glib/gregex.c

   1 /* GRegex -- regular expression API wrapper around PCRE.
   2  *
   3  * Copyright (C) 1999, 2000 Scott Wimer
   4  * Copyright (C) 2004, Matthias Clasen <mclasen@redhat.com>
   5  * Copyright (C) 2005 - 2007, Marco Barisione <marco@barisione.org>
   6  *
   7  * This library is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2.1 of the License, or (at your option) any later version.
  11  *
  12  * This library is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU Lesser General Public
  18  * License along with this library; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
  20  */
  21
  22 #include "config.h"
  23
  24 #include <string.h>
  25
  26 #ifdef USE_SYSTEM_PCRE
  27 #include <pcre.h>
  28 #else
  29 #include "pcre/pcre.h"
  30 #endif
  31
  32 #include "gtypes.h"
  33 #include "gregex.h"
  34 #include "glibintl.h"
  35 #include "glist.h"
  36 #include "gmessages.h"
  37 #include "gstrfuncs.h"
  38 #include "gatomic.h"
  39 #include "gthread.h"
  40
  41 /**
  42  * SECTION:gregex
  43  * @title: Perl-compatible regular expressions
  44  * @short_description: matches strings against regular expressions
  45  * @see_also: [Regular expression syntax][glib-regex-syntax]
  46  *
  47  * The g_regex_*() functions implement regular
  48  * expression pattern matching using syntax and semantics similar to
  49  * Perl regular expression.
  50  *
  51  * Some functions accept a @start_position argument, setting it differs
  52  * from just passing over a shortened string and setting #G_REGEX_MATCH_NOTBOL
  53  * in the case of a pattern that begins with any kind of lookbehind assertion.
  54  * For example, consider the pattern "\Biss\B" which finds occurrences of "iss"
  55  * in the middle of words. ("\B" matches only if the current position in the
  56  * subject is not a word boundary.) When applied to the string "Mississipi"
  57  * from the fourth byte, namely "issipi", it does not match, because "\B" is
  58  * always false at the start of the subject, which is deemed to be a word
  59  * boundary. However, if the entire string is passed , but with
  60  * @start_position set to 4, it finds the second occurrence of "iss" because
  61  * it is able to look behind the starting point to discover that it is
  62  * preceded by a letter.
  63  *
  64  * Note that, unless you set the #G_REGEX_RAW flag, all the strings passed
  65  * to these functions must be encoded in UTF-8. The lengths and the positions
  66  * inside the strings are in bytes and not in characters, so, for instance,
  67  * "\xc3\xa0" (i.e. "&agrave;") is two bytes long but it is treated as a
  68  * single character. If you set #G_REGEX_RAW the strings can be non-valid
  69  * UTF-8 strings and a byte is treated as a character, so "\xc3\xa0" is two
  70  * bytes and two characters long.
  71  *
  72  * When matching a pattern, "\n" matches only against a "\n" character in
  73  * the string, and "\r" matches only a "\r" character. To match any newline
  74  * sequence use "\R". This particular group matches either the two-character
  75  * sequence CR + LF ("\r\n"), or one of the single characters LF (linefeed,
  76  * U+000A, "\n"), VT vertical tab, U+000B, "\v"), FF (formfeed, U+000C, "\f"),
  77  * CR (carriage return, U+000D, "\r"), NEL (next line, U+0085), LS (line
  78  * separator, U+2028), or PS (paragraph separator, U+2029).
  79  *
  80  * The behaviour of the dot, circumflex, and dollar metacharacters are
  81  * affected by newline characters, the default is to recognize any newline
  82  * character (the same characters recognized by "\R"). This can be changed
  83  * with #G_REGEX_NEWLINE_CR, #G_REGEX_NEWLINE_LF and #G_REGEX_NEWLINE_CRLF
  84  * compile options, and with #G_REGEX_MATCH_NEWLINE_ANY,
  85  * #G_REGEX_MATCH_NEWLINE_CR, #G_REGEX_MATCH_NEWLINE_LF and
  86  * #G_REGEX_MATCH_NEWLINE_CRLF match options. These settings are also
  87  * relevant when compiling a pattern if #G_REGEX_EXTENDED is set, and an
  88  * unescaped "#" outside a character class is encountered. This indicates
  89  * a comment that lasts until after the next newline.
  90  *
  91  * When setting the %G_REGEX_JAVASCRIPT_COMPAT flag, pattern syntax and pattern
  92  * matching is changed to be compatible with the way that regular expressions
  93  * work in JavaScript. More precisely, a lonely ']' character in the pattern
  94  * is a syntax error; the '\x' escape only allows 0 to 2 hexadecimal digits, and
  95  * you must use the '\u' escape sequence with 4 hex digits to specify a unicode
  96  * codepoint instead of '\x' or 'x{....}'. If '\x' or '\u' are not followed by
  97  * the specified number of hex digits, they match 'x' and 'u' literally; also
  98  * '\U' always matches 'U' instead of being an error in the pattern. Finally,
  99  * pattern matching is modified so that back references to an unset subpattern
 100  * group produces a match with the empty string instead of an error. See
 101  * pcreapi(3) for more information.
 102  *
 103  * Creating and manipulating the same #GRegex structure from different
 104  * threads is not a problem as #GRegex does not modify its internal
 105  * state between creation and destruction, on the other hand #GMatchInfo
 106  * is not threadsafe.
 107  *
 108  * The regular expressions low-level functionalities are obtained through
 109  * the excellent
 110  * [PCRE](http://www.pcre.org/)
 111  * library written by Philip Hazel.
 112  */
 113
 114 /* Mask of all the possible values for GRegexCompileFlags. */
 115 #define G_REGEX_COMPILE_MASK (G_REGEX_CASELESS          | \
 116                               G_REGEX_MULTILINE         | \
 117                               G_REGEX_DOTALL            | \
 118                               G_REGEX_EXTENDED          | \
 119                               G_REGEX_ANCHORED          | \
 120                               G_REGEX_DOLLAR_ENDONLY    | \
 121                               G_REGEX_UNGREEDY          | \
 122                               G_REGEX_RAW               | \
 123                               G_REGEX_NO_AUTO_CAPTURE   | \
 124                               G_REGEX_OPTIMIZE          | \
 125                               G_REGEX_FIRSTLINE         | \
 126                               G_REGEX_DUPNAMES          | \
 127                               G_REGEX_NEWLINE_CR        | \
 128                               G_REGEX_NEWLINE_LF        | \
 129                               G_REGEX_NEWLINE_CRLF      | \
 130                               G_REGEX_NEWLINE_ANYCRLF   | \
 131                               G_REGEX_BSR_ANYCRLF       | \
 132                               G_REGEX_JAVASCRIPT_COMPAT)
 133
 134 /* Mask of all GRegexCompileFlags values that are (not) passed trough to PCRE */
 135 #define G_REGEX_COMPILE_PCRE_MASK (G_REGEX_COMPILE_MASK & ~G_REGEX_COMPILE_NONPCRE_MASK)
 136 #define G_REGEX_COMPILE_NONPCRE_MASK (G_REGEX_RAW              | \
 137                                       G_REGEX_OPTIMIZE)
 138
 139 /* Mask of all the possible values for GRegexMatchFlags. */
 140 #define G_REGEX_MATCH_MASK (G_REGEX_MATCH_ANCHORED         | \
 141                             G_REGEX_MATCH_NOTBOL           | \
 142                             G_REGEX_MATCH_NOTEOL           | \
 143                             G_REGEX_MATCH_NOTEMPTY         | \
 144                             G_REGEX_MATCH_PARTIAL          | \
 145                             G_REGEX_MATCH_NEWLINE_CR       | \
 146                             G_REGEX_MATCH_NEWLINE_LF       | \
 147                             G_REGEX_MATCH_NEWLINE_CRLF     | \
 148                             G_REGEX_MATCH_NEWLINE_ANY      | \
 149                             G_REGEX_MATCH_NEWLINE_ANYCRLF  | \
 150                             G_REGEX_MATCH_BSR_ANYCRLF      | \
 151                             G_REGEX_MATCH_BSR_ANY          | \
 152                             G_REGEX_MATCH_PARTIAL_SOFT     | \
 153                             G_REGEX_MATCH_PARTIAL_HARD     | \
 154                             G_REGEX_MATCH_NOTEMPTY_ATSTART)
 155
 156 /* we rely on these flags having the same values */
 157 G_STATIC_ASSERT (G_REGEX_CASELESS          == PCRE_CASELESS);
 158 G_STATIC_ASSERT (G_REGEX_MULTILINE         == PCRE_MULTILINE);
 159 G_STATIC_ASSERT (G_REGEX_DOTALL            == PCRE_DOTALL);
 160 G_STATIC_ASSERT (G_REGEX_EXTENDED          == PCRE_EXTENDED);
 161 G_STATIC_ASSERT (G_REGEX_ANCHORED          == PCRE_ANCHORED);
 162 G_STATIC_ASSERT (G_REGEX_DOLLAR_ENDONLY    == PCRE_DOLLAR_ENDONLY);
 163 G_STATIC_ASSERT (G_REGEX_UNGREEDY          == PCRE_UNGREEDY);
 164 G_STATIC_ASSERT (G_REGEX_NO_AUTO_CAPTURE   == PCRE_NO_AUTO_CAPTURE);
 165 G_STATIC_ASSERT (G_REGEX_FIRSTLINE         == PCRE_FIRSTLINE);
 166 G_STATIC_ASSERT (G_REGEX_DUPNAMES          == PCRE_DUPNAMES);
 167 G_STATIC_ASSERT (G_REGEX_NEWLINE_CR        == PCRE_NEWLINE_CR);
 168 G_STATIC_ASSERT (G_REGEX_NEWLINE_LF        == PCRE_NEWLINE_LF);
 169 G_STATIC_ASSERT (G_REGEX_NEWLINE_CRLF      == PCRE_NEWLINE_CRLF);
 170 G_STATIC_ASSERT (G_REGEX_NEWLINE_ANYCRLF   == PCRE_NEWLINE_ANYCRLF);
 171 G_STATIC_ASSERT (G_REGEX_BSR_ANYCRLF       == PCRE_BSR_ANYCRLF);
 172 G_STATIC_ASSERT (G_REGEX_JAVASCRIPT_COMPAT == PCRE_JAVASCRIPT_COMPAT);
 173
 174 G_STATIC_ASSERT (G_REGEX_MATCH_ANCHORED         == PCRE_ANCHORED);
 175 G_STATIC_ASSERT (G_REGEX_MATCH_NOTBOL           == PCRE_NOTBOL);
 176 G_STATIC_ASSERT (G_REGEX_MATCH_NOTEOL           == PCRE_NOTEOL);
 177 G_STATIC_ASSERT (G_REGEX_MATCH_NOTEMPTY         == PCRE_NOTEMPTY);
 178 G_STATIC_ASSERT (G_REGEX_MATCH_PARTIAL          == PCRE_PARTIAL);
 179 G_STATIC_ASSERT (G_REGEX_MATCH_NEWLINE_CR       == PCRE_NEWLINE_CR);
 180 G_STATIC_ASSERT (G_REGEX_MATCH_NEWLINE_LF       == PCRE_NEWLINE_LF);
 181 G_STATIC_ASSERT (G_REGEX_MATCH_NEWLINE_CRLF     == PCRE_NEWLINE_CRLF);
 182 G_STATIC_ASSERT (G_REGEX_MATCH_NEWLINE_ANY      == PCRE_NEWLINE_ANY);
 183 G_STATIC_ASSERT (G_REGEX_MATCH_NEWLINE_ANYCRLF  == PCRE_NEWLINE_ANYCRLF);
 184 G_STATIC_ASSERT (G_REGEX_MATCH_BSR_ANYCRLF      == PCRE_BSR_ANYCRLF);
 185 G_STATIC_ASSERT (G_REGEX_MATCH_BSR_ANY          == PCRE_BSR_UNICODE);
 186 G_STATIC_ASSERT (G_REGEX_MATCH_PARTIAL_SOFT     == PCRE_PARTIAL_SOFT);
 187 G_STATIC_ASSERT (G_REGEX_MATCH_PARTIAL_HARD     == PCRE_PARTIAL_HARD);
 188 G_STATIC_ASSERT (G_REGEX_MATCH_NOTEMPTY_ATSTART == PCRE_NOTEMPTY_ATSTART);
 189
 190 /* These PCRE flags are unused or not exposed publically in GRegexFlags, so
 191  * it should be ok to reuse them for different things.
 192  */
 193 G_STATIC_ASSERT (G_REGEX_OPTIMIZE          == PCRE_NO_UTF8_CHECK);
 194 G_STATIC_ASSERT (G_REGEX_RAW               == PCRE_UTF8);
 195
 196 /* if the string is in UTF-8 use g_utf8_ functions, else use
 197  * use just +/- 1. */
 198 #define NEXT_CHAR(re, s) (((re)->compile_opts & G_REGEX_RAW) ? \
 199                                 ((s) + 1) : \
 200                                 g_utf8_next_char (s))
 201 #define PREV_CHAR(re, s) (((re)->compile_opts & G_REGEX_RAW) ? \
 202                                 ((s) - 1) : \
 203                                 g_utf8_prev_char (s))
 204
 205 struct _GMatchInfo
 206 {
 207   volatile gint ref_count;      /* the ref count */
 208   GRegex *regex;                /* the regex */
 209   GRegexMatchFlags match_opts;  /* options used at match time on the regex */
 210   gint matches;                 /* number of matching sub patterns */
 211   gint pos;                     /* position in the string where last match left off */
 212   gint  n_offsets;              /* number of offsets */
 213   gint *offsets;                /* array of offsets paired 0,1 ; 2,3 ; 3,4 etc */
 214   gint *workspace;              /* workspace for pcre_dfa_exec() */
 215   gint n_workspace;             /* number of workspace elements */
 216   const gchar *string;          /* string passed to the match function */
 217   gssize string_len;            /* length of string */
 218 };
 219
 220 struct _GRegex
 221 {
 222   volatile gint ref_count;      /* the ref count for the immutable part */
 223   gchar *pattern;               /* the pattern */
 224   pcre *pcre_re;                /* compiled form of the pattern */
 225   GRegexCompileFlags compile_opts;      /* options used at compile time on the pattern */
 226   GRegexMatchFlags match_opts;  /* options used at match time on the regex */
 227   pcre_extra *extra;            /* data stored when G_REGEX_OPTIMIZE is used */
 228 };
 229
 230 /* TRUE if ret is an error code, FALSE otherwise. */
 231 #define IS_PCRE_ERROR(ret) ((ret) < PCRE_ERROR_NOMATCH && (ret) != PCRE_ERROR_PARTIAL)
 232
 233 typedef struct _InterpolationData InterpolationData;
 234 static gboolean  interpolation_list_needs_match (GList *list);
 235 static gboolean  interpolate_replacement        (const GMatchInfo *match_info,
 236                                                  GString *result,
 237                                                  gpointer data);
 238 static GList    *split_replacement              (const gchar *replacement,
 239                                                  GError **error);
 240 static void      free_interpolation_data        (InterpolationData *data);
 241
 242
 243 static const gchar *
 244 match_error (gint errcode)
 245 {
 246   switch (errcode)
 247     {
 248     case PCRE_ERROR_NOMATCH:
 249       /* not an error */
 250       break;
 251     case PCRE_ERROR_NULL:
 252       /* NULL argument, this should not happen in GRegex */
 253       g_warning ("A NULL argument was passed to PCRE");
 254       break;
 255     case PCRE_ERROR_BADOPTION:
 256       return "bad options";
 257     case PCRE_ERROR_BADMAGIC:
 258       return _("corrupted object");
 259     case PCRE_ERROR_UNKNOWN_OPCODE:
 260       return N_("internal error or corrupted object");
 261     case PCRE_ERROR_NOMEMORY:
 262       return _("out of memory");
 263     case PCRE_ERROR_NOSUBSTRING:
 264       /* not used by pcre_exec() */
 265       break;
 266     case PCRE_ERROR_MATCHLIMIT:
 267       return _("backtracking limit reached");
 268     case PCRE_ERROR_CALLOUT:
 269       /* callouts are not implemented */
 270       break;
 271     case PCRE_ERROR_BADUTF8:
 272     case PCRE_ERROR_BADUTF8_OFFSET:
 273       /* we do not check if strings are valid */
 274       break;
 275     case PCRE_ERROR_PARTIAL:
 276       /* not an error */
 277       break;
 278     case PCRE_ERROR_BADPARTIAL:
 279       return _("the pattern contains items not supported for partial matching");
 280     case PCRE_ERROR_INTERNAL:
 281       return _("internal error");
 282     case PCRE_ERROR_BADCOUNT:
 283       /* negative ovecsize, this should not happen in GRegex */
 284       g_warning ("A negative ovecsize was passed to PCRE");
 285       break;
 286     case PCRE_ERROR_DFA_UITEM:
 287       return _("the pattern contains items not supported for partial matching");
 288     case PCRE_ERROR_DFA_UCOND:
 289       return _("back references as conditions are not supported for partial matching");
 290     case PCRE_ERROR_DFA_UMLIMIT:
 291       /* the match_field field is not used in GRegex */
 292       break;
 293     case PCRE_ERROR_DFA_WSSIZE:
 294       /* handled expanding the workspace */
 295       break;
 296     case PCRE_ERROR_DFA_RECURSE:
 297     case PCRE_ERROR_RECURSIONLIMIT:
 298       return _("recursion limit reached");
 299     case PCRE_ERROR_BADNEWLINE:
 300       return _("invalid combination of newline flags");
 301     case PCRE_ERROR_BADOFFSET:
 302       return _("bad offset");
 303     case PCRE_ERROR_SHORTUTF8:
 304       return _("short utf8");
 305     case PCRE_ERROR_RECURSELOOP:
 306       return _("recursion loop");
 307     default:
 308       break;
 309     }
 310   return _("unknown error");
 311 }
 312
 313 static void
 314 translate_compile_error (gint *errcode, const gchar **errmsg)
 315 {
 316   /* Compile errors are created adding 100 to the error code returned
 317    * by PCRE.
