent < MAX_COMPOSITION_COMPONENTS;)
	    {
	      DECODE_EMACS_MULE_COMPOSITION_CHAR (c, bufp);
	      if (c < 0)
		break;
	      component[ncomponent++] = c;
	    }
	  if (ncomponent < 2)
	    return 0;
	  nchars = ncomponent;
	}
      else if (c == 0xFF)
	{
	  method = COMPOSITION_WITH_RULE;
	  src++;
	  DECODE_EMACS_MULE_COMPOSITION_CHAR (c, bufp);
	  if (c < 0)
	    return 0;
	  component[0] = c;
	  for (ncomponent = 1;
	       ncomponent < MAX_COMPOSITION_COMPONENTS * 2 - 1;)
	    {
	      DECODE_EMACS_MULE_COMPOSITION_RULE (c);
	      if (c < 0)
		break;
	      component[ncomponent++] = c;
	      DECODE_EMACS_MULE_COMPOSITION_CHAR (c, bufp);
	      if (c < 0)
		break;
	      component[ncomponent++] = c;
	    }
	  if (ncomponent < 3)
	    return 0;
	  nchars = (ncomponent + 1) / 2;
	}
      else
	return 0;
    }
931 932
  else
    return 0;
933 934 935 936 937 938

  if (buf == bufp || dst + (bufp - buf) <= (dst_bytes ? dst_end : src))
    {
      CODING_ADD_COMPOSITION_START (coding, coding->produced_char, method);
      for (i = 0; i < ncomponent; i++)
	CODING_ADD_COMPOSITION_COMPONENT (coding, component[i]);
939
      CODING_ADD_COMPOSITION_END (coding, coding->produced_char + nchars);
940 941 942 943 944 945 946 947 948 949 950 951 952
      if (buf < bufp)
	{
	  unsigned char *p = buf;
	  EMIT_BYTES (p, bufp);
	  *destination += bufp - buf;
	  coding->produced_char += nchars;
	}
      return (src - src_base);
    }
 label_end_of_loop:
  return -1;
}

953
/* See the above "GENERAL NOTES on `decode_coding_XXX ()' functions".  */
Karl Heuer's avatar
Karl Heuer committed
954

955 956 957
static void
decode_coding_emacs_mule (coding, source, destination, src_bytes, dst_bytes)
     struct coding_system *coding;
958 959
     const unsigned char *source;
     unsigned char *destination;
960 961
     int src_bytes, dst_bytes;
{
962 963
  const unsigned char *src = source;
  const unsigned char *src_end = source + src_bytes;
964 965 966 967 968 969
  unsigned char *dst = destination;
  unsigned char *dst_end = destination + dst_bytes;
  /* SRC_BASE remembers the start position in source in each loop.
     The loop will be exited when there's not enough source code, or
     when there's not enough destination area to produce a
     character.  */
970
  const unsigned char *src_base;
Karl Heuer's avatar
Karl Heuer committed
971

972
  coding->produced_char = 0;
973
  while ((src_base = src) < src_end)
974
    {
975 976
      unsigned char tmp[MAX_MULTIBYTE_LENGTH];
      const unsigned char *p;
977
      int bytes;
Kenichi Handa's avatar
Kenichi Handa committed
978

979 980
      if (*src == '\r')
	{
981
	  int c = *src++;
982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010

