gl/regex.h

   1 /* Definitions for data structures and routines for the regular
   2    expression library.
   3    Copyright (C) 1985, 1989, 1990, 1991, 1992, 1993, 1995, 1996, 1997, 1998,
   4    2000, 2001, 2002, 2003, 2005, 2006, 2009, 2010 Free Software Foundation,
   5    Inc.
   6    This file is part of the GNU C Library.
   7
   8    This program is free software; you can redistribute it and/or modify
   9    it under the terms of the GNU General Public License as published by
  10    the Free Software Foundation; either version 3, or (at your option)
  11    any later version.
  12
  13    This program is distributed in the hope that it will be useful,
  14    but WITHOUT ANY WARRANTY; without even the implied warranty of
  15    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  16    GNU General Public License for more details.
  17
  18    You should have received a copy of the GNU General Public License along
  19    with this program; if not, write to the Free Software Foundation,
  20    Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */
  21
  22 #ifndef _REGEX_H
  23 #define _REGEX_H 1
  24
  25 #include <sys/types.h>
  26
  27 /* Allow the use in C++ code.  */
  28 #ifdef __cplusplus
  29 extern "C" {
  30 #endif
  31
  32 /* Define __USE_GNU_REGEX to declare GNU extensions that violate the
  33    POSIX name space rules.  */
  34 #undef __USE_GNU_REGEX
  35 #if (defined _GNU_SOURCE                                        \
  36      || (!defined _POSIX_C_SOURCE && !defined _POSIX_SOURCE     \
  37          && !defined _XOPEN_SOURCE))
  38 # define __USE_GNU_REGEX 1
  39 #endif
  40
  41 #ifdef _REGEX_LARGE_OFFSETS
  42
  43 /* Use types and values that are wide enough to represent signed and
  44    unsigned byte offsets in memory.  This currently works only when
  45    the regex code is used outside of the GNU C library; it is not yet
  46    supported within glibc itself, and glibc users should not define
  47    _REGEX_LARGE_OFFSETS.  */
  48
  49 /* The type of the offset of a byte within a string.
  50    For historical reasons POSIX 1003.1-2004 requires that regoff_t be
  51    at least as wide as off_t.  However, many common POSIX platforms set
  52    regoff_t to the more-sensible ssize_t and the Open Group has
  53    signalled its intention to change the requirement to be that
  54    regoff_t be at least as wide as ptrdiff_t and ssize_t; see XBD ERN
  55    60 (2005-08-25).  We don't know of any hosts where ssize_t or
  56    ptrdiff_t is wider than ssize_t, so ssize_t is safe.  */
  57 typedef ssize_t regoff_t;
  58
  59 /* The type of nonnegative object indexes.  Traditionally, GNU regex
  60    uses 'int' for these.  Code that uses __re_idx_t should work
  61    regardless of whether the type is signed.  */
  62 typedef size_t __re_idx_t;
  63
  64 /* The type of object sizes.  */
  65 typedef size_t __re_size_t;
  66
  67 /* The type of object sizes, in places where the traditional code
  68    uses unsigned long int.  */
  69 typedef size_t __re_long_size_t;
  70
  71 #else
  72
  73 /* Use types that are binary-compatible with the traditional GNU regex
  74    implementation, which mishandles strings longer than INT_MAX.  */
  75
  76 typedef int regoff_t;
  77 typedef int __re_idx_t;
  78 typedef unsigned int __re_size_t;
  79 typedef unsigned long int __re_long_size_t;
  80
  81 #endif
  82
  83 /* The following two types have to be signed and unsigned integer type
  84    wide enough to hold a value of a pointer.  For most ANSI compilers
  85    ptrdiff_t and size_t should be likely OK.  Still size of these two
  86    types is 2 for Microsoft C.  Ugh... */
  87 typedef long int s_reg_t;
  88 typedef unsigned long int active_reg_t;
  89
  90 /* The following bits are used to determine the regexp syntax we
  91    recognize.  The set/not-set meanings are chosen so that Emacs syntax
  92    remains the value 0.  The bits are given in alphabetical order, and
  93    the definitions shifted by one from the previous bit; thus, when we
  94    add or remove a bit, only one other definition need change.  */
  95 typedef unsigned long int reg_syntax_t;
  96
  97 #ifdef __USE_GNU_REGEX
  98
  99 /* If this bit is not set, then \ inside a bracket expression is literal.