 318    * If errcode is known we put the translatable error message in
 319    * erromsg. If errcode is unknown we put the generic
 320    * G_REGEX_ERROR_COMPILE error code in errcode and keep the
 321    * untranslated error message returned by PCRE.
 322    * Note that there can be more PCRE errors with the same GRegexError
 323    * and that some PCRE errors are useless for us.
 324    */
 325   *errcode += 100;
 326
 327   switch (*errcode)
 328     {
 329     case G_REGEX_ERROR_STRAY_BACKSLASH:
 330       *errmsg = _("\\ at end of pattern");
 331       break;
 332     case G_REGEX_ERROR_MISSING_CONTROL_CHAR:
 333       *errmsg = _("\\c at end of pattern");
 334       break;
 335     case G_REGEX_ERROR_UNRECOGNIZED_ESCAPE:
 336       *errmsg = _("unrecognized character following \\");
 337       break;
 338     case G_REGEX_ERROR_QUANTIFIERS_OUT_OF_ORDER:
 339       *errmsg = _("numbers out of order in {} quantifier");
 340       break;
 341     case G_REGEX_ERROR_QUANTIFIER_TOO_BIG:
 342       *errmsg = _("number too big in {} quantifier");
 343       break;
 344     case G_REGEX_ERROR_UNTERMINATED_CHARACTER_CLASS:
 345       *errmsg = _("missing terminating ] for character class");
 346       break;
 347     case G_REGEX_ERROR_INVALID_ESCAPE_IN_CHARACTER_CLASS:
 348       *errmsg = _("invalid escape sequence in character class");
 349       break;
 350     case G_REGEX_ERROR_RANGE_OUT_OF_ORDER:
 351       *errmsg = _("range out of order in character class");
 352       break;
 353     case G_REGEX_ERROR_NOTHING_TO_REPEAT:
 354       *errmsg = _("nothing to repeat");
 355       break;
 356     case 111: /* internal error: unexpected repeat */
 357       *errcode = G_REGEX_ERROR_INTERNAL;
 358       *errmsg = _("unexpected repeat");
 359       break;
 360     case G_REGEX_ERROR_UNRECOGNIZED_CHARACTER:
 361       *errmsg = _("unrecognized character after (? or (?-");
 362       break;
 363     case G_REGEX_ERROR_POSIX_NAMED_CLASS_OUTSIDE_CLASS:
 364       *errmsg = _("POSIX named classes are supported only within a class");
 365       break;
 366     case G_REGEX_ERROR_UNMATCHED_PARENTHESIS:
 367       *errmsg = _("missing terminating )");
 368       break;
 369     case G_REGEX_ERROR_INEXISTENT_SUBPATTERN_REFERENCE:
 370       *errmsg = _("reference to non-existent subpattern");
 371       break;
 372     case G_REGEX_ERROR_UNTERMINATED_COMMENT:
 373       *errmsg = _("missing ) after comment");
 374       break;
 375     case G_REGEX_ERROR_EXPRESSION_TOO_LARGE:
 376       *errmsg = _("regular expression is too large");
 377       break;
 378     case G_REGEX_ERROR_MEMORY_ERROR:
 379       *errmsg = _("failed to get memory");
 380       break;
 381     case 122: /* unmatched parentheses */
 382       *errcode = G_REGEX_ERROR_UNMATCHED_PARENTHESIS;
 383       *errmsg = _(") without opening (");
 384       break;
 385     case 123: /* internal error: code overflow */
 386       *errcode = G_REGEX_ERROR_INTERNAL;
 387       *errmsg = _("code overflow");
 388       break;
 389     case 124: /* "unrecognized character after (?<\0 */
 390       *errcode = G_REGEX_ERROR_UNRECOGNIZED_CHARACTER;
 391       *errmsg = _("unrecognized character after (?<");
 392       break;
 393     case G_REGEX_ERROR_VARIABLE_LENGTH_LOOKBEHIND:
 394       *errmsg = _("lookbehind assertion is not fixed length");
 395       break;
 396     case G_REGEX_ERROR_MALFORMED_CONDITION:
 397       *errmsg = _("malformed number or name after (?(");
 398       break;
 399     case G_REGEX_ERROR_TOO_MANY_CONDITIONAL_BRANCHES:
 400       *errmsg = _("conditional group contains more than two branches");
 401       break;
 402     case G_REGEX_ERROR_ASSERTION_EXPECTED:
 403       *errmsg = _("assertion expected after (?(");
 404       break;
 405     case 129:
 406       *errcode = G_REGEX_ERROR_UNMATCHED_PARENTHESIS;
 407       /* translators: '(?R' and '(?[+-]digits' are both meant as (groups of)
 408        * sequences here, '(?-54' would be an example for the second group.
 409        */
 410       *errmsg = _("(?R or (?[+-]digits must be followed by )");
 411       break;
 412     case G_REGEX_ERROR_UNKNOWN_POSIX_CLASS_NAME:
 413       *errmsg = _("unknown POSIX class name");
 414       break;
 415     case G_REGEX_ERROR_POSIX_COLLATING_ELEMENTS_NOT_SUPPORTED:
 416       *errmsg = _("POSIX collating elements are not supported");
 417       break;
 418     case G_REGEX_ERROR_HEX_CODE_TOO_LARGE:
 419       *errmsg = _("character value in \\x{...} sequence is too large");
 420       break;
 421     case G_REGEX_ERROR_INVALID_CONDITION:
 422       *errmsg = _("invalid condition (?(0)");
 423       break;
 424     case G_REGEX_ERROR_SINGLE_BYTE_MATCH_IN_LOOKBEHIND:
 425       *errmsg = _("\\C not allowed in lookbehind assertion");
 426       break;
 427     case 137: /* PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0 */
 428       /* A number of Perl escapes are not handled by PCRE.
 429        * Therefore it explicitly raises ERR37.
 430        */
 431       *errcode = G_REGEX_ERROR_UNRECOGNIZED_ESCAPE;
 432       *errmsg = _("escapes \\L, \\l, \\N{name}, \\U, and \\u are not supported");
 433       break;
 434     case G_REGEX_ERROR_INFINITE_LOOP:
 435       *errmsg = _("recursive call could loop indefinitely");
 436       break;
 437     case 141: /* unrecognized character after (?P\0 */
 438       *errcode = G_REGEX_ERROR_UNRECOGNIZED_CHARACTER;
 439       *errmsg = _("unrecognized character after (?P");
 440       break;
 441     case G_REGEX_ERROR_MISSING_SUBPATTERN_NAME_TERMINATOR:
 442       *errmsg = _("missing terminator in subpattern name");
 443       break;
 444     case G_REGEX_ERROR_DUPLICATE_SUBPATTERN_NAME:
 445       *errmsg = _("two named subpatterns have the same name");
 446       break;
 447     case G_REGEX_ERROR_MALFORMED_PROPERTY:
 448       *errmsg = _("malformed \\P or \\p sequence");
 449       break;
 450     case G_REGEX_ERROR_UNKNOWN_PROPERTY:
 451       *errmsg = _("unknown property name after \\P or \\p");
 452       break;
 453     case G_REGEX_ERROR_SUBPATTERN_NAME_TOO_LONG:
 454       *errmsg = _("subpattern name is too long (maximum 32 characters)");
 455       break;
 456     case G_REGEX_ERROR_TOO_MANY_SUBPATTERNS:
 457       *errmsg = _("too many named subpatterns (maximum 10,000)");
 458       break;
 459     case G_REGEX_ERROR_INVALID_OCTAL_VALUE:
 460       *errmsg = _("octal value is greater than \\377");
 461       break;
 462     case 152: /* internal error: overran compiling workspace */
 463       *errcode = G_REGEX_ERROR_INTERNAL;
 464       *errmsg = _("overran compiling workspace");
 465       break;
 466     case 153: /* internal error: previously-checked referenced subpattern not found */
 467       *errcode = G_REGEX_ERROR_INTERNAL;
 468       *errmsg = _("previously-checked referenced subpattern not found");
 469       break;
 470     case G_REGEX_ERROR_TOO_MANY_BRANCHES_IN_DEFINE:
 471       *errmsg = _("DEFINE group contains more than one branch");
 472       break;
 473     case G_REGEX_ERROR_INCONSISTENT_NEWLINE_OPTIONS:
 474       *errmsg = _("inconsistent NEWLINE options");
 475       break;
 476     case G_REGEX_ERROR_MISSING_BACK_REFERENCE:
 477       *errmsg = _("\\g is not followed by a braced, angle-bracketed, or quoted name or "
 478                   "number, or by a plain number");
 479       break;
 480     case G_REGEX_ERROR_INVALID_RELATIVE_REFERENCE:
 481       *errmsg = _("a numbered reference must not be zero");
 482       break;
 483     case G_REGEX_ERROR_BACKTRACKING_CONTROL_VERB_ARGUMENT_FORBIDDEN:
 484       *errmsg = _("an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)");
 485       break;
 486     case G_REGEX_ERROR_UNKNOWN_BACKTRACKING_CONTROL_VERB:
 487       *errmsg = _("(*VERB) not recognized");
 488       break;
 489     case G_REGEX_ERROR_NUMBER_TOO_BIG:
 490       *errmsg = _("number is too big");
 491       break;
 492     case G_REGEX_ERROR_MISSING_SUBPATTERN_NAME:
 493       *errmsg = _("missing subpattern name after (?&");
 494       break;
 495     case G_REGEX_ERROR_MISSING_DIGIT:
 496       *errmsg = _("digit expected after (?+");
 497       break;
 498     case G_REGEX_ERROR_INVALID_DATA_CHARACTER:
 499       *errmsg = _("] is an invalid data character in JavaScript compatibility mode");
 500       break;
 501     case G_REGEX_ERROR_EXTRA_SUBPATTERN_NAME:
 502       *errmsg = _("different names for subpatterns of the same number are not allowed");
 503       break;
 504     case G_REGEX_ERROR_BACKTRACKING_CONTROL_VERB_ARGUMENT_REQUIRED:
 505       *errmsg = _("(*MARK) must have an argument");
 506       break;
 507     case G_REGEX_ERROR_INVALID_CONTROL_CHAR:
 508       *errmsg = _( "\\c must be followed by an ASCII character");
 509       break;
 510     case G_REGEX_ERROR_MISSING_NAME:
 511       *errmsg = _("\\k is not followed by a braced, angle-bracketed, or quoted name");
 512       break;
 513     case G_REGEX_ERROR_NOT_SUPPORTED_IN_CLASS:
 514       *errmsg = _("\\N is not supported in a class");
 515       break;
 516     case G_REGEX_ERROR_TOO_MANY_FORWARD_REFERENCES:
 517       *errmsg = _("too many forward references");
 518       break;
 519     case G_REGEX_ERROR_NAME_TOO_LONG:
 520       *errmsg = _("name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)");
 521       break;
 522     case G_REGEX_ERROR_CHARACTER_VALUE_TOO_LARGE:
 523       *errmsg = _("character value in \\u.... sequence is too large");
 524       break;
 525
 526     case 116: /* erroffset passed as NULL */
 527       /* This should not happen as we never pass a NULL erroffset */
 528       g_warning ("erroffset passed as NULL");
 529       *errcode = G_REGEX_ERROR_COMPILE;
 530       break;
 531     case 117: /* unknown option bit(s) set */
 532       /* This should not happen as we check options before passing them
 533        * to pcre_compile2() */
 534       g_warning ("unknown option bit(s) set");
 535       *errcode = G_REGEX_ERROR_COMPILE;
 536       break;
 537     case 132: /* this version of PCRE is compiled without UTF support */
 538     case 144: /* invalid UTF-8 string */
 539     case 145: /* support for \\P, \\p, and \\X has not been compiled */
 540     case 167: /* this version of PCRE is not compiled with Unicode property support */
 541     case 173: /* disallowed Unicode code point (>= 0xd800 && <= 0xdfff) */
 542     case 174: /* invalid UTF-16 string */
 543       /* These errors should not happen as we are using an UTF-8 and UCP-enabled PCRE
 544        * and we do not check if strings are valid */
 545     case 170: /* internal error: unknown opcode in find_fixedlength() */
 546       *errcode = G_REGEX_ERROR_INTERNAL;
 547       break;
 548
 549     default:
 550       *errcode = G_REGEX_ERROR_COMPILE;
 551     }
 552 }
 553
 554 /* GMatchInfo */
 555
 556 static GMatchInfo *
 557 match_info_new (const GRegex *regex,
 558                 const gchar  *string,
 559                 gint          string_len,
 560                 gint          start_position,
 561                 gint          match_options,
 562                 gboolean      is_dfa)
 563 {
 564   GMatchInfo *match_info;
 565
 566   if (string_len < 0)
 567     string_len = strlen (string);
 568
 569   match_info = g_new0 (GMatchInfo, 1);
 570   match_info->ref_count = 1;
 571   match_info->regex = g_regex_ref ((GRegex *)regex);
 572   match_info->string = string;
 573   match_info->string_len = string_len;
 574   match_info->matches = PCRE_ERROR_NOMATCH;
 575   match_info->pos = start_position;
 576   match_info->match_opts = match_options;
 577
 578   if (is_dfa)
 579     {
 580       /* These values should be enough for most cases, if they are not
 581        * enough g_regex_match_all_full() will expand them. */
 582       match_info->n_offsets = 24;
 583       match_info->n_workspace = 100;
 584       match_info->workspace = g_new (gint, match_info->n_workspace);
 585     }
 586   else
 587     {
 588       gint capture_count;
 589       pcre_fullinfo (regex->pcre_re, regex->extra,
 590                      PCRE_INFO_CAPTURECOUNT, &capture_count);
 591       match_info->n_offsets = (capture_count + 1) * 3;
 592     }
 593
 594   match_info->offsets = g_new0 (gint, match_info->n_offsets);
 595   /* Set an invalid position for the previous match. */
 596   match_info->offsets[0] = -1;
 597   match_info->offsets[1] = -1;
 598
 599   return match_info;
 600 }
 601
 602 /**
 603  * g_match_info_get_regex:
 604  * @match_info: a #GMatchInfo
 605  *
 606  * Returns #GRegex object used in @match_info. It belongs to Glib
 607  * and must not be freed. Use g_regex_ref() if you need to keep it
 608  * after you free @match_info object.
 609  *
 610  * Returns: #GRegex object used in @match_info
 611  *
 612  * Since: 2.14
 613  */
 614 GRegex *
 615 g_match_info_get_regex (const GMatchInfo *match_info)
 616 {
 617   g_return_val_if_fail (match_info != NULL, NULL);
 618   return match_info->regex;
 619 }
 620
 621 /**
 622  * g_match_info_get_string:
 623  * @match_info: a #GMatchInfo
 624  *
 625  * Returns the string searched with @match_info. This is the
 626  * string passed to g_regex_match() or g_regex_replace() so
 627  * you may not free it before calling this function.
 628  *
 629  * Returns: the string searched with @match_info
 630  *
 631  * Since: 2.14
 632  */
 633 const gchar *
 634 g_match_info_get_string (const GMatchInfo *match_info)
 635 {
 636   g_return_val_if_fail (match_info != NULL, NULL);
 637   return match_info->string;
 638 }
 639
 640 /**
 641  * g_match_info_ref:
 642  * @match_info: a #GMatchInfo
 643  *
 644  * Increases reference count of @match_info by 1.
 645  *
 646  * Returns: @match_info
 647  *
 648  * Since: 2.30
 649  */
 650 GMatchInfo       *
 651 g_match_info_ref (GMatchInfo *match_info)
 652 {
 653   g_return_val_if_fail (match_info != NULL, NULL);
 654   g_atomic_int_inc (&match_info->ref_count);
 655   return match_info;
 656 }
 657
 658 /**
 659  * g_match_info_unref:
 660  * @match_info: a #GMatchInfo
 661  *
 662  * Decreases reference count of @match_info by 1. When reference count drops
 663  * to zero, it frees all the memory associated with the match_info structure.
 664  *
 665  * Since: 2.30
 666  */
 667 void
 668 g_match_info_unref (GMatchInfo *match_info)
 669 {
 670   if (g_atomic_int_dec_and_test (&match_info->ref_count))
 671     {
 672       g_regex_unref (match_info->regex);
 673       g_free (match_info->offsets);
 674       g_free (match_info->workspace);
 675       g_free (match_info);
 676     }
 677 }
 678
 679 /**
 680  * g_match_info_free:
 681  * @match_info: (allow-none): a #GMatchInfo, or %NULL
 682  *
 683  * If @match_info is not %NULL, calls g_match_info_unref(); otherwise does
 684  * nothing.
 685  *
 686  * Since: 2.14
 687  */
 688 void
 689 g_match_info_free (GMatchInfo *match_info)
 690 {
 691   if (match_info == NULL)
 692     return;
 693
 694   g_match_info_unref (match_info);
 695 }
 696
 697 /**
 698  * g_match_info_next:
 699  * @match_info: a #GMatchInfo structure
 700  * @error: location to store the error occurring, or %NULL to ignore errors
 701  *
 702  * Scans for the next match using the same parameters of the previous
 703  * call to g_regex_match_full() or g_regex_match() that returned
 704  * @match_info.
 705  *
 706  * The match is done on the string passed to the match function, so you
 707  * cannot free it before calling this function.
 708  *
 709  * Returns: %TRUE is the string matched, %FALSE otherwise
 710  *
 711  * Since: 2.14
 712  */
 713 gboolean
 714 g_match_info_next (GMatchInfo  *match_info,
 715                    GError     **error)
 716 {
 717   gint prev_match_start;
 718   gint prev_match_end;
 719
 720   g_return_val_if_fail (match_info != NULL, FALSE);
 721   g_return_val_if_fail (error == NULL || *error == NULL, FALSE);
 722   g_return_val_if_fail (match_info->pos >= 0, FALSE);
 723
 724   prev_match_start = match_info->offsets[0];
 725   prev_match_end = match_info->offsets[1];
 726
 727   if (match_info->pos > match_info->string_len)
 728     {
 729       /* we have reached the end of the string */
 730       match_info->pos = -1;
 731       match_info->matches = PCRE_ERROR_NOMATCH;
 732       return FALSE;
 733     }
 734
 735   match_info->matches = pcre_exec (match_info->regex->pcre_re,
 736                                    match_info->regex->extra,
 737                                    match_info->string,
 738                                    match_info->string_len,
 739                                    match_info->pos,
 740                                    match_info->regex->match_opts | match_info->match_opts,
 741                                    match_info->offsets,
 742                                    match_info->n_offsets);
 743   if (IS_PCRE_ERROR (match_info->matches))
 744     {
 745       g_set_error (error, G_REGEX_ERROR, G_REGEX_ERROR_MATCH,
 746                    _("Error while matching regular expression %s: %s"),
 747                    match_info->regex->pattern, match_error (match_info->matches));
 748       return FALSE;
 749     }
 750
 751   /* avoid infinite loops if the pattern is an empty string or something
 752    * equivalent */
 753   if (match_info->pos == match_info->offsets[1])
 754     {
 755       if (match_info->pos > match_info->string_len)
 756         {
 757           /* we have reached the end of the string */
 758           match_info->pos = -1;
 759           match_info->matches = PCRE_ERROR_NOMATCH;
 760           return FALSE;
 761         }
 762
 763       match_info->pos = NEXT_CHAR (match_info->regex,
 764                                    &match_info->string[match_info->pos]) -
 765                                    match_info->string;
 766     }
 767   else
 768     {
 769       match_info->pos = match_info->offsets[1];
 770     }
 771
 772   /* it's possible to get two identical matches when we are matching
 773    * empty strings, for instance if the pattern is "(?=[A-Z0-9])" and
 774    * the string is "RegExTest" we have:
 775    *  - search at position 0: match from 0 to 0
 776    *  - search at position 1: match from 3 to 3
 777    *  - search at position 3: match from 3 to 3 (duplicate)
 778    *  - search at position 4: match from 5 to 5
 779    *  - search at position 5: match from 5 to 5 (duplicate)
 780    *  - search at position 6: no match -> stop
 781    * so we have to ignore the duplicates.
 782    * see bug #515944: http://bugzilla.gnome.org/show_bug.cgi?id=515944 */
 783   if (match_info->matches >= 0 &&
 784       prev_match_start == match_info->offsets[0] &&
 785       prev_match_end == match_info->offsets[1])
 786     {
 787       /* ignore this match and search the next one */
 788       return g_match_info_next (match_info, error);
 789     }
 790
 791   return match_info->matches >= 0;
 792 }
 793
 794 /**
 795  * g_match_info_matches:
 796  * @match_info: a #GMatchInfo structure
 797  *
 798  * Returns whether the previous match operation succeeded.
 799  *
 800  * Returns: %TRUE if the previous match operation succeeded,
 801  *   %FALSE otherwise
 802  *
 803  * Since: 2.14
 804  */
 805 gboolean
 806 g_match_info_matches (const GMatchInfo *match_info)
 807 {
 808   g_return_val_if_fail (match_info != NULL, FALSE);
 809
 810   return match_info->matches >= 0;
 811 }
 812
 813 /**
 814  * g_match_info_get_match_count:
 815  * @match_info: a #GMatchInfo structure
 816  *
 817  * Retrieves the number of matched substrings (including substring 0,
 818  * that is the whole matched text), so 1 is returned if the pattern
 819  * has no substrings in it and 0 is returned if the match failed.
 820  *
 821  * If the last match was obtained using the DFA algorithm, that is
 822  * using g_regex_match_all() or g_regex_match_all_full(), the retrieved
 823  * count is not that of the number of capturing parentheses but that of
 824  * the number of matched substrings.
 825  *
 826  * Returns: Number of matched substrings, or -1 if an error occurred
 827  *
 828  * Since: 2.14
 829  */
 830 gint
 831 g_match_info_get_match_count (const GMatchInfo *match_info)
 832 {
 833   g_return_val_if_fail (match_info, -1);
 834
 835   if (match_info->matches == PCRE_ERROR_NOMATCH)
 836     /* no match */
 837     return 0;
 838   else if (match_info->matches < PCRE_ERROR_NOMATCH)
 839     /* error */
 840     return -1;
 841   else
 842     /* match */
 843     return match_info->matches;
 844 }
 845
 846 /**
 847  * g_match_info_is_partial_match:
 848  * @match_info: a #GMatchInfo structure
 849  *
 850  * Usually if the string passed to g_regex_match*() matches as far as
 851  * it goes, but is too short to match the entire pattern, %FALSE is
 852  * returned. There are circumstances where it might be helpful to
 853  * distinguish this case from other cases in which there is no match.
 854  *
 855  * Consider, for example, an application where a human is required to
 856  * type in data for a field with specific formatting requirements. An
 857  * example might be a date in the form ddmmmyy, defined by the pattern
 858  * "^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$".
 859  * If the application sees the user’s keystrokes one by one, and can
 860  * check that what has been typed so far is potentially valid, it is
 861  * able to raise an error as soon as a mistake is made.
 862  *
 863  * GRegex supports the concept of partial matching by means of the
 864  * #G_REGEX_MATCH_PARTIAL_SOFT and #G_REGEX_MATCH_PARTIAL_HARD flags.
 865  * When they are used, the return code for
 866  * g_regex_match() or g_regex_match_full() is, as usual, %TRUE
 867  * for a complete match, %FALSE otherwise. But, when these functions
 868  * return %FALSE, you can check if the match was partial calling
 869  * g_match_info_is_partial_match().
 870  *
 871  * The difference between #G_REGEX_MATCH_PARTIAL_SOFT and
 872  * #G_REGEX_MATCH_PARTIAL_HARD is that when a partial match is encountered
 873  * with #G_REGEX_MATCH_PARTIAL_SOFT, matching continues to search for a
 874  * possible complete match, while with #G_REGEX_MATCH_PARTIAL_HARD matching
 875  * stops at the partial match.
 876  * When both #G_REGEX_MATCH_PARTIAL_SOFT and #G_REGEX_MATCH_PARTIAL_HARD
 877  * are set, the latter takes precedence.
 878  *
 879  * There were formerly some restrictions on the pattern for partial matching.
 880  * The restrictions no longer apply.
 881  *
 882  * See pcrepartial(3) for more information on partial matching.
 883  *
 884  * Returns: %TRUE if the match was partial, %FALSE otherwise
 885  *
 886  * Since: 2.14
 887  */
 888 gboolean
 889 g_match_info_is_partial_match (const GMatchInfo *match_info)
 890 {
 891   g_return_val_if_fail (match_info != NULL, FALSE);
 892
 893   return match_info->matches == PCRE_ERROR_PARTIAL;
 894 }
 895
 896 /**
 897  * g_match_info_expand_references:
 898  * @match_info: (allow-none): a #GMatchInfo or %NULL
 899  * @string_to_expand: the string to expand
 900  * @error: location to store the error occurring, or %NULL to ignore errors
 901  *
 902  * Returns a new string containing the text in @string_to_expand with
 903  * references and escape sequences expanded. References refer to the last
 904  * match done with @string against @regex and have the same syntax used by
 905  * g_regex_replace().
 906  *
 907  * The @string_to_expand must be UTF-8 encoded even if #G_REGEX_RAW was
 908  * passed to g_regex_new().
 909  *
 910  * The backreferences are extracted from the string passed to the match
 911  * function, so you cannot call this function after freeing the string.
 912  *
 913  * @match_info may be %NULL in which case @string_to_expand must not
 914  * contain references. For instance "foo\n" does not refer to an actual
 915  * pattern and '\n' merely will be replaced with \n character,
 916  * while to expand "\0" (whole match) one needs the result of a match.
 917  * Use g_regex_check_replacement() to find out whether @string_to_expand
 918  * contains references.
 919  *
 920  * Returns: (allow-none): the expanded string, or %NULL if an error occurred
 921  *
 922  * Since: 2.14
 923  */
 924 gchar *
 925 g_match_info_expand_references (const GMatchInfo  *match_info,
 926                                 const gchar       *string_to_expand,
 927                                 GError           **error)
 928 {
 929   GString *result;
 930   GList *list;
 931   GError *tmp_error = NULL;
 932
 933   g_return_val_if_fail (string_to_expand != NULL, NULL);
 934   g_return_val_if_fail (error == NULL || *error == NULL, NULL);
 935
 936   list = split_replacement (string_to_expand, &tmp_error);
 937   if (tmp_error != NULL)
 938     {
 939       g_propagate_error (error, tmp_error);
 940       return NULL;
 941     }
 942
 943   if (!match_info && interpolation_list_needs_match (list))
 944     {
 945       g_critical ("String '%s' contains references to the match, can't "
 946                   "expand references without GMatchInfo object",
 947                   string_to_expand);
 948       return NULL;
 949     }
 950
 951   result = g_string_sized_new (strlen (string_to_expand));
 952   interpolate_replacement (match_info, result, list);
 953
 954   g_list_free_full (list, (GDestroyNotify) free_interpolation_data);
 955
 956   return g_string_free (result, FALSE);
 957 }
 958
 959 /**
 960  * g_match_info_fetch:
 961  * @match_info: #GMatchInfo structure
 962  * @match_num: number of the sub expression
 963  *
 964  * Retrieves the text matching the @match_num'th capturing
 965  * parentheses. 0 is the full text of the match, 1 is the first paren
 966  * set, 2 the second, and so on.
 967  *
 968  * If @match_num is a valid sub pattern but it didn't match anything
 969  * (e.g. sub pattern 1, matching "b" against "(a)?b") then an empty
 970  * string is returned.
 971  *
 972  * If the match was obtained using the DFA algorithm, that is using
 973  * g_regex_match_all() or g_regex_match_all_full(), the retrieved
 974  * string is not that of a set of parentheses but that of a matched
 975  * substring. Substrings are matched in reverse order of length, so
 976  * 0 is the longest match.