	  if (coding->eol_type == CODING_EOL_CR)
	    c = '\n';
	  else if (coding->eol_type == CODING_EOL_CRLF)
	    {
	      ONE_MORE_BYTE (c);
	      if (c != '\n')
		{
		  src--;
		  c = '\r';
		}
	    }
	  *dst++ = c;
	  coding->produced_char++;
	  continue;
	}
      else if (*src == '\n')
	{
	  if ((coding->eol_type == CODING_EOL_CR
	       || coding->eol_type == CODING_EOL_CRLF)
	      && coding->mode & CODING_MODE_INHIBIT_INCONSISTENT_EOL)
	    {
	      coding->result = CODING_FINISH_INCONSISTENT_EOL;
	      goto label_end_of_loop;
	    }
	  *dst++ = *src++;
	  coding->produced_char++;
	  continue;
	}
1011
      else if (*src == 0x80 && coding->cmp_data)
1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027
	{
	  /* Start of composition data.  */
	  int consumed  = decode_composition_emacs_mule (coding, src, src_end,
							 &dst, dst_end,
							 dst_bytes);
	  if (consumed < 0)
	    goto label_end_of_loop;
	  else if (consumed > 0)
	    {
	      src += consumed;
	      continue;
	    }
	  bytes = CHAR_STRING (*src, tmp);
	  p = tmp;
	  src++;
	}
1028 1029 1030 1031
      else if (UNIBYTE_STR_AS_MULTIBYTE_P (src, src_end - src, bytes)
	       || (coding->flags /* We are recovering a file.  */
		   && src[0] == LEADING_CODE_8_BIT_CONTROL
		   && ! CHAR_HEAD_P (src[1])))
1032 1033 1034 1035 1036 1037
	{
	  p = src;
	  src += bytes;
	}
      else
	{
1038 1039 1040
	  int i, c;

	  bytes = BYTES_BY_CHAR_HEAD (*src);
1041
	  src++;
1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057
	  for (i = 1; i < bytes; i++)
	    {
	      ONE_MORE_BYTE (c);
	      if (CHAR_HEAD_P (c))
		break;
	    }
	  if (i < bytes)
	    {
	      bytes = CHAR_STRING (*src_base, tmp);
	      p = tmp;
	      src = src_base + 1;
	    }
	  else
	    {
	      p = src_base;
	    }
1058 1059 1060 1061
	}
      if (dst + bytes >= (dst_bytes ? dst_end : src))
	{
	  coding->result = CODING_FINISH_INSUFFICIENT_DST;
Karl Heuer's avatar
Karl Heuer committed
1062 1063
	  break;
	}
1064 1065
      while (bytes--) *dst++ = *p++;
      coding->produced_char++;
Karl Heuer's avatar
Karl Heuer committed
1066
    }
1067
 label_end_of_loop:
1068 1069
  coding->consumed = coding->consumed_char = src_base - source;
  coding->produced = dst - destination;
Karl Heuer's avatar
Karl Heuer committed
1070 1071
}

1072

1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121
/* Encode composition data stored at DATA into a special byte sequence
   starting by 0x80.  Update CODING->cmp_data_start and maybe
   CODING->cmp_data for the next call.  */

#define ENCODE_COMPOSITION_EMACS_MULE(coding, data)			\
  do {									\
    unsigned char buf[1024], *p0 = buf, *p;				\
    int len = data[0];							\
    int i;								\
    									\
    buf[0] = 0x80;							\
    buf[1] = 0xF0 + data[3];	/* METHOD */				\
    buf[3] = 0xA0 + (data[2] - data[1]); /* COMPOSED-CHARS */		\
    p = buf + 4;							\
    if (data[3] == COMPOSITION_WITH_RULE				\
	|| data[3] == COMPOSITION_WITH_RULE_ALTCHARS)			\
      {									\
	p += CHAR_STRING (data[4], p);					\
	for (i = 5; i < len; i += 2)					\
	  {								\
	    int gref, nref;						\
	     COMPOSITION_DECODE_RULE (data[i], gref, nref);		\
	    *p++ = 0x20 + gref;						\
	    *p++ = 0x20 + nref;						\
	    p += CHAR_STRING (data[i + 1], p);				\
	  }								\
      }									\
    else								\
      {									\
	for (i = 4; i < len; i++)					\
	  p += CHAR_STRING (data[i], p);				\
      }									\
    buf[2] = 0xA0 + (p - buf);	/* COMPONENTS-BYTES */			\
    									\
    if (dst + (p - buf) + 4 > (dst_bytes ? dst_end : src))		\
      {									\
	coding->result = CODING_FINISH_INSUFFICIENT_DST;		\
	goto label_end_of_loop;						\
      }									\
    while (p0 < p)							\
      *dst++ = *p0++;							\
    coding->cmp_data_start += data[0];					\
    if (coding->cmp_data_start == coding->cmp_data->used		\
	&& coding->cmp_data->next)					\
      {									\
	coding->cmp_data = coding->cmp_data->next;			\
	coding->cmp_data_start = 0;					\
      }									\
  } while (0)
1122