 100    If set, then such a \ quotes the following character.  */
 101 # define RE_BACKSLASH_ESCAPE_IN_LISTS ((unsigned long int) 1)
 102
 103 /* If this bit is not set, then + and ? are operators, and \+ and \? are
 104      literals.
 105    If set, then \+ and \? are operators and + and ? are literals.  */
 106 # define RE_BK_PLUS_QM (RE_BACKSLASH_ESCAPE_IN_LISTS << 1)
 107
 108 /* If this bit is set, then character classes are supported.  They are:
 109      [:alpha:], [:upper:], [:lower:],  [:digit:], [:alnum:], [:xdigit:],
 110      [:space:], [:print:], [:punct:], [:graph:], and [:cntrl:].
 111    If not set, then character classes are not supported.  */
 112 # define RE_CHAR_CLASSES (RE_BK_PLUS_QM << 1)
 113
 114 /* If this bit is set, then ^ and $ are always anchors (outside bracket
 115      expressions, of course).
 116    If this bit is not set, then it depends:
 117         ^  is an anchor if it is at the beginning of a regular
 118            expression or after an open-group or an alternation operator;
 119         $  is an anchor if it is at the end of a regular expression, or
 120            before a close-group or an alternation operator.
 121
 122    This bit could be (re)combined with RE_CONTEXT_INDEP_OPS, because
 123    POSIX draft 11.2 says that * etc. in leading positions is undefined.
 124    We already implemented a previous draft which made those constructs
 125    invalid, though, so we haven't changed the code back.  */
 126 # define RE_CONTEXT_INDEP_ANCHORS (RE_CHAR_CLASSES << 1)
 127
 128 /* If this bit is set, then special characters are always special
 129      regardless of where they are in the pattern.
 130    If this bit is not set, then special characters are special only in
 131      some contexts; otherwise they are ordinary.  Specifically,
 132      * + ? and intervals are only special when not after the beginning,
 133      open-group, or alternation operator.  */
 134 # define RE_CONTEXT_INDEP_OPS (RE_CONTEXT_INDEP_ANCHORS << 1)
 135
 136 /* If this bit is set, then *, +, ?, and { cannot be first in an re or
 137      immediately after an alternation or begin-group operator.  */
 138 # define RE_CONTEXT_INVALID_OPS (RE_CONTEXT_INDEP_OPS << 1)
 139
 140 /* If this bit is set, then . matches newline.
 141    If not set, then it doesn't.  */
 142 # define RE_DOT_NEWLINE (RE_CONTEXT_INVALID_OPS << 1)
 143
 144 /* If this bit is set, then . doesn't match NUL.
 145    If not set, then it does.  */
 146 # define RE_DOT_NOT_NULL (RE_DOT_NEWLINE << 1)
 147
 148 /* If this bit is set, nonmatching lists [^...] do not match newline.
 149    If not set, they do.  */
 150 # define RE_HAT_LISTS_NOT_NEWLINE (RE_DOT_NOT_NULL << 1)
 151
 152 /* If this bit is set, either \{...\} or {...} defines an
 153      interval, depending on RE_NO_BK_BRACES.
 154    If not set, \{, \}, {, and } are literals.  */
 155 # define RE_INTERVALS (RE_HAT_LISTS_NOT_NEWLINE << 1)
 156
 157 /* If this bit is set, +, ? and | aren't recognized as operators.
 158    If not set, they are.  */
 159 # define RE_LIMITED_OPS (RE_INTERVALS << 1)
 160
 161 /* If this bit is set, newline is an alternation operator.
 162    If not set, newline is literal.  */
 163 # define RE_NEWLINE_ALT (RE_LIMITED_OPS << 1)
 164
 165 /* If this bit is set, then `{...}' defines an interval, and \{ and \}
 166      are literals.
 167   If not set, then `\{...\}' defines an interval.  */
 168 # define RE_NO_BK_BRACES (RE_NEWLINE_ALT << 1)
 169
 170 /* If this bit is set, (...) defines a group, and \( and \) are literals.
 171    If not set, \(...\) defines a group, and ( and ) are literals.  */
 172 # define RE_NO_BK_PARENS (RE_NO_BK_BRACES << 1)
 173
 174 /* If this bit is set, then \<digit> matches <digit>.
