glib/gregex.c

   1 /* GRegex -- regular expression API wrapper around PCRE.
   2  *
   3  * Copyright (C) 1999, 2000 Scott Wimer
   4  * Copyright (C) 2004, Matthias Clasen <mclasen@redhat.com>
   5  * Copyright (C) 2005 - 2007, Marco Barisione <marco@barisione.org>
   6  *
   7  * This library is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2.1 of the License, or (at your option) any later version.
  11  *
  12  * This library is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU Lesser General Public
  18  * License along with this library; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
  20  */
  21
  22 #include "config.h"
  23
  24 #include <string.h>
  25
  26 #ifdef USE_SYSTEM_PCRE
  27 #include <pcre.h>
  28 #else
  29 #include "pcre/pcre.h"
  30 #endif
  31
  32 #include "gtypes.h"
  33 #include "gregex.h"
  34 #include "glibintl.h"
  35 #include "glist.h"
  36 #include "gmessages.h"
  37 #include "gstrfuncs.h"
  38 #include "gatomic.h"
  39 #include "gthread.h"
  40
  41 /**
  42  * SECTION:gregex
  43  * @title: Perl-compatible regular expressions
  44  * @short_description: matches strings against regular expressions
  45  * @see_also: <xref linkend="glib-regex-syntax"/>
  46  *
  47  * The <function>g_regex_*()</function> functions implement regular
  48  * expression pattern matching using syntax and semantics similar to
  49  * Perl regular expression.
  50  *
  51  * Some functions accept a @start_position argument, setting it differs
  52  * from just passing over a shortened string and setting #G_REGEX_MATCH_NOTBOL
  53  * in the case of a pattern that begins with any kind of lookbehind assertion.
  54  * For example, consider the pattern "\Biss\B" which finds occurrences of "iss"
  55  * in the middle of words. ("\B" matches only if the current position in the
  56  * subject is not a word boundary.) When applied to the string "Mississipi"
  57  * from the fourth byte, namely "issipi", it does not match, because "\B" is
  58  * always false at the start of the subject, which is deemed to be a word
  59  * boundary. However, if the entire string is passed , but with
  60  * @start_position set to 4, it finds the second occurrence of "iss" because
  61  * it is able to look behind the starting point to discover that it is
  62  * preceded by a letter.
  63  *
  64  * Note that, unless you set the #G_REGEX_RAW flag, all the strings passed
  65  * to these functions must be encoded in UTF-8. The lengths and the positions
  66  * inside the strings are in bytes and not in characters, so, for instance,
  67  * "\xc3\xa0" (i.e. "&agrave;") is two bytes long but it is treated as a
  68  * single character. If you set #G_REGEX_RAW the strings can be non-valid
  69  * UTF-8 strings and a byte is treated as a character, so "\xc3\xa0" is two
  70  * bytes and two characters long.
  71  *
  72  * When matching a pattern, "\n" matches only against a "\n" character in
  73  * the string, and "\r" matches only a "\r" character. To match any newline
  74  * sequence use "\R". This particular group matches either the two-character
  75  * sequence CR + LF ("\r\n"), or one of the single characters LF (linefeed,
  76  * U+000A, "\n"), VT vertical tab, U+000B, "\v"), FF (formfeed, U+000C, "\f"),
  77  * CR (carriage return, U+000D, "\r"), NEL (next line, U+0085), LS (line
  78  * separator, U+2028), or PS (paragraph separator, U+2029).
  79  *
  80  * The behaviour of the dot, circumflex, and dollar metacharacters are
  81  * affected by newline characters, the default is to recognize any newline
  82  * character (the same characters recognized by "\R"). This can be changed
  83  * with #G_REGEX_NEWLINE_CR, #G_REGEX_NEWLINE_LF and #G_REGEX_NEWLINE_CRLF
  84  * compile options, and with #G_REGEX_MATCH_NEWLINE_ANY,
  85  * #G_REGEX_MATCH_NEWLINE_CR, #G_REGEX_MATCH_NEWLINE_LF and
  86  * #G_REGEX_MATCH_NEWLINE_CRLF match options. These settings are also
  87  * relevant when compiling a pattern if #G_REGEX_EXTENDED is set, and an
  88  * unescaped "#" outside a character class is encountered. This indicates
  89  * a comment that lasts until after the next newline.
  90  *
  91  * When setting the %G_REGEX_JAVASCRIPT_COMPAT flag, pattern syntax and pattern
  92  * matching is changed to be compatible with the way that regular expressions
  93  * work in JavaScript. More precisely, a lonely ']' character in the pattern
  94  * is a syntax error; the '\x' escape only allows 0 to 2 hexadecimal digits, and
  95  * you must use the '\u' escape sequence with 4 hex digits to specify a unicode
  96  * codepoint instead of '\x' or 'x{....}'. If '\x' or '\u' are not followed by
  97  * the specified number of hex digits, they match 'x' and 'u' literally; also
  98  * '\U' always matches 'U' instead of being an error in the pattern. Finally,
  99  * pattern matching is modified so that back references to an unset subpattern
 100  * group produces a match with the empty string instead of an error. See
 101  * <ulink>man:pcreapi(3)</ulink> for more information.
 102  *
 103  * Creating and manipulating the same #GRegex structure from different
 104  * threads is not a problem as #GRegex does not modify its internal
 105  * state between creation and destruction, on the other hand #GMatchInfo
 106  * is not threadsafe.
 107  *
 108  * The regular expressions low-level functionalities are obtained through
 109  * the excellent <ulink url="http://www.pcre.org/">PCRE</ulink> library
 110  * 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 */
 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: (allow-none): 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 <ulink>man:pcrepartial</ulink> 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: (allow-none): 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: (allow-none): 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: (allow-none): 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) (allow-none): pointer to location where to store
1014  *     the start position, or %NULL
1015  * @end_pos: (out) (allow-none): 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&lt;X&gt;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: (allow-none): 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) (allow-none): pointer to location where to store
1139  *     the start position, or %NULL
1140  * @end_pos: (out) (allow-none): 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&lt;X&gt;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  * g_regex_new:
1271  * @pattern: the regular expression
1272  * @compile_options: compile options for the regular expression, or 0
1273  * @match_options: match options for the regular expression, or 0
1274  * @error: return location for a #GError
1275  *
1276  * Compiles the regular expression to an internal form, and does
1277  * the initial setup of the #GRegex structure.
1278  *
1279  * Returns: a #GRegex structure. Call g_regex_unref() when you
1280  *   are done with it
1281  *
1282  * Since: 2.14
1283  */
1284 GRegex *
1285 g_regex_new (const gchar         *pattern,
1286              GRegexCompileFlags   compile_options,
1287              GRegexMatchFlags     match_options,
1288              GError             **error)
1289 {
1290   GRegex *regex;
1291   pcre *re;
1292   const gchar *errmsg;
1293   gint erroffset;
1294   gint errcode;
1295   gboolean optimize = FALSE;
1296   static volatile gsize initialised = 0;
1297   unsigned long int pcre_compile_options;
1298   GRegexCompileFlags nonpcre_compile_options;
1299
1300   g_return_val_if_fail (pattern != NULL, NULL);
1301   g_return_val_if_fail (error == NULL || *error == NULL, NULL);
1302   g_return_val_if_fail ((compile_options & ~G_REGEX_COMPILE_MASK) == 0, NULL);
1303   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
1304
1305   if (g_once_init_enter (&initialised))
1306     {
1307       int supports_utf8, supports_ucp;
1308
1309       pcre_config (PCRE_CONFIG_UTF8, &supports_utf8);
1310       if (!supports_utf8)
1311         g_critical (_("PCRE library is compiled without UTF8 support"));
1312
1313       pcre_config (PCRE_CONFIG_UNICODE_PROPERTIES, &supports_ucp);
1314       if (!supports_ucp)
1315         g_critical (_("PCRE library is compiled without UTF8 properties support"));
1316
1317       g_once_init_leave (&initialised, supports_utf8 && supports_ucp ? 1 : 2);
1318     }
1319
1320   if (G_UNLIKELY (initialised != 1))
1321     {
1322       g_set_error_literal (error, G_REGEX_ERROR, G_REGEX_ERROR_COMPILE,
1323                            _("PCRE library is compiled with incompatible options"));
1324       return NULL;
1325     }
1326
1327   nonpcre_compile_options = compile_options & G_REGEX_COMPILE_NONPCRE_MASK;
1328
1329   /* G_REGEX_OPTIMIZE has the same numeric value of PCRE_NO_UTF8_CHECK,
1330    * as we do not need to wrap PCRE_NO_UTF8_CHECK. */
1331   if (compile_options & G_REGEX_OPTIMIZE)
1332     optimize = TRUE;
1333
1334   /* In GRegex the string are, by default, UTF-8 encoded. PCRE
1335    * instead uses UTF-8 only if required with PCRE_UTF8. */
1336   if (compile_options & G_REGEX_RAW)
1337     {
1338       /* disable utf-8 */
1339       compile_options &= ~G_REGEX_RAW;
1340     }
1341   else
1342     {
1343       /* enable utf-8 */
1344       compile_options |= PCRE_UTF8 | PCRE_NO_UTF8_CHECK;
1345       match_options |= PCRE_NO_UTF8_CHECK;
1346     }
1347
1348   /* PCRE_NEWLINE_ANY is the default for the internal PCRE but
1349    * not for the system one. */
1350   if (!(compile_options & G_REGEX_NEWLINE_CR) &&
1351       !(compile_options & G_REGEX_NEWLINE_LF))
1352     {
1353       compile_options |= PCRE_NEWLINE_ANY;
1354     }
1355
1356   compile_options |= PCRE_UCP;
1357
1358   /* PCRE_BSR_UNICODE is the default for the internal PCRE but
1359    * possibly not for the system one.
1360    */
1361   if (~compile_options & G_REGEX_BSR_ANYCRLF)
1362     compile_options |= PCRE_BSR_UNICODE;
1363
1364   /* compile the pattern */
1365   re = pcre_compile2 (pattern, compile_options, &errcode,
1366                       &errmsg, &erroffset, NULL);
1367
1368   /* if the compilation failed, set the error member and return
1369    * immediately */
1370   if (re == NULL)
1371     {
1372       GError *tmp_error;
1373
1374       /* Translate the PCRE error code to GRegexError and use a translated
1375        * error message if possible */
1376       translate_compile_error (&errcode, &errmsg);
1377
1378       /* PCRE uses byte offsets but we want to show character offsets */
1379       erroffset = g_utf8_pointer_to_offset (pattern, &pattern[erroffset]);
1380
1381       tmp_error = g_error_new (G_REGEX_ERROR, errcode,
1382                                _("Error while compiling regular "
1383                                  "expression %s at char %d: %s"),
1384                                pattern, erroffset, errmsg);
1385       g_propagate_error (error, tmp_error);
1386
1387       return NULL;
1388     }
1389
1390   /* For options set at the beginning of the pattern, pcre puts them into
1391    * compile options, e.g. "(?i)foo" will make the pcre structure store
1392    * PCRE_CASELESS even though it wasn't explicitly given for compilation. */
1393   pcre_fullinfo (re, NULL, PCRE_INFO_OPTIONS, &pcre_compile_options);
1394   compile_options = pcre_compile_options & G_REGEX_COMPILE_PCRE_MASK;
1395
1396   /* Don't leak PCRE_NEWLINE_ANY, which is part of PCRE_NEWLINE_ANYCRLF */
1397   if ((pcre_compile_options & PCRE_NEWLINE_ANYCRLF) != PCRE_NEWLINE_ANYCRLF)
1398     compile_options &= ~PCRE_NEWLINE_ANY;
1399
1400   compile_options |= nonpcre_compile_options;
1401
1402   if (!(compile_options & G_REGEX_DUPNAMES))
1403     {
1404       gboolean jchanged = FALSE;
1405       pcre_fullinfo (re, NULL, PCRE_INFO_JCHANGED, &jchanged);
1406       if (jchanged)
1407         compile_options |= G_REGEX_DUPNAMES;
1408     }
1409
1410   regex = g_new0 (GRegex, 1);
1411   regex->ref_count = 1;
1412   regex->pattern = g_strdup (pattern);
1413   regex->pcre_re = re;
1414   regex->compile_opts = compile_options;
1415   regex->match_opts = match_options;
1416
1417   if (optimize)
1418     {
1419       regex->extra = pcre_study (regex->pcre_re, 0, &errmsg);
1420       if (errmsg != NULL)
1421         {
1422           GError *tmp_error = g_error_new (G_REGEX_ERROR,
1423                                            G_REGEX_ERROR_OPTIMIZE,
1424                                            _("Error while optimizing "
1425                                              "regular expression %s: %s"),
1426                                            regex->pattern,
1427                                            errmsg);
1428           g_propagate_error (error, tmp_error);
1429
1430           g_regex_unref (regex);
1431           return NULL;
1432         }
1433     }
1434
1435   return regex;
1436 }
1437
1438 /**
1439  * g_regex_get_pattern:
1440  * @regex: a #GRegex structure
1441  *
1442  * Gets the pattern string associated with @regex, i.e. a copy of
1443  * the string passed to g_regex_new().
1444  *
1445  * Returns: the pattern of @regex
1446  *
1447  * Since: 2.14
1448  */
1449 const gchar *
1450 g_regex_get_pattern (const GRegex *regex)
1451 {
1452   g_return_val_if_fail (regex != NULL, NULL);
1453
1454   return regex->pattern;
1455 }
1456
1457 /**
1458  * g_regex_get_max_backref:
1459  * @regex: a #GRegex
1460  *
1461  * Returns the number of the highest back reference
1462  * in the pattern, or 0 if the pattern does not contain
1463  * back references.
1464  *
1465  * Returns: the number of the highest back reference
1466  *
1467  * Since: 2.14
1468  */
1469 gint
1470 g_regex_get_max_backref (const GRegex *regex)
1471 {
1472   gint value;
1473
1474   pcre_fullinfo (regex->pcre_re, regex->extra,
1475                  PCRE_INFO_BACKREFMAX, &value);
1476
1477   return value;
1478 }
1479
1480 /**
1481  * g_regex_get_capture_count:
1482  * @regex: a #GRegex
1483  *
1484  * Returns the number of capturing subpatterns in the pattern.
1485  *
1486  * Returns: the number of capturing subpatterns
1487  *
1488  * Since: 2.14
1489  */
1490 gint
1491 g_regex_get_capture_count (const GRegex *regex)
1492 {
1493   gint value;
1494
1495   pcre_fullinfo (regex->pcre_re, regex->extra,
1496                  PCRE_INFO_CAPTURECOUNT, &value);
1497
1498   return value;
1499 }
1500
1501 /**
1502  * g_regex_get_has_cr_or_lf:
1503  * @regex: a #GRegex structure
1504  *
1505  * Checks whether the pattern contains explicit CR or LF references.
1506  *
1507  * Returns: %TRUE if the pattern contains explicit CR or LF references
1508  *
1509  * Since: 2.34
1510  */
1511 gboolean
1512 g_regex_get_has_cr_or_lf (const GRegex *regex)
1513 {
1514   gint value;
1515
1516   pcre_fullinfo (regex->pcre_re, regex->extra,
1517                  PCRE_INFO_HASCRORLF, &value);
1518
1519   return !!value;
1520 }
1521
1522 /**
1523  * g_regex_get_max_lookbehind:
1524  * @regex: a #GRegex structure
1525  *
1526  * Gets the number of characters in the longest lookbehind assertion in the
1527  * pattern. This information is useful when doing multi-segment matching using
1528  * the partial matching facilities.
1529  *
1530  * Returns: the number of characters in the longest lookbehind assertion.
1531  *
1532  * Since: 2.38
1533  */
1534 gint
1535 g_regex_get_max_lookbehind (const GRegex *regex)
1536 {
1537   gint max_lookbehind;
1538
1539   pcre_fullinfo (regex->pcre_re, regex->extra,
1540                  PCRE_INFO_MAXLOOKBEHIND, &max_lookbehind);
1541
1542   return max_lookbehind;
1543 }
1544
1545 /**
1546  * g_regex_get_compile_flags:
1547  * @regex: a #GRegex
1548  *
1549  * Returns the compile options that @regex was created with.
1550  *
1551  * Returns: flags from #GRegexCompileFlags
1552  *
1553  * Since: 2.26
1554  */
1555 GRegexCompileFlags
1556 g_regex_get_compile_flags (const GRegex *regex)
1557 {
1558   g_return_val_if_fail (regex != NULL, 0);
1559
1560   return regex->compile_opts;
1561 }
1562
1563 /**
1564  * g_regex_get_match_flags:
1565  * @regex: a #GRegex
1566  *
1567  * Returns the match options that @regex was created with.
1568  *
1569  * Returns: flags from #GRegexMatchFlags
1570  *
1571  * Since: 2.26
1572  */
1573 GRegexMatchFlags
1574 g_regex_get_match_flags (const GRegex *regex)
1575 {
1576   g_return_val_if_fail (regex != NULL, 0);
1577
1578   return regex->match_opts & G_REGEX_MATCH_MASK;
1579 }
1580
1581 /**
1582  * g_regex_match_simple:
1583  * @pattern: the regular expression
1584  * @string: the string to scan for matches
1585  * @compile_options: compile options for the regular expression, or 0
1586  * @match_options: match options, or 0
1587  *
1588  * Scans for a match in @string for @pattern.
1589  *
1590  * This function is equivalent to g_regex_match() but it does not
1591  * require to compile the pattern with g_regex_new(), avoiding some
1592  * lines of code when you need just to do a match without extracting
1593  * substrings, capture counts, and so on.
1594  *
1595  * If this function is to be called on the same @pattern more than
1596  * once, it's more efficient to compile the pattern once with
1597  * g_regex_new() and then use g_regex_match().
1598  *
1599  * Returns: %TRUE if the string matched, %FALSE otherwise
1600  *
1601  * Since: 2.14
1602  */
1603 gboolean
1604 g_regex_match_simple (const gchar        *pattern,
1605                       const gchar        *string,
1606                       GRegexCompileFlags  compile_options,
1607                       GRegexMatchFlags    match_options)
1608 {
1609   GRegex *regex;
1610   gboolean result;
1611
1612   regex = g_regex_new (pattern, compile_options, 0, NULL);
1613   if (!regex)
1614     return FALSE;
1615   result = g_regex_match_full (regex, string, -1, 0, match_options, NULL, NULL);
1616   g_regex_unref (regex);
1617   return result;
1618 }
1619
1620 /**
1621  * g_regex_match:
1622  * @regex: a #GRegex structure from g_regex_new()
1623  * @string: the string to scan for matches
1624  * @match_options: match options
1625  * @match_info: (out) (allow-none): pointer to location where to store
1626  *     the #GMatchInfo, or %NULL if you do not need it
1627  *
1628  * Scans for a match in string for the pattern in @regex.
1629  * The @match_options are combined with the match options specified
1630  * when the @regex structure was created, letting you have more
1631  * flexibility in reusing #GRegex structures.
1632  *
1633  * A #GMatchInfo structure, used to get information on the match,
1634  * is stored in @match_info if not %NULL. Note that if @match_info
1635  * is not %NULL then it is created even if the function returns %FALSE,
1636  * i.e. you must free it regardless if regular expression actually matched.
1637  *
1638  * To retrieve all the non-overlapping matches of the pattern in
1639  * string you can use g_match_info_next().
1640  *
1641  * |[
1642  * static void
1643  * print_uppercase_words (const gchar *string)
1644  * {
1645  *   /&ast; Print all uppercase-only words. &ast;/
1646  *   GRegex *regex;
1647  *   GMatchInfo *match_info;
1648  *   &nbsp;
1649  *   regex = g_regex_new ("[A-Z]+", 0, 0, NULL);
1650  *   g_regex_match (regex, string, 0, &amp;match_info);
1651  *   while (g_match_info_matches (match_info))
1652  *     {
1653  *       gchar *word = g_match_info_fetch (match_info, 0);
1654  *       g_print ("Found: %s\n", word);
1655  *       g_free (word);
1656  *       g_match_info_next (match_info, NULL);
1657  *     }
1658  *   g_match_info_free (match_info);
1659  *   g_regex_unref (regex);
1660  * }
1661  * ]|
1662  *
1663  * @string is not copied and is used in #GMatchInfo internally. If
1664  * you use any #GMatchInfo method (except g_match_info_free()) after
1665  * freeing or modifying @string then the behaviour is undefined.
1666  *
1667  * Returns: %TRUE is the string matched, %FALSE otherwise
1668  *
1669  * Since: 2.14
1670  */
1671 gboolean
1672 g_regex_match (const GRegex      *regex,
1673                const gchar       *string,
1674                GRegexMatchFlags   match_options,
1675                GMatchInfo       **match_info)
1676 {
1677   return g_regex_match_full (regex, string, -1, 0, match_options,
1678                              match_info, NULL);
1679 }
1680
1681 /**
1682  * g_regex_match_full:
1683  * @regex: a #GRegex structure from g_regex_new()
1684  * @string: (array length=string_len): the string to scan for matches
1685  * @string_len: the length of @string, or -1 if @string is nul-terminated
1686  * @start_position: starting index of the string to match
1687  * @match_options: match options
1688  * @match_info: (out) (allow-none): pointer to location where to store
1689  *     the #GMatchInfo, or %NULL if you do not need it
1690  * @error: location to store the error occurring, or %NULL to ignore errors
1691  *
1692  * Scans for a match in string for the pattern in @regex.
1693  * The @match_options are combined with the match options specified
1694  * when the @regex structure was created, letting you have more
1695  * flexibility in reusing #GRegex structures.
1696  *
1697  * Setting @start_position differs from just passing over a shortened
1698  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern
1699  * that begins with any kind of lookbehind assertion, such as "\b".
1700  *
1701  * A #GMatchInfo structure, used to get information on the match, is
1702  * stored in @match_info if not %NULL. Note that if @match_info is
1703  * not %NULL then it is created even if the function returns %FALSE,
1704  * i.e. you must free it regardless if regular expression actually
1705  * matched.
1706  *
1707  * @string is not copied and is used in #GMatchInfo internally. If
1708  * you use any #GMatchInfo method (except g_match_info_free()) after
1709  * freeing or modifying @string then the behaviour is undefined.
1710  *
1711  * To retrieve all the non-overlapping matches of the pattern in
1712  * string you can use g_match_info_next().
1713  *
1714  * |[
1715  * static void
1716  * print_uppercase_words (const gchar *string)
1717  * {
1718  *   /&ast; Print all uppercase-only words. &ast;/
1719  *   GRegex *regex;
1720  *   GMatchInfo *match_info;
1721  *   GError *error = NULL;
1722  *   &nbsp;
1723  *   regex = g_regex_new ("[A-Z]+", 0, 0, NULL);
1724  *   g_regex_match_full (regex, string, -1, 0, 0, &amp;match_info, &amp;error);
1725  *   while (g_match_info_matches (match_info))
1726  *     {
1727  *       gchar *word = g_match_info_fetch (match_info, 0);
1728  *       g_print ("Found: %s\n", word);
1729  *       g_free (word);
1730  *       g_match_info_next (match_info, &amp;error);
1731  *     }
1732  *   g_match_info_free (match_info);
1733  *   g_regex_unref (regex);
1734  *   if (error != NULL)
1735  *     {
1736  *       g_printerr ("Error while matching: %s\n", error->message);
1737  *       g_error_free (error);
1738  *     }
1739  * }
1740  * ]|
1741  *
1742  * Returns: %TRUE is the string matched, %FALSE otherwise
1743  *
1744  * Since: 2.14
1745  */
1746 gboolean
1747 g_regex_match_full (const GRegex      *regex,
1748                     const gchar       *string,
1749                     gssize             string_len,
1750                     gint               start_position,
1751                     GRegexMatchFlags   match_options,
1752                     GMatchInfo       **match_info,
1753                     GError           **error)
1754 {
1755   GMatchInfo *info;
1756   gboolean match_ok;
1757
1758   g_return_val_if_fail (regex != NULL, FALSE);
1759   g_return_val_if_fail (string != NULL, FALSE);
1760   g_return_val_if_fail (start_position >= 0, FALSE);
1761   g_return_val_if_fail (error == NULL || *error == NULL, FALSE);
1762   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, FALSE);
1763
1764   info = match_info_new (regex, string, string_len, start_position,
1765                          match_options, FALSE);
1766   match_ok = g_match_info_next (info, error);
1767   if (match_info != NULL)
1768     *match_info = info;
1769   else
1770     g_match_info_free (info);
1771
1772   return match_ok;
1773 }
1774
1775 /**
1776  * g_regex_match_all:
1777  * @regex: a #GRegex structure from g_regex_new()
1778  * @string: the string to scan for matches
1779  * @match_options: match options
1780  * @match_info: (out) (allow-none): pointer to location where to store
1781  *     the #GMatchInfo, or %NULL if you do not need it
1782  *
1783  * Using the standard algorithm for regular expression matching only
1784  * the longest match in the string is retrieved. This function uses
1785  * a different algorithm so it can retrieve all the possible matches.
1786  * For more documentation see g_regex_match_all_full().
1787  *
1788  * A #GMatchInfo structure, used to get information on the match, is
1789  * stored in @match_info if not %NULL. Note that if @match_info is
1790  * not %NULL then it is created even if the function returns %FALSE,
1791  * i.e. you must free it regardless if regular expression actually
1792  * matched.
1793  *
1794  * @string is not copied and is used in #GMatchInfo internally. If
1795  * you use any #GMatchInfo method (except g_match_info_free()) after
1796  * freeing or modifying @string then the behaviour is undefined.
1797  *
1798  * Returns: %TRUE is the string matched, %FALSE otherwise
1799  *
1800  * Since: 2.14
1801  */
1802 gboolean
1803 g_regex_match_all (const GRegex      *regex,
1804                    const gchar       *string,
1805                    GRegexMatchFlags   match_options,
1806                    GMatchInfo       **match_info)
1807 {
1808   return g_regex_match_all_full (regex, string, -1, 0, match_options,
1809                                  match_info, NULL);
1810 }
1811
1812 /**
1813  * g_regex_match_all_full:
1814  * @regex: a #GRegex structure from g_regex_new()
1815  * @string: (array length=string_len): the string to scan for matches
1816  * @string_len: the length of @string, or -1 if @string is nul-terminated
1817  * @start_position: starting index of the string to match
1818  * @match_options: match options
1819  * @match_info: (out) (allow-none): pointer to location where to store
1820  *     the #GMatchInfo, or %NULL if you do not need it
1821  * @error: location to store the error occurring, or %NULL to ignore errors
1822  *
1823  * Using the standard algorithm for regular expression matching only
1824  * the longest match in the string is retrieved, it is not possible
1825  * to obtain all the available matches. For instance matching
1826  * "&lt;a&gt; &lt;b&gt; &lt;c&gt;" against the pattern "&lt;.*&gt;"
1827  * you get "&lt;a&gt; &lt;b&gt; &lt;c&gt;".
1828  *
1829  * This function uses a different algorithm (called DFA, i.e. deterministic
1830  * finite automaton), so it can retrieve all the possible matches, all
1831  * starting at the same point in the string. For instance matching
1832  * "&lt;a&gt; &lt;b&gt; &lt;c&gt;" against the pattern "&lt;.*&gt;"
1833  * you would obtain three matches: "&lt;a&gt; &lt;b&gt; &lt;c&gt;",
1834  * "&lt;a&gt; &lt;b&gt;" and "&lt;a&gt;".
1835  *
1836  * The number of matched strings is retrieved using
1837  * g_match_info_get_match_count(). To obtain the matched strings and
1838  * their position you can use, respectively, g_match_info_fetch() and
1839  * g_match_info_fetch_pos(). Note that the strings are returned in
1840  * reverse order of length; that is, the longest matching string is
1841  * given first.
1842  *
1843  * Note that the DFA algorithm is slower than the standard one and it
1844  * is not able to capture substrings, so backreferences do not work.
1845  *
1846  * Setting @start_position differs from just passing over a shortened
1847  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern
1848  * that begins with any kind of lookbehind assertion, such as "\b".
1849  *
1850  * A #GMatchInfo structure, used to get information on the match, is
1851  * stored in @match_info if not %NULL. Note that if @match_info is
1852  * not %NULL then it is created even if the function returns %FALSE,
1853  * i.e. you must free it regardless if regular expression actually
1854  * matched.
1855  *
1856  * @string is not copied and is used in #GMatchInfo internally. If
1857  * you use any #GMatchInfo method (except g_match_info_free()) after
1858  * freeing or modifying @string then the behaviour is undefined.
1859  *
1860  * Returns: %TRUE is the string matched, %FALSE otherwise
1861  *
1862  * Since: 2.14
1863  */
1864 gboolean
1865 g_regex_match_all_full (const GRegex      *regex,
1866                         const gchar       *string,
1867                         gssize             string_len,
1868                         gint               start_position,
1869                         GRegexMatchFlags   match_options,
1870                         GMatchInfo       **match_info,
1871                         GError           **error)
1872 {
1873   GMatchInfo *info;
1874   gboolean done;
1875
1876   g_return_val_if_fail (regex != NULL, FALSE);
1877   g_return_val_if_fail (string != NULL, FALSE);
1878   g_return_val_if_fail (start_position >= 0, FALSE);
1879   g_return_val_if_fail (error == NULL || *error == NULL, FALSE);
1880   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, FALSE);
1881
1882   info = match_info_new (regex, string, string_len, start_position,
1883                          match_options, TRUE);
1884
1885   done = FALSE;
1886   while (!done)
1887     {
1888       done = TRUE;
1889       info->matches = pcre_dfa_exec (regex->pcre_re, regex->extra,
1890                                      info->string, info->string_len,
1891                                      info->pos,
1892                                      regex->match_opts | match_options,
1893                                      info->offsets, info->n_offsets,
1894                                      info->workspace, info->n_workspace);
1895       if (info->matches == PCRE_ERROR_DFA_WSSIZE)
1896         {
1897           /* info->workspace is too small. */
1898           info->n_workspace *= 2;
1899           info->workspace = g_realloc (info->workspace,
1900                                        info->n_workspace * sizeof (gint));
1901           done = FALSE;
1902         }
1903       else if (info->matches == 0)
1904         {
1905           /* info->offsets is too small. */
1906           info->n_offsets *= 2;
1907           info->offsets = g_realloc (info->offsets,
1908                                      info->n_offsets * sizeof (gint));
1909           done = FALSE;
1910         }
1911       else if (IS_PCRE_ERROR (info->matches))
1912         {
1913           g_set_error (error, G_REGEX_ERROR, G_REGEX_ERROR_MATCH,
1914                        _("Error while matching regular expression %s: %s"),
1915                        regex->pattern, match_error (info->matches));
1916         }
1917     }
1918
1919   /* set info->pos to -1 so that a call to g_match_info_next() fails. */
1920   info->pos = -1;
1921
1922   if (match_info != NULL)
1923     *match_info = info;
1924   else
1925     g_match_info_free (info);
1926
1927   return info->matches >= 0;
1928 }
1929
1930 /**
1931  * g_regex_get_string_number:
1932  * @regex: #GRegex structure
1933  * @name: name of the subexpression
1934  *
1935  * Retrieves the number of the subexpression named @name.
1936  *
1937  * Returns: The number of the subexpression or -1 if @name
1938  *   does not exists
1939  *
1940  * Since: 2.14
1941  */
1942 gint
1943 g_regex_get_string_number (const GRegex *regex,
1944                            const gchar  *name)
1945 {
1946   gint num;
1947
1948   g_return_val_if_fail (regex != NULL, -1);
1949   g_return_val_if_fail (name != NULL, -1);
1950
1951   num = pcre_get_stringnumber (regex->pcre_re, name);
1952   if (num == PCRE_ERROR_NOSUBSTRING)
1953     num = -1;
1954
1955   return num;
1956 }
1957
1958 /**
1959  * g_regex_split_simple:
1960  * @pattern: the regular expression
1961  * @string: the string to scan for matches
1962  * @compile_options: compile options for the regular expression, or 0
1963  * @match_options: match options, or 0
1964  *
1965  * Breaks the string on the pattern, and returns an array of
1966  * the tokens. If the pattern contains capturing parentheses,
1967  * then the text for each of the substrings will also be returned.