1123

1124
static void encode_eol P_ ((struct coding_system *, const unsigned char *,
1125 1126 1127 1128 1129
			    unsigned char *, int, int));

static void
encode_coding_emacs_mule (coding, source, destination, src_bytes, dst_bytes)
     struct coding_system *coding;
1130 1131
     const unsigned char *source;
     unsigned char *destination;
1132 1133
     int src_bytes, dst_bytes;
{
1134 1135
  const unsigned char *src = source;
  const unsigned char *src_end = source + src_bytes;
1136 1137
  unsigned char *dst = destination;
  unsigned char *dst_end = destination + dst_bytes;
1138
  const unsigned char *src_base;
1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179
  int c;
  int char_offset;
  int *data;

  Lisp_Object translation_table;

  translation_table = Qnil;

  /* Optimization for the case that there's no composition.  */
  if (!coding->cmp_data || coding->cmp_data->used == 0)
    {
      encode_eol (coding, source, destination, src_bytes, dst_bytes);
      return;
    }

  char_offset = coding->cmp_data->char_offset;
  data = coding->cmp_data->data + coding->cmp_data_start;
  while (1)
    {
      src_base = src;

      /* If SRC starts a composition, encode the information about the
	 composition in advance.  */
      if (coding->cmp_data_start < coding->cmp_data->used
	  && char_offset + coding->consumed_char == data[1])
	{
	  ENCODE_COMPOSITION_EMACS_MULE (coding, data);
	  char_offset = coding->cmp_data->char_offset;
	  data = coding->cmp_data->data + coding->cmp_data_start;
	}

      ONE_MORE_CHAR (c);
      if (c == '\n' && (coding->eol_type == CODING_EOL_CRLF
			|| coding->eol_type == CODING_EOL_CR))
	{
	  if (coding->eol_type == CODING_EOL_CRLF)
	    EMIT_TWO_BYTES ('\r', c);
	  else
	    EMIT_ONE_BYTE ('\r');
	}
      else if (SINGLE_BYTE_CHAR_P (c))
1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195
	{
	  if (coding->flags && ! ASCII_BYTE_P (c))
	    {
	      /* As we are auto saving, retain the multibyte form for
		 8-bit chars.  */
	      unsigned char buf[MAX_MULTIBYTE_LENGTH];
	      int bytes = CHAR_STRING (c, buf);

	      if (bytes == 1)
		EMIT_ONE_BYTE (buf[0]);
	      else
		EMIT_TWO_BYTES (buf[0], buf[1]);
	    }
	  else
	    EMIT_ONE_BYTE (c);
	}
1196 1197 1198 1199 1200 1201 1202 1203 1204
      else
	EMIT_BYTES (src_base, src);
      coding->consumed_char++;
    }
 label_end_of_loop:
  coding->consumed = src_base - source;
  coding->produced = coding->produced_char = dst - destination;
  return;
}
1205

Karl Heuer's avatar
Karl Heuer committed
1206 1207 1208 1209

/*** 3. ISO2022 handlers ***/

/* The following note describes the coding system ISO2022 briefly.
1210
   Since the intention of this note is to help understand the
Dave Love's avatar
Dave Love committed
1211
   functions in this file, some parts are NOT ACCURATE or are OVERLY
1212
   SIMPLIFIED.  For thorough understanding, please refer to the
Dave Love's avatar
Dave Love committed
1213 1214
   original document of ISO2022.  This is equivalent to the standard
   ECMA-35, obtainable from <URL:http://www.ecma.ch/> (*).
Karl Heuer's avatar
Karl Heuer committed
1215 1216