 977  *
 978  * The string is fetched from the string passed to the match function,
 979  * so you cannot call this function after freeing the string.
 980  *
 981  * Returns: (allow-none): The matched substring, or %NULL if an error
 982  *     occurred. You have to free the string yourself
 983  *
 984  * Since: 2.14
 985  */
 986 gchar *
 987 g_match_info_fetch (const GMatchInfo *match_info,
 988                     gint              match_num)
 989 {
 990   /* we cannot use pcre_get_substring() because it allocates the
 991    * string using pcre_malloc(). */
 992   gchar *match = NULL;
 993   gint start, end;
 994
 995   g_return_val_if_fail (match_info != NULL, NULL);
 996   g_return_val_if_fail (match_num >= 0, NULL);
 997
 998   /* match_num does not exist or it didn't matched, i.e. matching "b"
 999    * against "(a)?b" then group 0 is empty. */
1000   if (!g_match_info_fetch_pos (match_info, match_num, &start, &end))
1001     match = NULL;
1002   else if (start == -1)
1003     match = g_strdup ("");
1004   else
1005     match = g_strndup (&match_info->string[start], end - start);
1006
1007   return match;
1008 }
1009
1010 /**
1011  * g_match_info_fetch_pos:
1012  * @match_info: #GMatchInfo structure
1013  * @match_num: number of the sub expression
1014  * @start_pos: (out) (allow-none): pointer to location where to store
1015  *     the start position, or %NULL
1016  * @end_pos: (out) (allow-none): pointer to location where to store
1017  *     the end position, or %NULL
1018  *
1019  * Retrieves the position in bytes of the @match_num'th capturing
1020  * parentheses. 0 is the full text of the match, 1 is the first
1021  * paren set, 2 the second, and so on.
1022  *
1023  * If @match_num is a valid sub pattern but it didn't match anything
1024  * (e.g. sub pattern 1, matching "b" against "(a)?b") then @start_pos
1025  * and @end_pos are set to -1 and %TRUE is returned.
1026  *
1027  * If the match was obtained using the DFA algorithm, that is using
1028  * g_regex_match_all() or g_regex_match_all_full(), the retrieved
1029  * position is not that of a set of parentheses but that of a matched
1030  * substring. Substrings are matched in reverse order of length, so
1031  * 0 is the longest match.
1032  *
1033  * Returns: %TRUE if the position was fetched, %FALSE otherwise. If
1034  *   the position cannot be fetched, @start_pos and @end_pos are left
1035  *   unchanged
1036  *
1037  * Since: 2.14
1038  */
1039 gboolean
1040 g_match_info_fetch_pos (const GMatchInfo *match_info,
1041                         gint              match_num,
1042                         gint             *start_pos,
1043                         gint             *end_pos)
1044 {
1045   g_return_val_if_fail (match_info != NULL, FALSE);
1046   g_return_val_if_fail (match_num >= 0, FALSE);
1047
1048   /* make sure the sub expression number they're requesting is less than
1049    * the total number of sub expressions that were matched. */
1050   if (match_num >= match_info->matches)
1051     return FALSE;
1052
1053   if (start_pos != NULL)
1054     *start_pos = match_info->offsets[2 * match_num];
1055
1056   if (end_pos != NULL)
1057     *end_pos = match_info->offsets[2 * match_num + 1];
1058
1059   return TRUE;
1060 }
1061
1062 /*
1063  * Returns number of first matched subpattern with name @name.
1064  * There may be more than one in case when DUPNAMES is used,
1065  * and not all subpatterns with that name match;
1066  * pcre_get_stringnumber() does not work in that case.
1067  */
1068 static gint
1069 get_matched_substring_number (const GMatchInfo *match_info,
1070                               const gchar      *name)
1071 {
1072   gint entrysize;
1073   gchar *first, *last;
1074   guchar *entry;
1075
1076   if (!(match_info->regex->compile_opts & G_REGEX_DUPNAMES))
1077     return pcre_get_stringnumber (match_info->regex->pcre_re, name);
1078
1079   /* This code is copied from pcre_get.c: get_first_set() */
1080   entrysize = pcre_get_stringtable_entries (match_info->regex->pcre_re,
1081                                             name,
1082                                             &first,
1083                                             &last);
1084
1085   if (entrysize <= 0)
1086     return entrysize;
1087
1088   for (entry = (guchar*) first; entry <= (guchar*) last; entry += entrysize)
1089     {
1090       gint n = (entry[0] << 8) + entry[1];
1091       if (match_info->offsets[n*2] >= 0)
1092         return n;
1093     }
1094
1095   return (first[0] << 8) + first[1];
1096 }
1097
1098 /**
1099  * g_match_info_fetch_named:
1100  * @match_info: #GMatchInfo structure
1101  * @name: name of the subexpression
1102  *
1103  * Retrieves the text matching the capturing parentheses named @name.
1104  *
1105  * If @name is a valid sub pattern name but it didn't match anything
1106  * (e.g. sub pattern "X", matching "b" against "(?P<X>a)?b")
1107  * then an empty string is returned.
1108  *
1109  * The string is fetched from the string passed to the match function,
1110  * so you cannot call this function after freeing the string.
1111  *
1112  * Returns: (allow-none): The matched substring, or %NULL if an error
1113  *     occurred. You have to free the string yourself
1114  *
1115  * Since: 2.14
1116  */
1117 gchar *
1118 g_match_info_fetch_named (const GMatchInfo *match_info,
1119                           const gchar      *name)
1120 {
1121   /* we cannot use pcre_get_named_substring() because it allocates the
1122    * string using pcre_malloc(). */
1123   gint num;
1124
1125   g_return_val_if_fail (match_info != NULL, NULL);
1126   g_return_val_if_fail (name != NULL, NULL);
1127
1128   num = get_matched_substring_number (match_info, name);
1129   if (num < 0)
1130     return NULL;
1131   else
1132     return g_match_info_fetch (match_info, num);
1133 }
1134
1135 /**
1136  * g_match_info_fetch_named_pos:
1137  * @match_info: #GMatchInfo structure
1138  * @name: name of the subexpression
1139  * @start_pos: (out) (allow-none): pointer to location where to store
1140  *     the start position, or %NULL
1141  * @end_pos: (out) (allow-none): pointer to location where to store
1142  *     the end position, or %NULL
1143  *
1144  * Retrieves the position in bytes of the capturing parentheses named @name.
1145  *
1146  * If @name is a valid sub pattern name but it didn't match anything
1147  * (e.g. sub pattern "X", matching "b" against "(?P<X>a)?b")
1148  * then @start_pos and @end_pos are set to -1 and %TRUE is returned.
1149  *
1150  * Returns: %TRUE if the position was fetched, %FALSE otherwise.
1151  *     If the position cannot be fetched, @start_pos and @end_pos
1152  *     are left unchanged.
1153  *
1154  * Since: 2.14
1155  */
1156 gboolean
1157 g_match_info_fetch_named_pos (const GMatchInfo *match_info,
1158                               const gchar      *name,
1159                               gint             *start_pos,
1160                               gint             *end_pos)
1161 {
1162   gint num;
1163
1164   g_return_val_if_fail (match_info != NULL, FALSE);
1165   g_return_val_if_fail (name != NULL, FALSE);
1166
1167   num = get_matched_substring_number (match_info, name);
1168   if (num < 0)
1169     return FALSE;
1170
1171   return g_match_info_fetch_pos (match_info, num, start_pos, end_pos);
1172 }
1173
1174 /**
1175  * g_match_info_fetch_all:
1176  * @match_info: a #GMatchInfo structure
1177  *
1178  * Bundles up pointers to each of the matching substrings from a match
1179  * and stores them in an array of gchar pointers. The first element in
1180  * the returned array is the match number 0, i.e. the entire matched
1181  * text.
1182  *
1183  * If a sub pattern didn't match anything (e.g. sub pattern 1, matching
1184  * "b" against "(a)?b") then an empty string is inserted.
1185  *
1186  * If the last match was obtained using the DFA algorithm, that is using
1187  * g_regex_match_all() or g_regex_match_all_full(), the retrieved
1188  * strings are not that matched by sets of parentheses but that of the
1189  * matched substring. Substrings are matched in reverse order of length,
1190  * so the first one is the longest match.
1191  *
1192  * The strings are fetched from the string passed to the match function,
1193  * so you cannot call this function after freeing the string.
1194  *
1195  * Returns: (transfer full): a %NULL-terminated array of gchar *
1196  *     pointers.  It must be freed using g_strfreev(). If the previous
1197  *     match failed %NULL is returned
1198  *
1199  * Since: 2.14
1200  */
1201 gchar **
1202 g_match_info_fetch_all (const GMatchInfo *match_info)
1203 {
1204   /* we cannot use pcre_get_substring_list() because the returned value
1205    * isn't suitable for g_strfreev(). */
1206   gchar **result;
1207   gint i;
1208
1209   g_return_val_if_fail (match_info != NULL, NULL);
1210
1211   if (match_info->matches < 0)
1212     return NULL;
1213
1214   result = g_new (gchar *, match_info->matches + 1);
1215   for (i = 0; i < match_info->matches; i++)
1216     result[i] = g_match_info_fetch (match_info, i);
1217   result[i] = NULL;
1218
1219   return result;
1220 }
1221
1222
1223 /* GRegex */
1224
1225 G_DEFINE_QUARK (g-regex-error-quark, g_regex_error)
1226
1227 /**
1228  * g_regex_ref:
1229  * @regex: a #GRegex
1230  *
1231  * Increases reference count of @regex by 1.
1232  *
1233  * Returns: @regex
1234  *
1235  * Since: 2.14
1236  */
1237 GRegex *
1238 g_regex_ref (GRegex *regex)
1239 {
1240   g_return_val_if_fail (regex != NULL, NULL);
1241   g_atomic_int_inc (&regex->ref_count);
1242   return regex;
1243 }
1244
1245 /**
1246  * g_regex_unref:
1247  * @regex: a #GRegex
1248  *
1249  * Decreases reference count of @regex by 1. When reference count drops
1250  * to zero, it frees all the memory associated with the regex structure.
1251  *
1252  * Since: 2.14
1253  */
1254 void
1255 g_regex_unref (GRegex *regex)
1256 {
1257   g_return_if_fail (regex != NULL);
1258
1259   if (g_atomic_int_dec_and_test (&regex->ref_count))
1260     {
1261       g_free (regex->pattern);
1262       if (regex->pcre_re != NULL)
1263         pcre_free (regex->pcre_re);
1264       if (regex->extra != NULL)
1265         pcre_free (regex->extra);
1266       g_free (regex);
1267     }
1268 }
1269
1270 /*
1271  * @match_options: (inout) (optional):
1272  */
1273 static pcre *regex_compile (const gchar         *pattern,
1274                             GRegexCompileFlags   compile_options,
1275                             GRegexCompileFlags  *compile_options_out,
1276                             GRegexMatchFlags    *match_options,
1277                             GError             **error);
1278
1279 /**
1280  * g_regex_new:
1281  * @pattern: the regular expression
1282  * @compile_options: compile options for the regular expression, or 0
1283  * @match_options: match options for the regular expression, or 0
1284  * @error: return location for a #GError
1285  *
1286  * Compiles the regular expression to an internal form, and does
1287  * the initial setup of the #GRegex structure.
1288  *
1289  * Returns: (nullable): a #GRegex structure or %NULL if an error occured. Call
1290  *   g_regex_unref() when you are done with it
1291  *
1292  * Since: 2.14
1293  */
1294 GRegex *
1295 g_regex_new (const gchar         *pattern,
1296              GRegexCompileFlags   compile_options,
1297              GRegexMatchFlags     match_options,
1298              GError             **error)
1299 {
1300   GRegex *regex;
1301   pcre *re;
1302   const gchar *errmsg;
1303   gboolean optimize = FALSE;
1304   static volatile gsize initialised = 0;
1305
1306   g_return_val_if_fail (pattern != NULL, NULL);
1307   g_return_val_if_fail (error == NULL || *error == NULL, NULL);
1308   g_return_val_if_fail ((compile_options & ~G_REGEX_COMPILE_MASK) == 0, NULL);
1309   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
1310
1311   if (g_once_init_enter (&initialised))
1312     {
1313       int supports_utf8, supports_ucp;
1314
1315       pcre_config (PCRE_CONFIG_UTF8, &supports_utf8);
1316       if (!supports_utf8)
1317         g_critical (_("PCRE library is compiled without UTF8 support"));
1318
1319       pcre_config (PCRE_CONFIG_UNICODE_PROPERTIES, &supports_ucp);
1320       if (!supports_ucp)
1321         g_critical (_("PCRE library is compiled without UTF8 properties support"));
1322
1323       g_once_init_leave (&initialised, supports_utf8 && supports_ucp ? 1 : 2);
1324     }
1325
1326   if (G_UNLIKELY (initialised != 1))
1327     {
1328       g_set_error_literal (error, G_REGEX_ERROR, G_REGEX_ERROR_COMPILE,
1329                            _("PCRE library is compiled with incompatible options"));
1330       return NULL;
1331     }
1332
1333   /* G_REGEX_OPTIMIZE has the same numeric value of PCRE_NO_UTF8_CHECK,
1334    * as we do not need to wrap PCRE_NO_UTF8_CHECK. */
1335   if (compile_options & G_REGEX_OPTIMIZE)
1336     optimize = TRUE;
1337
1338   re = regex_compile (pattern, compile_options, &compile_options,
1339                       &match_options, error);
1340
1341   if (re == NULL)
1342     return NULL;
1343
1344   regex = g_new0 (GRegex, 1);
1345   regex->ref_count = 1;
1346   regex->pattern = g_strdup (pattern);
1347   regex->pcre_re = re;
1348   regex->compile_opts = compile_options;
1349   regex->match_opts = match_options;
1350
1351   if (optimize)
1352     {
1353       regex->extra = pcre_study (regex->pcre_re, 0, &errmsg);
1354       if (errmsg != NULL)
1355         {
1356           GError *tmp_error = g_error_new (G_REGEX_ERROR,
1357                                            G_REGEX_ERROR_OPTIMIZE,
1358                                            _("Error while optimizing "
1359                                              "regular expression %s: %s"),
1360                                            regex->pattern,
1361                                            errmsg);
1362           g_propagate_error (error, tmp_error);
1363
1364           g_regex_unref (regex);
1365           return NULL;
1366         }
1367     }
1368
1369   return regex;
1370 }
1371
1372 static pcre *
1373 regex_compile (const gchar         *pattern,
1374                GRegexCompileFlags   compile_options,
1375                GRegexCompileFlags  *compile_options_out,
1376                GRegexMatchFlags    *match_options,
1377                GError             **error)
1378 {
1379   pcre *re;
1380   const gchar *errmsg;
1381   gint erroffset;
1382   gint errcode;
1383   GRegexCompileFlags nonpcre_compile_options;
1384   unsigned long int pcre_compile_options;
1385
1386   nonpcre_compile_options = compile_options & G_REGEX_COMPILE_NONPCRE_MASK;
1387
1388   /* In GRegex the string are, by default, UTF-8 encoded. PCRE
1389    * instead uses UTF-8 only if required with PCRE_UTF8. */
1390   if (compile_options & G_REGEX_RAW)
1391     {
1392       /* disable utf-8 */
1393       compile_options &= ~G_REGEX_RAW;
1394     }
1395   else
1396     {
1397       /* enable utf-8 */
1398       compile_options |= PCRE_UTF8 | PCRE_NO_UTF8_CHECK;
1399
1400       if (match_options != NULL)
1401         *match_options |= PCRE_NO_UTF8_CHECK;
1402     }
1403
1404   /* PCRE_NEWLINE_ANY is the default for the internal PCRE but
1405    * not for the system one. */
1406   if (!(compile_options & G_REGEX_NEWLINE_CR) &&
1407       !(compile_options & G_REGEX_NEWLINE_LF))
1408     {
1409       compile_options |= PCRE_NEWLINE_ANY;
1410     }
1411
1412   compile_options |= PCRE_UCP;
1413
1414   /* PCRE_BSR_UNICODE is the default for the internal PCRE but
1415    * possibly not for the system one.
1416    */
1417   if (~compile_options & G_REGEX_BSR_ANYCRLF)
1418     compile_options |= PCRE_BSR_UNICODE;
1419
1420   /* compile the pattern */
1421   re = pcre_compile2 (pattern, compile_options, &errcode,
1422                       &errmsg, &erroffset, NULL);
1423
1424   /* if the compilation failed, set the error member and return
1425    * immediately */
1426   if (re == NULL)
1427     {
1428       GError *tmp_error;
1429
1430       /* Translate the PCRE error code to GRegexError and use a translated
1431        * error message if possible */
1432       translate_compile_error (&errcode, &errmsg);
1433
1434       /* PCRE uses byte offsets but we want to show character offsets */
1435       erroffset = g_utf8_pointer_to_offset (pattern, &pattern[erroffset]);
1436
1437       tmp_error = g_error_new (G_REGEX_ERROR, errcode,
1438                                _("Error while compiling regular "
1439                                  "expression %s at char %d: %s"),
1440                                pattern, erroffset, errmsg);
1441       g_propagate_error (error, tmp_error);
1442
1443       return NULL;
1444     }
1445
1446   /* For options set at the beginning of the pattern, pcre puts them into
1447    * compile options, e.g. "(?i)foo" will make the pcre structure store
1448    * PCRE_CASELESS even though it wasn't explicitly given for compilation. */
1449   pcre_fullinfo (re, NULL, PCRE_INFO_OPTIONS, &pcre_compile_options);
1450   compile_options = pcre_compile_options & G_REGEX_COMPILE_PCRE_MASK;
1451
1452   /* Don't leak PCRE_NEWLINE_ANY, which is part of PCRE_NEWLINE_ANYCRLF */
1453   if ((pcre_compile_options & PCRE_NEWLINE_ANYCRLF) != PCRE_NEWLINE_ANYCRLF)
1454     compile_options &= ~PCRE_NEWLINE_ANY;
1455
1456   compile_options |= nonpcre_compile_options;
1457
1458   if (!(compile_options & G_REGEX_DUPNAMES))
1459     {
1460       gboolean jchanged = FALSE;
1461       pcre_fullinfo (re, NULL, PCRE_INFO_JCHANGED, &jchanged);
1462       if (jchanged)
1463         compile_options |= G_REGEX_DUPNAMES;
1464     }
1465
1466   if (compile_options_out != 0)
1467     *compile_options_out = compile_options;
1468
1469   return re;
1470 }
1471
1472 /**
1473  * g_regex_get_pattern:
1474  * @regex: a #GRegex structure
1475  *
1476  * Gets the pattern string associated with @regex, i.e. a copy of
1477  * the string passed to g_regex_new().
1478  *
1479  * Returns: the pattern of @regex
1480  *
1481  * Since: 2.14
1482  */
1483 const gchar *
1484 g_regex_get_pattern (const GRegex *regex)
1485 {
1486   g_return_val_if_fail (regex != NULL, NULL);
1487
1488   return regex->pattern;
1489 }
1490
1491 /**
1492  * g_regex_get_max_backref:
1493  * @regex: a #GRegex
1494  *
1495  * Returns the number of the highest back reference
1496  * in the pattern, or 0 if the pattern does not contain
1497  * back references.
1498  *
1499  * Returns: the number of the highest back reference
1500  *
1501  * Since: 2.14
1502  */
1503 gint
1504 g_regex_get_max_backref (const GRegex *regex)
1505 {
1506   gint value;
1507
1508   pcre_fullinfo (regex->pcre_re, regex->extra,
1509                  PCRE_INFO_BACKREFMAX, &value);
1510
1511   return value;
1512 }
1513
1514 /**
1515  * g_regex_get_capture_count:
1516  * @regex: a #GRegex
1517  *
1518  * Returns the number of capturing subpatterns in the pattern.
1519  *
1520  * Returns: the number of capturing subpatterns
1521  *
1522  * Since: 2.14
1523  */
1524 gint
1525 g_regex_get_capture_count (const GRegex *regex)
1526 {
1527   gint value;
1528
1529   pcre_fullinfo (regex->pcre_re, regex->extra,
1530                  PCRE_INFO_CAPTURECOUNT, &value);
1531
1532   return value;
1533 }
1534
1535 /**
1536  * g_regex_get_has_cr_or_lf:
1537  * @regex: a #GRegex structure
1538  *
1539  * Checks whether the pattern contains explicit CR or LF references.
1540  *
1541  * Returns: %TRUE if the pattern contains explicit CR or LF references
1542  *
1543  * Since: 2.34
1544  */
1545 gboolean
1546 g_regex_get_has_cr_or_lf (const GRegex *regex)
1547 {
1548   gint value;
1549
1550   pcre_fullinfo (regex->pcre_re, regex->extra,
1551                  PCRE_INFO_HASCRORLF, &value);
1552
1553   return !!value;
1554 }
1555
1556 /**
1557  * g_regex_get_max_lookbehind:
1558  * @regex: a #GRegex structure
1559  *
1560  * Gets the number of characters in the longest lookbehind assertion in the
1561  * pattern. This information is useful when doing multi-segment matching using
1562  * the partial matching facilities.
1563  *
1564  * Returns: the number of characters in the longest lookbehind assertion.
1565  *
1566  * Since: 2.38
1567  */
1568 gint
1569 g_regex_get_max_lookbehind (const GRegex *regex)
1570 {
1571   gint max_lookbehind;
1572
1573   pcre_fullinfo (regex->pcre_re, regex->extra,
1574                  PCRE_INFO_MAXLOOKBEHIND, &max_lookbehind);
1575
1576   return max_lookbehind;
1577 }
1578
1579 /**
1580  * g_regex_get_compile_flags:
1581  * @regex: a #GRegex
1582  *
1583  * Returns the compile options that @regex was created with.
1584  *
1585  * Returns: flags from #GRegexCompileFlags
1586  *
1587  * Since: 2.26
1588  */
1589 GRegexCompileFlags
1590 g_regex_get_compile_flags (const GRegex *regex)
1591 {
1592   g_return_val_if_fail (regex != NULL, 0);
1593
1594   return regex->compile_opts;
1595 }
1596
1597 /**
1598  * g_regex_get_match_flags:
1599  * @regex: a #GRegex
1600  *
1601  * Returns the match options that @regex was created with.
1602  *
1603  * Returns: flags from #GRegexMatchFlags
1604  *
1605  * Since: 2.26
1606  */
1607 GRegexMatchFlags
1608 g_regex_get_match_flags (const GRegex *regex)
1609 {
1610   g_return_val_if_fail (regex != NULL, 0);
1611
1612   return regex->match_opts & G_REGEX_MATCH_MASK;
1613 }
1614
1615 /**
1616  * g_regex_match_simple:
1617  * @pattern: the regular expression
1618  * @string: the string to scan for matches
1619  * @compile_options: compile options for the regular expression, or 0
1620  * @match_options: match options, or 0
1621  *
1622  * Scans for a match in @string for @pattern.
1623  *
1624  * This function is equivalent to g_regex_match() but it does not
1625  * require to compile the pattern with g_regex_new(), avoiding some
1626  * lines of code when you need just to do a match without extracting
1627  * substrings, capture counts, and so on.
1628  *
1629  * If this function is to be called on the same @pattern more than
1630  * once, it's more efficient to compile the pattern once with
1631  * g_regex_new() and then use g_regex_match().
1632  *
1633  * Returns: %TRUE if the string matched, %FALSE otherwise
1634  *
1635  * Since: 2.14
1636  */
1637 gboolean
1638 g_regex_match_simple (const gchar        *pattern,
1639                       const gchar        *string,
1640                       GRegexCompileFlags  compile_options,
1641                       GRegexMatchFlags    match_options)
1642 {
1643   GRegex *regex;
1644   gboolean result;
1645
1646   regex = g_regex_new (pattern, compile_options, 0, NULL);
1647   if (!regex)
1648     return FALSE;
1649   result = g_regex_match_full (regex, string, -1, 0, match_options, NULL, NULL);
1650   g_regex_unref (regex);
1651   return result;
1652 }
1653
1654 /**
1655  * g_regex_match:
1656  * @regex: a #GRegex structure from g_regex_new()
1657  * @string: the string to scan for matches
1658  * @match_options: match options
1659  * @match_info: (out) (allow-none): pointer to location where to store
1660  *     the #GMatchInfo, or %NULL if you do not need it
1661  *
1662  * Scans for a match in string for the pattern in @regex.
1663  * The @match_options are combined with the match options specified
1664  * when the @regex structure was created, letting you have more
1665  * flexibility in reusing #GRegex structures.
1666  *
1667  * A #GMatchInfo structure, used to get information on the match,
1668  * is stored in @match_info if not %NULL. Note that if @match_info
1669  * is not %NULL then it is created even if the function returns %FALSE,
1670  * i.e. you must free it regardless if regular expression actually matched.
1671  *
1672  * To retrieve all the non-overlapping matches of the pattern in
1673  * string you can use g_match_info_next().
1674  *
1675  * |[<!-- language="C" -->
1676  * static void
1677  * print_uppercase_words (const gchar *string)
1678  * {
1679  *   // Print all uppercase-only words.
1680  *   GRegex *regex;
1681  *   GMatchInfo *match_info;
1682  *
1683  *   regex = g_regex_new ("[A-Z]+", 0, 0, NULL);
1684  *   g_regex_match (regex, string, 0, &match_info);
1685  *   while (g_match_info_matches (match_info))
1686  *     {
1687  *       gchar *word = g_match_info_fetch (match_info, 0);
1688  *       g_print ("Found: %s\n", word);
1689  *       g_free (word);
1690  *       g_match_info_next (match_info, NULL);
1691  *     }
1692  *   g_match_info_free (match_info);
1693  *   g_regex_unref (regex);
1694  * }
1695  * ]|
1696  *
1697  * @string is not copied and is used in #GMatchInfo internally. If
1698  * you use any #GMatchInfo method (except g_match_info_free()) after
1699  * freeing or modifying @string then the behaviour is undefined.
1700  *
1701  * Returns: %TRUE is the string matched, %FALSE otherwise
1702  *
1703  * Since: 2.14
1704  */
1705 gboolean
1706 g_regex_match (const GRegex      *regex,
1707                const gchar       *string,
1708                GRegexMatchFlags   match_options,
1709                GMatchInfo       **match_info)
1710 {
1711   return g_regex_match_full (regex, string, -1, 0, match_options,
1712                              match_info, NULL);
1713 }
1714
1715 /**
1716  * g_regex_match_full:
1717  * @regex: a #GRegex structure from g_regex_new()
1718  * @string: (array length=string_len): the string to scan for matches
1719  * @string_len: the length of @string, or -1 if @string is nul-terminated
1720  * @start_position: starting index of the string to match, in bytes
1721  * @match_options: match options
1722  * @match_info: (out) (allow-none): pointer to location where to store
1723  *     the #GMatchInfo, or %NULL if you do not need it
1724  * @error: location to store the error occurring, or %NULL to ignore errors
1725  *
1726  * Scans for a match in string for the pattern in @regex.
1727  * The @match_options are combined with the match options specified
1728  * when the @regex structure was created, letting you have more
1729  * flexibility in reusing #GRegex structures.
1730  *
1731  * Setting @start_position differs from just passing over a shortened
1732  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern
1733  * that begins with any kind of lookbehind assertion, such as "\b".
1734  *
1735  * A #GMatchInfo structure, used to get information on the match, is
1736  * stored in @match_info if not %NULL. Note that if @match_info is
1737  * not %NULL then it is created even if the function returns %FALSE,
1738  * i.e. you must free it regardless if regular expression actually
1739  * matched.
1740  *
1741  * @string is not copied and is used in #GMatchInfo internally. If
1742  * you use any #GMatchInfo method (except g_match_info_free()) after
1743  * freeing or modifying @string then the behaviour is undefined.
1744  *
1745  * To retrieve all the non-overlapping matches of the pattern in
1746  * string you can use g_match_info_next().
1747  *
1748  * |[<!-- language="C" -->
1749  * static void
1750  * print_uppercase_words (const gchar *string)
1751  * {
1752  *   // Print all uppercase-only words.
1753  *   GRegex *regex;
1754  *   GMatchInfo *match_info;
1755  *   GError *error = NULL;
1756  *
1757  *   regex = g_regex_new ("[A-Z]+", 0, 0, NULL);
1758  *   g_regex_match_full (regex, string, -1, 0, 0, &match_info, &error);
1759  *   while (g_match_info_matches (match_info))
1760  *     {
1761  *       gchar *word = g_match_info_fetch (match_info, 0);
1762  *       g_print ("Found: %s\n", word);
1763  *       g_free (word);
1764  *       g_match_info_next (match_info, &error);
1765  *     }
1766  *   g_match_info_free (match_info);
1767  *   g_regex_unref (regex);
1768  *   if (error != NULL)
1769  *     {
1770  *       g_printerr ("Error while matching: %s\n", error->message);
1771  *       g_error_free (error);
1772  *     }
1773  * }
1774  * ]|
1775  *
1776  * Returns: %TRUE is the string matched, %FALSE otherwise
1777  *
1778  * Since: 2.14
1779  */
1780 gboolean
1781 g_regex_match_full (const GRegex      *regex,
1782                     const gchar       *string,
1783                     gssize             string_len,
1784                     gint               start_position,
1785                     GRegexMatchFlags   match_options,
1786                     GMatchInfo       **match_info,
1787                     GError           **error)
1788 {
1789   GMatchInfo *info;
1790   gboolean match_ok;
1791
1792   g_return_val_if_fail (regex != NULL, FALSE);
1793   g_return_val_if_fail (string != NULL, FALSE);
1794   g_return_val_if_fail (start_position >= 0, FALSE);
1795   g_return_val_if_fail (error == NULL || *error == NULL, FALSE);
1796   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, FALSE);
1797
1798   info = match_info_new (regex, string, string_len, start_position,
1799                          match_options, FALSE);
1800   match_ok = g_match_info_next (info, error);
1801   if (match_info != NULL)
1802     *match_info = info;
1803   else
1804     g_match_info_free (info);
1805
1806   return match_ok;
1807 }
1808
1809 /**
1810  * g_regex_match_all:
1811  * @regex: a #GRegex structure from g_regex_new()
1812  * @string: the string to scan for matches
1813  * @match_options: match options
1814  * @match_info: (out) (allow-none): pointer to location where to store
1815  *     the #GMatchInfo, or %NULL if you do not need it
1816  *
1817  * Using the standard algorithm for regular expression matching only
1818  * the longest match in the string is retrieved. This function uses
1819  * a different algorithm so it can retrieve all the possible matches.
1820  * For more documentation see g_regex_match_all_full().
1821  *
1822  * A #GMatchInfo structure, used to get information on the match, is
1823  * stored in @match_info if not %NULL. Note that if @match_info is
1824  * not %NULL then it is created even if the function returns %FALSE,
1825  * i.e. you must free it regardless if regular expression actually
1826  * matched.
1827  *
1828  * @string is not copied and is used in #GMatchInfo internally. If
1829  * you use any #GMatchInfo method (except g_match_info_free()) after
1830  * freeing or modifying @string then the behaviour is undefined.
1831  *
1832  * Returns: %TRUE is the string matched, %FALSE otherwise
1833  *
1834  * Since: 2.14
1835  */
1836 gboolean
1837 g_regex_match_all (const GRegex      *regex,
1838                    const gchar       *string,
1839                    GRegexMatchFlags   match_options,
1840                    GMatchInfo       **match_info)
1841 {
1842   return g_regex_match_all_full (regex, string, -1, 0, match_options,
1843                                  match_info, NULL);
1844 }
1845
1846 /**
1847  * g_regex_match_all_full:
1848  * @regex: a #GRegex structure from g_regex_new()
1849  * @string: (array length=string_len): the string to scan for matches
1850  * @string_len: the length of @string, or -1 if @string is nul-terminated
1851  * @start_position: starting index of the string to match, in bytes
1852  * @match_options: match options
1853  * @match_info: (out) (allow-none): pointer to location where to store
1854  *     the #GMatchInfo, or %NULL if you do not need it
1855  * @error: location to store the error occurring, or %NULL to ignore errors
1856  *
1857  * Using the standard algorithm for regular expression matching only
1858  * the longest match in the string is retrieved, it is not possible
1859  * to obtain all the available matches. For instance matching
1860  * "<a> <b> <c>" against the pattern "<.*>"
1861  * you get "<a> <b> <c>".
1862  *
1863  * This function uses a different algorithm (called DFA, i.e. deterministic
1864  * finite automaton), so it can retrieve all the possible matches, all
1865  * starting at the same point in the string. For instance matching
1866  * "<a> <b> <c>" against the pattern "<.*>;"
1867  * you would obtain three matches: "<a> <b> <c>",
1868  * "<a> <b>" and "<a>".
1869  *
1870  * The number of matched strings is retrieved using
1871  * g_match_info_get_match_count(). To obtain the matched strings and
1872  * their position you can use, respectively, g_match_info_fetch() and
1873  * g_match_info_fetch_pos(). Note that the strings are returned in
1874  * reverse order of length; that is, the longest matching string is
1875  * given first.
1876  *
1877  * Note that the DFA algorithm is slower than the standard one and it
1878  * is not able to capture substrings, so backreferences do not work.
1879  *
1880  * Setting @start_position differs from just passing over a shortened
1881  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern
1882  * that begins with any kind of lookbehind assertion, such as "\b".
1883  *
1884  * A #GMatchInfo structure, used to get information on the match, is
1885  * stored in @match_info if not %NULL. Note that if @match_info is
1886  * not %NULL then it is created even if the function returns %FALSE,
1887  * i.e. you must free it regardless if regular expression actually
1888  * matched.
1889  *
1890  * @string is not copied and is used in #GMatchInfo internally. If
1891  * you use any #GMatchInfo method (except g_match_info_free()) after
1892  * freeing or modifying @string then the behaviour is undefined.
1893  *
1894  * Returns: %TRUE is the string matched, %FALSE otherwise
1895  *
1896  * Since: 2.14
1897  */
1898 gboolean
1899 g_regex_match_all_full (const GRegex      *regex,
1900                         const gchar       *string,
1901                         gssize             string_len,
1902                         gint               start_position,
1903                         GRegexMatchFlags   match_options,
1904                         GMatchInfo       **match_info,
1905                         GError           **error)
1906 {
1907   GMatchInfo *info;
1908   gboolean done;
1909   pcre *pcre_re;
1910   pcre_extra *extra;
1911
1912   g_return_val_if_fail (regex != NULL, FALSE);
1913   g_return_val_if_fail (string != NULL, FALSE);
1914   g_return_val_if_fail (start_position >= 0, FALSE);
1915   g_return_val_if_fail (error == NULL || *error == NULL, FALSE);
1916   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, FALSE);
1917
1918 #ifdef PCRE_NO_AUTO_POSSESS
1919   /* For PCRE >= 8.34 we need to turn off PCRE_NO_AUTO_POSSESS, which
1920    * is an optimization for normal regex matching, but results in omitting
1921    * some shorter matches here, and an observable behaviour change.
1922    *
1923    * DFA matching is rather niche, and very rarely used according to
1924    * codesearch.debian.net, so don't bother caching the recompiled RE. */
1925   pcre_re = regex_compile (regex->pattern,
1926                            regex->compile_opts | PCRE_NO_AUTO_POSSESS,
1927                            NULL, NULL, error);
1928
1929   if (pcre_re == NULL)
1930     return FALSE;
1931
1932   /* Not bothering to cache the optimization data either, with similar
1933    * reasoning */
1934   extra = NULL;
1935 #else
1936   /* For PCRE < 8.33 the precompiled regex is fine. */
1937   pcre_re = regex->pcre_re;
1938   extra = regex->extra;
1939 #endif
1940
1941   info = match_info_new (regex, string, string_len, start_position,
1942                          match_options, TRUE);
1943
1944   done = FALSE;
1945   while (!done)
1946     {
1947       done = TRUE;
1948       info->matches = pcre_dfa_exec (pcre_re, extra,
1949                                      info->string, info->string_len,
1950                                      info->pos,
1951                                      regex->match_opts | match_options,
1952                                      info->offsets, info->n_offsets,
1953                                      info->workspace, info->n_workspace);
1954       if (info->matches == PCRE_ERROR_DFA_WSSIZE)
1955         {
1956           /* info->workspace is too small. */
1957           info->n_workspace *= 2;
1958           info->workspace = g_realloc (info->workspace,
1959                                        info->n_workspace * sizeof (gint));
1960           done = FALSE;
1961         }
1962       else if (info->matches == 0)
1963         {
1964           /* info->offsets is too small. */
1965           info->n_offsets *= 2;
1966           info->offsets = g_realloc (info->offsets,
1967                                      info->n_offsets * sizeof (gint));
1968           done = FALSE;
1969         }
1970       else if (IS_PCRE_ERROR (info->matches))
1971         {
1972           g_set_error (error, G_REGEX_ERROR, G_REGEX_ERROR_MATCH,
1973                        _("Error while matching regular expression %s: %s"),
1974                        regex->pattern, match_error (info->matches));
1975         }
1976     }
1977
1978 #ifdef PCRE_NO_AUTO_POSSESS
1979   pcre_free (pcre_re);
1980 #endif
1981
1982   /* set info->pos to -1 so that a call to g_match_info_next() fails. */
1983   info->pos = -1;
1984
1985   if (match_info != NULL)
1986     *match_info = info;
1987   else
1988     g_match_info_free (info);
1989
1990   return info->matches >= 0;
1991 }
1992
1993 /**
1994  * g_regex_get_string_number:
1995  * @regex: #GRegex structure
1996  * @name: name of the subexpression
1997  *
1998  * Retrieves the number of the subexpression named @name.
1999  *
2000  * Returns: The number of the subexpression or -1 if @name
2001  *   does not exists
2002  *
2003  * Since: 2.14
2004  */
2005 gint
2006 g_regex_get_string_number (const GRegex *regex,
2007                            const gchar  *name)
2008 {
2009   gint num;
2010
2011   g_return_val_if_fail (regex != NULL, -1);
2012   g_return_val_if_fail (name != NULL, -1);
2013
2014   num = pcre_get_stringnumber (regex->pcre_re, name);
2015   if (num == PCRE_ERROR_NOSUBSTRING)
2016     num = -1;
2017
2018   return num;
2019 }
2020
2021 /**
2022  * g_regex_split_simple:
2023  * @pattern: the regular expression
2024  * @string: the string to scan for matches
2025  * @compile_options: compile options for the regular expression, or 0
2026  * @match_options: match options, or 0
2027  *
2028  * Breaks the string on the pattern, and returns an array of
2029  * the tokens. If the pattern contains capturing parentheses,
2030  * then the text for each of the substrings will also be returned.
2031  * If the pattern does not match anywhere in the string, then the
2032  * whole string is returned as the first token.
2033  *
2034  * This function is equivalent to g_regex_split() but it does
2035  * not require to compile the pattern with g_regex_new(), avoiding
2036  * some lines of code when you need just to do a split without
2037  * extracting substrings, capture counts, and so on.
2038  *
2039  * If this function is to be called on the same @pattern more than
2040  * once, it's more efficient to compile the pattern once with
2041  * g_regex_new() and then use g_regex_split().
2042  *
2043  * As a special case, the result of splitting the empty string ""
2044  * is an empty vector, not a vector containing a single string.
2045  * The reason for this special case is that being able to represent
2046  * a empty vector is typically more useful than consistent handling
2047  * of empty elements. If you do need to represent empty elements,
2048  * you'll need to check for the empty string before calling this
2049  * function.
2050  *
2051  * A pattern that can match empty strings splits @string into
2052  * separate characters wherever it matches the empty string between
2053  * characters. For example splitting "ab c" using as a separator
2054  * "\s*", you will get "a", "b" and "c".
2055  *
2056  * Returns: (transfer full): a %NULL-terminated array of strings. Free
2057  * it using g_strfreev()
2058  *
2059  * Since: 2.14
2060  **/
2061 gchar **
2062 g_regex_split_simple (const gchar        *pattern,
2063                       const gchar        *string,
2064                       GRegexCompileFlags  compile_options,
2065                       GRegexMatchFlags    match_options)
2066 {
2067   GRegex *regex;
2068   gchar **result;
2069
2070   regex = g_regex_new (pattern, compile_options, 0, NULL);
2071   if (!regex)
2072     return NULL;
2073
2074   result = g_regex_split_full (regex, string, -1, 0, match_options, 0, NULL);
2075   g_regex_unref (regex);
2076   return result;
2077 }
2078
2079 /**
2080  * g_regex_split:
2081  * @regex: a #GRegex structure
2082  * @string: the string to split with the pattern
2083  * @match_options: match time option flags
2084  *
2085  * Breaks the string on the pattern, and returns an array of the tokens.
2086  * If the pattern contains capturing parentheses, then the text for each
2087  * of the substrings will also be returned. If the pattern does not match
2088  * anywhere in the string, then the whole string is returned as the first
2089  * token.
2090  *
2091  * As a special case, the result of splitting the empty string "" is an
2092  * empty vector, not a vector containing a single string. The reason for
2093  * this special case is that being able to represent a empty vector is
2094  * typically more useful than consistent handling of empty elements. If
2095  * you do need to represent empty elements, you'll need to check for the
2096  * empty string before calling this function.
2097  *
2098  * A pattern that can match empty strings splits @string into separate
2099  * characters wherever it matches the empty string between characters.
2100  * For example splitting "ab c" using as a separator "\s*", you will get
2101  * "a", "b" and "c".
2102  *
2103  * Returns: (transfer full): a %NULL-terminated gchar ** array. Free
2104  * it using g_strfreev()
2105  *
2106  * Since: 2.14
2107  **/
2108 gchar **
2109 g_regex_split (const GRegex     *regex,
2110                const gchar      *string,
2111                GRegexMatchFlags  match_options)
2112 {
2113   return g_regex_split_full (regex, string, -1, 0,
2114                              match_options, 0, NULL);
2115 }
2116
2117 /**
2118  * g_regex_split_full:
2119  * @regex: a #GRegex structure
2120  * @string: (array length=string_len): the string to split with the pattern
2121  * @string_len: the length of @string, or -1 if @string is nul-terminated
2122  * @start_position: starting index of the string to match, in bytes
2123  * @match_options: match time option flags
2124  * @max_tokens: the maximum number of tokens to split @string into.
2125  *   If this is less than 1, the string is split completely
2126  * @error: return location for a #GError
2127  *
2128  * Breaks the string on the pattern, and returns an array of the tokens.
2129  * If the pattern contains capturing parentheses, then the text for each
2130  * of the substrings will also be returned. If the pattern does not match
2131  * anywhere in the string, then the whole string is returned as the first
2132  * token.
2133  *
2134  * As a special case, the result of splitting the empty string "" is an
2135  * empty vector, not a vector containing a single string. The reason for
2136  * this special case is that being able to represent a empty vector is
2137  * typically more useful than consistent handling of empty elements. If
2138  * you do need to represent empty elements, you'll need to check for the
2139  * empty string before calling this function.
2140  *
2141  * A pattern that can match empty strings splits @string into separate
2142  * characters wherever it matches the empty string between characters.
2143  * For example splitting "ab c" using as a separator "\s*", you will get
2144  * "a", "b" and "c".
2145  *
2146  * Setting @start_position differs from just passing over a shortened
2147  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern
2148  * that begins with any kind of lookbehind assertion, such as "\b".
2149  *
2150  * Returns: (transfer full): a %NULL-terminated gchar ** array. Free
2151  * it using g_strfreev()
2152  *
2153  * Since: 2.14
2154  **/
2155 gchar **
2156 g_regex_split_full (const GRegex      *regex,
2157                     const gchar       *string,
2158                     gssize             string_len,
2159                     gint               start_position,
2160                     GRegexMatchFlags   match_options,
2161                     gint               max_tokens,
2162                     GError           **error)
2163 {
2164   GError *tmp_error = NULL;
2165   GMatchInfo *match_info;
2166   GList *list, *last;
2167   gint i;
2168   gint token_count;
2169   gboolean match_ok;
2170   /* position of the last separator. */
2171   gint last_separator_end;
2172   /* was the last match 0 bytes long? */
2173   gboolean last_match_is_empty;
2174   /* the returned array of char **s */
2175   gchar **string_list;
2176
2177   g_return_val_if_fail (regex != NULL, NULL);
2178   g_return_val_if_fail (string != NULL, NULL);
2179   g_return_val_if_fail (start_position >= 0, NULL);
2180   g_return_val_if_fail (error == NULL || *error == NULL, NULL);
2181   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
2182
2183   if (max_tokens <= 0)
2184     max_tokens = G_MAXINT;
2185
2186   if (string_len < 0)
2187     string_len = strlen (string);
2188
2189   /* zero-length string */
2190   if (string_len - start_position == 0)
2191     return g_new0 (gchar *, 1);
2192
2193   if (max_tokens == 1)
2194     {
2195       string_list = g_new0 (gchar *, 2);
2196       string_list[0] = g_strndup (&string[start_position],
2197                                   string_len - start_position);
2198       return string_list;
2199     }
2200
2201   list = NULL;
2202   token_count = 0;
2203   last_separator_end = start_position;
2204   last_match_is_empty = FALSE;
2205
2206   match_ok = g_regex_match_full (regex, string, string_len, start_position,
2207                                  match_options, &match_info, &tmp_error);
2208
2209   while (tmp_error == NULL)
2210     {
2211       if (match_ok)
2212         {
2213           last_match_is_empty =
2214                     (match_info->offsets[0] == match_info->offsets[1]);
2215
2216           /* we need to skip empty separators at the same position of the end
2217            * of another separator. e.g. the string is "a b" and the separator
2218            * is " *", so from 1 to 2 we have a match and at position 2 we have
2219            * an empty match. */
2220           if (last_separator_end != match_info->offsets[1])
2221             {
2222               gchar *token;
2223               gint match_count;
2224
2225               token = g_strndup (string + last_separator_end,
2226                                  match_info->offsets[0] - last_separator_end);
2227               list = g_list_prepend (list, token);
2228               token_count++;
2229
2230               /* if there were substrings, these need to be added to
2231                * the list. */
2232               match_count = g_match_info_get_match_count (match_info);
2233               if (match_count > 1)
2234                 {
2235                   for (i = 1; i < match_count; i++)
2236                     list = g_list_prepend (list, g_match_info_fetch (match_info, i));
2237                 }
2238             }
2239         }
2240       else
2241         {
2242           /* if there was no match, copy to end of string. */
2243           if (!last_match_is_empty)
2244             {
2245               gchar *token = g_strndup (string + last_separator_end,
2246                                         match_info->string_len - last_separator_end);
2247               list = g_list_prepend (list, token);
2248             }
2249           /* no more tokens, end the loop. */
2250           break;
2251         }
2252
2253       /* -1 to leave room for the last part. */
2254       if (token_count >= max_tokens - 1)
2255         {
2256           /* we have reached the maximum number of tokens, so we copy
2257            * the remaining part of the string. */
2258           if (last_match_is_empty)
2259             {
2260               /* the last match was empty, so we have moved one char
2261                * after the real position to avoid empty matches at the
2262                * same position. */
2263               match_info->pos = PREV_CHAR (regex, &string[match_info->pos]) - string;
2264             }
2265           /* the if is needed in the case we have terminated the available
2266            * tokens, but we are at the end of the string, so there are no
2267            * characters left to copy. */
2268           if (string_len > match_info->pos)
2269             {
2270               gchar *token = g_strndup (string + match_info->pos,
2271                                         string_len - match_info->pos);
2272               list = g_list_prepend (list, token);
2273             }
2274           /* end the loop. */
2275           break;
2276         }
2277
2278       last_separator_end = match_info->pos;
2279       if (last_match_is_empty)
2280         /* if the last match was empty, g_match_info_next() has moved
2281          * forward to avoid infinite loops, but we still need to copy that
2282          * character. */
2283         last_separator_end = PREV_CHAR (regex, &string[last_separator_end]) - string;
2284
2285       match_ok = g_match_info_next (match_info, &tmp_error);
2286     }
2287   g_match_info_free (match_info);
2288   if (tmp_error != NULL)
2289     {
2290       g_propagate_error (error, tmp_error);
2291       g_list_free_full (list, g_free);
2292       match_info->pos = -1;
2293       return NULL;
2294     }
2295
2296   string_list = g_new (gchar *, g_list_length (list) + 1);
2297   i = 0;
2298   for (last = g_list_last (list); last; last = g_list_previous (last))
2299     string_list[i++] = last->data;
2300   string_list[i] = NULL;
2301   g_list_free (list);
2302
2303   return string_list;
2304 }
2305
2306 enum
2307 {
2308   REPL_TYPE_STRING,
2309   REPL_TYPE_CHARACTER,
2310   REPL_TYPE_SYMBOLIC_REFERENCE,
2311   REPL_TYPE_NUMERIC_REFERENCE,
2312   REPL_TYPE_CHANGE_CASE
2313 };
2314
2315 typedef enum
2316 {
2317   CHANGE_CASE_NONE         = 1 << 0,
2318   CHANGE_CASE_UPPER        = 1 << 1,
2319   CHANGE_CASE_LOWER        = 1 << 2,
2320   CHANGE_CASE_UPPER_SINGLE = 1 << 3,
2321   CHANGE_CASE_LOWER_SINGLE = 1 << 4,
2322   CHANGE_CASE_SINGLE_MASK  = CHANGE_CASE_UPPER_SINGLE | CHANGE_CASE_LOWER_SINGLE,
2323   CHANGE_CASE_LOWER_MASK   = CHANGE_CASE_LOWER | CHANGE_CASE_LOWER_SINGLE,
2324   CHANGE_CASE_UPPER_MASK   = CHANGE_CASE_UPPER | CHANGE_CASE_UPPER_SINGLE
2325 } ChangeCase;
2326
2327 struct _InterpolationData
2328 {
2329   gchar     *text;
2330   gint       type;
2331   gint       num;
2332   gchar      c;
2333   ChangeCase change_case;
2334 };
2335
2336 static void
2337 free_interpolation_data (InterpolationData *data)
2338 {
2339   g_free (data->text);
2340   g_free (data);
2341 }
2342
2343 static const gchar *
2344 expand_escape (const gchar        *replacement,
2345                const gchar        *p,
2346                InterpolationData  *data,
2347                GError            **error)
2348 {
2349   const gchar *q, *r;
2350   gint x, d, h, i;
2351   const gchar *error_detail;
2352   gint base = 0;
2353   GError *tmp_error = NULL;
2354
2355   p++;
2356   switch (*p)
2357     {
2358     case 't':
2359       p++;
2360       data->c = '\t';
2361       data->type = REPL_TYPE_CHARACTER;
2362       break;
2363     case 'n':
2364       p++;
2365       data->c = '\n';
2366       data->type = REPL_TYPE_CHARACTER;
2367       break;
2368     case 'v':
2369       p++;
2370       data->c = '\v';
2371       data->type = REPL_TYPE_CHARACTER;
2372       break;
2373     case 'r':
2374       p++;
2375       data->c = '\r';
2376       data->type = REPL_TYPE_CHARACTER;
2377       break;
2378     case 'f':
2379       p++;
2380       data->c = '\f';
2381       data->type = REPL_TYPE_CHARACTER;
2382       break;
2383     case 'a':
2384       p++;
2385       data->c = '\a';
2386       data->type = REPL_TYPE_CHARACTER;
2387       break;
2388     case 'b':
2389       p++;
2390       data->c = '\b';
2391       data->type = REPL_TYPE_CHARACTER;
2392       break;
2393     case '\\':
2394       p++;
2395       data->c = '\\';
2396       data->type = REPL_TYPE_CHARACTER;
2397       break;
2398     case 'x':
2399       p++;
2400       x = 0;
2401       if (*p == '{')
2402         {
2403           p++;
2404           do
2405             {
2406               h = g_ascii_xdigit_value (*p);
2407               if (h < 0)
2408                 {
2409                   error_detail = _("hexadecimal digit or '}' expected");
2410                   goto error;
2411                 }
2412               x = x * 16 + h;
2413               p++;
2414             }
2415           while (*p != '}');
2416           p++;
2417         }
2418       else
2419         {
2420           for (i = 0; i < 2; i++)
2421             {
2422               h = g_ascii_xdigit_value (*p);
2423               if (h < 0)
2424                 {
2425                   error_detail = _("hexadecimal digit expected");
2426                   goto error;
2427                 }
2428               x = x * 16 + h;
2429               p++;
2430             }
2431         }
2432       data->type = REPL_TYPE_STRING;
2433       data->text = g_new0 (gchar, 8);
2434       g_unichar_to_utf8 (x, data->text);
2435       break;
2436     case 'l':
2437       p++;
2438       data->type = REPL_TYPE_CHANGE_CASE;
2439       data->change_case = CHANGE_CASE_LOWER_SINGLE;
2440       break;
2441     case 'u':
2442       p++;
2443       data->type = REPL_TYPE_CHANGE_CASE;
2444       data->change_case = CHANGE_CASE_UPPER_SINGLE;
2445       break;
2446     case 'L':
2447       p++;
2448       data->type = REPL_TYPE_CHANGE_CASE;
2449       data->change_case = CHANGE_CASE_LOWER;
2450       break;
2451     case 'U':
2452       p++;
2453       data->type = REPL_TYPE_CHANGE_CASE;
2454       data->change_case = CHANGE_CASE_UPPER;
2455       break;
2456     case 'E':
2457       p++;
2458       data->type = REPL_TYPE_CHANGE_CASE;
2459       data->change_case = CHANGE_CASE_NONE;
2460       break;
2461     case 'g':
2462       p++;
2463       if (*p != '<')
2464         {
2465           error_detail = _("missing '<' in symbolic reference");
2466           goto error;
2467         }
2468       q = p + 1;
2469       do
2470         {
2471           p++;
2472           if (!*p)
2473             {
2474               error_detail = _("unfinished symbolic reference");
2475               goto error;
2476             }
2477         }
2478       while (*p != '>');
2479       if (p - q == 0)
2480         {
2481           error_detail = _("zero-length symbolic reference");
2482           goto error;
2483         }
2484       if (g_ascii_isdigit (*q))
2485         {
2486           x = 0;
2487           do
2488             {
2489               h = g_ascii_digit_value (*q);
2490               if (h < 0)
2491                 {
2492                   error_detail = _("digit expected");
2493                   p = q;
2494                   goto error;
2495                 }
2496               x = x * 10 + h;
2497               q++;
2498             }
2499           while (q != p);
2500           data->num = x;
2501           data->type = REPL_TYPE_NUMERIC_REFERENCE;
2502         }
2503       else
2504         {
2505           r = q;
2506           do
2507             {
2508               if (!g_ascii_isalnum (*r))
2509                 {
2510                   error_detail = _("illegal symbolic reference");
2511                   p = r;
2512                   goto error;
2513                 }
2514               r++;
2515             }
2516           while (r != p);
2517           data->text = g_strndup (q, p - q);
2518           data->type = REPL_TYPE_SYMBOLIC_REFERENCE;
2519         }
2520       p++;
2521       break;
2522     case '0':
2523       /* if \0 is followed by a number is an octal number representing a
2524        * character, else it is a numeric reference. */
2525       if (g_ascii_digit_value (*g_utf8_next_char (p)) >= 0)
2526         {
2527           base = 8;
2528           p = g_utf8_next_char (p);
2529         }
2530     case '1':
2531     case '2':
2532     case '3':
2533     case '4':
2534     case '5':
2535     case '6':
2536     case '7':
2537     case '8':
2538     case '9':
2539       x = 0;
2540       d = 0;
2541       for (i = 0; i < 3; i++)
2542         {
2543           h = g_ascii_digit_value (*p);
2544           if (h < 0)
2545             break;
2546           if (h > 7)
2547             {
2548               if (base == 8)
2549                 break;
2550               else
2551                 base = 10;
2552             }
2553           if (i == 2 && base == 10)
2554             break;
2555           x = x * 8 + h;
2556           d = d * 10 + h;
2557           p++;
2558         }
2559       if (base == 8 || i == 3)
2560         {
2561           data->type = REPL_TYPE_STRING;
2562           data->text = g_new0 (gchar, 8);
2563           g_unichar_to_utf8 (x, data->text);
2564         }
2565       else
2566         {
2567           data->type = REPL_TYPE_NUMERIC_REFERENCE;
2568           data->num = d;
2569         }
2570       break;
2571     case 0:
2572       error_detail = _("stray final '\\'");
2573       goto error;
2574       break;
2575     default:
2576       error_detail = _("unknown escape sequence");
2577       goto error;
2578     }
2579
2580   return p;
2581
2582  error:
2583   /* G_GSSIZE_FORMAT doesn't work with gettext, so we use %lu */
2584   tmp_error = g_error_new (G_REGEX_ERROR,
2585                            G_REGEX_ERROR_REPLACE,
2586                            _("Error while parsing replacement "
2587                              "text \"%s\" at char %lu: %s"),
2588                            replacement,
2589                            (gulong)(p - replacement),
2590                            error_detail);
2591   g_propagate_error (error, tmp_error);
2592
2593   return NULL;
2594 }
2595
2596 static GList *
2597 split_replacement (const gchar  *replacement,
2598                    GError      **error)
2599 {
2600   GList *list = NULL;
2601   InterpolationData *data;
2602   const gchar *p, *start;
2603
2604   start = p = replacement;
2605   while (*p)
2606     {
2607       if (*p == '\\')
2608         {
2609           data = g_new0 (InterpolationData, 1);
2610           start = p = expand_escape (replacement, p, data, error);
2611           if (p == NULL)
2612             {
2613               g_list_free_full (list, (GDestroyNotify) free_interpolation_data);
2614               free_interpolation_data (data);
2615
2616               return NULL;
2617             }
2618           list = g_list_prepend (list, data);
2619         }
2620       else
2621         {
2622           p++;
2623           if (*p == '\\' || *p == '\0')
2624             {
2625               if (p - start > 0)
2626                 {
2627                   data = g_new0 (InterpolationData, 1);
2628                   data->text = g_strndup (start, p - start);
2629                   data->type = REPL_TYPE_STRING;
2630                   list = g_list_prepend (list, data);
2631                 }
2632             }
2633         }
2634     }
2635
2636   return g_list_reverse (list);
2637 }
2638
2639 /* Change the case of c based on change_case. */
2640 #define CHANGE_CASE(c, change_case) \
2641         (((change_case) & CHANGE_CASE_LOWER_MASK) ? \
2642                 g_unichar_tolower (c) : \
2643                 g_unichar_toupper (c))
2644
2645 static void
2646 string_append (GString     *string,
2647                const gchar *text,
2648                ChangeCase  *change_case)
2649 {
2650   gunichar c;
2651
2652   if (text[0] == '\0')
2653     return;
2654
2655   if (*change_case == CHANGE_CASE_NONE)
2656     {
2657       g_string_append (string, text);
2658     }
2659   else if (*change_case & CHANGE_CASE_SINGLE_MASK)
2660     {
2661       c = g_utf8_get_char (text);
2662       g_string_append_unichar (string, CHANGE_CASE (c, *change_case));
2663       g_string_append (string, g_utf8_next_char (text));
2664       *change_case = CHANGE_CASE_NONE;
2665     }
2666   else
2667     {
2668       while (*text != '\0')
2669         {
2670           c = g_utf8_get_char (text);
2671           g_string_append_unichar (string, CHANGE_CASE (c, *change_case));
2672           text = g_utf8_next_char (text);
2673         }
2674     }
2675 }
2676
2677 static gboolean
2678 interpolate_replacement (const GMatchInfo *match_info,
2679                          GString          *result,
2680                          gpointer          data)
2681 {
2682   GList *list;
2683   InterpolationData *idata;
2684   gchar *match;
2685   ChangeCase change_case = CHANGE_CASE_NONE;
2686
2687   for (list = data; list; list = list->next)
2688     {
2689       idata = list->data;
2690       switch (idata->type)
2691         {
2692         case REPL_TYPE_STRING:
2693           string_append (result, idata->text, &change_case);
2694           break;
2695         case REPL_TYPE_CHARACTER:
2696           g_string_append_c (result, CHANGE_CASE (idata->c, change_case));
2697           if (change_case & CHANGE_CASE_SINGLE_MASK)
2698             change_case = CHANGE_CASE_NONE;
2699           break;
2700         case REPL_TYPE_NUMERIC_REFERENCE:
2701           match = g_match_info_fetch (match_info, idata->num);
2702           if (match)
2703             {
2704               string_append (result, match, &change_case);
2705               g_free (match);
2706             }
2707           break;
2708         case REPL_TYPE_SYMBOLIC_REFERENCE:
2709           match = g_match_info_fetch_named (match_info, idata->text);
2710           if (match)
2711             {
2712               string_append (result, match, &change_case);
2713               g_free (match);
2714             }
2715           break;
2716         case REPL_TYPE_CHANGE_CASE:
2717           change_case = idata->change_case;
2718           break;
2719         }
2720     }
2721
2722   return FALSE;
2723 }
2724
2725 /* whether actual match_info is needed for replacement, i.e.