1968  * If the pattern does not match anywhere in the string, then the
1969  * whole string is returned as the first token.
1970  *
1971  * This function is equivalent to g_regex_split() but it does
1972  * not require to compile the pattern with g_regex_new(), avoiding
1973  * some lines of code when you need just to do a split without
1974  * extracting substrings, capture counts, and so on.
1975  *
1976  * If this function is to be called on the same @pattern more than
1977  * once, it's more efficient to compile the pattern once with
1978  * g_regex_new() and then use g_regex_split().
1979  *
1980  * As a special case, the result of splitting the empty string ""
1981  * is an empty vector, not a vector containing a single string.
1982  * The reason for this special case is that being able to represent
1983  * a empty vector is typically more useful than consistent handling
1984  * of empty elements. If you do need to represent empty elements,
1985  * you'll need to check for the empty string before calling this
1986  * function.
1987  *
1988  * A pattern that can match empty strings splits @string into
1989  * separate characters wherever it matches the empty string between
1990  * characters. For example splitting "ab c" using as a separator
1991  * "\s*", you will get "a", "b" and "c".
1992  *
1993  * Returns: (transfer full): a %NULL-terminated array of strings. Free
1994  * it using g_strfreev()
1995  *
1996  * Since: 2.14
1997  **/
1998 gchar **
1999 g_regex_split_simple (const gchar        *pattern,
2000                       const gchar        *string,
2001                       GRegexCompileFlags  compile_options,
2002                       GRegexMatchFlags    match_options)
2003 {
2004   GRegex *regex;
2005   gchar **result;
2006
2007   regex = g_regex_new (pattern, compile_options, 0, NULL);
2008   if (!regex)
2009     return NULL;
2010
2011   result = g_regex_split_full (regex, string, -1, 0, match_options, 0, NULL);
2012   g_regex_unref (regex);
2013   return result;
2014 }
2015
2016 /**
2017  * g_regex_split:
2018  * @regex: a #GRegex structure
2019  * @string: the string to split with the pattern
2020  * @match_options: match time option flags
2021  *
2022  * Breaks the string on the pattern, and returns an array of the tokens.
2023  * If the pattern contains capturing parentheses, then the text for each
2024  * of the substrings will also be returned. If the pattern does not match
2025  * anywhere in the string, then the whole string is returned as the first
2026  * token.
2027  *
2028  * As a special case, the result of splitting the empty string "" is an
2029  * empty vector, not a vector containing a single string. The reason for
2030  * this special case is that being able to represent a empty vector is
2031  * typically more useful than consistent handling of empty elements. If
2032  * you do need to represent empty elements, you'll need to check for the
2033  * empty string before calling this function.
2034  *
2035  * A pattern that can match empty strings splits @string into separate
2036  * characters wherever it matches the empty string between characters.
2037  * For example splitting "ab c" using as a separator "\s*", you will get
2038  * "a", "b" and "c".
2039  *
2040  * Returns: (transfer full): a %NULL-terminated gchar ** array. Free
2041  * it using g_strfreev()
2042  *
2043  * Since: 2.14
2044  **/
2045 gchar **
2046 g_regex_split (const GRegex     *regex,
2047                const gchar      *string,
2048                GRegexMatchFlags  match_options)
2049 {
2050   return g_regex_split_full (regex, string, -1, 0,
2051                              match_options, 0, NULL);
2052 }
2053
2054 /**
2055  * g_regex_split_full:
2056  * @regex: a #GRegex structure
2057  * @string: (array length=string_len): the string to split with the pattern
2058  * @string_len: the length of @string, or -1 if @string is nul-terminated
2059  * @start_position: starting index of the string to match
2060  * @match_options: match time option flags
2061  * @max_tokens: the maximum number of tokens to split @string into.
2062  *   If this is less than 1, the string is split completely
2063  * @error: return location for a #GError
2064  *
2065  * Breaks the string on the pattern, and returns an array of the tokens.
2066  * If the pattern contains capturing parentheses, then the text for each
2067  * of the substrings will also be returned. If the pattern does not match
2068  * anywhere in the string, then the whole string is returned as the first
2069  * token.
2070  *
2071  * As a special case, the result of splitting the empty string "" is an
2072  * empty vector, not a vector containing a single string. The reason for
2073  * this special case is that being able to represent a empty vector is
2074  * typically more useful than consistent handling of empty elements. If
2075  * you do need to represent empty elements, you'll need to check for the
2076  * empty string before calling this function.
2077  *
2078  * A pattern that can match empty strings splits @string into separate
2079  * characters wherever it matches the empty string between characters.
2080  * For example splitting "ab c" using as a separator "\s*", you will get
2081  * "a", "b" and "c".
2082  *
2083  * Setting @start_position differs from just passing over a shortened
2084  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern
2085  * that begins with any kind of lookbehind assertion, such as "\b".