   ISO2022 provides many mechanisms to encode several character sets
Dave Love's avatar
Dave Love committed
1217
   in 7-bit and 8-bit environments.  For 7-bit environments, all text
1218 1219
   is encoded using bytes less than 128.  This may make the encoded
   text a little bit longer, but the text passes more easily through
Dave Love's avatar
Dave Love committed
1220
   several types of gateway, some of which strip off the MSB (Most
Dave Love's avatar
Dave Love committed
1221
   Significant Bit).
1222

Dave Love's avatar
Dave Love committed
1223 1224
   There are two kinds of character sets: control character sets and
   graphic character sets.  The former contain control characters such
Karl Heuer's avatar
Karl Heuer committed
1225
   as `newline' and `escape' to provide control functions (control
1226
   functions are also provided by escape sequences).  The latter
Dave Love's avatar
Dave Love committed
1227
   contain graphic characters such as 'A' and '-'.  Emacs recognizes
Karl Heuer's avatar
Karl Heuer committed
1228 1229 1230
   two control character sets and many graphic character sets.

   Graphic character sets are classified into one of the following
1231 1232 1233 1234 1235 1236 1237 1238
   four classes, according to the number of bytes (DIMENSION) and
   number of characters in one dimension (CHARS) of the set:
   - DIMENSION1_CHARS94
   - DIMENSION1_CHARS96
   - DIMENSION2_CHARS94
   - DIMENSION2_CHARS96

   In addition, each character set is assigned an identification tag,
Dave Love's avatar
Dave Love committed
1239
   unique for each set, called the "final character" (denoted as <F>
1240 1241 1242
   hereafter).  The <F> of each character set is decided by ECMA(*)
   when it is registered in ISO.  The code range of <F> is 0x30..0x7F
   (0x30..0x3F are for private use only).
Karl Heuer's avatar
Karl Heuer committed
1243 1244 1245

   Note (*): ECMA = European Computer Manufacturers Association

Dave Love's avatar
Dave Love committed
1246
   Here are examples of graphic character sets [NAME(<F>)]:
Karl Heuer's avatar
Karl Heuer committed
1247 1248 1249 1250 1251
	o DIMENSION1_CHARS94 -- ASCII('B'), right-half-of-JISX0201('I'), ...
	o DIMENSION1_CHARS96 -- right-half-of-ISO8859-1('A'), ...
	o DIMENSION2_CHARS94 -- GB2312('A'), JISX0208('B'), ...
	o DIMENSION2_CHARS96 -- none for the moment

1252
   A code area (1 byte=8 bits) is divided into 4 areas, C0, GL, C1, and GR.
Karl Heuer's avatar
Karl Heuer committed
1253 1254 1255 1256 1257 1258
	C0 [0x00..0x1F] -- control character plane 0
	GL [0x20..0x7F] -- graphic character plane 0
	C1 [0x80..0x9F] -- control character plane 1
	GR [0xA0..0xFF] -- graphic character plane 1

   A control character set is directly designated and invoked to C0 or
1259 1260 1261 1262 1263 1264 1265 1266
   C1 by an escape sequence.  The most common case is that:
   - ISO646's  control character set is designated/invoked to C0, and
   - ISO6429's control character set is designated/invoked to C1,
   and usually these designations/invocations are omitted in encoded
   text.  In a 7-bit environment, only C0 can be used, and a control
   character for C1 is encoded by an appropriate escape sequence to
   fit into the environment.  All control characters for C1 are
   defined to have corresponding escape sequences.
Karl Heuer's avatar
Karl Heuer committed
1267 1268 1269 1270 1271