 175    If not set, then \<digit> is a back-reference.  */
 176 # define RE_NO_BK_REFS (RE_NO_BK_PARENS << 1)
 177
 178 /* If this bit is set, then | is an alternation operator, and \| is literal.
 179    If not set, then \| is an alternation operator, and | is literal.  */
 180 # define RE_NO_BK_VBAR (RE_NO_BK_REFS << 1)
 181
 182 /* If this bit is set, then an ending range point collating higher
 183      than the starting range point, as in [z-a], is invalid.
 184    If not set, then when ending range point collates higher than the
 185      starting range point, the range is ignored.  */
 186 # define RE_NO_EMPTY_RANGES (RE_NO_BK_VBAR << 1)
 187
 188 /* If this bit is set, then an unmatched ) is ordinary.
 189    If not set, then an unmatched ) is invalid.  */
 190 # define RE_UNMATCHED_RIGHT_PAREN_ORD (RE_NO_EMPTY_RANGES << 1)
 191
 192 /* If this bit is set, succeed as soon as we match the whole pattern,
 193    without further backtracking.  */
 194 # define RE_NO_POSIX_BACKTRACKING (RE_UNMATCHED_RIGHT_PAREN_ORD << 1)
 195
 196 /* If this bit is set, do not process the GNU regex operators.
 197    If not set, then the GNU regex operators are recognized. */
 198 # define RE_NO_GNU_OPS (RE_NO_POSIX_BACKTRACKING << 1)
 199
 200 /* If this bit is set, turn on internal regex debugging.
 201    If not set, and debugging was on, turn it off.
 202    This only works if regex.c is compiled -DDEBUG.
 203    We define this bit always, so that all that's needed to turn on
 204    debugging is to recompile regex.c; the calling code can always have
 205    this bit set, and it won't affect anything in the normal case. */
 206 # define RE_DEBUG (RE_NO_GNU_OPS << 1)
 207
 208 /* If this bit is set, a syntactically invalid interval is treated as
 209    a string of ordinary characters.  For example, the ERE 'a{1' is
 210    treated as 'a\{1'.  */
 211 # define RE_INVALID_INTERVAL_ORD (RE_DEBUG << 1)
 212
 213 /* If this bit is set, then ignore case when matching.
 214    If not set, then case is significant.  */
 215 # define RE_ICASE (RE_INVALID_INTERVAL_ORD << 1)
 216
 217 /* This bit is used internally like RE_CONTEXT_INDEP_ANCHORS but only
 218    for ^, because it is difficult to scan the regex backwards to find
 219    whether ^ should be special.  */
 220 # define RE_CARET_ANCHORS_HERE (RE_ICASE << 1)
 221
 222 /* If this bit is set, then \{ cannot be first in an bre or
 223    immediately after an alternation or begin-group operator.  */
 224 # define RE_CONTEXT_INVALID_DUP (RE_CARET_ANCHORS_HERE << 1)
 225
 226 /* If this bit is set, then no_sub will be set to 1 during
 227    re_compile_pattern.  */
 228 # define RE_NO_SUB (RE_CONTEXT_INVALID_DUP << 1)
 229
 230 #endif /* defined __USE_GNU_REGEX */
 231
 232 /* This global variable defines the particular regexp syntax to use (for
 233    some interfaces).  When a regexp is compiled, the syntax used is
 234    stored in the pattern buffer, so changing this does not affect
 235    already-compiled regexps.  */
 236 extern reg_syntax_t re_syntax_options;
 237 \f
 238 #ifdef __USE_GNU_REGEX
 239 /* Define combinations of the above bits for the standard possibilities.