2086  *
2087  * Returns: (transfer full): a %NULL-terminated gchar ** array. Free
2088  * it using g_strfreev()
2089  *
2090  * Since: 2.14
2091  **/
2092 gchar **
2093 g_regex_split_full (const GRegex      *regex,
2094                     const gchar       *string,
2095                     gssize             string_len,
2096                     gint               start_position,
2097                     GRegexMatchFlags   match_options,
2098                     gint               max_tokens,
2099                     GError           **error)
2100 {
2101   GError *tmp_error = NULL;
2102   GMatchInfo *match_info;
2103   GList *list, *last;
2104   gint i;
2105   gint token_count;
2106   gboolean match_ok;
2107   /* position of the last separator. */
2108   gint last_separator_end;
2109   /* was the last match 0 bytes long? */
2110   gboolean last_match_is_empty;
2111   /* the returned array of char **s */
2112   gchar **string_list;
2113
2114   g_return_val_if_fail (regex != NULL, NULL);
2115   g_return_val_if_fail (string != NULL, NULL);
2116   g_return_val_if_fail (start_position >= 0, NULL);
2117   g_return_val_if_fail (error == NULL || *error == NULL, NULL);
2118   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
2119
2120   if (max_tokens <= 0)
2121     max_tokens = G_MAXINT;
2122
2123   if (string_len < 0)
2124     string_len = strlen (string);
2125
2126   /* zero-length string */
2127   if (string_len - start_position == 0)
2128     return g_new0 (gchar *, 1);
2129
2130   if (max_tokens == 1)
2131     {
2132       string_list = g_new0 (gchar *, 2);
2133       string_list[0] = g_strndup (&string[start_position],
2134                                   string_len - start_position);
2135       return string_list;
2136     }
2137
2138   list = NULL;
2139   token_count = 0;
2140   last_separator_end = start_position;
2141   last_match_is_empty = FALSE;
2142
2143   match_ok = g_regex_match_full (regex, string, string_len, start_position,
2144                                  match_options, &match_info, &tmp_error);
2145
2146   while (tmp_error == NULL)
2147     {
2148       if (match_ok)
2149         {
2150           last_match_is_empty =
2151                     (match_info->offsets[0] == match_info->offsets[1]);
2152
2153           /* we need to skip empty separators at the same position of the end
2154            * of another separator. e.g. the string is "a b" and the separator
2155            * is " *", so from 1 to 2 we have a match and at position 2 we have
2156            * an empty match. */
2157           if (last_separator_end != match_info->offsets[1])
2158             {
2159               gchar *token;
2160               gint match_count;
2161
2162               token = g_strndup (string + last_separator_end,
2163                                  match_info->offsets[0] - last_separator_end);
2164               list = g_list_prepend (list, token);
2165               token_count++;
2166
2167               /* if there were substrings, these need to be added to
2168                * the list. */
2169               match_count = g_match_info_get_match_count (match_info);
2170               if (match_count > 1)
2171                 {
2172                   for (i = 1; i < match_count; i++)
2173                     list = g_list_prepend (list, g_match_info_fetch (match_info, i));
2174                 }
2175             }
2176         }
2177       else
2178         {
2179           /* if there was no match, copy to end of string. */
2180           if (!last_match_is_empty)
2181             {
2182               gchar *token = g_strndup (string + last_separator_end,
2183                                         match_info->string_len - last_separator_end);
2184               list = g_list_prepend (list, token);
2185             }
2186           /* no more tokens, end the loop. */
2187           break;
2188         }
2189
2190       /* -1 to leave room for the last part. */
2191       if (token_count >= max_tokens - 1)
2192         {
2193           /* we have reached the maximum number of tokens, so we copy
2194            * the remaining part of the string. */
2195           if (last_match_is_empty)
2196             {
2197               /* the last match was empty, so we have moved one char
2198                * after the real position to avoid empty matches at the
2199                * same position. */
2200               match_info->pos = PREV_CHAR (regex, &string[match_info->pos]) - string;
2201             }
2202           /* the if is needed in the case we have terminated the available
2203            * tokens, but we are at the end of the string, so there are no
2204            * characters left to copy. */
2205           if (string_len > match_info->pos)
2206             {
2207               gchar *token = g_strndup (string + match_info->pos,
2208                                         string_len - match_info->pos);
2209               list = g_list_prepend (list, token);
2210             }
2211           /* end the loop. */
2212           break;
2213         }
2214
2215       last_separator_end = match_info->pos;
2216       if (last_match_is_empty)
2217         /* if the last match was empty, g_match_info_next() has moved
2218          * forward to avoid infinite loops, but we still need to copy that
2219          * character. */
2220         last_separator_end = PREV_CHAR (regex, &string[last_separator_end]) - string;
2221
2222       match_ok = g_match_info_next (match_info, &tmp_error);
2223     }
2224   g_match_info_free (match_info);
2225   if (tmp_error != NULL)
2226     {
2227       g_propagate_error (error, tmp_error);
2228       g_list_free_full (list, g_free);
2229       match_info->pos = -1;
2230       return NULL;
2231     }
2232
2233   string_list = g_new (gchar *, g_list_length (list) + 1);
2234   i = 0;
2235   for (last = g_list_last (list); last; last = g_list_previous (last))
2236     string_list[i++] = last->data;
2237   string_list[i] = NULL;
2238   g_list_free (list);
2239
2240   return string_list;
2241 }
2242
2243 enum
2244 {
2245   REPL_TYPE_STRING,
2246   REPL_TYPE_CHARACTER,
2247   REPL_TYPE_SYMBOLIC_REFERENCE,
2248   REPL_TYPE_NUMERIC_REFERENCE,
2249   REPL_TYPE_CHANGE_CASE
2250 };
2251
2252 typedef enum
2253 {
2254   CHANGE_CASE_NONE         = 1 << 0,
2255   CHANGE_CASE_UPPER        = 1 << 1,
2256   CHANGE_CASE_LOWER        = 1 << 2,
2257   CHANGE_CASE_UPPER_SINGLE = 1 << 3,
2258   CHANGE_CASE_LOWER_SINGLE = 1 << 4,
2259   CHANGE_CASE_SINGLE_MASK  = CHANGE_CASE_UPPER_SINGLE | CHANGE_CASE_LOWER_SINGLE,
2260   CHANGE_CASE_LOWER_MASK   = CHANGE_CASE_LOWER | CHANGE_CASE_LOWER_SINGLE,
2261   CHANGE_CASE_UPPER_MASK   = CHANGE_CASE_UPPER | CHANGE_CASE_UPPER_SINGLE
2262 } ChangeCase;
2263
2264 struct _InterpolationData
2265 {
2266   gchar     *text;
2267   gint       type;
2268   gint       num;
2269   gchar      c;
2270   ChangeCase change_case;
2271 };
2272
2273 static void
2274 free_interpolation_data (InterpolationData *data)
2275 {
2276   g_free (data->text);
2277   g_free (data);
2278 }
2279
2280 static const gchar *
2281 expand_escape (const gchar        *replacement,
2282                const gchar        *p,
2283                InterpolationData  *data,
2284                GError            **error)
2285 {
2286   const gchar *q, *r;
2287   gint x, d, h, i;
2288   const gchar *error_detail;
2289   gint base = 0;