   A graphic character set is at first designated to one of four
   graphic registers (G0 through G3), then these graphic registers are
   invoked to GL or GR.  These designations and invocations can be
   done independently.  The most common case is that G0 is invoked to
1272 1273 1274
   GL, G1 is invoked to GR, and ASCII is designated to G0.  Usually
   these invocations and designations are omitted in encoded text.
   In a 7-bit environment, only GL can be used.
Karl Heuer's avatar
Karl Heuer committed
1275

1276 1277 1278 1279
   When a graphic character set of CHARS94 is invoked to GL, codes
   0x20 and 0x7F of the GL area work as control characters SPACE and
   DEL respectively, and codes 0xA0 and 0xFF of the GR area should not
   be used.
Karl Heuer's avatar
Karl Heuer committed
1280 1281 1282

   There are two ways of invocation: locking-shift and single-shift.
   With locking-shift, the invocation lasts until the next different
1283 1284 1285 1286
   invocation, whereas with single-shift, the invocation affects the
   following character only and doesn't affect the locking-shift
   state.  Invocations are done by the following control characters or
   escape sequences:
Karl Heuer's avatar
Karl Heuer committed
1287 1288

   ----------------------------------------------------------------------
1289
   abbrev  function	             cntrl escape seq	description
Karl Heuer's avatar
Karl Heuer committed
1290
   ----------------------------------------------------------------------
1291 1292 1293 1294 1295 1296 1297 1298 1299
   SI/LS0  (shift-in)		     0x0F  none		invoke G0 into GL
   SO/LS1  (shift-out)		     0x0E  none		invoke G1 into GL
   LS2     (locking-shift-2)	     none  ESC 'n'	invoke G2 into GL
   LS3     (locking-shift-3)	     none  ESC 'o'	invoke G3 into GL
   LS1R    (locking-shift-1 right)   none  ESC '~'      invoke G1 into GR (*)
   LS2R    (locking-shift-2 right)   none  ESC '}'      invoke G2 into GR (*)
   LS3R    (locking-shift 3 right)   none  ESC '|'      invoke G3 into GR (*)
   SS2     (single-shift-2)	     0x8E  ESC 'N'	invoke G2 for one char
   SS3     (single-shift-3)	     0x8F  ESC 'O'	invoke G3 for one char
Karl Heuer's avatar
Karl Heuer committed
1300
   ----------------------------------------------------------------------
1301 1302 1303 1304
   (*) These are not used by any known coding system.

   Control characters for these functions are defined by macros
   ISO_CODE_XXX in `coding.h'.
Karl Heuer's avatar
Karl Heuer committed
1305

1306
   Designations are done by the following escape sequences:
Karl Heuer's avatar
Karl Heuer committed
1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328
   ----------------------------------------------------------------------
   escape sequence	description
   ----------------------------------------------------------------------
   ESC '(' <F>		designate DIMENSION1_CHARS94<F> to G0
   ESC ')' <F>		designate DIMENSION1_CHARS94<F> to G1
   ESC '*' <F>		designate DIMENSION1_CHARS94<F> to G2
   ESC '+' <F>		designate DIMENSION1_CHARS94<F> to G3
   ESC ',' <F>		designate DIMENSION1_CHARS96<F> to G0 (*)
   ESC '-' <F>		designate DIMENSION1_CHARS96<F> to G1
   ESC '.' <F>		designate DIMENSION1_CHARS96<F> to G2
   ESC '/' <F>		designate DIMENSION1_CHARS96<F> to G3
   ESC '$' '(' <F>	designate DIMENSION2_CHARS94<F> to G0 (**)
   ESC '$' ')' <F>	designate DIMENSION2_CHARS94<F> to G1
   ESC '$' '*' <F>	designate DIMENSION2_CHARS94<F> to G2
   ESC '$' '+' <F>	designate DIMENSION2_CHARS94<F> to G3
   ESC '$' ',' <F>	designate DIMENSION2_CHARS96<F> to G0 (*)
   ESC '$' '-' <F>	designate DIMENSION2_CHARS96<F> to G1
   ESC '$' '.' <F>	designate DIMENSION2_CHARS96<F> to G2
   ESC '$' '/' <F>	designate DIMENSION2_CHARS96<F> to G3
   ----------------------------------------------------------------------