2726  * whether there are references
2727  */
2728 static gboolean
2729 interpolation_list_needs_match (GList *list)
2730 {
2731   while (list != NULL)
2732     {
2733       InterpolationData *data = list->data;
2734
2735       if (data->type == REPL_TYPE_SYMBOLIC_REFERENCE ||
2736           data->type == REPL_TYPE_NUMERIC_REFERENCE)
2737         {
2738           return TRUE;
2739         }
2740
2741       list = list->next;
2742     }
2743
2744   return FALSE;
2745 }
2746
2747 /**
2748  * g_regex_replace:
2749  * @regex: a #GRegex structure
2750  * @string: (array length=string_len): the string to perform matches against
2751  * @string_len: the length of @string, or -1 if @string is nul-terminated
2752  * @start_position: starting index of the string to match, in bytes
2753  * @replacement: text to replace each match with
2754  * @match_options: options for the match
2755  * @error: location to store the error occurring, or %NULL to ignore errors
2756  *
2757  * Replaces all occurrences of the pattern in @regex with the
2758  * replacement text. Backreferences of the form '\number' or
2759  * '\g<number>' in the replacement text are interpolated by the
2760  * number-th captured subexpression of the match, '\g<name>' refers
2761  * to the captured subexpression with the given name. '\0' refers
2762  * to the complete match, but '\0' followed by a number is the octal
2763  * representation of a character. To include a literal '\' in the
2764  * replacement, write '\\'.
2765  *
2766  * There are also escapes that changes the case of the following text:
2767  *
2768  * - \l: Convert to lower case the next character
2769  * - \u: Convert to upper case the next character
2770  * - \L: Convert to lower case till \E
2771  * - \U: Convert to upper case till \E
2772  * - \E: End case modification
2773  *
2774  * If you do not need to use backreferences use g_regex_replace_literal().
2775  *
2776  * The @replacement string must be UTF-8 encoded even if #G_REGEX_RAW was
2777  * passed to g_regex_new(). If you want to use not UTF-8 encoded stings
2778  * you can use g_regex_replace_literal().
2779  *
2780  * Setting @start_position differs from just passing over a shortened
2781  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern that
2782  * begins with any kind of lookbehind assertion, such as "\b".
2783  *
2784  * Returns: a newly allocated string containing the replacements
2785  *
2786  * Since: 2.14
2787  */
2788 gchar *
2789 g_regex_replace (const GRegex      *regex,
2790                  const gchar       *string,
2791                  gssize             string_len,
2792                  gint               start_position,
2793                  const gchar       *replacement,
2794                  GRegexMatchFlags   match_options,
2795                  GError           **error)
2796 {
2797   gchar *result;
2798   GList *list;
2799   GError *tmp_error = NULL;
2800
2801   g_return_val_if_fail (regex != NULL, NULL);
2802   g_return_val_if_fail (string != NULL, NULL);
2803   g_return_val_if_fail (start_position >= 0, NULL);
2804   g_return_val_if_fail (replacement != NULL, NULL);
2805   g_return_val_if_fail (error == NULL || *error == NULL, NULL);
2806   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
2807
2808   list = split_replacement (replacement, &tmp_error);
2809   if (tmp_error != NULL)
2810     {
2811       g_propagate_error (error, tmp_error);
2812       return NULL;
2813     }
2814
2815   result = g_regex_replace_eval (regex,
2816                                  string, string_len, start_position,
2817                                  match_options,
2818                                  interpolate_replacement,
2819                                  (gpointer)list,
2820                                  &tmp_error);
2821   if (tmp_error != NULL)
2822     g_propagate_error (error, tmp_error);
2823
2824   g_list_free_full (list, (GDestroyNotify) free_interpolation_data);
2825
2826   return result;
2827 }
2828
2829 static gboolean
2830 literal_replacement (const GMatchInfo *match_info,
2831                      GString          *result,
2832                      gpointer          data)
2833 {
2834   g_string_append (result, data);
2835   return FALSE;
2836 }
2837
2838 /**
2839  * g_regex_replace_literal:
2840  * @regex: a #GRegex structure
2841  * @string: (array length=string_len): the string to perform matches against
2842  * @string_len: the length of @string, or -1 if @string is nul-terminated
2843  * @start_position: starting index of the string to match, in bytes
2844  * @replacement: text to replace each match with
2845  * @match_options: options for the match
2846  * @error: location to store the error occurring, or %NULL to ignore errors
2847  *
2848  * Replaces all occurrences of the pattern in @regex with the
2849  * replacement text. @replacement is replaced literally, to
2850  * include backreferences use g_regex_replace().
2851  *
2852  * Setting @start_position differs from just passing over a
2853  * shortened string and setting #G_REGEX_MATCH_NOTBOL in the
2854  * case of a pattern that begins with any kind of lookbehind
2855  * assertion, such as "\b".
2856  *
2857  * Returns: a newly allocated string containing the replacements
2858  *
2859  * Since: 2.14
2860  */
2861 gchar *
2862 g_regex_replace_literal (const GRegex      *regex,
2863                          const gchar       *string,
2864                          gssize             string_len,
2865                          gint               start_position,
2866                          const gchar       *replacement,
2867                          GRegexMatchFlags   match_options,
2868                          GError           **error)
2869 {
2870   g_return_val_if_fail (replacement != NULL, NULL);
2871   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
2872
2873   return g_regex_replace_eval (regex,
2874                                string, string_len, start_position,
2875                                match_options,
2876                                literal_replacement,
2877                                (gpointer)replacement,
2878                                error);
2879 }
2880
2881 /**
2882  * g_regex_replace_eval:
2883  * @regex: a #GRegex structure from g_regex_new()
2884  * @string: (array length=string_len): string to perform matches against
2885  * @string_len: the length of @string, or -1 if @string is nul-terminated
2886  * @start_position: starting index of the string to match, in bytes
2887  * @match_options: options for the match
2888  * @eval: a function to call for each match
2889  * @user_data: user data to pass to the function
2890  * @error: location to store the error occurring, or %NULL to ignore errors
2891  *
2892  * Replaces occurrences of the pattern in regex with the output of
2893  * @eval for that occurrence.
2894  *
2895  * Setting @start_position differs from just passing over a shortened
2896  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern
2897  * that begins with any kind of lookbehind assertion, such as "\b".
2898  *
2899  * The following example uses g_regex_replace_eval() to replace multiple
2900  * strings at once:
2901  * |[<!-- language="C" -->
2902  * static gboolean
2903  * eval_cb (const GMatchInfo *info,
2904  *          GString          *res,
2905  *          gpointer          data)
2906  * {
2907  *   gchar *match;
2908  *   gchar *r;
2909  *
2910  *    match = g_match_info_fetch (info, 0);
2911  *    r = g_hash_table_lookup ((GHashTable *)data, match);
2912  *    g_string_append (res, r);
2913  *    g_free (match);
2914  *
2915  *    return FALSE;
2916  * }
2917  *
2918  * ...
2919  *
2920  * GRegex *reg;
2921  * GHashTable *h;
2922  * gchar *res;
2923  *
2924  * h = g_hash_table_new (g_str_hash, g_str_equal);
2925  *
2926  * g_hash_table_insert (h, "1", "ONE");
2927  * g_hash_table_insert (h, "2", "TWO");
2928  * g_hash_table_insert (h, "3", "THREE");
2929  * g_hash_table_insert (h, "4", "FOUR");
2930  *
2931  * reg = g_regex_new ("1|2|3|4", 0, 0, NULL);
2932  * res = g_regex_replace_eval (reg, text, -1, 0, 0, eval_cb, h, NULL);
2933  * g_hash_table_destroy (h);
2934  *
2935  * ...
2936  * ]|
2937  *
2938  * Returns: a newly allocated string containing the replacements
2939  *
2940  * Since: 2.14
2941  */
2942 gchar *
2943 g_regex_replace_eval (const GRegex        *regex,
2944                       const gchar         *string,
2945                       gssize               string_len,
2946                       gint                 start_position,
2947                       GRegexMatchFlags     match_options,
2948                       GRegexEvalCallback   eval,
2949                       gpointer             user_data,
2950                       GError             **error)
2951 {
2952   GMatchInfo *match_info;
2953   GString *result;
2954   gint str_pos = 0;
2955   gboolean done = FALSE;
2956   GError *tmp_error = NULL;
2957
2958   g_return_val_if_fail (regex != NULL, NULL);
2959   g_return_val_if_fail (string != NULL, NULL);
2960   g_return_val_if_fail (start_position >= 0, NULL);
2961   g_return_val_if_fail (eval != NULL, NULL);
2962   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
2963
2964   if (string_len < 0)
2965     string_len = strlen (string);
2966
2967   result = g_string_sized_new (string_len);
2968
2969   /* run down the string making matches. */
2970   g_regex_match_full (regex, string, string_len, start_position,
2971                       match_options, &match_info, &tmp_error);
2972   while (!done && g_match_info_matches (match_info))
2973     {
2974       g_string_append_len (result,
2975                            string + str_pos,
2976                            match_info->offsets[0] - str_pos);
2977       done = (*eval) (match_info, result, user_data);
2978       str_pos = match_info->offsets[1];
2979       g_match_info_next (match_info, &tmp_error);
2980     }
2981   g_match_info_free (match_info);
2982   if (tmp_error != NULL)
2983     {
2984       g_propagate_error (error, tmp_error);
2985       g_string_free (result, TRUE);
2986       return NULL;
2987     }
2988
2989   g_string_append_len (result, string + str_pos, string_len - str_pos);
2990   return g_string_free (result, FALSE);
2991 }
2992
2993 /**
2994  * g_regex_check_replacement:
2995  * @replacement: the replacement string
2996  * @has_references: (out) (allow-none): location to store information about
2997  *   references in @replacement or %NULL
2998  * @error: location to store error
2999  *
3000  * Checks whether @replacement is a valid replacement string
3001  * (see g_regex_replace()), i.e. that all escape sequences in
3002  * it are valid.
3003  *
3004  * If @has_references is not %NULL then @replacement is checked
3005  * for pattern references. For instance, replacement text 'foo\n'
3006  * does not contain references and may be evaluated without information
3007  * about actual match, but '\0\1' (whole match followed by first
3008  * subpattern) requires valid #GMatchInfo object.
3009  *
3010  * Returns: whether @replacement is a valid replacement string
3011  *
3012  * Since: 2.14
3013  */
3014 gboolean
3015 g_regex_check_replacement (const gchar  *replacement,
3016                            gboolean     *has_references,
3017                            GError      **error)
3018 {
3019   GList *list;
3020   GError *tmp = NULL;
3021
3022   list = split_replacement (replacement, &tmp);
3023
3024   if (tmp)
3025   {
3026     g_propagate_error (error, tmp);
3027     return FALSE;
3028   }
3029
3030   if (has_references)
3031     *has_references = interpolation_list_needs_match (list);
3032
3033   g_list_free_full (list, (GDestroyNotify) free_interpolation_data);
3034
3035   return TRUE;
3036 }
3037
3038 /**
3039  * g_regex_escape_nul:
3040  * @string: the string to escape
3041  * @length: the length of @string
3042  *
3043  * Escapes the nul characters in @string to "\x00".  It can be used
3044  * to compile a regex with embedded nul characters.
3045  *
3046  * For completeness, @length can be -1 for a nul-terminated string.
3047  * In this case the output string will be of course equal to @string.
3048  *
3049  * Returns: a newly-allocated escaped string
3050  *
3051  * Since: 2.30
3052  */
3053 gchar *
3054 g_regex_escape_nul (const gchar *string,
3055                     gint         length)
3056 {
3057   GString *escaped;
3058   const gchar *p, *piece_start, *end;
3059   gint backslashes;
3060
3061   g_return_val_if_fail (string != NULL, NULL);
3062
3063   if (length < 0)
3064     return g_strdup (string);
3065
3066   end = string + length;
3067   p = piece_start = string;
3068   escaped = g_string_sized_new (length + 1);
3069
3070   backslashes = 0;
3071   while (p < end)
3072     {
3073       switch (*p)
3074         {
3075         case '\0':
3076           if (p != piece_start)
3077             {
3078               /* copy the previous piece. */
3079               g_string_append_len (escaped, piece_start, p - piece_start);
3080             }
3081           if ((backslashes & 1) == 0)
3082             g_string_append_c (escaped, '\\');
3083           g_string_append_c (escaped, 'x');
3084           g_string_append_c (escaped, '0');
3085           g_string_append_c (escaped, '0');
3086           piece_start = ++p;
3087           backslashes = 0;
3088           break;
3089         case '\\':
3090           backslashes++;
3091           ++p;
3092           break;
3093         default:
3094           backslashes = 0;
3095           p = g_utf8_next_char (p);
3096           break;
3097         }
3098     }
3099
3100   if (piece_start < end)
3101     g_string_append_len (escaped, piece_start, end - piece_start);
3102
3103   return g_string_free (escaped, FALSE);
3104 }
3105
3106 /**
3107  * g_regex_escape_string:
3108  * @string: (array length=length): the string to escape
3109  * @length: the length of @string, or -1 if @string is nul-terminated
3110  *
3111  * Escapes the special characters used for regular expressions
3112  * in @string, for instance "a.b*c" becomes "a\.b\*c". This
3113  * function is useful to dynamically generate regular expressions.
3114  *
3115  * @string can contain nul characters that are replaced with "\0",
3116  * in this case remember to specify the correct length of @string
3117  * in @length.
3118  *
3119  * Returns: a newly-allocated escaped string
3120  *
3121  * Since: 2.14
3122  */
3123 gchar *
3124 g_regex_escape_string (const gchar *string,
3125                        gint         length)
3126 {
3127   GString *escaped;
3128   const char *p, *piece_start, *end;
3129
3130   g_return_val_if_fail (string != NULL, NULL);
3131
3132   if (length < 0)
3133     length = strlen (string);
3134
3135   end = string + length;
3136   p = piece_start = string;
3137   escaped = g_string_sized_new (length + 1);
3138
3139   while (p < end)
3140     {
3141       switch (*p)
3142         {
3143         case '\0':
3144         case '\\':
3145         case '|':
3146         case '(':
3147         case ')':
3148         case '[':
3149         case ']':
3150         case '{':
3151         case '}':
3152         case '^':
3153         case '$':
3154         case '*':
3155         case '+':
3156         case '?':
3157         case '.':
3158           if (p != piece_start)
3159             /* copy the previous piece. */
3160             g_string_append_len (escaped, piece_start, p - piece_start);
3161           g_string_append_c (escaped, '\\');
3162           if (*p == '\0')
3163             g_string_append_c (escaped, '0');
3164           else
3165             g_string_append_c (escaped, *p);
3166           piece_start = ++p;
3167           break;
3168         default:
3169           p = g_utf8_next_char (p);
3170           break;
3171         }
3172   }
3173
3174   if (piece_start < end)
3175     g_string_append_len (escaped, piece_start, end - piece_start);
3176
3177   return g_string_free (escaped, FALSE);
3178 }