 240    (The [[[ comments delimit what gets put into the Texinfo file, so
 241    don't delete them!)  */
 242 /* [[[begin syntaxes]]] */
 243 # define RE_SYNTAX_EMACS 0
 244
 245 # define RE_SYNTAX_AWK                                                  \
 246   (RE_BACKSLASH_ESCAPE_IN_LISTS   | RE_DOT_NOT_NULL                     \
 247    | RE_NO_BK_PARENS              | RE_NO_BK_REFS                       \
 248    | RE_NO_BK_VBAR                | RE_NO_EMPTY_RANGES                  \
 249    | RE_DOT_NEWLINE               | RE_CONTEXT_INDEP_ANCHORS            \
 250    | RE_UNMATCHED_RIGHT_PAREN_ORD | RE_NO_GNU_OPS)
 251
 252 # define RE_SYNTAX_GNU_AWK                                              \
 253   ((RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS | RE_DEBUG) \
 254    & ~(RE_DOT_NOT_NULL | RE_INTERVALS | RE_CONTEXT_INDEP_OPS            \
 255        | RE_CONTEXT_INVALID_OPS ))
 256
 257 # define RE_SYNTAX_POSIX_AWK                                            \
 258   (RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS              \
 259    | RE_INTERVALS           | RE_NO_GNU_OPS)
 260
 261 # define RE_SYNTAX_GREP                                                 \
 262   (RE_BK_PLUS_QM              | RE_CHAR_CLASSES                         \
 263    | RE_HAT_LISTS_NOT_NEWLINE | RE_INTERVALS                            \
 264    | RE_NEWLINE_ALT)
 265
 266 # define RE_SYNTAX_EGREP                                                \
 267   (RE_CHAR_CLASSES        | RE_CONTEXT_INDEP_ANCHORS                    \
 268    | RE_CONTEXT_INDEP_OPS | RE_HAT_LISTS_NOT_NEWLINE                    \
 269    | RE_NEWLINE_ALT       | RE_NO_BK_PARENS                             \
 270    | RE_NO_BK_VBAR)
 271
 272 # define RE_SYNTAX_POSIX_EGREP                                          \
 273   (RE_SYNTAX_EGREP | RE_INTERVALS | RE_NO_BK_BRACES                     \
 274    | RE_INVALID_INTERVAL_ORD)
 275
 276 /* P1003.2/D11.2, section 4.20.7.1, lines 5078ff.  */
 277 # define RE_SYNTAX_ED RE_SYNTAX_POSIX_BASIC
 278
 279 # define RE_SYNTAX_SED RE_SYNTAX_POSIX_BASIC
 280
 281 /* Syntax bits common to both basic and extended POSIX regex syntax.  */
 282 # define _RE_SYNTAX_POSIX_COMMON                                        \
 283   (RE_CHAR_CLASSES | RE_DOT_NEWLINE      | RE_DOT_NOT_NULL              \
 284    | RE_INTERVALS  | RE_NO_EMPTY_RANGES)
 285
 286 # define RE_SYNTAX_POSIX_BASIC                                          \
 287   (_RE_SYNTAX_POSIX_COMMON | RE_BK_PLUS_QM | RE_CONTEXT_INVALID_DUP)
 288
 289 /* Differs from ..._POSIX_BASIC only in that RE_BK_PLUS_QM becomes
 290    RE_LIMITED_OPS, i.e., \? \+ \| are not recognized.  Actually, this
 291    isn't minimal, since other operators, such as \`, aren't disabled.  */
 292 # define RE_SYNTAX_POSIX_MINIMAL_BASIC                                  \
 293   (_RE_SYNTAX_POSIX_COMMON | RE_LIMITED_OPS)
 294
 295 # define RE_SYNTAX_POSIX_EXTENDED                                       \
 296   (_RE_SYNTAX_POSIX_COMMON  | RE_CONTEXT_INDEP_ANCHORS                  \
 297    | RE_CONTEXT_INDEP_OPS   | RE_NO_BK_BRACES                           \
 298    | RE_NO_BK_PARENS        | RE_NO_BK_VBAR                             \
 299    | RE_CONTEXT_INVALID_OPS | RE_UNMATCHED_RIGHT_PAREN_ORD)
 300
 301 /* Differs from ..._POSIX_EXTENDED in that RE_CONTEXT_INDEP_OPS is
 302    removed and RE_NO_BK_REFS is added.  */
 303 # define RE_SYNTAX_POSIX_MINIMAL_EXTENDED                               \
 304   (_RE_SYNTAX_POSIX_COMMON  | RE_CONTEXT_INDEP_ANCHORS                  \
 305    | RE_CONTEXT_INVALID_OPS | RE_NO_BK_BRACES                           \
 306    | RE_NO_BK_PARENS        | RE_NO_BK_REFS                             \
 307    | RE_NO_BK_VBAR          | RE_UNMATCHED_RIGHT_PAREN_ORD)
 308 /* [[[end syntaxes]]] */
 309
 310 #endif /* defined __USE_GNU_REGEX */
 311 \f
 312 #ifdef __USE_GNU_REGEX
 313
 314 /* Maximum number of duplicates an interval can allow.  POSIX-conforming
 315    systems might define this in <limits.h>, but we want our
 316    value, so remove any previous define.  */
 317 # ifdef RE_DUP_MAX
 318 #  undef RE_DUP_MAX
 319 # endif
 320
 321 /* RE_DUP_MAX is 2**15 - 1 because an earlier implementation stored
 322    the counter as a 2-byte signed integer.  This is no longer true, so
 323    RE_DUP_MAX could be increased to (INT_MAX / 10 - 1), or to
 324    ((SIZE_MAX - 2) / 10 - 1) if _REGEX_LARGE_OFFSETS is defined.
 325    However, there would be a huge performance problem if someone
 326    actually used a pattern like a\{214748363\}, so RE_DUP_MAX retains
 327    its historical value.  */
 328 # define RE_DUP_MAX (0x7fff)
 329
 330 #endif /* defined __USE_GNU_REGEX */
 331
 332
 333 /* POSIX `cflags' bits (i.e., information for `regcomp').  */
 334
 335 /* If this bit is set, then use extended regular expression syntax.
 336    If not set, then use basic regular expression syntax.  */
 337 #define REG_EXTENDED 1
 338
 339 /* If this bit is set, then ignore case when matching.
 340    If not set, then case is significant.  */
 341 #define REG_ICASE (1 << 1)
 342
 343 /* If this bit is set, then anchors do not match at newline
 344      characters in the string.
 345    If not set, then anchors do match at newlines.  */
 346 #define REG_NEWLINE (1 << 2)
 347
 348 /* If this bit is set, then report only success or fail in regexec.
 349    If not set, then returns differ between not matching and errors.  */
 350 #define REG_NOSUB (1 << 3)
 351
 352
 353 /* POSIX `eflags' bits (i.e., information for regexec).  */
 354
 355 /* If this bit is set, then the beginning-of-line operator doesn't match
 356      the beginning of the string (presumably because it's not the
 357      beginning of a line).
 358    If not set, then the beginning-of-line operator does match the
 359      beginning of the string.  */
 360 #define REG_NOTBOL 1
 361
 362 /* Like REG_NOTBOL, except for the end-of-line.  */
 363 #define REG_NOTEOL (1 << 1)
 364
 365 /* Use PMATCH[0] to delimit the start and end of the search in the
 366    buffer.  */
 367 #define REG_STARTEND (1 << 2)
 368
 369
 370 /* If any error codes are removed, changed, or added, update the
 371    `__re_error_msgid' table in regcomp.c.  */
 372
 373 typedef enum
 374 {
 375   _REG_ENOSYS = -1,     /* This will never happen for this implementation.  */
 376   _REG_NOERROR = 0,     /* Success.  */
 377   _REG_NOMATCH,         /* Didn't find a match (for regexec).  */
 378
 379   /* POSIX regcomp return error codes.  (In the order listed in the
 380      standard.)  */
 381   _REG_BADPAT,          /* Invalid pattern.  */
 382   _REG_ECOLLATE,        /* Invalid collating element.  */
 383   _REG_ECTYPE,          /* Invalid character class name.  */
 384   _REG_EESCAPE,         /* Trailing backslash.  */
 385   _REG_ESUBREG,         /* Invalid back reference.  */
 386   _REG_EBRACK,          /* Unmatched left bracket.  */
 387   _REG_EPAREN,          /* Parenthesis imbalance.  */
 388   _REG_EBRACE,          /* Unmatched \{.  */
 389   _REG_BADBR,           /* Invalid contents of \{\}.  */
 390   _REG_ERANGE,          /* Invalid range end.  */
 391   _REG_ESPACE,          /* Ran out of memory.  */
 392   _REG_BADRPT,          /* No preceding re for repetition op.  */
 393
 394   /* Error codes we've added.  */
 395   _REG_EEND,            /* Premature end.  */
 396   _REG_ESIZE,           /* Compiled pattern bigger than 2^16 bytes.  */
 397   _REG_ERPAREN          /* Unmatched ) or \); not returned from regcomp.  */
 398 } reg_errcode_t;
 399
 400 #ifdef _XOPEN_SOURCE
 401 # define REG_ENOSYS     _REG_ENOSYS
 402 #endif
 403 #define REG_NOERROR     _REG_NOERROR
 404 #define REG_NOMATCH     _REG_NOMATCH
 405 #define REG_BADPAT      _REG_BADPAT
 406 #define REG_ECOLLATE    _REG_ECOLLATE
 407 #define REG_ECTYPE      _REG_ECTYPE
 408 #define REG_EESCAPE     _REG_EESCAPE
 409 #define REG_ESUBREG     _REG_ESUBREG
 410 #define REG_EBRACK      _REG_EBRACK
 411 #define REG_EPAREN      _REG_EPAREN
 412 #define REG_EBRACE      _REG_EBRACE
 413 #define REG_BADBR       _REG_BADBR
 414 #define REG_ERANGE      _REG_ERANGE
 415 #define REG_ESPACE      _REG_ESPACE
 416 #define REG_BADRPT      _REG_BADRPT
 417 #define REG_EEND        _REG_EEND
 418 #define REG_ESIZE       _REG_ESIZE
 419 #define REG_ERPAREN     _REG_ERPAREN
 420 \f
 421 /* struct re_pattern_buffer normally uses member names like `buffer'
 422    that POSIX does not allow.  In POSIX mode these members have names
 423    with leading `re_' (e.g., `re_buffer').  */
 424 #ifdef __USE_GNU_REGEX
 425 # define _REG_RE_NAME(id) id
 426 # define _REG_RM_NAME(id) id
 427 #else
 428 # define _REG_RE_NAME(id) re_##id
 429 # define _REG_RM_NAME(id) rm_##id
 430 #endif
 431
 432 /* The user can specify the type of the re_translate member by
 433    defining the macro RE_TRANSLATE_TYPE, which defaults to unsigned
 434    char *.  This pollutes the POSIX name space, so in POSIX mode just
 435    use unsigned char *.  */
 436 #ifdef __USE_GNU_REGEX
 437 # ifndef RE_TRANSLATE_TYPE
 438 #  define RE_TRANSLATE_TYPE unsigned char *
 439 # endif
 440 # define REG_TRANSLATE_TYPE RE_TRANSLATE_TYPE
 441 #else
 442 # define REG_TRANSLATE_TYPE unsigned char *
 443 #endif
 444
 445 /* This data structure represents a compiled pattern.  Before calling
 446    the pattern compiler, the fields `buffer', `allocated', `fastmap',
 447    `translate', and `no_sub' can be set.  After the pattern has been
 448    compiled, the `re_nsub' field is available.  All other fields are
 449    private to the regex routines.  */
 450
 451 struct re_pattern_buffer
 452 {
 453   /* Space that holds the compiled pattern.  It is declared as
 454      `unsigned char *' because its elements are sometimes used as
 455      array indexes.  */
 456   unsigned char *_REG_RE_NAME (buffer);
 457
 458   /* Number of bytes to which `buffer' points.  */
 459   __re_long_size_t _REG_RE_NAME (allocated);
 460
 461   /* Number of bytes actually used in `buffer'.  */
 462   __re_long_size_t _REG_RE_NAME (used);
 463
 464   /* Syntax setting with which the pattern was compiled.  */
 465   reg_syntax_t _REG_RE_NAME (syntax);
 466
 467   /* Pointer to a fastmap, if any, otherwise zero.  re_search uses the
 468      fastmap, if there is one, to skip over impossible starting points
 469      for matches.  */
 470   char *_REG_RE_NAME (fastmap);
 471
 472   /* Either a translate table to apply to all characters before
 473      comparing them, or zero for no translation.  The translation is
 474      applied to a pattern when it is compiled and to a string when it
 475      is matched.  */
 476   REG_TRANSLATE_TYPE _REG_RE_NAME (translate);
 477
 478   /* Number of subexpressions found by the compiler.  */
 479   size_t re_nsub;
 480
 481   /* Zero if this pattern cannot match the empty string, one else.
 482      Well, in truth it's used only in `re_search_2', to see whether or
 483      not we should use the fastmap, so we don't set this absolutely
 484      perfectly; see `re_compile_fastmap' (the `duplicate' case).  */
 485   unsigned int _REG_RE_NAME (can_be_null) : 1;
 486
 487   /* If REGS_UNALLOCATED, allocate space in the `regs' structure
 488      for `max (RE_NREGS, re_nsub + 1)' groups.