   In this list, "DIMENSION1_CHARS94<F>" means a graphic character set
1329
   of dimension 1, chars 94, and final character <F>, etc...
Karl Heuer's avatar
Karl Heuer committed
1330 1331 1332

   Note (*): Although these designations are not allowed in ISO2022,
   Emacs accepts them on decoding, and produces them on encoding
1333
   CHARS96 character sets in a coding system which is characterized as
Karl Heuer's avatar
Karl Heuer committed
1334 1335 1336
   7-bit environment, non-locking-shift, and non-single-shift.

   Note (**): If <F> is '@', 'A', or 'B', the intermediate character
1337
   '(' can be omitted.  We refer to this as "short-form" hereafter.
Karl Heuer's avatar
Karl Heuer committed
1338

Dave Love's avatar
Dave Love committed
1339
   Now you may notice that there are a lot of ways of encoding the
1340 1341
   same multilingual text in ISO2022.  Actually, there exist many
   coding systems such as Compound Text (used in X11's inter client
Dave Love's avatar
Dave Love committed
1342 1343
   communication, ISO-2022-JP (used in Japanese Internet), ISO-2022-KR
   (used in Korean Internet), EUC (Extended UNIX Code, used in Asian
Karl Heuer's avatar
Karl Heuer committed
1344 1345 1346 1347 1348 1349
   localized platforms), and all of these are variants of ISO2022.

   In addition to the above, Emacs handles two more kinds of escape
   sequences: ISO6429's direction specification and Emacs' private
   sequence for specifying character composition.

1350
   ISO6429's direction specification takes the following form:
Karl Heuer's avatar
Karl Heuer committed
1351 1352 1353 1354 1355
	o CSI ']'      -- end of the current direction
	o CSI '0' ']'  -- end of the current direction
	o CSI '1' ']'  -- start of left-to-right text
	o CSI '2' ']'  -- start of right-to-left text
   The control character CSI (0x9B: control sequence introducer) is
1356 1357 1358
   abbreviated to the escape sequence ESC '[' in a 7-bit environment.

   Character composition specification takes the following form:
Kenichi Handa's avatar
Kenichi Handa committed
1359 1360 1361 1362 1363
	o ESC '0' -- start relative composition
	o ESC '1' -- end composition
	o ESC '2' -- start rule-base composition (*)
	o ESC '3' -- start relative composition with alternate chars  (**)
	o ESC '4' -- start rule-base composition with alternate chars  (**)
1364
  Since these are not standard escape sequences of any ISO standard,
Dave Love's avatar
Dave Love committed
1365
  the use of them with these meanings is restricted to Emacs only.
Kenichi Handa's avatar
Kenichi Handa committed
1366

Dave Love's avatar
Dave Love committed
1367
  (*) This form is used only in Emacs 20.5 and older versions,
1368
  but the newer versions can safely decode it.
Dave Love's avatar
Dave Love committed
1369
  (**) This form is used only in Emacs 21.1 and newer versions,
1370
  and the older versions can't decode it.
Kenichi Handa's avatar
Kenichi Handa committed
1371

Dave Love's avatar
Dave Love committed
1372
  Here's a list of example usages of these composition escape
1373
  sequences (categorized by `enum composition_method').
Kenichi Handa's avatar
Kenichi Handa committed
1374

1375
  COMPOSITION_RELATIVE:
Kenichi Handa's avatar
Kenichi Handa committed
1376
	ESC 0 CHAR [ CHAR ] ESC 1
Dave Love's avatar
Dave Love committed
1377
  COMPOSITION_WITH_RULE:
Kenichi Handa's avatar
Kenichi Handa committed
1378
	ESC 2 CHAR [ RULE CHAR ] ESC 1
1379
  COMPOSITION_WITH_ALTCHARS: