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 License
  18  * along with this library; if not, see <http://www.gnu.org/licenses/>.
  19  */
  20
  21 #include "config.h"
  22
  23 #include <string.h>
  24
  25 #ifdef USE_SYSTEM_PCRE
  26 #include <pcre.h>
  27 #else
  28 #include "pcre/pcre.h"
  29 #endif
  30
  31 #include "gtypes.h"
  32 #include "gregex.h"
  33 #include "glibintl.h"
  34 #include "glist.h"
  35 #include "gmessages.h"
  36 #include "gstrfuncs.h"
  37 #include "gatomic.h"
  38 #include "gthread.h"
  39
  40 /**
  41  * SECTION:gregex
  42  * @title: Perl-compatible regular expressions
  43  * @short_description: matches strings against regular expressions
  44  * @see_also: [Regular expression syntax][glib-regex-syntax]
  45  *
  46  * The g_regex_*() functions implement regular
  47  * expression pattern matching using syntax and semantics similar to
  48  * Perl regular expression.
  49  *
  50  * Some functions accept a @start_position argument, setting it differs
  51  * from just passing over a shortened string and setting #G_REGEX_MATCH_NOTBOL
  52  * in the case of a pattern that begins with any kind of lookbehind assertion.
  53  * For example, consider the pattern "\Biss\B" which finds occurrences of "iss"
  54  * in the middle of words. ("\B" matches only if the current position in the
  55  * subject is not a word boundary.) When applied to the string "Mississipi"
  56  * from the fourth byte, namely "issipi", it does not match, because "\B" is
  57  * always false at the start of the subject, which is deemed to be a word
  58  * boundary. However, if the entire string is passed , but with
  59  * @start_position set to 4, it finds the second occurrence of "iss" because
  60  * it is able to look behind the starting point to discover that it is
  61  * preceded by a letter.
  62  *
  63  * Note that, unless you set the #G_REGEX_RAW flag, all the strings passed
  64  * to these functions must be encoded in UTF-8. The lengths and the positions
  65  * inside the strings are in bytes and not in characters, so, for instance,
  66  * "\xc3\xa0" (i.e. "à") is two bytes long but it is treated as a
  67  * single character. If you set #G_REGEX_RAW the strings can be non-valid
  68  * UTF-8 strings and a byte is treated as a character, so "\xc3\xa0" is two
  69  * bytes and two characters long.
  70  *
  71  * When matching a pattern, "\n" matches only against a "\n" character in
  72  * the string, and "\r" matches only a "\r" character. To match any newline
  73  * sequence use "\R". This particular group matches either the two-character
  74  * sequence CR + LF ("\r\n"), or one of the single characters LF (linefeed,
  75  * U+000A, "\n"), VT vertical tab, U+000B, "\v"), FF (formfeed, U+000C, "\f"),
  76  * CR (carriage return, U+000D, "\r"), NEL (next line, U+0085), LS (line
  77  * separator, U+2028), or PS (paragraph separator, U+2029).
  78  *
  79  * The behaviour of the dot, circumflex, and dollar metacharacters are
  80  * affected by newline characters, the default is to recognize any newline
  81  * character (the same characters recognized by "\R"). This can be changed
  82  * with #G_REGEX_NEWLINE_CR, #G_REGEX_NEWLINE_LF and #G_REGEX_NEWLINE_CRLF
  83  * compile options, and with #G_REGEX_MATCH_NEWLINE_ANY,
  84  * #G_REGEX_MATCH_NEWLINE_CR, #G_REGEX_MATCH_NEWLINE_LF and
  85  * #G_REGEX_MATCH_NEWLINE_CRLF match options. These settings are also
  86  * relevant when compiling a pattern if #G_REGEX_EXTENDED is set, and an
  87  * unescaped "#" outside a character class is encountered. This indicates
  88  * a comment that lasts until after the next newline.
  89  *
  90  * When setting the %G_REGEX_JAVASCRIPT_COMPAT flag, pattern syntax and pattern
  91  * matching is changed to be compatible with the way that regular expressions
  92  * work in JavaScript. More precisely, a lonely ']' character in the pattern
  93  * is a syntax error; the '\x' escape only allows 0 to 2 hexadecimal digits, and
  94  * you must use the '\u' escape sequence with 4 hex digits to specify a unicode
  95  * codepoint instead of '\x' or 'x{....}'. If '\x' or '\u' are not followed by
  96  * the specified number of hex digits, they match 'x' and 'u' literally; also
  97  * '\U' always matches 'U' instead of being an error in the pattern. Finally,
  98  * pattern matching is modified so that back references to an unset subpattern
  99  * group produces a match with the empty string instead of an error. See
 100  * pcreapi(3) for more information.
 101  *
 102  * Creating and manipulating the same #GRegex structure from different
 103  * threads is not a problem as #GRegex does not modify its internal
 104  * state between creation and destruction, on the other hand #GMatchInfo
 105  * is not threadsafe.
 106  *
 107  * The regular expressions low-level functionalities are obtained through
 108  * the excellent
 109  * [PCRE](http://www.pcre.org/)
 110  * library written by Philip Hazel.
 111  */
 112
 113 /* Mask of all the possible values for GRegexCompileFlags. */
 114 #define G_REGEX_COMPILE_MASK (G_REGEX_CASELESS          | \
 115                               G_REGEX_MULTILINE         | \
 116                               G_REGEX_DOTALL            | \
 117                               G_REGEX_EXTENDED          | \
 118                               G_REGEX_ANCHORED          | \
 119                               G_REGEX_DOLLAR_ENDONLY    | \
 120                               G_REGEX_UNGREEDY          | \
 121                               G_REGEX_RAW               | \
 122                               G_REGEX_NO_AUTO_CAPTURE   | \
 123                               G_REGEX_OPTIMIZE          | \
 124                               G_REGEX_FIRSTLINE         | \
 125                               G_REGEX_DUPNAMES          | \
 126                               G_REGEX_NEWLINE_CR        | \
 127                               G_REGEX_NEWLINE_LF        | \
 128                               G_REGEX_NEWLINE_CRLF      | \
 129                               G_REGEX_NEWLINE_ANYCRLF   | \
 130                               G_REGEX_BSR_ANYCRLF       | \
 131                               G_REGEX_JAVASCRIPT_COMPAT)
 132
 133 /* Mask of all GRegexCompileFlags values that are (not) passed trough to PCRE */
 134 #define G_REGEX_COMPILE_PCRE_MASK (G_REGEX_COMPILE_MASK & ~G_REGEX_COMPILE_NONPCRE_MASK)
 135 #define G_REGEX_COMPILE_NONPCRE_MASK (G_REGEX_RAW              | \
 136                                       G_REGEX_OPTIMIZE)
 137
 138 /* Mask of all the possible values for GRegexMatchFlags. */
 139 #define G_REGEX_MATCH_MASK (G_REGEX_MATCH_ANCHORED         | \
 140                             G_REGEX_MATCH_NOTBOL           | \
 141                             G_REGEX_MATCH_NOTEOL           | \
 142                             G_REGEX_MATCH_NOTEMPTY         | \
 143                             G_REGEX_MATCH_PARTIAL          | \
 144                             G_REGEX_MATCH_NEWLINE_CR       | \
 145                             G_REGEX_MATCH_NEWLINE_LF       | \
 146                             G_REGEX_MATCH_NEWLINE_CRLF     | \
 147                             G_REGEX_MATCH_NEWLINE_ANY      | \
 148                             G_REGEX_MATCH_NEWLINE_ANYCRLF  | \
 149                             G_REGEX_MATCH_BSR_ANYCRLF      | \
 150                             G_REGEX_MATCH_BSR_ANY          | \
 151                             G_REGEX_MATCH_PARTIAL_SOFT     | \
 152                             G_REGEX_MATCH_PARTIAL_HARD     | \
 153                             G_REGEX_MATCH_NOTEMPTY_ATSTART)
 154
 155 /* we rely on these flags having the same values */
 156 G_STATIC_ASSERT (G_REGEX_CASELESS          == PCRE_CASELESS);
 157 G_STATIC_ASSERT (G_REGEX_MULTILINE         == PCRE_MULTILINE);
 158 G_STATIC_ASSERT (G_REGEX_DOTALL            == PCRE_DOTALL);
 159 G_STATIC_ASSERT (G_REGEX_EXTENDED          == PCRE_EXTENDED);
 160 G_STATIC_ASSERT (G_REGEX_ANCHORED          == PCRE_ANCHORED);
 161 G_STATIC_ASSERT (G_REGEX_DOLLAR_ENDONLY    == PCRE_DOLLAR_ENDONLY);
 162 G_STATIC_ASSERT (G_REGEX_UNGREEDY          == PCRE_UNGREEDY);
 163 G_STATIC_ASSERT (G_REGEX_NO_AUTO_CAPTURE   == PCRE_NO_AUTO_CAPTURE);
 164 G_STATIC_ASSERT (G_REGEX_FIRSTLINE         == PCRE_FIRSTLINE);
 165 G_STATIC_ASSERT (G_REGEX_DUPNAMES          == PCRE_DUPNAMES);
 166 G_STATIC_ASSERT (G_REGEX_NEWLINE_CR        == PCRE_NEWLINE_CR);
 167 G_STATIC_ASSERT (G_REGEX_NEWLINE_LF        == PCRE_NEWLINE_LF);
 168 G_STATIC_ASSERT (G_REGEX_NEWLINE_CRLF      == PCRE_NEWLINE_CRLF);
 169 G_STATIC_ASSERT (G_REGEX_NEWLINE_ANYCRLF   == PCRE_NEWLINE_ANYCRLF);
 170 G_STATIC_ASSERT (G_REGEX_BSR_ANYCRLF       == PCRE_BSR_ANYCRLF);
 171 G_STATIC_ASSERT (G_REGEX_JAVASCRIPT_COMPAT == PCRE_JAVASCRIPT_COMPAT);
 172
 173 G_STATIC_ASSERT (G_REGEX_MATCH_ANCHORED         == PCRE_ANCHORED);
 174 G_STATIC_ASSERT (G_REGEX_MATCH_NOTBOL           == PCRE_NOTBOL);
 175 G_STATIC_ASSERT (G_REGEX_MATCH_NOTEOL           == PCRE_NOTEOL);
 176 G_STATIC_ASSERT (G_REGEX_MATCH_NOTEMPTY         == PCRE_NOTEMPTY);
 177 G_STATIC_ASSERT (G_REGEX_MATCH_PARTIAL          == PCRE_PARTIAL);
 178 G_STATIC_ASSERT (G_REGEX_MATCH_NEWLINE_CR       == PCRE_NEWLINE_CR);
 179 G_STATIC_ASSERT (G_REGEX_MATCH_NEWLINE_LF       == PCRE_NEWLINE_LF);
 180 G_STATIC_ASSERT (G_REGEX_MATCH_NEWLINE_CRLF     == PCRE_NEWLINE_CRLF);
 181 G_STATIC_ASSERT (G_REGEX_MATCH_NEWLINE_ANY      == PCRE_NEWLINE_ANY);
 182 G_STATIC_ASSERT (G_REGEX_MATCH_NEWLINE_ANYCRLF  == PCRE_NEWLINE_ANYCRLF);
 183 G_STATIC_ASSERT (G_REGEX_MATCH_BSR_ANYCRLF      == PCRE_BSR_ANYCRLF);
 184 G_STATIC_ASSERT (G_REGEX_MATCH_BSR_ANY          == PCRE_BSR_UNICODE);
 185 G_STATIC_ASSERT (G_REGEX_MATCH_PARTIAL_SOFT     == PCRE_PARTIAL_SOFT);
 186 G_STATIC_ASSERT (G_REGEX_MATCH_PARTIAL_HARD     == PCRE_PARTIAL_HARD);
 187 G_STATIC_ASSERT (G_REGEX_MATCH_NOTEMPTY_ATSTART == PCRE_NOTEMPTY_ATSTART);
 188
 189 /* These PCRE flags are unused or not exposed publically in GRegexFlags, so
 190  * it should be ok to reuse them for different things.
 191  */
 192 G_STATIC_ASSERT (G_REGEX_OPTIMIZE          == PCRE_NO_UTF8_CHECK);
 193 G_STATIC_ASSERT (G_REGEX_RAW               == PCRE_UTF8);
 194
 195 /* if the string is in UTF-8 use g_utf8_ functions, else use
 196  * use just +/- 1. */
 197 #define NEXT_CHAR(re, s) (((re)->compile_opts & G_REGEX_RAW) ? \
 198                                 ((s) + 1) : \
 199                                 g_utf8_next_char (s))
 200 #define PREV_CHAR(re, s) (((re)->compile_opts & G_REGEX_RAW) ? \
 201                                 ((s) - 1) : \
 202                                 g_utf8_prev_char (s))
 203
 204 struct _GMatchInfo
 205 {
 206   volatile gint ref_count;      /* the ref count */
 207   GRegex *regex;                /* the regex */
 208   GRegexMatchFlags match_opts;  /* options used at match time on the regex */
 209   gint matches;                 /* number of matching sub patterns */
 210   gint pos;                     /* position in the string where last match left off */
 211   gint  n_offsets;              /* number of offsets */
 212   gint *offsets;                /* array of offsets paired 0,1 ; 2,3 ; 3,4 etc */
 213   gint *workspace;              /* workspace for pcre_dfa_exec() */
 214   gint n_workspace;             /* number of workspace elements */
 215   const gchar *string;          /* string passed to the match function */
 216   gssize string_len;            /* length of string, in bytes */
 217 };
 218
 219 struct _GRegex
 220 {
 221   volatile gint ref_count;      /* the ref count for the immutable part */
 222   gchar *pattern;               /* the pattern */
 223   pcre *pcre_re;                /* compiled form of the pattern */
 224   GRegexCompileFlags compile_opts;      /* options used at compile time on the pattern */
 225   GRegexMatchFlags match_opts;  /* options used at match time on the regex */
 226   pcre_extra *extra;            /* data stored when G_REGEX_OPTIMIZE is used */
 227 };
 228
 229 /* TRUE if ret is an error code, FALSE otherwise. */
 230 #define IS_PCRE_ERROR(ret) ((ret) < PCRE_ERROR_NOMATCH && (ret) != PCRE_ERROR_PARTIAL)
 231
 232 typedef struct _InterpolationData InterpolationData;
 233 static gboolean  interpolation_list_needs_match (GList *list);
 234 static gboolean  interpolate_replacement        (const GMatchInfo *match_info,
 235                                                  GString *result,
 236                                                  gpointer data);
 237 static GList    *split_replacement              (const gchar *replacement,
 238                                                  GError **error);
 239 static void      free_interpolation_data        (InterpolationData *data);
 240
 241
 242 static const gchar *
 243 match_error (gint errcode)
 244 {
 245   switch (errcode)
 246     {
 247     case PCRE_ERROR_NOMATCH:
 248       /* not an error */
 249       break;
 250     case PCRE_ERROR_NULL:
 251       /* NULL argument, this should not happen in GRegex */
 252       g_warning ("A NULL argument was passed to PCRE");
 253       break;
 254     case PCRE_ERROR_BADOPTION:
 255       return "bad options";
 256     case PCRE_ERROR_BADMAGIC:
 257       return _("corrupted object");
 258     case PCRE_ERROR_UNKNOWN_OPCODE:
 259       return N_("internal error or corrupted object");
 260     case PCRE_ERROR_NOMEMORY:
 261       return _("out of memory");
 262     case PCRE_ERROR_NOSUBSTRING:
 263       /* not used by pcre_exec() */
 264       break;
 265     case PCRE_ERROR_MATCHLIMIT:
 266       return _("backtracking limit reached");
 267     case PCRE_ERROR_CALLOUT:
 268       /* callouts are not implemented */
 269       break;
 270     case PCRE_ERROR_BADUTF8:
 271     case PCRE_ERROR_BADUTF8_OFFSET:
 272       /* we do not check if strings are valid */
 273       break;
 274     case PCRE_ERROR_PARTIAL:
 275       /* not an error */
 276       break;
 277     case PCRE_ERROR_BADPARTIAL:
 278       return _("the pattern contains items not supported for partial matching");
 279     case PCRE_ERROR_INTERNAL:
 280       return _("internal error");
 281     case PCRE_ERROR_BADCOUNT:
 282       /* negative ovecsize, this should not happen in GRegex */
 283       g_warning ("A negative ovecsize was passed to PCRE");
 284       break;
 285     case PCRE_ERROR_DFA_UITEM:
 286       return _("the pattern contains items not supported for partial matching");
 287     case PCRE_ERROR_DFA_UCOND:
 288       return _("back references as conditions are not supported for partial matching");
 289     case PCRE_ERROR_DFA_UMLIMIT:
 290       /* the match_field field is not used in GRegex */
 291       break;
 292     case PCRE_ERROR_DFA_WSSIZE:
 293       /* handled expanding the workspace */
 294       break;
 295     case PCRE_ERROR_DFA_RECURSE:
 296     case PCRE_ERROR_RECURSIONLIMIT:
 297       return _("recursion limit reached");
 298     case PCRE_ERROR_BADNEWLINE:
 299       return _("invalid combination of newline flags");
 300     case PCRE_ERROR_BADOFFSET:
 301       return _("bad offset");
 302     case PCRE_ERROR_SHORTUTF8:
 303       return _("short utf8");
 304     case PCRE_ERROR_RECURSELOOP:
 305       return _("recursion loop");
 306     default:
 307       break;
 308     }
 309   return _("unknown error");
 310 }
 311
 312 static void
 313 translate_compile_error (gint *errcode, const gchar **errmsg)
 314 {
 315   /* Compile errors are created adding 100 to the error code returned
 316    * by PCRE.
 317    * If errcode is known we put the translatable error message in
 318    * erromsg. If errcode is unknown we put the generic
 319    * G_REGEX_ERROR_COMPILE error code in errcode and keep the
 320    * untranslated error message returned by PCRE.
 321    * Note that there can be more PCRE errors with the same GRegexError
 322    * and that some PCRE errors are useless for us.
 323    */
 324   *errcode += 100;
 325
 326   switch (*errcode)
 327     {
 328     case G_REGEX_ERROR_STRAY_BACKSLASH:
 329       *errmsg = _("\\ at end of pattern");
 330       break;
 331     case G_REGEX_ERROR_MISSING_CONTROL_CHAR:
 332       *errmsg = _("\\c at end of pattern");
 333       break;
 334     case G_REGEX_ERROR_UNRECOGNIZED_ESCAPE:
 335       *errmsg = _("unrecognized character following \\");
 336       break;
 337     case G_REGEX_ERROR_QUANTIFIERS_OUT_OF_ORDER:
 338       *errmsg = _("numbers out of order in {} quantifier");
 339       break;
 340     case G_REGEX_ERROR_QUANTIFIER_TOO_BIG:
 341       *errmsg = _("number too big in {} quantifier");
 342       break;
 343     case G_REGEX_ERROR_UNTERMINATED_CHARACTER_CLASS:
 344       *errmsg = _("missing terminating ] for character class");
 345       break;
 346     case G_REGEX_ERROR_INVALID_ESCAPE_IN_CHARACTER_CLASS:
 347       *errmsg = _("invalid escape sequence in character class");
 348       break;
 349     case G_REGEX_ERROR_RANGE_OUT_OF_ORDER:
 350       *errmsg = _("range out of order in character class");
 351       break;
 352     case G_REGEX_ERROR_NOTHING_TO_REPEAT:
 353       *errmsg = _("nothing to repeat");
 354       break;
 355     case 111: /* internal error: unexpected repeat */
 356       *errcode = G_REGEX_ERROR_INTERNAL;
 357       *errmsg = _("unexpected repeat");
 358       break;
 359     case G_REGEX_ERROR_UNRECOGNIZED_CHARACTER:
 360       *errmsg = _("unrecognized character after (? or (?-");
 361       break;
 362     case G_REGEX_ERROR_POSIX_NAMED_CLASS_OUTSIDE_CLASS:
 363       *errmsg = _("POSIX named classes are supported only within a class");
 364       break;
 365     case G_REGEX_ERROR_UNMATCHED_PARENTHESIS:
 366       *errmsg = _("missing terminating )");
 367       break;
 368     case G_REGEX_ERROR_INEXISTENT_SUBPATTERN_REFERENCE:
 369       *errmsg = _("reference to non-existent subpattern");
 370       break;
 371     case G_REGEX_ERROR_UNTERMINATED_COMMENT:
 372       *errmsg = _("missing ) after comment");
 373       break;
 374     case G_REGEX_ERROR_EXPRESSION_TOO_LARGE:
 375       *errmsg = _("regular expression is too large");
 376       break;
 377     case G_REGEX_ERROR_MEMORY_ERROR:
 378       *errmsg = _("failed to get memory");
 379       break;
 380     case 122: /* unmatched parentheses */
 381       *errcode = G_REGEX_ERROR_UNMATCHED_PARENTHESIS;
 382       *errmsg = _(") without opening (");
 383       break;
 384     case 123: /* internal error: code overflow */
 385       *errcode = G_REGEX_ERROR_INTERNAL;
 386       *errmsg = _("code overflow");
 387       break;
 388     case 124: /* "unrecognized character after (?<\0 */
 389       *errcode = G_REGEX_ERROR_UNRECOGNIZED_CHARACTER;
 390       *errmsg = _("unrecognized character after (?<");
 391       break;
 392     case G_REGEX_ERROR_VARIABLE_LENGTH_LOOKBEHIND:
 393       *errmsg = _("lookbehind assertion is not fixed length");
 394       break;
 395     case G_REGEX_ERROR_MALFORMED_CONDITION:
 396       *errmsg = _("malformed number or name after (?(");
 397       break;
 398     case G_REGEX_ERROR_TOO_MANY_CONDITIONAL_BRANCHES:
 399       *errmsg = _("conditional group contains more than two branches");
 400       break;
 401     case G_REGEX_ERROR_ASSERTION_EXPECTED:
 402       *errmsg = _("assertion expected after (?(");
 403       break;
 404     case 129:
 405       *errcode = G_REGEX_ERROR_UNMATCHED_PARENTHESIS;
 406       /* translators: '(?R' and '(?[+-]digits' are both meant as (groups of)
 407        * sequences here, '(?-54' would be an example for the second group.
 408        */
 409       *errmsg = _("(?R or (?[+-]digits must be followed by )");
 410       break;
 411     case G_REGEX_ERROR_UNKNOWN_POSIX_CLASS_NAME:
 412       *errmsg = _("unknown POSIX class name");
 413       break;
 414     case G_REGEX_ERROR_POSIX_COLLATING_ELEMENTS_NOT_SUPPORTED:
 415       *errmsg = _("POSIX collating elements are not supported");
 416       break;
 417     case G_REGEX_ERROR_HEX_CODE_TOO_LARGE:
 418       *errmsg = _("character value in \\x{...} sequence is too large");
 419       break;
 420     case G_REGEX_ERROR_INVALID_CONDITION:
 421       *errmsg = _("invalid condition (?(0)");
 422       break;
 423     case G_REGEX_ERROR_SINGLE_BYTE_MATCH_IN_LOOKBEHIND:
 424       *errmsg = _("\\C not allowed in lookbehind assertion");
 425       break;
 426     case 137: /* PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0 */
 427       /* A number of Perl escapes are not handled by PCRE.
 428        * Therefore it explicitly raises ERR37.
 429        */
 430       *errcode = G_REGEX_ERROR_UNRECOGNIZED_ESCAPE;
 431       *errmsg = _("escapes \\L, \\l, \\N{name}, \\U, and \\u are not supported");
 432       break;
 433     case G_REGEX_ERROR_INFINITE_LOOP:
 434       *errmsg = _("recursive call could loop indefinitely");
 435       break;
 436     case 141: /* unrecognized character after (?P\0 */
 437       *errcode = G_REGEX_ERROR_UNRECOGNIZED_CHARACTER;
 438       *errmsg = _("unrecognized character after (?P");
 439       break;
 440     case G_REGEX_ERROR_MISSING_SUBPATTERN_NAME_TERMINATOR:
 441       *errmsg = _("missing terminator in subpattern name");
 442       break;
 443     case G_REGEX_ERROR_DUPLICATE_SUBPATTERN_NAME:
 444       *errmsg = _("two named subpatterns have the same name");
 445       break;
 446     case G_REGEX_ERROR_MALFORMED_PROPERTY:
 447       *errmsg = _("malformed \\P or \\p sequence");
 448       break;
 449     case G_REGEX_ERROR_UNKNOWN_PROPERTY:
 450       *errmsg = _("unknown property name after \\P or \\p");
 451       break;
 452     case G_REGEX_ERROR_SUBPATTERN_NAME_TOO_LONG:
 453       *errmsg = _("subpattern name is too long (maximum 32 characters)");
 454       break;
 455     case G_REGEX_ERROR_TOO_MANY_SUBPATTERNS:
 456       *errmsg = _("too many named subpatterns (maximum 10,000)");
 457       break;
 458     case G_REGEX_ERROR_INVALID_OCTAL_VALUE:
 459       *errmsg = _("octal value is greater than \\377");
 460       break;
 461     case 152: /* internal error: overran compiling workspace */
 462       *errcode = G_REGEX_ERROR_INTERNAL;
 463       *errmsg = _("overran compiling workspace");
 464       break;
 465     case 153: /* internal error: previously-checked referenced subpattern not found */
 466       *errcode = G_REGEX_ERROR_INTERNAL;
 467       *errmsg = _("previously-checked referenced subpattern not found");
 468       break;
 469     case G_REGEX_ERROR_TOO_MANY_BRANCHES_IN_DEFINE:
 470       *errmsg = _("DEFINE group contains more than one branch");
 471       break;
 472     case G_REGEX_ERROR_INCONSISTENT_NEWLINE_OPTIONS:
 473       *errmsg = _("inconsistent NEWLINE options");
 474       break;
 475     case G_REGEX_ERROR_MISSING_BACK_REFERENCE:
 476       *errmsg = _("\\g is not followed by a braced, angle-bracketed, or quoted name or "
 477                   "number, or by a plain number");
 478       break;
 479     case G_REGEX_ERROR_INVALID_RELATIVE_REFERENCE:
 480       *errmsg = _("a numbered reference must not be zero");
 481       break;
 482     case G_REGEX_ERROR_BACKTRACKING_CONTROL_VERB_ARGUMENT_FORBIDDEN:
 483       *errmsg = _("an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)");
 484       break;
 485     case G_REGEX_ERROR_UNKNOWN_BACKTRACKING_CONTROL_VERB:
 486       *errmsg = _("(*VERB) not recognized");
 487       break;
 488     case G_REGEX_ERROR_NUMBER_TOO_BIG:
 489       *errmsg = _("number is too big");
 490       break;
 491     case G_REGEX_ERROR_MISSING_SUBPATTERN_NAME:
 492       *errmsg = _("missing subpattern name after (?&");
 493       break;
 494     case G_REGEX_ERROR_MISSING_DIGIT:
 495       *errmsg = _("digit expected after (?+");
 496       break;
 497     case G_REGEX_ERROR_INVALID_DATA_CHARACTER:
 498       *errmsg = _("] is an invalid data character in JavaScript compatibility mode");
 499       break;
 500     case G_REGEX_ERROR_EXTRA_SUBPATTERN_NAME:
 501       *errmsg = _("different names for subpatterns of the same number are not allowed");
 502       break;
 503     case G_REGEX_ERROR_BACKTRACKING_CONTROL_VERB_ARGUMENT_REQUIRED:
 504       *errmsg = _("(*MARK) must have an argument");
 505       break;
 506     case G_REGEX_ERROR_INVALID_CONTROL_CHAR:
 507       *errmsg = _( "\\c must be followed by an ASCII character");
 508       break;
 509     case G_REGEX_ERROR_MISSING_NAME:
 510       *errmsg = _("\\k is not followed by a braced, angle-bracketed, or quoted name");
 511       break;
 512     case G_REGEX_ERROR_NOT_SUPPORTED_IN_CLASS:
 513       *errmsg = _("\\N is not supported in a class");
 514       break;
 515     case G_REGEX_ERROR_TOO_MANY_FORWARD_REFERENCES:
 516       *errmsg = _("too many forward references");
 517       break;
 518     case G_REGEX_ERROR_NAME_TOO_LONG:
 519       *errmsg = _("name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)");
 520       break;
 521     case G_REGEX_ERROR_CHARACTER_VALUE_TOO_LARGE:
 522       *errmsg = _("character value in \\u.... sequence is too large");
 523       break;
 524
 525     case 116: /* erroffset passed as NULL */
 526       /* This should not happen as we never pass a NULL erroffset */
 527       g_warning ("erroffset passed as NULL");
 528       *errcode = G_REGEX_ERROR_COMPILE;
 529       break;
 530     case 117: /* unknown option bit(s) set */
 531       /* This should not happen as we check options before passing them
 532        * to pcre_compile2() */
 533       g_warning ("unknown option bit(s) set");
 534       *errcode = G_REGEX_ERROR_COMPILE;
 535       break;
 536     case 132: /* this version of PCRE is compiled without UTF support */
 537     case 144: /* invalid UTF-8 string */
 538     case 145: /* support for \\P, \\p, and \\X has not been compiled */
 539     case 167: /* this version of PCRE is not compiled with Unicode property support */
 540     case 173: /* disallowed Unicode code point (>= 0xd800 && <= 0xdfff) */
 541     case 174: /* invalid UTF-16 string */
 542       /* These errors should not happen as we are using an UTF-8 and UCP-enabled PCRE
 543        * and we do not check if strings are valid */
 544     case 170: /* internal error: unknown opcode in find_fixedlength() */
 545       *errcode = G_REGEX_ERROR_INTERNAL;
 546       break;
 547
 548     default:
 549       *errcode = G_REGEX_ERROR_COMPILE;
 550     }
 551 }
 552
 553 /* GMatchInfo */
 554
 555 static GMatchInfo *
 556 match_info_new (const GRegex *regex,
 557                 const gchar  *string,
 558                 gint          string_len,
 559                 gint          start_position,
 560                 gint          match_options,
 561                 gboolean      is_dfa)
 562 {
 563   GMatchInfo *match_info;
 564
 565   if (string_len < 0)
 566     string_len = strlen (string);
 567
 568   match_info = g_new0 (GMatchInfo, 1);
 569   match_info->ref_count = 1;
 570   match_info->regex = g_regex_ref ((GRegex *)regex);
 571   match_info->string = string;
 572   match_info->string_len = string_len;
 573   match_info->matches = PCRE_ERROR_NOMATCH;
 574   match_info->pos = start_position;
 575   match_info->match_opts = match_options;
 576
 577   if (is_dfa)
 578     {
 579       /* These values should be enough for most cases, if they are not
 580        * enough g_regex_match_all_full() will expand them. */
 581       match_info->n_offsets = 24;
 582       match_info->n_workspace = 100;
 583       match_info->workspace = g_new (gint, match_info->n_workspace);
 584     }
 585   else
 586     {
 587       gint capture_count;
 588       pcre_fullinfo (regex->pcre_re, regex->extra,
 589                      PCRE_INFO_CAPTURECOUNT, &capture_count);
 590       match_info->n_offsets = (capture_count + 1) * 3;
 591     }
 592
 593   match_info->offsets = g_new0 (gint, match_info->n_offsets);
 594   /* Set an invalid position for the previous match. */
 595   match_info->offsets[0] = -1;
 596   match_info->offsets[1] = -1;
 597
 598   return match_info;
 599 }
 600
 601 /**
 602  * g_match_info_get_regex:
 603  * @match_info: a #GMatchInfo
 604  *
 605  * Returns #GRegex object used in @match_info. It belongs to Glib
 606  * and must not be freed. Use g_regex_ref() if you need to keep it
 607  * after you free @match_info object.
 608  *
 609  * Returns: #GRegex object used in @match_info
 610  *
 611  * Since: 2.14
 612  */
 613 GRegex *
 614 g_match_info_get_regex (const GMatchInfo *match_info)
 615 {
 616   g_return_val_if_fail (match_info != NULL, NULL);
 617   return match_info->regex;
 618 }
 619
 620 /**
 621  * g_match_info_get_string:
 622  * @match_info: a #GMatchInfo
 623  *
 624  * Returns the string searched with @match_info. This is the
 625  * string passed to g_regex_match() or g_regex_replace() so
 626  * you may not free it before calling this function.
 627  *
 628  * Returns: the string searched with @match_info
 629  *
 630  * Since: 2.14
 631  */
 632 const gchar *
 633 g_match_info_get_string (const GMatchInfo *match_info)
 634 {
 635   g_return_val_if_fail (match_info != NULL, NULL);
 636   return match_info->string;
 637 }
 638
 639 /**
 640  * g_match_info_ref:
 641  * @match_info: a #GMatchInfo
 642  *
 643  * Increases reference count of @match_info by 1.
 644  *
 645  * Returns: @match_info
 646  *
 647  * Since: 2.30
 648  */
 649 GMatchInfo       *
 650 g_match_info_ref (GMatchInfo *match_info)
 651 {
 652   g_return_val_if_fail (match_info != NULL, NULL);
 653   g_atomic_int_inc (&match_info->ref_count);
 654   return match_info;
 655 }
 656
 657 /**
 658  * g_match_info_unref:
 659  * @match_info: a #GMatchInfo
 660  *
 661  * Decreases reference count of @match_info by 1. When reference count drops
 662  * to zero, it frees all the memory associated with the match_info structure.
 663  *
 664  * Since: 2.30
 665  */
 666 void
 667 g_match_info_unref (GMatchInfo *match_info)
 668 {
 669   if (g_atomic_int_dec_and_test (&match_info->ref_count))
 670     {
 671       g_regex_unref (match_info->regex);
 672       g_free (match_info->offsets);
 673       g_free (match_info->workspace);
 674       g_free (match_info);
 675     }
 676 }
 677
 678 /**
 679  * g_match_info_free:
 680  * @match_info: (nullable): a #GMatchInfo, or %NULL
 681  *
 682  * If @match_info is not %NULL, calls g_match_info_unref(); otherwise does
 683  * nothing.
 684  *
 685  * Since: 2.14
 686  */
 687 void
 688 g_match_info_free (GMatchInfo *match_info)
 689 {
 690   if (match_info == NULL)
 691     return;
 692
 693   g_match_info_unref (match_info);
 694 }
 695
 696 /**
 697  * g_match_info_next:
 698  * @match_info: a #GMatchInfo structure
 699  * @error: location to store the error occurring, or %NULL to ignore errors
 700  *
 701  * Scans for the next match using the same parameters of the previous
 702  * call to g_regex_match_full() or g_regex_match() that returned
 703  * @match_info.
 704  *
 705  * The match is done on the string passed to the match function, so you
 706  * cannot free it before calling this function.
 707  *
 708  * Returns: %TRUE is the string matched, %FALSE otherwise
 709  *
 710  * Since: 2.14
 711  */
 712 gboolean
 713 g_match_info_next (GMatchInfo  *match_info,
 714                    GError     **error)
 715 {
 716   gint prev_match_start;
 717   gint prev_match_end;
 718
 719   g_return_val_if_fail (match_info != NULL, FALSE);
 720   g_return_val_if_fail (error == NULL || *error == NULL, FALSE);
 721   g_return_val_if_fail (match_info->pos >= 0, FALSE);
 722
 723   prev_match_start = match_info->offsets[0];
 724   prev_match_end = match_info->offsets[1];
 725
 726   if (match_info->pos > match_info->string_len)
 727     {
 728       /* we have reached the end of the string */
 729       match_info->pos = -1;
 730       match_info->matches = PCRE_ERROR_NOMATCH;
 731       return FALSE;
 732     }
 733
 734   match_info->matches = pcre_exec (match_info->regex->pcre_re,
 735                                    match_info->regex->extra,
 736                                    match_info->string,
 737                                    match_info->string_len,
 738                                    match_info->pos,
 739                                    match_info->regex->match_opts | match_info->match_opts,
 740                                    match_info->offsets,
 741                                    match_info->n_offsets);
 742   if (IS_PCRE_ERROR (match_info->matches))
 743     {
 744       g_set_error (error, G_REGEX_ERROR, G_REGEX_ERROR_MATCH,
 745                    _("Error while matching regular expression %s: %s"),
 746                    match_info->regex->pattern, match_error (match_info->matches));
 747       return FALSE;
 748     }
 749
 750   /* avoid infinite loops if the pattern is an empty string or something
 751    * equivalent */
 752   if (match_info->pos == match_info->offsets[1])
 753     {
 754       if (match_info->pos > match_info->string_len)
 755         {
 756           /* we have reached the end of the string */
 757           match_info->pos = -1;
 758           match_info->matches = PCRE_ERROR_NOMATCH;
 759           return FALSE;
 760         }
 761
 762       match_info->pos = NEXT_CHAR (match_info->regex,
 763                                    &match_info->string[match_info->pos]) -
 764                                    match_info->string;
 765     }
 766   else
 767     {
 768       match_info->pos = match_info->offsets[1];
 769     }
 770
 771   /* it's possible to get two identical matches when we are matching
 772    * empty strings, for instance if the pattern is "(?=[A-Z0-9])" and
 773    * the string is "RegExTest" we have:
 774    *  - search at position 0: match from 0 to 0
 775    *  - search at position 1: match from 3 to 3
 776    *  - search at position 3: match from 3 to 3 (duplicate)
 777    *  - search at position 4: match from 5 to 5
 778    *  - search at position 5: match from 5 to 5 (duplicate)
 779    *  - search at position 6: no match -> stop
 780    * so we have to ignore the duplicates.
 781    * see bug #515944: http://bugzilla.gnome.org/show_bug.cgi?id=515944 */
 782   if (match_info->matches >= 0 &&
 783       prev_match_start == match_info->offsets[0] &&
 784       prev_match_end == match_info->offsets[1])
 785     {
 786       /* ignore this match and search the next one */
 787       return g_match_info_next (match_info, error);
 788     }
 789
 790   return match_info->matches >= 0;
 791 }
 792
 793 /**
 794  * g_match_info_matches:
 795  * @match_info: a #GMatchInfo structure
 796  *
 797  * Returns whether the previous match operation succeeded.
 798  *
 799  * Returns: %TRUE if the previous match operation succeeded,
 800  *   %FALSE otherwise
 801  *
 802  * Since: 2.14
 803  */
 804 gboolean
 805 g_match_info_matches (const GMatchInfo *match_info)
 806 {
 807   g_return_val_if_fail (match_info != NULL, FALSE);
 808
 809   return match_info->matches >= 0;
 810 }
 811
 812 /**
 813  * g_match_info_get_match_count:
 814  * @match_info: a #GMatchInfo structure
 815  *
 816  * Retrieves the number of matched substrings (including substring 0,
 817  * that is the whole matched text), so 1 is returned if the pattern
 818  * has no substrings in it and 0 is returned if the match failed.
 819  *
 820  * If the last match was obtained using the DFA algorithm, that is
 821  * using g_regex_match_all() or g_regex_match_all_full(), the retrieved
 822  * count is not that of the number of capturing parentheses but that of
 823  * the number of matched substrings.
 824  *
 825  * Returns: Number of matched substrings, or -1 if an error occurred
 826  *
 827  * Since: 2.14
 828  */
 829 gint
 830 g_match_info_get_match_count (const GMatchInfo *match_info)
 831 {
 832   g_return_val_if_fail (match_info, -1);
 833
 834   if (match_info->matches == PCRE_ERROR_NOMATCH)
 835     /* no match */
 836     return 0;
 837   else if (match_info->matches < PCRE_ERROR_NOMATCH)
 838     /* error */
 839     return -1;
 840   else
 841     /* match */
 842     return match_info->matches;
 843 }
 844
 845 /**
 846  * g_match_info_is_partial_match:
 847  * @match_info: a #GMatchInfo structure
 848  *
 849  * Usually if the string passed to g_regex_match*() matches as far as
 850  * it goes, but is too short to match the entire pattern, %FALSE is
 851  * returned. There are circumstances where it might be helpful to
 852  * distinguish this case from other cases in which there is no match.
 853  *
 854  * Consider, for example, an application where a human is required to
 855  * type in data for a field with specific formatting requirements. An
 856  * example might be a date in the form ddmmmyy, defined by the pattern
 857  * "^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$".
 858  * If the application sees the user’s keystrokes one by one, and can
 859  * check that what has been typed so far is potentially valid, it is
 860  * able to raise an error as soon as a mistake is made.
 861  *
 862  * GRegex supports the concept of partial matching by means of the
 863  * #G_REGEX_MATCH_PARTIAL_SOFT and #G_REGEX_MATCH_PARTIAL_HARD flags.
 864  * When they are used, the return code for
 865  * g_regex_match() or g_regex_match_full() is, as usual, %TRUE
 866  * for a complete match, %FALSE otherwise. But, when these functions
 867  * return %FALSE, you can check if the match was partial calling
 868  * g_match_info_is_partial_match().
 869  *
 870  * The difference between #G_REGEX_MATCH_PARTIAL_SOFT and
 871  * #G_REGEX_MATCH_PARTIAL_HARD is that when a partial match is encountered
 872  * with #G_REGEX_MATCH_PARTIAL_SOFT, matching continues to search for a
 873  * possible complete match, while with #G_REGEX_MATCH_PARTIAL_HARD matching
 874  * stops at the partial match.
 875  * When both #G_REGEX_MATCH_PARTIAL_SOFT and #G_REGEX_MATCH_PARTIAL_HARD
 876  * are set, the latter takes precedence.
 877  *
 878  * There were formerly some restrictions on the pattern for partial matching.
 879  * The restrictions no longer apply.
 880  *
 881  * See pcrepartial(3) for more information on partial matching.
 882  *
 883  * Returns: %TRUE if the match was partial, %FALSE otherwise
 884  *
 885  * Since: 2.14
 886  */
 887 gboolean
 888 g_match_info_is_partial_match (const GMatchInfo *match_info)
 889 {
 890   g_return_val_if_fail (match_info != NULL, FALSE);
 891
 892   return match_info->matches == PCRE_ERROR_PARTIAL;
 893 }
 894
 895 /**
 896  * g_match_info_expand_references:
 897  * @match_info: (nullable): a #GMatchInfo or %NULL
 898  * @string_to_expand: the string to expand
 899  * @error: location to store the error occurring, or %NULL to ignore errors
 900  *
 901  * Returns a new string containing the text in @string_to_expand with
 902  * references and escape sequences expanded. References refer to the last
 903  * match done with @string against @regex and have the same syntax used by
 904  * g_regex_replace().
 905  *
 906  * The @string_to_expand must be UTF-8 encoded even if #G_REGEX_RAW was
 907  * passed to g_regex_new().
 908  *
 909  * The backreferences are extracted from the string passed to the match
 910  * function, so you cannot call this function after freeing the string.
 911  *
 912  * @match_info may be %NULL in which case @string_to_expand must not
 913  * contain references. For instance "foo\n" does not refer to an actual
 914  * pattern and '\n' merely will be replaced with \n character,
 915  * while to expand "\0" (whole match) one needs the result of a match.
 916  * Use g_regex_check_replacement() to find out whether @string_to_expand
 917  * contains references.
 918  *
 919  * Returns: (nullable): the expanded string, or %NULL if an error occurred
 920  *
 921  * Since: 2.14
 922  */
 923 gchar *
 924 g_match_info_expand_references (const GMatchInfo  *match_info,
 925                                 const gchar       *string_to_expand,
 926                                 GError           **error)
 927 {
 928   GString *result;
 929   GList *list;
 930   GError *tmp_error = NULL;
 931
 932   g_return_val_if_fail (string_to_expand != NULL, NULL);
 933   g_return_val_if_fail (error == NULL || *error == NULL, NULL);
 934
 935   list = split_replacement (string_to_expand, &tmp_error);
 936   if (tmp_error != NULL)
 937     {
 938       g_propagate_error (error, tmp_error);
 939       return NULL;
 940     }
 941
 942   if (!match_info && interpolation_list_needs_match (list))
 943     {
 944       g_critical ("String '%s' contains references to the match, can't "
 945                   "expand references without GMatchInfo object",
 946                   string_to_expand);
 947       return NULL;
 948     }
 949
 950   result = g_string_sized_new (strlen (string_to_expand));
 951   interpolate_replacement (match_info, result, list);
 952
 953   g_list_free_full (list, (GDestroyNotify) free_interpolation_data);
 954
 955   return g_string_free (result, FALSE);
 956 }
 957
 958 /**
 959  * g_match_info_fetch:
 960  * @match_info: #GMatchInfo structure
 961  * @match_num: number of the sub expression
 962  *
 963  * Retrieves the text matching the @match_num'th capturing
 964  * parentheses. 0 is the full text of the match, 1 is the first paren
 965  * set, 2 the second, and so on.
 966  *
 967  * If @match_num is a valid sub pattern but it didn't match anything
 968  * (e.g. sub pattern 1, matching "b" against "(a)?b") then an empty
 969  * string is returned.
 970  *
 971  * If the match was obtained using the DFA algorithm, that is using
 972  * g_regex_match_all() or g_regex_match_all_full(), the retrieved
 973  * string is not that of a set of parentheses but that of a matched
 974  * substring. Substrings are matched in reverse order of length, so
 975  * 0 is the longest match.
 976  *
 977  * The string is fetched from the string passed to the match function,
 978  * so you cannot call this function after freeing the string.
 979  *
 980  * Returns: (nullable): The matched substring, or %NULL if an error
 981  *     occurred. You have to free the string yourself
 982  *
 983  * Since: 2.14
 984  */
 985 gchar *
 986 g_match_info_fetch (const GMatchInfo *match_info,
 987                     gint              match_num)
 988 {
 989   /* we cannot use pcre_get_substring() because it allocates the
 990    * string using pcre_malloc(). */
 991   gchar *match = NULL;
 992   gint start, end;
 993
 994   g_return_val_if_fail (match_info != NULL, NULL);
 995   g_return_val_if_fail (match_num >= 0, NULL);
 996
 997   /* match_num does not exist or it didn't matched, i.e. matching "b"
 998    * against "(a)?b" then group 0 is empty. */
 999   if (!g_match_info_fetch_pos (match_info, match_num, &start, &end))
1000     match = NULL;
1001   else if (start == -1)
1002     match = g_strdup ("");
1003   else
1004     match = g_strndup (&match_info->string[start], end - start);
1005
1006   return match;
1007 }
1008
1009 /**
1010  * g_match_info_fetch_pos:
1011  * @match_info: #GMatchInfo structure
1012  * @match_num: number of the sub expression
1013  * @start_pos: (out) (optional): pointer to location where to store
1014  *     the start position, or %NULL
1015  * @end_pos: (out) (optional): pointer to location where to store
1016  *     the end position, or %NULL
1017  *
1018  * Retrieves the position in bytes of the @match_num'th capturing
1019  * parentheses. 0 is the full text of the match, 1 is the first
1020  * paren set, 2 the second, and so on.
1021  *
1022  * If @match_num is a valid sub pattern but it didn't match anything
1023  * (e.g. sub pattern 1, matching "b" against "(a)?b") then @start_pos
1024  * and @end_pos are set to -1 and %TRUE is returned.
1025  *
1026  * If the match was obtained using the DFA algorithm, that is using
1027  * g_regex_match_all() or g_regex_match_all_full(), the retrieved
1028  * position is not that of a set of parentheses but that of a matched
1029  * substring. Substrings are matched in reverse order of length, so
1030  * 0 is the longest match.
1031  *
1032  * Returns: %TRUE if the position was fetched, %FALSE otherwise. If
1033  *   the position cannot be fetched, @start_pos and @end_pos are left
1034  *   unchanged
1035  *
1036  * Since: 2.14
1037  */
1038 gboolean
1039 g_match_info_fetch_pos (const GMatchInfo *match_info,
1040                         gint              match_num,
1041                         gint             *start_pos,
1042                         gint             *end_pos)
1043 {
1044   g_return_val_if_fail (match_info != NULL, FALSE);
1045   g_return_val_if_fail (match_num >= 0, FALSE);
1046
1047   /* make sure the sub expression number they're requesting is less than
1048    * the total number of sub expressions that were matched. */
1049   if (match_num >= match_info->matches)
1050     return FALSE;
1051
1052   if (start_pos != NULL)
1053     *start_pos = match_info->offsets[2 * match_num];
1054
1055   if (end_pos != NULL)
1056     *end_pos = match_info->offsets[2 * match_num + 1];
1057
1058   return TRUE;
1059 }
1060
1061 /*
1062  * Returns number of first matched subpattern with name @name.
1063  * There may be more than one in case when DUPNAMES is used,
1064  * and not all subpatterns with that name match;
1065  * pcre_get_stringnumber() does not work in that case.
1066  */
1067 static gint
1068 get_matched_substring_number (const GMatchInfo *match_info,
1069                               const gchar      *name)
1070 {
1071   gint entrysize;
1072   gchar *first, *last;
1073   guchar *entry;
1074
1075   if (!(match_info->regex->compile_opts & G_REGEX_DUPNAMES))
1076     return pcre_get_stringnumber (match_info->regex->pcre_re, name);
1077
1078   /* This code is copied from pcre_get.c: get_first_set() */
1079   entrysize = pcre_get_stringtable_entries (match_info->regex->pcre_re,
1080                                             name,
1081                                             &first,
1082                                             &last);
1083
1084   if (entrysize <= 0)
1085     return entrysize;
1086
1087   for (entry = (guchar*) first; entry <= (guchar*) last; entry += entrysize)
1088     {
1089       gint n = (entry[0] << 8) + entry[1];
1090       if (match_info->offsets[n*2] >= 0)
1091         return n;
1092     }
1093
1094   return (first[0] << 8) + first[1];
1095 }
1096
1097 /**
1098  * g_match_info_fetch_named:
1099  * @match_info: #GMatchInfo structure
1100  * @name: name of the subexpression
1101  *
1102  * Retrieves the text matching the capturing parentheses named @name.
1103  *
1104  * If @name is a valid sub pattern name but it didn't match anything
1105  * (e.g. sub pattern "X", matching "b" against "(?P<X>a)?b")
1106  * then an empty string is returned.
1107  *
1108  * The string is fetched from the string passed to the match function,
1109  * so you cannot call this function after freeing the string.
1110  *
1111  * Returns: (nullable): The matched substring, or %NULL if an error
1112  *     occurred. You have to free the string yourself
1113  *
1114  * Since: 2.14
1115  */
1116 gchar *
1117 g_match_info_fetch_named (const GMatchInfo *match_info,
1118                           const gchar      *name)
1119 {
1120   /* we cannot use pcre_get_named_substring() because it allocates the
1121    * string using pcre_malloc(). */
1122   gint num;
1123
1124   g_return_val_if_fail (match_info != NULL, NULL);
1125   g_return_val_if_fail (name != NULL, NULL);
1126
1127   num = get_matched_substring_number (match_info, name);
1128   if (num < 0)
1129     return NULL;
1130   else
1131     return g_match_info_fetch (match_info, num);
1132 }
1133
1134 /**
1135  * g_match_info_fetch_named_pos:
1136  * @match_info: #GMatchInfo structure
1137  * @name: name of the subexpression
1138  * @start_pos: (out) (optional): pointer to location where to store
1139  *     the start position, or %NULL
1140  * @end_pos: (out) (optional): pointer to location where to store
1141  *     the end position, or %NULL
1142  *
1143  * Retrieves the position in bytes of the capturing parentheses named @name.
1144  *
1145  * If @name is a valid sub pattern name but it didn't match anything
1146  * (e.g. sub pattern "X", matching "b" against "(?P<X>a)?b")
1147  * then @start_pos and @end_pos are set to -1 and %TRUE is returned.
1148  *
1149  * Returns: %TRUE if the position was fetched, %FALSE otherwise.
1150  *     If the position cannot be fetched, @start_pos and @end_pos
1151  *     are left unchanged.
1152  *
1153  * Since: 2.14
1154  */
1155 gboolean
1156 g_match_info_fetch_named_pos (const GMatchInfo *match_info,
1157                               const gchar      *name,
1158                               gint             *start_pos,
1159                               gint             *end_pos)
1160 {
1161   gint num;
1162
1163   g_return_val_if_fail (match_info != NULL, FALSE);
1164   g_return_val_if_fail (name != NULL, FALSE);
1165
1166   num = get_matched_substring_number (match_info, name);
1167   if (num < 0)
1168     return FALSE;
1169
1170   return g_match_info_fetch_pos (match_info, num, start_pos, end_pos);
1171 }
1172
1173 /**
1174  * g_match_info_fetch_all:
1175  * @match_info: a #GMatchInfo structure
1176  *
1177  * Bundles up pointers to each of the matching substrings from a match
1178  * and stores them in an array of gchar pointers. The first element in
1179  * the returned array is the match number 0, i.e. the entire matched
1180  * text.
1181  *
1182  * If a sub pattern didn't match anything (e.g. sub pattern 1, matching
1183  * "b" against "(a)?b") then an empty string is inserted.
1184  *
1185  * If the last match was obtained using the DFA algorithm, that is using
1186  * g_regex_match_all() or g_regex_match_all_full(), the retrieved
1187  * strings are not that matched by sets of parentheses but that of the
1188  * matched substring. Substrings are matched in reverse order of length,
1189  * so the first one is the longest match.
1190  *
1191  * The strings are fetched from the string passed to the match function,
1192  * so you cannot call this function after freeing the string.
1193  *
1194  * Returns: (transfer full): a %NULL-terminated array of gchar *
1195  *     pointers.  It must be freed using g_strfreev(). If the previous
1196  *     match failed %NULL is returned
1197  *
1198  * Since: 2.14
1199  */
1200 gchar **
1201 g_match_info_fetch_all (const GMatchInfo *match_info)
1202 {
1203   /* we cannot use pcre_get_substring_list() because the returned value
1204    * isn't suitable for g_strfreev(). */
1205   gchar **result;
1206   gint i;
1207
1208   g_return_val_if_fail (match_info != NULL, NULL);
1209
1210   if (match_info->matches < 0)
1211     return NULL;
1212
1213   result = g_new (gchar *, match_info->matches + 1);
1214   for (i = 0; i < match_info->matches; i++)
1215     result[i] = g_match_info_fetch (match_info, i);
1216   result[i] = NULL;
1217
1218   return result;
1219 }
1220
1221
1222 /* GRegex */
1223
1224 G_DEFINE_QUARK (g-regex-error-quark, g_regex_error)
1225
1226 /**
1227  * g_regex_ref:
1228  * @regex: a #GRegex
1229  *
1230  * Increases reference count of @regex by 1.
1231  *
1232  * Returns: @regex
1233  *
1234  * Since: 2.14
1235  */
1236 GRegex *
1237 g_regex_ref (GRegex *regex)
1238 {
1239   g_return_val_if_fail (regex != NULL, NULL);
1240   g_atomic_int_inc (&regex->ref_count);
1241   return regex;
1242 }
1243
1244 /**
1245  * g_regex_unref:
1246  * @regex: a #GRegex
1247  *
1248  * Decreases reference count of @regex by 1. When reference count drops
1249  * to zero, it frees all the memory associated with the regex structure.
1250  *
1251  * Since: 2.14
1252  */
1253 void
1254 g_regex_unref (GRegex *regex)
1255 {
1256   g_return_if_fail (regex != NULL);
1257
1258   if (g_atomic_int_dec_and_test (&regex->ref_count))
1259     {
1260       g_free (regex->pattern);
1261       if (regex->pcre_re != NULL)
1262         pcre_free (regex->pcre_re);
1263       if (regex->extra != NULL)
1264         pcre_free (regex->extra);
1265       g_free (regex);
1266     }
1267 }
1268
1269 /*
1270  * @match_options: (inout) (optional):
1271  */
1272 static pcre *regex_compile (const gchar         *pattern,
1273                             GRegexCompileFlags   compile_options,
1274                             GRegexCompileFlags  *compile_options_out,
1275                             GRegexMatchFlags    *match_options,
1276                             GError             **error);
1277
1278 /**
1279  * g_regex_new:
1280  * @pattern: the regular expression
1281  * @compile_options: compile options for the regular expression, or 0
1282  * @match_options: match options for the regular expression, or 0
1283  * @error: return location for a #GError
1284  *
1285  * Compiles the regular expression to an internal form, and does
1286  * the initial setup of the #GRegex structure.
1287  *
1288  * Returns: (nullable): a #GRegex structure or %NULL if an error occured. Call
1289  *   g_regex_unref() when you are done with it
1290  *
1291  * Since: 2.14
1292  */
1293 GRegex *
1294 g_regex_new (const gchar         *pattern,
1295              GRegexCompileFlags   compile_options,
1296              GRegexMatchFlags     match_options,
1297              GError             **error)
1298 {
1299   GRegex *regex;
1300   pcre *re;
1301   const gchar *errmsg;
1302   gboolean optimize = FALSE;
1303   static volatile gsize initialised = 0;
1304
1305   g_return_val_if_fail (pattern != NULL, NULL);
1306   g_return_val_if_fail (error == NULL || *error == NULL, NULL);
1307   g_return_val_if_fail ((compile_options & ~G_REGEX_COMPILE_MASK) == 0, NULL);
1308   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
1309
1310   if (g_once_init_enter (&initialised))
1311     {
1312       int supports_utf8, supports_ucp;
1313
1314       pcre_config (PCRE_CONFIG_UTF8, &supports_utf8);
1315       if (!supports_utf8)
1316         g_critical (_("PCRE library is compiled without UTF8 support"));
1317
1318       pcre_config (PCRE_CONFIG_UNICODE_PROPERTIES, &supports_ucp);
1319       if (!supports_ucp)
1320         g_critical (_("PCRE library is compiled without UTF8 properties support"));
1321
1322       g_once_init_leave (&initialised, supports_utf8 && supports_ucp ? 1 : 2);
1323     }
1324
1325   if (G_UNLIKELY (initialised != 1))
1326     {
1327       g_set_error_literal (error, G_REGEX_ERROR, G_REGEX_ERROR_COMPILE,
1328                            _("PCRE library is compiled with incompatible options"));
1329       return NULL;
1330     }
1331
1332   /* G_REGEX_OPTIMIZE has the same numeric value of PCRE_NO_UTF8_CHECK,
1333    * as we do not need to wrap PCRE_NO_UTF8_CHECK. */
1334   if (compile_options & G_REGEX_OPTIMIZE)
1335     optimize = TRUE;
1336
1337   re = regex_compile (pattern, compile_options, &compile_options,
1338                       &match_options, error);
1339
1340   if (re == NULL)
1341     return NULL;
1342
1343   regex = g_new0 (GRegex, 1);
1344   regex->ref_count = 1;
1345   regex->pattern = g_strdup (pattern);
1346   regex->pcre_re = re;
1347   regex->compile_opts = compile_options;
1348   regex->match_opts = match_options;
1349
1350   if (optimize)
1351     {
1352       regex->extra = pcre_study (regex->pcre_re, 0, &errmsg);
1353       if (errmsg != NULL)
1354         {
1355           GError *tmp_error = g_error_new (G_REGEX_ERROR,
1356                                            G_REGEX_ERROR_OPTIMIZE,
1357                                            _("Error while optimizing "
1358                                              "regular expression %s: %s"),
1359                                            regex->pattern,
1360                                            errmsg);
1361           g_propagate_error (error, tmp_error);
1362
1363           g_regex_unref (regex);
1364           return NULL;
1365         }
1366     }
1367
1368   return regex;
1369 }
1370
1371 static pcre *
1372 regex_compile (const gchar         *pattern,
1373                GRegexCompileFlags   compile_options,
1374                GRegexCompileFlags  *compile_options_out,
1375                GRegexMatchFlags    *match_options,
1376                GError             **error)
1377 {
1378   pcre *re;
1379   const gchar *errmsg;
1380   gint erroffset;
1381   gint errcode;
1382   GRegexCompileFlags nonpcre_compile_options;
1383   unsigned long int pcre_compile_options;
1384
1385   nonpcre_compile_options = compile_options & G_REGEX_COMPILE_NONPCRE_MASK;
1386
1387   /* In GRegex the string are, by default, UTF-8 encoded. PCRE
1388    * instead uses UTF-8 only if required with PCRE_UTF8. */
1389   if (compile_options & G_REGEX_RAW)
1390     {
1391       /* disable utf-8 */
1392       compile_options &= ~G_REGEX_RAW;
1393     }
1394   else
1395     {
1396       /* enable utf-8 */
1397       compile_options |= PCRE_UTF8 | PCRE_NO_UTF8_CHECK;
1398
1399       if (match_options != NULL)
1400         *match_options |= PCRE_NO_UTF8_CHECK;
1401     }
1402
1403   /* PCRE_NEWLINE_ANY is the default for the internal PCRE but
1404    * not for the system one. */
1405   if (!(compile_options & G_REGEX_NEWLINE_CR) &&
1406       !(compile_options & G_REGEX_NEWLINE_LF))
1407     {
1408       compile_options |= PCRE_NEWLINE_ANY;
1409     }
1410
1411   compile_options |= PCRE_UCP;
1412
1413   /* PCRE_BSR_UNICODE is the default for the internal PCRE but
1414    * possibly not for the system one.
1415    */
1416   if (~compile_options & G_REGEX_BSR_ANYCRLF)
1417     compile_options |= PCRE_BSR_UNICODE;
1418
1419   /* compile the pattern */
1420   re = pcre_compile2 (pattern, compile_options, &errcode,
1421                       &errmsg, &erroffset, NULL);
1422
1423   /* if the compilation failed, set the error member and return
1424    * immediately */
1425   if (re == NULL)
1426     {
1427       GError *tmp_error;
1428
1429       /* Translate the PCRE error code to GRegexError and use a translated
1430        * error message if possible */
1431       translate_compile_error (&errcode, &errmsg);
1432
1433       /* PCRE uses byte offsets but we want to show character offsets */
1434       erroffset = g_utf8_pointer_to_offset (pattern, &pattern[erroffset]);
1435
1436       tmp_error = g_error_new (G_REGEX_ERROR, errcode,
1437                                _("Error while compiling regular "
1438                                  "expression %s at char %d: %s"),
1439                                pattern, erroffset, errmsg);
1440       g_propagate_error (error, tmp_error);
1441
1442       return NULL;
1443     }
1444
1445   /* For options set at the beginning of the pattern, pcre puts them into
1446    * compile options, e.g. "(?i)foo" will make the pcre structure store
1447    * PCRE_CASELESS even though it wasn't explicitly given for compilation. */
1448   pcre_fullinfo (re, NULL, PCRE_INFO_OPTIONS, &pcre_compile_options);
1449   compile_options = pcre_compile_options & G_REGEX_COMPILE_PCRE_MASK;
1450
1451   /* Don't leak PCRE_NEWLINE_ANY, which is part of PCRE_NEWLINE_ANYCRLF */
1452   if ((pcre_compile_options & PCRE_NEWLINE_ANYCRLF) != PCRE_NEWLINE_ANYCRLF)
1453     compile_options &= ~PCRE_NEWLINE_ANY;
1454
1455   compile_options |= nonpcre_compile_options;
1456
1457   if (!(compile_options & G_REGEX_DUPNAMES))
1458     {
1459       gboolean jchanged = FALSE;
1460       pcre_fullinfo (re, NULL, PCRE_INFO_JCHANGED, &jchanged);
1461       if (jchanged)
1462         compile_options |= G_REGEX_DUPNAMES;
1463     }
1464
1465   if (compile_options_out != 0)
1466     *compile_options_out = compile_options;
1467
1468   return re;
1469 }
1470
1471 /**
1472  * g_regex_get_pattern:
1473  * @regex: a #GRegex structure
1474  *
1475  * Gets the pattern string associated with @regex, i.e. a copy of
1476  * the string passed to g_regex_new().
1477  *
1478  * Returns: the pattern of @regex
1479  *
1480  * Since: 2.14
1481  */
1482 const gchar *
1483 g_regex_get_pattern (const GRegex *regex)
1484 {
1485   g_return_val_if_fail (regex != NULL, NULL);
1486
1487   return regex->pattern;
1488 }
1489
1490 /**
1491  * g_regex_get_max_backref:
1492  * @regex: a #GRegex
1493  *
1494  * Returns the number of the highest back reference
1495  * in the pattern, or 0 if the pattern does not contain
1496  * back references.
1497  *
1498  * Returns: the number of the highest back reference
1499  *
1500  * Since: 2.14
1501  */
1502 gint
1503 g_regex_get_max_backref (const GRegex *regex)
1504 {
1505   gint value;
1506
1507   pcre_fullinfo (regex->pcre_re, regex->extra,
1508                  PCRE_INFO_BACKREFMAX, &value);
1509
1510   return value;
1511 }
1512
1513 /**
1514  * g_regex_get_capture_count:
1515  * @regex: a #GRegex
1516  *
1517  * Returns the number of capturing subpatterns in the pattern.
1518  *
1519  * Returns: the number of capturing subpatterns
1520  *
1521  * Since: 2.14
1522  */
1523 gint
1524 g_regex_get_capture_count (const GRegex *regex)
1525 {
1526   gint value;
1527
1528   pcre_fullinfo (regex->pcre_re, regex->extra,
1529                  PCRE_INFO_CAPTURECOUNT, &value);
1530
1531   return value;
1532 }
1533
1534 /**
1535  * g_regex_get_has_cr_or_lf:
1536  * @regex: a #GRegex structure
1537  *
1538  * Checks whether the pattern contains explicit CR or LF references.
1539  *
1540  * Returns: %TRUE if the pattern contains explicit CR or LF references
1541  *
1542  * Since: 2.34
1543  */
1544 gboolean
1545 g_regex_get_has_cr_or_lf (const GRegex *regex)
1546 {
1547   gint value;
1548
1549   pcre_fullinfo (regex->pcre_re, regex->extra,
1550                  PCRE_INFO_HASCRORLF, &value);
1551
1552   return !!value;
1553 }
1554
1555 /**
1556  * g_regex_get_max_lookbehind:
1557  * @regex: a #GRegex structure
1558  *
1559  * Gets the number of characters in the longest lookbehind assertion in the
1560  * pattern. This information is useful when doing multi-segment matching using
1561  * the partial matching facilities.
1562  *
1563  * Returns: the number of characters in the longest lookbehind assertion.
1564  *
1565  * Since: 2.38
1566  */
1567 gint
1568 g_regex_get_max_lookbehind (const GRegex *regex)
1569 {
1570   gint max_lookbehind;
1571
1572   pcre_fullinfo (regex->pcre_re, regex->extra,
1573                  PCRE_INFO_MAXLOOKBEHIND, &max_lookbehind);
1574
1575   return max_lookbehind;
1576 }
1577
1578 /**
1579  * g_regex_get_compile_flags:
1580  * @regex: a #GRegex
1581  *
1582  * Returns the compile options that @regex was created with.
1583  *
1584  * Depending on the version of PCRE that is used, this may or may not
1585  * include flags set by option expressions such as `(?i)` found at the
1586  * top-level within the compiled pattern.
1587  *
1588  * Returns: flags from #GRegexCompileFlags
1589  *
1590  * Since: 2.26
1591  */
1592 GRegexCompileFlags
1593 g_regex_get_compile_flags (const GRegex *regex)
1594 {
1595   g_return_val_if_fail (regex != NULL, 0);
1596
1597   return regex->compile_opts;
1598 }
1599
1600 /**
1601  * g_regex_get_match_flags:
1602  * @regex: a #GRegex
1603  *
1604  * Returns the match options that @regex was created with.
1605  *
1606  * Returns: flags from #GRegexMatchFlags
1607  *
1608  * Since: 2.26
1609  */
1610 GRegexMatchFlags
1611 g_regex_get_match_flags (const GRegex *regex)
1612 {
1613   g_return_val_if_fail (regex != NULL, 0);
1614
1615   return regex->match_opts & G_REGEX_MATCH_MASK;
1616 }
1617
1618 /**
1619  * g_regex_match_simple:
1620  * @pattern: the regular expression
1621  * @string: the string to scan for matches
1622  * @compile_options: compile options for the regular expression, or 0
1623  * @match_options: match options, or 0
1624  *
1625  * Scans for a match in @string for @pattern.
1626  *
1627  * This function is equivalent to g_regex_match() but it does not
1628  * require to compile the pattern with g_regex_new(), avoiding some
1629  * lines of code when you need just to do a match without extracting
1630  * substrings, capture counts, and so on.
1631  *
1632  * If this function is to be called on the same @pattern more than
1633  * once, it's more efficient to compile the pattern once with
1634  * g_regex_new() and then use g_regex_match().
1635  *
1636  * Returns: %TRUE if the string matched, %FALSE otherwise
1637  *
1638  * Since: 2.14
1639  */
1640 gboolean
1641 g_regex_match_simple (const gchar        *pattern,
1642                       const gchar        *string,
1643                       GRegexCompileFlags  compile_options,
1644                       GRegexMatchFlags    match_options)
1645 {
1646   GRegex *regex;
1647   gboolean result;
1648
1649   regex = g_regex_new (pattern, compile_options, 0, NULL);
1650   if (!regex)
1651     return FALSE;
1652   result = g_regex_match_full (regex, string, -1, 0, match_options, NULL, NULL);
1653   g_regex_unref (regex);
1654   return result;
1655 }
1656
1657 /**
1658  * g_regex_match:
1659  * @regex: a #GRegex structure from g_regex_new()
1660  * @string: the string to scan for matches
1661  * @match_options: match options
1662  * @match_info: (out) (optional): pointer to location where to store
1663  *     the #GMatchInfo, or %NULL if you do not need it
1664  *
1665  * Scans for a match in @string for the pattern in @regex.
1666  * The @match_options are combined with the match options specified
1667  * when the @regex structure was created, letting you have more
1668  * flexibility in reusing #GRegex structures.
1669  *
1670  * Unless %G_REGEX_RAW is specified in the options, @string must be valid UTF-8.
1671  *
1672  * A #GMatchInfo structure, used to get information on the match,
1673  * is stored in @match_info if not %NULL. Note that if @match_info
1674  * is not %NULL then it is created even if the function returns %FALSE,
1675  * i.e. you must free it regardless if regular expression actually matched.
1676  *
1677  * To retrieve all the non-overlapping matches of the pattern in
1678  * string you can use g_match_info_next().
1679  *
1680  * |[<!-- language="C" -->
1681  * static void
1682  * print_uppercase_words (const gchar *string)
1683  * {
1684  *   // Print all uppercase-only words.
1685  *   GRegex *regex;
1686  *   GMatchInfo *match_info;
1687  *
1688  *   regex = g_regex_new ("[A-Z]+", 0, 0, NULL);
1689  *   g_regex_match (regex, string, 0, &match_info);
1690  *   while (g_match_info_matches (match_info))
1691  *     {
1692  *       gchar *word = g_match_info_fetch (match_info, 0);
1693  *       g_print ("Found: %s\n", word);
1694  *       g_free (word);
1695  *       g_match_info_next (match_info, NULL);
1696  *     }
1697  *   g_match_info_free (match_info);
1698  *   g_regex_unref (regex);
1699  * }
1700  * ]|
1701  *
1702  * @string is not copied and is used in #GMatchInfo internally. If
1703  * you use any #GMatchInfo method (except g_match_info_free()) after
1704  * freeing or modifying @string then the behaviour is undefined.
1705  *
1706  * Returns: %TRUE is the string matched, %FALSE otherwise
1707  *
1708  * Since: 2.14
1709  */
1710 gboolean
1711 g_regex_match (const GRegex      *regex,
1712                const gchar       *string,
1713                GRegexMatchFlags   match_options,
1714                GMatchInfo       **match_info)
1715 {
1716   return g_regex_match_full (regex, string, -1, 0, match_options,
1717                              match_info, NULL);
1718 }
1719
1720 /**
1721  * g_regex_match_full:
1722  * @regex: a #GRegex structure from g_regex_new()
1723  * @string: (array length=string_len): the string to scan for matches
1724  * @string_len: the length of @string, in bytes, or -1 if @string is nul-terminated
1725  * @start_position: starting index of the string to match, in bytes
1726  * @match_options: match options
1727  * @match_info: (out) (optional): pointer to location where to store
1728  *     the #GMatchInfo, or %NULL if you do not need it
1729  * @error: location to store the error occurring, or %NULL to ignore errors
1730  *
1731  * Scans for a match in @string for the pattern in @regex.
1732  * The @match_options are combined with the match options specified
1733  * when the @regex structure was created, letting you have more
1734  * flexibility in reusing #GRegex structures.
1735  *
1736  * Setting @start_position differs from just passing over a shortened
1737  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern
1738  * that begins with any kind of lookbehind assertion, such as "\b".
1739  *
1740  * Unless %G_REGEX_RAW is specified in the options, @string must be valid UTF-8.
1741  *
1742  * A #GMatchInfo structure, used to get information on the match, is
1743  * stored in @match_info if not %NULL. Note that if @match_info is
1744  * not %NULL then it is created even if the function returns %FALSE,
1745  * i.e. you must free it regardless if regular expression actually
1746  * matched.
1747  *
1748  * @string is not copied and is used in #GMatchInfo internally. If
1749  * you use any #GMatchInfo method (except g_match_info_free()) after
1750  * freeing or modifying @string then the behaviour is undefined.
1751  *
1752  * To retrieve all the non-overlapping matches of the pattern in
1753  * string you can use g_match_info_next().
1754  *
1755  * |[<!-- language="C" -->
1756  * static void
1757  * print_uppercase_words (const gchar *string)
1758  * {
1759  *   // Print all uppercase-only words.
1760  *   GRegex *regex;
1761  *   GMatchInfo *match_info;
1762  *   GError *error = NULL;
1763  *
1764  *   regex = g_regex_new ("[A-Z]+", 0, 0, NULL);
1765  *   g_regex_match_full (regex, string, -1, 0, 0, &match_info, &error);
1766  *   while (g_match_info_matches (match_info))
1767  *     {
1768  *       gchar *word = g_match_info_fetch (match_info, 0);
1769  *       g_print ("Found: %s\n", word);
1770  *       g_free (word);
1771  *       g_match_info_next (match_info, &error);
1772  *     }
1773  *   g_match_info_free (match_info);
1774  *   g_regex_unref (regex);
1775  *   if (error != NULL)
1776  *     {
1777  *       g_printerr ("Error while matching: %s\n", error->message);
1778  *       g_error_free (error);
1779  *     }
1780  * }
1781  * ]|
1782  *
1783  * Returns: %TRUE is the string matched, %FALSE otherwise
1784  *
1785  * Since: 2.14
1786  */
1787 gboolean
1788 g_regex_match_full (const GRegex      *regex,
1789                     const gchar       *string,
1790                     gssize             string_len,
1791                     gint               start_position,
1792                     GRegexMatchFlags   match_options,
1793                     GMatchInfo       **match_info,
1794                     GError           **error)
1795 {
1796   GMatchInfo *info;
1797   gboolean match_ok;
1798
1799   g_return_val_if_fail (regex != NULL, FALSE);
1800   g_return_val_if_fail (string != NULL, FALSE);
1801   g_return_val_if_fail (start_position >= 0, FALSE);
1802   g_return_val_if_fail (error == NULL || *error == NULL, FALSE);
1803   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, FALSE);
1804
1805   info = match_info_new (regex, string, string_len, start_position,
1806                          match_options, FALSE);
1807   match_ok = g_match_info_next (info, error);
1808   if (match_info != NULL)
1809     *match_info = info;
1810   else
1811     g_match_info_free (info);
1812
1813   return match_ok;
1814 }
1815
1816 /**
1817  * g_regex_match_all:
1818  * @regex: a #GRegex structure from g_regex_new()
1819  * @string: the string to scan for matches
1820  * @match_options: match options
1821  * @match_info: (out) (optional): pointer to location where to store
1822  *     the #GMatchInfo, or %NULL if you do not need it
1823  *
1824  * Using the standard algorithm for regular expression matching only
1825  * the longest match in the string is retrieved. This function uses
1826  * a different algorithm so it can retrieve all the possible matches.
1827  * For more documentation see g_regex_match_all_full().
1828  *
1829  * A #GMatchInfo structure, used to get information on the match, is
1830  * stored in @match_info if not %NULL. Note that if @match_info is
1831  * not %NULL then it is created even if the function returns %FALSE,
1832  * i.e. you must free it regardless if regular expression actually
1833  * matched.
1834  *
1835  * @string is not copied and is used in #GMatchInfo internally. If
1836  * you use any #GMatchInfo method (except g_match_info_free()) after
1837  * freeing or modifying @string then the behaviour is undefined.
1838  *
1839  * Returns: %TRUE is the string matched, %FALSE otherwise
1840  *
1841  * Since: 2.14
1842  */
1843 gboolean
1844 g_regex_match_all (const GRegex      *regex,
1845                    const gchar       *string,
1846                    GRegexMatchFlags   match_options,
1847                    GMatchInfo       **match_info)
1848 {
1849   return g_regex_match_all_full (regex, string, -1, 0, match_options,
1850                                  match_info, NULL);
1851 }
1852
1853 /**
1854  * g_regex_match_all_full:
1855  * @regex: a #GRegex structure from g_regex_new()
1856  * @string: (array length=string_len): the string to scan for matches
1857  * @string_len: the length of @string, in bytes, or -1 if @string is nul-terminated
1858  * @start_position: starting index of the string to match, in bytes
1859  * @match_options: match options
1860  * @match_info: (out) (optional): pointer to location where to store
1861  *     the #GMatchInfo, or %NULL if you do not need it
1862  * @error: location to store the error occurring, or %NULL to ignore errors
1863  *
1864  * Using the standard algorithm for regular expression matching only
1865  * the longest match in the @string is retrieved, it is not possible
1866  * to obtain all the available matches. For instance matching
1867  * "<a> <b> <c>" against the pattern "<.*>"
1868  * you get "<a> <b> <c>".
1869  *
1870  * This function uses a different algorithm (called DFA, i.e. deterministic
1871  * finite automaton), so it can retrieve all the possible matches, all
1872  * starting at the same point in the string. For instance matching
1873  * "<a> <b> <c>" against the pattern "<.*>;"
1874  * you would obtain three matches: "<a> <b> <c>",
1875  * "<a> <b>" and "<a>".
1876  *
1877  * The number of matched strings is retrieved using
1878  * g_match_info_get_match_count(). To obtain the matched strings and
1879  * their position you can use, respectively, g_match_info_fetch() and
1880  * g_match_info_fetch_pos(). Note that the strings are returned in
1881  * reverse order of length; that is, the longest matching string is
1882  * given first.
1883  *
1884  * Note that the DFA algorithm is slower than the standard one and it
1885  * is not able to capture substrings, so backreferences do not work.
1886  *
1887  * Setting @start_position differs from just passing over a shortened
1888  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern
1889  * that begins with any kind of lookbehind assertion, such as "\b".
1890  *
1891  * Unless %G_REGEX_RAW is specified in the options, @string must be valid UTF-8.
1892  *
1893  * A #GMatchInfo structure, used to get information on the match, is
1894  * stored in @match_info if not %NULL. Note that if @match_info is
1895  * not %NULL then it is created even if the function returns %FALSE,
1896  * i.e. you must free it regardless if regular expression actually
1897  * matched.
1898  *
1899  * @string is not copied and is used in #GMatchInfo internally. If
1900  * you use any #GMatchInfo method (except g_match_info_free()) after
1901  * freeing or modifying @string then the behaviour is undefined.
1902  *
1903  * Returns: %TRUE is the string matched, %FALSE otherwise
1904  *
1905  * Since: 2.14
1906  */
1907 gboolean
1908 g_regex_match_all_full (const GRegex      *regex,
1909                         const gchar       *string,
1910                         gssize             string_len,
1911                         gint               start_position,
1912                         GRegexMatchFlags   match_options,
1913                         GMatchInfo       **match_info,
1914                         GError           **error)
1915 {
1916   GMatchInfo *info;
1917   gboolean done;
1918   pcre *pcre_re;
1919   pcre_extra *extra;
1920   gboolean retval;
1921
1922   g_return_val_if_fail (regex != NULL, FALSE);
1923   g_return_val_if_fail (string != NULL, FALSE);
1924   g_return_val_if_fail (start_position >= 0, FALSE);
1925   g_return_val_if_fail (error == NULL || *error == NULL, FALSE);
1926   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, FALSE);
1927
1928 #ifdef PCRE_NO_AUTO_POSSESS
1929   /* For PCRE >= 8.34 we need to turn off PCRE_NO_AUTO_POSSESS, which
1930    * is an optimization for normal regex matching, but results in omitting
1931    * some shorter matches here, and an observable behaviour change.
1932    *
1933    * DFA matching is rather niche, and very rarely used according to
1934    * codesearch.debian.net, so don't bother caching the recompiled RE. */
1935   pcre_re = regex_compile (regex->pattern,
1936                            regex->compile_opts | PCRE_NO_AUTO_POSSESS,
1937                            NULL, NULL, error);
1938
1939   if (pcre_re == NULL)
1940     return FALSE;
1941
1942   /* Not bothering to cache the optimization data either, with similar
1943    * reasoning */
1944   extra = NULL;
1945 #else
1946   /* For PCRE < 8.33 the precompiled regex is fine. */
1947   pcre_re = regex->pcre_re;
1948   extra = regex->extra;
1949 #endif
1950
1951   info = match_info_new (regex, string, string_len, start_position,
1952                          match_options, TRUE);
1953
1954   done = FALSE;
1955   while (!done)
1956     {
1957       done = TRUE;
1958       info->matches = pcre_dfa_exec (pcre_re, extra,
1959                                      info->string, info->string_len,
1960                                      info->pos,
1961                                      regex->match_opts | match_options,
1962                                      info->offsets, info->n_offsets,
1963                                      info->workspace, info->n_workspace);
1964       if (info->matches == PCRE_ERROR_DFA_WSSIZE)
1965         {
1966           /* info->workspace is too small. */
1967           info->n_workspace *= 2;
1968           info->workspace = g_realloc (info->workspace,
1969                                        info->n_workspace * sizeof (gint));
1970           done = FALSE;
1971         }
1972       else if (info->matches == 0)
1973         {
1974           /* info->offsets is too small. */
1975           info->n_offsets *= 2;
1976           info->offsets = g_realloc (info->offsets,
1977                                      info->n_offsets * sizeof (gint));
1978           done = FALSE;
1979         }
1980       else if (IS_PCRE_ERROR (info->matches))
1981         {
1982           g_set_error (error, G_REGEX_ERROR, G_REGEX_ERROR_MATCH,
1983                        _("Error while matching regular expression %s: %s"),
1984                        regex->pattern, match_error (info->matches));
1985         }
1986     }
1987
1988 #ifdef PCRE_NO_AUTO_POSSESS
1989   pcre_free (pcre_re);
1990 #endif
1991
1992   /* set info->pos to -1 so that a call to g_match_info_next() fails. */
1993   info->pos = -1;
1994   retval = info->matches >= 0;
1995
1996   if (match_info != NULL)
1997     *match_info = info;
1998   else
1999     g_match_info_free (info);
2000
2001   return retval;
2002 }
2003
2004 /**
2005  * g_regex_get_string_number:
2006  * @regex: #GRegex structure
2007  * @name: name of the subexpression
2008  *
2009  * Retrieves the number of the subexpression named @name.
2010  *
2011  * Returns: The number of the subexpression or -1 if @name
2012  *   does not exists
2013  *
2014  * Since: 2.14
2015  */
2016 gint
2017 g_regex_get_string_number (const GRegex *regex,
2018                            const gchar  *name)
2019 {
2020   gint num;
2021
2022   g_return_val_if_fail (regex != NULL, -1);
2023   g_return_val_if_fail (name != NULL, -1);
2024
2025   num = pcre_get_stringnumber (regex->pcre_re, name);
2026   if (num == PCRE_ERROR_NOSUBSTRING)
2027     num = -1;
2028
2029   return num;
2030 }
2031
2032 /**
2033  * g_regex_split_simple:
2034  * @pattern: the regular expression
2035  * @string: the string to scan for matches
2036  * @compile_options: compile options for the regular expression, or 0
2037  * @match_options: match options, or 0
2038  *
2039  * Breaks the string on the pattern, and returns an array of
2040  * the tokens. If the pattern contains capturing parentheses,
2041  * then the text for each of the substrings will also be returned.
2042  * If the pattern does not match anywhere in the string, then the
2043  * whole string is returned as the first token.
2044  *
2045  * This function is equivalent to g_regex_split() but it does
2046  * not require to compile the pattern with g_regex_new(), avoiding
2047  * some lines of code when you need just to do a split without
2048  * extracting substrings, capture counts, and so on.
2049  *
2050  * If this function is to be called on the same @pattern more than
2051  * once, it's more efficient to compile the pattern once with
2052  * g_regex_new() and then use g_regex_split().
2053  *
2054  * As a special case, the result of splitting the empty string ""
2055  * is an empty vector, not a vector containing a single string.
2056  * The reason for this special case is that being able to represent
2057  * a empty vector is typically more useful than consistent handling
2058  * of empty elements. If you do need to represent empty elements,
2059  * you'll need to check for the empty string before calling this
2060  * function.
2061  *
2062  * A pattern that can match empty strings splits @string into
2063  * separate characters wherever it matches the empty string between
2064  * characters. For example splitting "ab c" using as a separator
2065  * "\s*", you will get "a", "b" and "c".
2066  *
2067  * Returns: (transfer full): a %NULL-terminated array of strings. Free
2068  * it using g_strfreev()
2069  *
2070  * Since: 2.14
2071  **/
2072 gchar **
2073 g_regex_split_simple (const gchar        *pattern,
2074                       const gchar        *string,
2075                       GRegexCompileFlags  compile_options,
2076                       GRegexMatchFlags    match_options)
2077 {
2078   GRegex *regex;
2079   gchar **result;
2080
2081   regex = g_regex_new (pattern, compile_options, 0, NULL);
2082   if (!regex)
2083     return NULL;
2084
2085   result = g_regex_split_full (regex, string, -1, 0, match_options, 0, NULL);
2086   g_regex_unref (regex);
2087   return result;
2088 }
2089
2090 /**
2091  * g_regex_split:
2092  * @regex: a #GRegex structure
2093  * @string: the string to split with the pattern
2094  * @match_options: match time option flags
2095  *
2096  * Breaks the string on the pattern, and returns an array of the tokens.
2097  * If the pattern contains capturing parentheses, then the text for each
2098  * of the substrings will also be returned. If the pattern does not match
2099  * anywhere in the string, then the whole string is returned as the first
2100  * token.
2101  *
2102  * As a special case, the result of splitting the empty string "" is an
2103  * empty vector, not a vector containing a single string. The reason for
2104  * this special case is that being able to represent a empty vector is
2105  * typically more useful than consistent handling of empty elements. If
2106  * you do need to represent empty elements, you'll need to check for the
2107  * empty string before calling this function.
2108  *
2109  * A pattern that can match empty strings splits @string into separate
2110  * characters wherever it matches the empty string between characters.
2111  * For example splitting "ab c" using as a separator "\s*", you will get
2112  * "a", "b" and "c".
2113  *
2114  * Returns: (transfer full): a %NULL-terminated gchar ** array. Free
2115  * it using g_strfreev()
2116  *
2117  * Since: 2.14
2118  **/
2119 gchar **
2120 g_regex_split (const GRegex     *regex,
2121                const gchar      *string,
2122                GRegexMatchFlags  match_options)
2123 {
2124   return g_regex_split_full (regex, string, -1, 0,
2125                              match_options, 0, NULL);
2126 }
2127
2128 /**
2129  * g_regex_split_full:
2130  * @regex: a #GRegex structure
2131  * @string: (array length=string_len): the string to split with the pattern
2132  * @string_len: the length of @string, in bytes, or -1 if @string is nul-terminated
2133  * @start_position: starting index of the string to match, in bytes
2134  * @match_options: match time option flags
2135  * @max_tokens: the maximum number of tokens to split @string into.
2136  *   If this is less than 1, the string is split completely
2137  * @error: return location for a #GError
2138  *
2139  * Breaks the string on the pattern, and returns an array of the tokens.
2140  * If the pattern contains capturing parentheses, then the text for each
2141  * of the substrings will also be returned. If the pattern does not match
2142  * anywhere in the string, then the whole string is returned as the first
2143  * token.
2144  *
2145  * As a special case, the result of splitting the empty string "" is an
2146  * empty vector, not a vector containing a single string. The reason for
2147  * this special case is that being able to represent a empty vector is
2148  * typically more useful than consistent handling of empty elements. If
2149  * you do need to represent empty elements, you'll need to check for the
2150  * empty string before calling this function.
2151  *
2152  * A pattern that can match empty strings splits @string into separate
2153  * characters wherever it matches the empty string between characters.
2154  * For example splitting "ab c" using as a separator "\s*", you will get
2155  * "a", "b" and "c".
2156  *
2157  * Setting @start_position differs from just passing over a shortened
2158  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern
2159  * that begins with any kind of lookbehind assertion, such as "\b".
2160  *
2161  * Returns: (transfer full): a %NULL-terminated gchar ** array. Free
2162  * it using g_strfreev()
2163  *
2164  * Since: 2.14
2165  **/
2166 gchar **
2167 g_regex_split_full (const GRegex      *regex,
2168                     const gchar       *string,
2169                     gssize             string_len,
2170                     gint               start_position,
2171                     GRegexMatchFlags   match_options,
2172                     gint               max_tokens,
2173                     GError           **error)
2174 {
2175   GError *tmp_error = NULL;
2176   GMatchInfo *match_info;
2177   GList *list, *last;
2178   gint i;
2179   gint token_count;
2180   gboolean match_ok;
2181   /* position of the last separator. */
2182   gint last_separator_end;
2183   /* was the last match 0 bytes long? */
2184   gboolean last_match_is_empty;
2185   /* the returned array of char **s */
2186   gchar **string_list;
2187
2188   g_return_val_if_fail (regex != NULL, NULL);
2189   g_return_val_if_fail (string != NULL, NULL);
2190   g_return_val_if_fail (start_position >= 0, NULL);
2191   g_return_val_if_fail (error == NULL || *error == NULL, NULL);
2192   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
2193
2194   if (max_tokens <= 0)
2195     max_tokens = G_MAXINT;
2196
2197   if (string_len < 0)
2198     string_len = strlen (string);
2199
2200   /* zero-length string */
2201   if (string_len - start_position == 0)
2202     return g_new0 (gchar *, 1);
2203
2204   if (max_tokens == 1)
2205     {
2206       string_list = g_new0 (gchar *, 2);
2207       string_list[0] = g_strndup (&string[start_position],
2208                                   string_len - start_position);
2209       return string_list;
2210     }
2211
2212   list = NULL;
2213   token_count = 0;
2214   last_separator_end = start_position;
2215   last_match_is_empty = FALSE;
2216
2217   match_ok = g_regex_match_full (regex, string, string_len, start_position,
2218                                  match_options, &match_info, &tmp_error);
2219
2220   while (tmp_error == NULL)
2221     {
2222       if (match_ok)
2223         {
2224           last_match_is_empty =
2225                     (match_info->offsets[0] == match_info->offsets[1]);
2226
2227           /* we need to skip empty separators at the same position of the end
2228            * of another separator. e.g. the string is "a b" and the separator
2229            * is " *", so from 1 to 2 we have a match and at position 2 we have
2230            * an empty match. */
2231           if (last_separator_end != match_info->offsets[1])
2232             {
2233               gchar *token;
2234               gint match_count;
2235
2236               token = g_strndup (string + last_separator_end,
2237                                  match_info->offsets[0] - last_separator_end);
2238               list = g_list_prepend (list, token);
2239               token_count++;
2240
2241               /* if there were substrings, these need to be added to
2242                * the list. */
2243               match_count = g_match_info_get_match_count (match_info);
2244               if (match_count > 1)
2245                 {
2246                   for (i = 1; i < match_count; i++)
2247                     list = g_list_prepend (list, g_match_info_fetch (match_info, i));
2248                 }
2249             }
2250         }
2251       else
2252         {
2253           /* if there was no match, copy to end of string. */
2254           if (!last_match_is_empty)
2255             {
2256               gchar *token = g_strndup (string + last_separator_end,
2257                                         match_info->string_len - last_separator_end);
2258               list = g_list_prepend (list, token);
2259             }
2260           /* no more tokens, end the loop. */
2261           break;
2262         }
2263
2264       /* -1 to leave room for the last part. */
2265       if (token_count >= max_tokens - 1)
2266         {
2267           /* we have reached the maximum number of tokens, so we copy
2268            * the remaining part of the string. */
2269           if (last_match_is_empty)
2270             {
2271               /* the last match was empty, so we have moved one char
2272                * after the real position to avoid empty matches at the
2273                * same position. */
2274               match_info->pos = PREV_CHAR (regex, &string[match_info->pos]) - string;
2275             }
2276           /* the if is needed in the case we have terminated the available
2277            * tokens, but we are at the end of the string, so there are no
2278            * characters left to copy. */
2279           if (string_len > match_info->pos)
2280             {
2281               gchar *token = g_strndup (string + match_info->pos,
2282                                         string_len - match_info->pos);
2283               list = g_list_prepend (list, token);
2284             }
2285           /* end the loop. */
2286           break;
2287         }
2288
2289       last_separator_end = match_info->pos;
2290       if (last_match_is_empty)
2291         /* if the last match was empty, g_match_info_next() has moved
2292          * forward to avoid infinite loops, but we still need to copy that
2293          * character. */
2294         last_separator_end = PREV_CHAR (regex, &string[last_separator_end]) - string;
2295
2296       match_ok = g_match_info_next (match_info, &tmp_error);
2297     }
2298   g_match_info_free (match_info);
2299   if (tmp_error != NULL)
2300     {
2301       g_propagate_error (error, tmp_error);
2302       g_list_free_full (list, g_free);
2303       return NULL;
2304     }
2305
2306   string_list = g_new (gchar *, g_list_length (list) + 1);
2307   i = 0;
2308   for (last = g_list_last (list); last; last = g_list_previous (last))
2309     string_list[i++] = last->data;
2310   string_list[i] = NULL;
2311   g_list_free (list);
2312
2313   return string_list;
2314 }
2315
2316 enum
2317 {
2318   REPL_TYPE_STRING,
2319   REPL_TYPE_CHARACTER,
2320   REPL_TYPE_SYMBOLIC_REFERENCE,
2321   REPL_TYPE_NUMERIC_REFERENCE,
2322   REPL_TYPE_CHANGE_CASE
2323 };
2324
2325 typedef enum
2326 {
2327   CHANGE_CASE_NONE         = 1 << 0,
2328   CHANGE_CASE_UPPER        = 1 << 1,
2329   CHANGE_CASE_LOWER        = 1 << 2,
2330   CHANGE_CASE_UPPER_SINGLE = 1 << 3,
2331   CHANGE_CASE_LOWER_SINGLE = 1 << 4,
2332   CHANGE_CASE_SINGLE_MASK  = CHANGE_CASE_UPPER_SINGLE | CHANGE_CASE_LOWER_SINGLE,
2333   CHANGE_CASE_LOWER_MASK   = CHANGE_CASE_LOWER | CHANGE_CASE_LOWER_SINGLE,
2334   CHANGE_CASE_UPPER_MASK   = CHANGE_CASE_UPPER | CHANGE_CASE_UPPER_SINGLE
2335 } ChangeCase;
2336
2337 struct _InterpolationData
2338 {
2339   gchar     *text;
2340   gint       type;
2341   gint       num;
2342   gchar      c;
2343   ChangeCase change_case;
2344 };
2345
2346 static void
2347 free_interpolation_data (InterpolationData *data)
2348 {
2349   g_free (data->text);
2350   g_free (data);
2351 }
2352
2353 static const gchar *
2354 expand_escape (const gchar        *replacement,
2355                const gchar        *p,
2356                InterpolationData  *data,
2357                GError            **error)
2358 {
2359   const gchar *q, *r;
2360   gint x, d, h, i;
2361   const gchar *error_detail;
2362   gint base = 0;
2363   GError *tmp_error = NULL;
2364
2365   p++;
2366   switch (*p)
2367     {
2368     case 't':
2369       p++;
2370       data->c = '\t';
2371       data->type = REPL_TYPE_CHARACTER;
2372       break;
2373     case 'n':
2374       p++;
2375       data->c = '\n';
2376       data->type = REPL_TYPE_CHARACTER;
2377       break;
2378     case 'v':
2379       p++;
2380       data->c = '\v';
2381       data->type = REPL_TYPE_CHARACTER;
2382       break;
2383     case 'r':
2384       p++;
2385       data->c = '\r';
2386       data->type = REPL_TYPE_CHARACTER;
2387       break;
2388     case 'f':
2389       p++;
2390       data->c = '\f';
2391       data->type = REPL_TYPE_CHARACTER;
2392       break;
2393     case 'a':
2394       p++;
2395       data->c = '\a';
2396       data->type = REPL_TYPE_CHARACTER;
2397       break;
2398     case 'b':
2399       p++;
2400       data->c = '\b';
2401       data->type = REPL_TYPE_CHARACTER;
2402       break;
2403     case '\\':
2404       p++;
2405       data->c = '\\';
2406       data->type = REPL_TYPE_CHARACTER;
2407       break;
2408     case 'x':
2409       p++;
2410       x = 0;
2411       if (*p == '{')
2412         {
2413           p++;
2414           do
2415             {
2416               h = g_ascii_xdigit_value (*p);
2417               if (h < 0)
2418                 {
2419                   error_detail = _("hexadecimal digit or “}” expected");
2420                   goto error;
2421                 }
2422               x = x * 16 + h;
2423               p++;
2424             }
2425           while (*p != '}');
2426           p++;
2427         }
2428       else
2429         {
2430           for (i = 0; i < 2; i++)
2431             {
2432               h = g_ascii_xdigit_value (*p);
2433               if (h < 0)
2434                 {
2435                   error_detail = _("hexadecimal digit expected");
2436                   goto error;
2437                 }
2438               x = x * 16 + h;
2439               p++;
2440             }
2441         }
2442       data->type = REPL_TYPE_STRING;
2443       data->text = g_new0 (gchar, 8);
2444       g_unichar_to_utf8 (x, data->text);
2445       break;
2446     case 'l':
2447       p++;
2448       data->type = REPL_TYPE_CHANGE_CASE;
2449       data->change_case = CHANGE_CASE_LOWER_SINGLE;
2450       break;
2451     case 'u':
2452       p++;
2453       data->type = REPL_TYPE_CHANGE_CASE;
2454       data->change_case = CHANGE_CASE_UPPER_SINGLE;
2455       break;
2456     case 'L':
2457       p++;
2458       data->type = REPL_TYPE_CHANGE_CASE;
2459       data->change_case = CHANGE_CASE_LOWER;
2460       break;
2461     case 'U':
2462       p++;
2463       data->type = REPL_TYPE_CHANGE_CASE;
2464       data->change_case = CHANGE_CASE_UPPER;
2465       break;
2466     case 'E':
2467       p++;
2468       data->type = REPL_TYPE_CHANGE_CASE;
2469       data->change_case = CHANGE_CASE_NONE;
2470       break;
2471     case 'g':
2472       p++;
2473       if (*p != '<')
2474         {
2475           error_detail = _("missing “<” in symbolic reference");
2476           goto error;
2477         }
2478       q = p + 1;
2479       do
2480         {
2481           p++;
2482           if (!*p)
2483             {
2484               error_detail = _("unfinished symbolic reference");
2485               goto error;
2486             }
2487         }
2488       while (*p != '>');
2489       if (p - q == 0)
2490         {
2491           error_detail = _("zero-length symbolic reference");
2492           goto error;
2493         }
2494       if (g_ascii_isdigit (*q))
2495         {
2496           x = 0;
2497           do
2498             {
2499               h = g_ascii_digit_value (*q);
2500               if (h < 0)
2501                 {
2502                   error_detail = _("digit expected");
2503                   p = q;
2504                   goto error;
2505                 }
2506               x = x * 10 + h;
2507               q++;
2508             }
2509           while (q != p);
2510           data->num = x;
2511           data->type = REPL_TYPE_NUMERIC_REFERENCE;
2512         }
2513       else
2514         {
2515           r = q;
2516           do
2517             {
2518               if (!g_ascii_isalnum (*r))
2519                 {
2520                   error_detail = _("illegal symbolic reference");
2521                   p = r;
2522                   goto error;
2523                 }
2524               r++;
2525             }
2526           while (r != p);
2527           data->text = g_strndup (q, p - q);
2528           data->type = REPL_TYPE_SYMBOLIC_REFERENCE;
2529         }
2530       p++;
2531       break;
2532     case '0':
2533       /* if \0 is followed by a number is an octal number representing a
2534        * character, else it is a numeric reference. */
2535       if (g_ascii_digit_value (*g_utf8_next_char (p)) >= 0)
2536         {
2537           base = 8;
2538           p = g_utf8_next_char (p);
2539         }
2540     case '1':
2541     case '2':
2542     case '3':
2543     case '4':
2544     case '5':
2545     case '6':
2546     case '7':
2547     case '8':
2548     case '9':
2549       x = 0;
2550       d = 0;
2551       for (i = 0; i < 3; i++)
2552         {
2553           h = g_ascii_digit_value (*p);
2554           if (h < 0)
2555             break;
2556           if (h > 7)
2557             {
2558               if (base == 8)
2559                 break;
2560               else
2561                 base = 10;
2562             }
2563           if (i == 2 && base == 10)
2564             break;
2565           x = x * 8 + h;
2566           d = d * 10 + h;
2567           p++;
2568         }
2569       if (base == 8 || i == 3)
2570         {
2571           data->type = REPL_TYPE_STRING;
2572           data->text = g_new0 (gchar, 8);
2573           g_unichar_to_utf8 (x, data->text);
2574         }
2575       else
2576         {
2577           data->type = REPL_TYPE_NUMERIC_REFERENCE;
2578           data->num = d;
2579         }
2580       break;
2581     case 0:
2582       error_detail = _("stray final “\\”");
2583       goto error;
2584       break;
2585     default:
2586       error_detail = _("unknown escape sequence");
2587       goto error;
2588     }
2589
2590   return p;
2591
2592  error:
2593   /* G_GSSIZE_FORMAT doesn't work with gettext, so we use %lu */
2594   tmp_error = g_error_new (G_REGEX_ERROR,
2595                            G_REGEX_ERROR_REPLACE,
2596                            _("Error while parsing replacement "
2597                              "text “%s” at char %lu: %s"),
2598                            replacement,
2599                            (gulong)(p - replacement),
2600                            error_detail);
2601   g_propagate_error (error, tmp_error);
2602
2603   return NULL;
2604 }
2605
2606 static GList *
2607 split_replacement (const gchar  *replacement,
2608                    GError      **error)
2609 {
2610   GList *list = NULL;
2611   InterpolationData *data;
2612   const gchar *p, *start;
2613
2614   start = p = replacement;
2615   while (*p)
2616     {
2617       if (*p == '\\')
2618         {
2619           data = g_new0 (InterpolationData, 1);
2620           start = p = expand_escape (replacement, p, data, error);
2621           if (p == NULL)
2622             {
2623               g_list_free_full (list, (GDestroyNotify) free_interpolation_data);
2624               free_interpolation_data (data);
2625
2626               return NULL;
2627             }
2628           list = g_list_prepend (list, data);
2629         }
2630       else
2631         {
2632           p++;
2633           if (*p == '\\' || *p == '\0')
2634             {
2635               if (p - start > 0)
2636                 {
2637                   data = g_new0 (InterpolationData, 1);
2638                   data->text = g_strndup (start, p - start);
2639                   data->type = REPL_TYPE_STRING;
2640                   list = g_list_prepend (list, data);
2641                 }
2642             }
2643         }
2644     }
2645
2646   return g_list_reverse (list);
2647 }
2648
2649 /* Change the case of c based on change_case. */
2650 #define CHANGE_CASE(c, change_case) \
2651         (((change_case) & CHANGE_CASE_LOWER_MASK) ? \
2652                 g_unichar_tolower (c) : \
2653                 g_unichar_toupper (c))
2654
2655 static void
2656 string_append (GString     *string,
2657                const gchar *text,
2658                ChangeCase  *change_case)
2659 {
2660   gunichar c;
2661
2662   if (text[0] == '\0')
2663     return;
2664
2665   if (*change_case == CHANGE_CASE_NONE)
2666     {
2667       g_string_append (string, text);
2668     }
2669   else if (*change_case & CHANGE_CASE_SINGLE_MASK)
2670     {
2671       c = g_utf8_get_char (text);
2672       g_string_append_unichar (string, CHANGE_CASE (c, *change_case));
2673       g_string_append (string, g_utf8_next_char (text));
2674       *change_case = CHANGE_CASE_NONE;
2675     }
2676   else
2677     {
2678       while (*text != '\0')
2679         {
2680           c = g_utf8_get_char (text);
2681           g_string_append_unichar (string, CHANGE_CASE (c, *change_case));
2682           text = g_utf8_next_char (text);
2683         }
2684     }
2685 }
2686
2687 static gboolean
2688 interpolate_replacement (const GMatchInfo *match_info,
2689                          GString          *result,
2690                          gpointer          data)
2691 {
2692   GList *list;
2693   InterpolationData *idata;
2694   gchar *match;
2695   ChangeCase change_case = CHANGE_CASE_NONE;
2696
2697   for (list = data; list; list = list->next)
2698     {
2699       idata = list->data;
2700       switch (idata->type)
2701         {
2702         case REPL_TYPE_STRING:
2703           string_append (result, idata->text, &change_case);
2704           break;
2705         case REPL_TYPE_CHARACTER:
2706           g_string_append_c (result, CHANGE_CASE (idata->c, change_case));
2707           if (change_case & CHANGE_CASE_SINGLE_MASK)
2708             change_case = CHANGE_CASE_NONE;
2709           break;
2710         case REPL_TYPE_NUMERIC_REFERENCE:
2711           match = g_match_info_fetch (match_info, idata->num);
2712           if (match)
2713             {
2714               string_append (result, match, &change_case);
2715               g_free (match);
2716             }
2717           break;
2718         case REPL_TYPE_SYMBOLIC_REFERENCE:
2719           match = g_match_info_fetch_named (match_info, idata->text);
2720           if (match)
2721             {
2722               string_append (result, match, &change_case);
2723               g_free (match);
2724             }
2725           break;
2726         case REPL_TYPE_CHANGE_CASE:
2727           change_case = idata->change_case;
2728           break;
2729         }
2730     }
2731
2732   return FALSE;
2733 }
2734
2735 /* whether actual match_info is needed for replacement, i.e.
2736  * whether there are references
2737  */
2738 static gboolean
2739 interpolation_list_needs_match (GList *list)
2740 {
2741   while (list != NULL)
2742     {
2743       InterpolationData *data = list->data;
2744
2745       if (data->type == REPL_TYPE_SYMBOLIC_REFERENCE ||
2746           data->type == REPL_TYPE_NUMERIC_REFERENCE)
2747         {
2748           return TRUE;
2749         }
2750
2751       list = list->next;
2752     }
2753
2754   return FALSE;
2755 }
2756
2757 /**
2758  * g_regex_replace:
2759  * @regex: a #GRegex structure
2760  * @string: (array length=string_len): the string to perform matches against
2761  * @string_len: the length of @string, in bytes, or -1 if @string is nul-terminated
2762  * @start_position: starting index of the string to match, in bytes
2763  * @replacement: text to replace each match with
2764  * @match_options: options for the match
2765  * @error: location to store the error occurring, or %NULL to ignore errors
2766  *
2767  * Replaces all occurrences of the pattern in @regex with the
2768  * replacement text. Backreferences of the form '\number' or
2769  * '\g<number>' in the replacement text are interpolated by the
2770  * number-th captured subexpression of the match, '\g<name>' refers
2771  * to the captured subexpression with the given name. '\0' refers
2772  * to the complete match, but '\0' followed by a number is the octal
2773  * representation of a character. To include a literal '\' in the
2774  * replacement, write '\\\\'.
2775  *
2776  * There are also escapes that changes the case of the following text:
2777  *
2778  * - \l: Convert to lower case the next character
2779  * - \u: Convert to upper case the next character
2780  * - \L: Convert to lower case till \E
2781  * - \U: Convert to upper case till \E
2782  * - \E: End case modification
2783  *
2784  * If you do not need to use backreferences use g_regex_replace_literal().
2785  *
2786  * The @replacement string must be UTF-8 encoded even if #G_REGEX_RAW was
2787  * passed to g_regex_new(). If you want to use not UTF-8 encoded stings
2788  * you can use g_regex_replace_literal().
2789  *
2790  * Setting @start_position differs from just passing over a shortened
2791  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern that
2792  * begins with any kind of lookbehind assertion, such as "\b".
2793  *
2794  * Returns: a newly allocated string containing the replacements
2795  *
2796  * Since: 2.14
2797  */
2798 gchar *
2799 g_regex_replace (const GRegex      *regex,
2800                  const gchar       *string,
2801                  gssize             string_len,
2802                  gint               start_position,
2803                  const gchar       *replacement,
2804                  GRegexMatchFlags   match_options,
2805                  GError           **error)
2806 {
2807   gchar *result;
2808   GList *list;
2809   GError *tmp_error = NULL;
2810
2811   g_return_val_if_fail (regex != NULL, NULL);
2812   g_return_val_if_fail (string != NULL, NULL);
2813   g_return_val_if_fail (start_position >= 0, NULL);
2814   g_return_val_if_fail (replacement != NULL, NULL);
2815   g_return_val_if_fail (error == NULL || *error == NULL, NULL);
2816   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
2817
2818   list = split_replacement (replacement, &tmp_error);
2819   if (tmp_error != NULL)
2820     {
2821       g_propagate_error (error, tmp_error);
2822       return NULL;
2823     }
2824
2825   result = g_regex_replace_eval (regex,
2826                                  string, string_len, start_position,
2827                                  match_options,
2828                                  interpolate_replacement,
2829                                  (gpointer)list,
2830                                  &tmp_error);
2831   if (tmp_error != NULL)
2832     g_propagate_error (error, tmp_error);
2833
2834   g_list_free_full (list, (GDestroyNotify) free_interpolation_data);
2835
2836   return result;
2837 }
2838
2839 static gboolean
2840 literal_replacement (const GMatchInfo *match_info,
2841                      GString          *result,
2842                      gpointer          data)
2843 {
2844   g_string_append (result, data);
2845   return FALSE;
2846 }
2847
2848 /**
2849  * g_regex_replace_literal:
2850  * @regex: a #GRegex structure
2851  * @string: (array length=string_len): the string to perform matches against
2852  * @string_len: the length of @string, in bytes, or -1 if @string is nul-terminated
2853  * @start_position: starting index of the string to match, in bytes
2854  * @replacement: text to replace each match with
2855  * @match_options: options for the match
2856  * @error: location to store the error occurring, or %NULL to ignore errors
2857  *
2858  * Replaces all occurrences of the pattern in @regex with the
2859  * replacement text. @replacement is replaced literally, to
2860  * include backreferences use g_regex_replace().
2861  *
2862  * Setting @start_position differs from just passing over a
2863  * shortened string and setting #G_REGEX_MATCH_NOTBOL in the
2864  * case of a pattern that begins with any kind of lookbehind
2865  * assertion, such as "\b".
2866  *
2867  * Returns: a newly allocated string containing the replacements
2868  *
2869  * Since: 2.14
2870  */
2871 gchar *
2872 g_regex_replace_literal (const GRegex      *regex,
2873                          const gchar       *string,
2874                          gssize             string_len,
2875                          gint               start_position,
2876                          const gchar       *replacement,
2877                          GRegexMatchFlags   match_options,
2878                          GError           **error)
2879 {
2880   g_return_val_if_fail (replacement != NULL, NULL);
2881   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
2882
2883   return g_regex_replace_eval (regex,
2884                                string, string_len, start_position,
2885                                match_options,
2886                                literal_replacement,
2887                                (gpointer)replacement,
2888                                error);
2889 }
2890
2891 /**
2892  * g_regex_replace_eval:
2893  * @regex: a #GRegex structure from g_regex_new()
2894  * @string: (array length=string_len): string to perform matches against
2895  * @string_len: the length of @string, in bytes, or -1 if @string is nul-terminated
2896  * @start_position: starting index of the string to match, in bytes
2897  * @match_options: options for the match
2898  * @eval: a function to call for each match
2899  * @user_data: user data to pass to the function
2900  * @error: location to store the error occurring, or %NULL to ignore errors
2901  *
2902  * Replaces occurrences of the pattern in regex with the output of
2903  * @eval for that occurrence.
2904  *
2905  * Setting @start_position differs from just passing over a shortened
2906  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern
2907  * that begins with any kind of lookbehind assertion, such as "\b".
2908  *
2909  * The following example uses g_regex_replace_eval() to replace multiple
2910  * strings at once:
2911  * |[<!-- language="C" -->
2912  * static gboolean
2913  * eval_cb (const GMatchInfo *info,
2914  *          GString          *res,
2915  *          gpointer          data)
2916  * {
2917  *   gchar *match;
2918  *   gchar *r;
2919  *
2920  *    match = g_match_info_fetch (info, 0);
2921  *    r = g_hash_table_lookup ((GHashTable *)data, match);
2922  *    g_string_append (res, r);
2923  *    g_free (match);
2924  *
2925  *    return FALSE;
2926  * }
2927  *
2928  * ...
2929  *
2930  * GRegex *reg;
2931  * GHashTable *h;
2932  * gchar *res;
2933  *
2934  * h = g_hash_table_new (g_str_hash, g_str_equal);
2935  *
2936  * g_hash_table_insert (h, "1", "ONE");
2937  * g_hash_table_insert (h, "2", "TWO");
2938  * g_hash_table_insert (h, "3", "THREE");
2939  * g_hash_table_insert (h, "4", "FOUR");
2940  *
2941  * reg = g_regex_new ("1|2|3|4", 0, 0, NULL);
2942  * res = g_regex_replace_eval (reg, text, -1, 0, 0, eval_cb, h, NULL);
2943  * g_hash_table_destroy (h);
2944  *
2945  * ...
2946  * ]|
2947  *
2948  * Returns: a newly allocated string containing the replacements
2949  *
2950  * Since: 2.14
2951  */
2952 gchar *
2953 g_regex_replace_eval (const GRegex        *regex,
2954                       const gchar         *string,
2955                       gssize               string_len,
2956                       gint                 start_position,
2957                       GRegexMatchFlags     match_options,
2958                       GRegexEvalCallback   eval,
2959                       gpointer             user_data,
2960                       GError             **error)
2961 {
2962   GMatchInfo *match_info;
2963   GString *result;
2964   gint str_pos = 0;
2965   gboolean done = FALSE;
2966   GError *tmp_error = NULL;
2967
2968   g_return_val_if_fail (regex != NULL, NULL);
2969   g_return_val_if_fail (string != NULL, NULL);
2970   g_return_val_if_fail (start_position >= 0, NULL);
2971   g_return_val_if_fail (eval != NULL, NULL);
2972   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
2973
2974   if (string_len < 0)
2975     string_len = strlen (string);
2976
2977   result = g_string_sized_new (string_len);
2978
2979   /* run down the string making matches. */
2980   g_regex_match_full (regex, string, string_len, start_position,
2981                       match_options, &match_info, &tmp_error);
2982   while (!done && g_match_info_matches (match_info))
2983     {
2984       g_string_append_len (result,
2985                            string + str_pos,
2986                            match_info->offsets[0] - str_pos);
2987       done = (*eval) (match_info, result, user_data);
2988       str_pos = match_info->offsets[1];
2989       g_match_info_next (match_info, &tmp_error);
2990     }
2991   g_match_info_free (match_info);
2992   if (tmp_error != NULL)
2993     {
2994       g_propagate_error (error, tmp_error);
2995       g_string_free (result, TRUE);
2996       return NULL;
2997     }
2998
2999   g_string_append_len (result, string + str_pos, string_len - str_pos);
3000   return g_string_free (result, FALSE);
3001 }
3002
3003 /**
3004  * g_regex_check_replacement:
3005  * @replacement: the replacement string
3006  * @has_references: (out) (optional): location to store information about
3007  *   references in @replacement or %NULL
3008  * @error: location to store error
3009  *
3010  * Checks whether @replacement is a valid replacement string
3011  * (see g_regex_replace()), i.e. that all escape sequences in
3012  * it are valid.
3013  *
3014  * If @has_references is not %NULL then @replacement is checked
3015  * for pattern references. For instance, replacement text 'foo\n'
3016  * does not contain references and may be evaluated without information
3017  * about actual match, but '\0\1' (whole match followed by first
3018  * subpattern) requires valid #GMatchInfo object.
3019  *
3020  * Returns: whether @replacement is a valid replacement string
3021  *
3022  * Since: 2.14
3023  */
3024 gboolean
3025 g_regex_check_replacement (const gchar  *replacement,
3026                            gboolean     *has_references,
3027                            GError      **error)
3028 {
3029   GList *list;
3030   GError *tmp = NULL;
3031
3032   list = split_replacement (replacement, &tmp);
3033
3034   if (tmp)
3035   {
3036     g_propagate_error (error, tmp);
3037     return FALSE;
3038   }
3039
3040   if (has_references)
3041     *has_references = interpolation_list_needs_match (list);
3042
3043   g_list_free_full (list, (GDestroyNotify) free_interpolation_data);
3044
3045   return TRUE;
3046 }
3047
3048 /**
3049  * g_regex_escape_nul:
3050  * @string: the string to escape
3051  * @length: the length of @string
3052  *
3053  * Escapes the nul characters in @string to "\x00".  It can be used
3054  * to compile a regex with embedded nul characters.
3055  *
3056  * For completeness, @length can be -1 for a nul-terminated string.
3057  * In this case the output string will be of course equal to @string.
3058  *
3059  * Returns: a newly-allocated escaped string
3060  *
3061  * Since: 2.30
3062  */
3063 gchar *
3064 g_regex_escape_nul (const gchar *string,
3065                     gint         length)
3066 {
3067   GString *escaped;
3068   const gchar *p, *piece_start, *end;
3069   gint backslashes;
3070
3071   g_return_val_if_fail (string != NULL, NULL);
3072
3073   if (length < 0)
3074     return g_strdup (string);
3075
3076   end = string + length;
3077   p = piece_start = string;
3078   escaped = g_string_sized_new (length + 1);
3079
3080   backslashes = 0;
3081   while (p < end)
3082     {
3083       switch (*p)
3084         {
3085         case '\0':
3086           if (p != piece_start)
3087             {
3088               /* copy the previous piece. */
3089               g_string_append_len (escaped, piece_start, p - piece_start);
3090             }
3091           if ((backslashes & 1) == 0)
3092             g_string_append_c (escaped, '\\');
3093           g_string_append_c (escaped, 'x');
3094           g_string_append_c (escaped, '0');
3095           g_string_append_c (escaped, '0');
3096           piece_start = ++p;
3097           backslashes = 0;
3098           break;
3099         case '\\':
3100           backslashes++;
3101           ++p;
3102           break;
3103         default:
3104           backslashes = 0;
3105           p = g_utf8_next_char (p);
3106           break;
3107         }
3108     }
3109
3110   if (piece_start < end)
3111     g_string_append_len (escaped, piece_start, end - piece_start);
3112
3113   return g_string_free (escaped, FALSE);
3114 }
3115
3116 /**
3117  * g_regex_escape_string:
3118  * @string: (array length=length): the string to escape
3119  * @length: the length of @string, in bytes, or -1 if @string is nul-terminated
3120  *
3121  * Escapes the special characters used for regular expressions
3122  * in @string, for instance "a.b*c" becomes "a\.b\*c". This
3123  * function is useful to dynamically generate regular expressions.
3124  *
3125  * @string can contain nul characters that are replaced with "\0",
3126  * in this case remember to specify the correct length of @string
3127  * in @length.
3128  *
3129  * Returns: a newly-allocated escaped string
3130  *
3131  * Since: 2.14
3132  */
3133 gchar *
3134 g_regex_escape_string (const gchar *string,
3135                        gint         length)
3136 {
3137   GString *escaped;
3138   const char *p, *piece_start, *end;
3139
3140   g_return_val_if_fail (string != NULL, NULL);
3141
3142   if (length < 0)
3143     length = strlen (string);
3144
3145   end = string + length;
3146   p = piece_start = string;
3147   escaped = g_string_sized_new (length + 1);
3148
3149   while (p < end)
3150     {
3151       switch (*p)
3152         {
3153         case '\0':
3154         case '\\':
3155         case '|':
3156         case '(':
3157         case ')':
3158         case '[':
3159         case ']':
3160         case '{':
3161         case '}':
3162         case '^':
3163         case '$':
3164         case '*':
3165         case '+':
3166         case '?':
3167         case '.':
3168           if (p != piece_start)
3169             /* copy the previous piece. */
3170             g_string_append_len (escaped, piece_start, p - piece_start);
3171           g_string_append_c (escaped, '\\');
3172           if (*p == '\0')
3173             g_string_append_c (escaped, '0');
3174           else
3175             g_string_append_c (escaped, *p);
3176           piece_start = ++p;
3177           break;
3178         default:
3179           p = g_utf8_next_char (p);
3180           break;
3181         }
3182   }
3183
3184   if (piece_start < end)
3185     g_string_append_len (escaped, piece_start, end - piece_start);
3186
3187   return g_string_free (escaped, FALSE);
3188 }