 489      If REGS_REALLOCATE, reallocate space if necessary.
 490      If REGS_FIXED, use what's there.  */
 491 #ifdef __USE_GNU_REGEX
 492 # define REGS_UNALLOCATED 0
 493 # define REGS_REALLOCATE 1
 494 # define REGS_FIXED 2
 495 #endif
 496   unsigned int _REG_RE_NAME (regs_allocated) : 2;
 497
 498   /* Set to zero when `regex_compile' compiles a pattern; set to one
 499      by `re_compile_fastmap' if it updates the fastmap.  */
 500   unsigned int _REG_RE_NAME (fastmap_accurate) : 1;
 501
 502   /* If set, `re_match_2' does not return information about
 503      subexpressions.  */
 504   unsigned int _REG_RE_NAME (no_sub) : 1;
 505
 506   /* If set, a beginning-of-line anchor doesn't match at the beginning
 507      of the string.  */
 508   unsigned int _REG_RE_NAME (not_bol) : 1;
 509
 510   /* Similarly for an end-of-line anchor.  */
 511   unsigned int _REG_RE_NAME (not_eol) : 1;
 512
 513   /* If true, an anchor at a newline matches.  */
 514   unsigned int _REG_RE_NAME (newline_anchor) : 1;
 515
 516 /* [[[end pattern_buffer]]] */
 517 };
 518
 519 typedef struct re_pattern_buffer regex_t;
 520 \f
 521 /* This is the structure we store register match data in.  See
 522    regex.texinfo for a full description of what registers match.  */
 523 struct re_registers
 524 {
 525   __re_size_t _REG_RM_NAME (num_regs);
 526   regoff_t *_REG_RM_NAME (start);
 527   regoff_t *_REG_RM_NAME (end);
 528 };
 529
 530
 531 /* If `regs_allocated' is REGS_UNALLOCATED in the pattern buffer,
 532    `re_match_2' returns information about at least this many registers
 533    the first time a `regs' structure is passed.  */
 534 #if !defined RE_NREGS && defined __USE_GNU_REGEX
 535 # define RE_NREGS 30
 536 #endif
 537
 538
 539 /* POSIX specification for registers.  Aside from the different names than
 540    `re_registers', POSIX uses an array of structures, instead of a
 541    structure of arrays.  */
 542 typedef struct
 543 {
 544   regoff_t rm_so;  /* Byte offset from string's start to substring's start.  */
 545   regoff_t rm_eo;  /* Byte offset from string's start to substring's end.  */
 546 } regmatch_t;
 547 \f
 548 /* Declarations for routines.  */
 549
 550 /* Sets the current default syntax to SYNTAX, and return the old syntax.
 551    You can also simply assign to the `re_syntax_options' variable.  */
 552 extern reg_syntax_t re_set_syntax (reg_syntax_t __syntax);
 553
 554 /* Compile the regular expression PATTERN, with length LENGTH
 555    and syntax given by the global `re_syntax_options', into the buffer
 556    BUFFER.  Return NULL if successful, and an error string if not.  */
 557 extern const char *re_compile_pattern (const char *__pattern, size_t __length,
 558                                        struct re_pattern_buffer *__buffer);
 559
 560
 561 /* Compile a fastmap for the compiled pattern in BUFFER; used to
 562    accelerate searches.  Return 0 if successful and -2 if was an
 563    internal error.  */
 564 extern int re_compile_fastmap (struct re_pattern_buffer *__buffer);
 565
 566
 567 /* Search in the string STRING (with length LENGTH) for the pattern
 568    compiled into BUFFER.  Start searching at position START, for RANGE
 569    characters.  Return the starting position of the match, -1 for no
 570    match, or -2 for an internal error.  Also return register
 571    information in REGS (if REGS and BUFFER->no_sub are nonzero).  */
 572 extern regoff_t re_search (struct re_pattern_buffer *__buffer,
 573                            const char *__string, __re_idx_t __length,
 574                            __re_idx_t __start, regoff_t __range,
 575                            struct re_registers *__regs);
 576
 577
 578 /* Like `re_search', but search in the concatenation of STRING1 and
 579    STRING2.  Also, stop searching at index START + STOP.  */
 580 extern regoff_t re_search_2 (struct re_pattern_buffer *__buffer,
 581                              const char *__string1, __re_idx_t __length1,
 582                              const char *__string2, __re_idx_t __length2,
 583                              __re_idx_t __start, regoff_t __range,
 584                              struct re_registers *__regs,
 585                              __re_idx_t __stop);
 586
 587
 588 /* Like `re_search', but return how many characters in STRING the regexp
 589    in BUFFER matched, starting at position START.  */
 590 extern regoff_t re_match (struct re_pattern_buffer *__buffer,
 591                           const char *__string, __re_idx_t __length,
 592                           __re_idx_t __start, struct re_registers *__regs);
 593
 594
 595 /* Relates to `re_match' as `re_search_2' relates to `re_search'.  */
 596 extern regoff_t re_match_2 (struct re_pattern_buffer *__buffer,
 597                             const char *__string1, __re_idx_t __length1,
 598                             const char *__string2, __re_idx_t __length2,
 599                             __re_idx_t __start, struct re_registers *__regs,
 600                             __re_idx_t __stop);
 601
 602
 603 /* Set REGS to hold NUM_REGS registers, storing them in STARTS and
 604    ENDS.  Subsequent matches using BUFFER and REGS will use this memory
 605    for recording register information.  STARTS and ENDS must be
 606    allocated with malloc, and must each be at least `NUM_REGS * sizeof
 607    (regoff_t)' bytes long.
 608
 609    If NUM_REGS == 0, then subsequent matches should allocate their own
 610    register data.
 611
 612    Unless this function is called, the first search or match using
 613    PATTERN_BUFFER will allocate its own register data, without
 614    freeing the old data.  */
 615 extern void re_set_registers (struct re_pattern_buffer *__buffer,
 616                               struct re_registers *__regs,
 617                               __re_size_t __num_regs,
 618                               regoff_t *__starts, regoff_t *__ends);
 619
 620 #if defined _REGEX_RE_COMP || defined _LIBC
 621 # ifndef _CRAY
 622 /* 4.2 bsd compatibility.  */
 623 extern char *re_comp (const char *);
 624 extern int re_exec (const char *);
 625 # endif
 626 #endif
 627
 628 /* GCC 2.95 and later have "__restrict"; C99 compilers have
 629    "restrict", and "configure" may have defined "restrict".
 630    Other compilers use __restrict, __restrict__, and _Restrict, and
 631    'configure' might #define 'restrict' to those words, so pick a
 632    different name.  */
 633 #ifndef _Restrict_
 634 # if 199901L <= __STDC_VERSION__
 635 #  define _Restrict_ restrict
 636 # elif 2 < __GNUC__ || (2 == __GNUC__ && 95 <= __GNUC_MINOR__)
 637 #  define _Restrict_ __restrict
 638 # else
 639 #  define _Restrict_
 640 # endif
 641 #endif
 642 /* gcc 3.1 and up support the [restrict] syntax.  Don't trust
 643    sys/cdefs.h's definition of __restrict_arr, though, as it
 644    mishandles gcc -ansi -pedantic.  */
 645 #ifndef _Restrict_arr_
 646 # if ((199901L <= __STDC_VERSION__                                      \
 647        || ((3 < __GNUC__ || (3 == __GNUC__ && 1 <= __GNUC_MINOR__))     \
 648            && !__STRICT_ANSI__))                                        \
 649       && !defined __GNUG__)
 650 #  define _Restrict_arr_ _Restrict_
 651 # else
 652 #  define _Restrict_arr_
 653 # endif
 654 #endif
 655
 656 /* POSIX compatibility.  */
 657 extern int regcomp (regex_t *_Restrict_ __preg,
 658                     const char *_Restrict_ __pattern,
 659                     int __cflags);
 660
 661 extern int regexec (const regex_t *_Restrict_ __preg,
 662                     const char *_Restrict_ __string, size_t __nmatch,
 663                     regmatch_t __pmatch[_Restrict_arr_],
 664                     int __eflags);
 665
 666 extern size_t regerror (int __errcode, const regex_t *_Restrict_ __preg,
 667                         char *_Restrict_ __errbuf, size_t __errbuf_size);
 668
 669 extern void regfree (regex_t *__preg);
 670
 671
 672 #ifdef __cplusplus
 673 }
 674 #endif  /* C++ */
 675
 676 #endif /* regex.h */