2290   GError *tmp_error = NULL;
2291
2292   p++;
2293   switch (*p)
2294     {
2295     case 't':
2296       p++;
2297       data->c = '\t';
2298       data->type = REPL_TYPE_CHARACTER;
2299       break;
2300     case 'n':
2301       p++;
2302       data->c = '\n';
2303       data->type = REPL_TYPE_CHARACTER;
2304       break;
2305     case 'v':
2306       p++;
2307       data->c = '\v';
2308       data->type = REPL_TYPE_CHARACTER;
2309       break;
2310     case 'r':
2311       p++;
2312       data->c = '\r';
2313       data->type = REPL_TYPE_CHARACTER;
2314       break;
2315     case 'f':
2316       p++;
2317       data->c = '\f';
2318       data->type = REPL_TYPE_CHARACTER;
2319       break;
2320     case 'a':
2321       p++;
2322       data->c = '\a';
2323       data->type = REPL_TYPE_CHARACTER;
2324       break;
2325     case 'b':
2326       p++;
2327       data->c = '\b';
2328       data->type = REPL_TYPE_CHARACTER;
2329       break;
2330     case '\\':
2331       p++;
2332       data->c = '\\';
2333       data->type = REPL_TYPE_CHARACTER;
2334       break;
2335     case 'x':
2336       p++;
2337       x = 0;
2338       if (*p == '{')
2339         {
2340           p++;
2341           do
2342             {
2343               h = g_ascii_xdigit_value (*p);
2344               if (h < 0)
2345                 {
2346                   error_detail = _("hexadecimal digit or '}' expected");
2347                   goto error;
2348                 }
2349               x = x * 16 + h;
2350               p++;
2351             }
2352           while (*p != '}');
2353           p++;
2354         }
2355       else
2356         {
2357           for (i = 0; i < 2; i++)
2358             {
2359               h = g_ascii_xdigit_value (*p);
2360               if (h < 0)
2361                 {
2362                   error_detail = _("hexadecimal digit expected");
2363                   goto error;
2364                 }
2365               x = x * 16 + h;
2366               p++;
2367             }
2368         }
2369       data->type = REPL_TYPE_STRING;
2370       data->text = g_new0 (gchar, 8);
2371       g_unichar_to_utf8 (x, data->text);
2372       break;
2373     case 'l':
2374       p++;
2375       data->type = REPL_TYPE_CHANGE_CASE;
2376       data->change_case = CHANGE_CASE_LOWER_SINGLE;
2377       break;
2378     case 'u':
2379       p++;
2380       data->type = REPL_TYPE_CHANGE_CASE;
2381       data->change_case = CHANGE_CASE_UPPER_SINGLE;
2382       break;
2383     case 'L':
2384       p++;
2385       data->type = REPL_TYPE_CHANGE_CASE;
2386       data->change_case = CHANGE_CASE_LOWER;
2387       break;
2388     case 'U':
2389       p++;
2390       data->type = REPL_TYPE_CHANGE_CASE;
2391       data->change_case = CHANGE_CASE_UPPER;
2392       break;
2393     case 'E':
2394       p++;
2395       data->type = REPL_TYPE_CHANGE_CASE;
2396       data->change_case = CHANGE_CASE_NONE;
2397       break;
2398     case 'g':
2399       p++;
2400       if (*p != '<')
2401         {
2402           error_detail = _("missing '<' in symbolic reference");
2403           goto error;
2404         }
2405       q = p + 1;
2406       do
2407         {
2408           p++;
2409           if (!*p)
2410             {
2411               error_detail = _("unfinished symbolic reference");
2412               goto error;
2413             }
2414         }
2415       while (*p != '>');
2416       if (p - q == 0)
2417         {
2418           error_detail = _("zero-length symbolic reference");
2419           goto error;
2420         }
2421       if (g_ascii_isdigit (*q))
2422         {
2423           x = 0;
2424           do
2425             {
2426               h = g_ascii_digit_value (*q);
2427               if (h < 0)
2428                 {
2429                   error_detail = _("digit expected");
2430                   p = q;
2431                   goto error;
2432                 }
2433               x = x * 10 + h;
2434               q++;
2435             }
2436           while (q != p);
2437           data->num = x;
2438           data->type = REPL_TYPE_NUMERIC_REFERENCE;
2439         }
2440       else
2441         {
2442           r = q;
2443           do
2444             {
2445               if (!g_ascii_isalnum (*r))
2446                 {
2447                   error_detail = _("illegal symbolic reference");
2448                   p = r;
2449                   goto error;
2450                 }
2451               r++;
2452             }
2453           while (r != p);
2454           data->text = g_strndup (q, p - q);
2455           data->type = REPL_TYPE_SYMBOLIC_REFERENCE;
2456         }
2457       p++;
2458       break;
2459     case '0':
2460       /* if \0 is followed by a number is an octal number representing a
2461        * character, else it is a numeric reference. */
2462       if (g_ascii_digit_value (*g_utf8_next_char (p)) >= 0)
2463         {
2464           base = 8;
2465           p = g_utf8_next_char (p);
2466         }
2467     case '1':
2468     case '2':
2469     case '3':
2470     case '4':
2471     case '5':
2472     case '6':
2473     case '7':
2474     case '8':
2475     case '9':
2476       x = 0;
2477       d = 0;
2478       for (i = 0; i < 3; i++)
2479         {
2480           h = g_ascii_digit_value (*p);
2481           if (h < 0)
2482             break;
2483           if (h > 7)
2484             {
2485               if (base == 8)
2486                 break;
2487               else
2488                 base = 10;
2489             }
2490           if (i == 2 && base == 10)
2491             break;
2492           x = x * 8 + h;
2493           d = d * 10 + h;
2494           p++;
2495         }
2496       if (base == 8 || i == 3)
2497         {
2498           data->type = REPL_TYPE_STRING;
2499           data->text = g_new0 (gchar, 8);
2500           g_unichar_to_utf8 (x, data->text);
2501         }
2502       else
2503         {
2504           data->type = REPL_TYPE_NUMERIC_REFERENCE;
2505           data->num = d;
2506         }
2507       break;
2508     case 0:
2509       error_detail = _("stray final '\\'");
2510       goto error;
2511       break;
2512     default:
2513       error_detail = _("unknown escape sequence");
2514       goto error;
2515     }
2516
2517   return p;
2518
2519  error:
2520   /* G_GSSIZE_FORMAT doesn't work with gettext, so we use %lu */
2521   tmp_error = g_error_new (G_REGEX_ERROR,
2522                            G_REGEX_ERROR_REPLACE,
2523                            _("Error while parsing replacement "
2524                              "text \"%s\" at char %lu: %s"),
2525                            replacement,
2526                            (gulong)(p - replacement),
2527                            error_detail);
2528   g_propagate_error (error, tmp_error);
2529
2530   return NULL;
2531 }
2532
2533 static GList *
2534 split_replacement (const gchar  *replacement,
2535                    GError      **error)
2536 {
2537   GList *list = NULL;
2538   InterpolationData *data;
2539   const gchar *p, *start;
2540
2541   start = p = replacement;
2542   while (*p)
2543     {
2544       if (*p == '\\')
2545         {
2546           data = g_new0 (InterpolationData, 1);
2547           start = p = expand_escape (replacement, p, data, error);
2548           if (p == NULL)
2549             {
2550               g_list_free_full (list, (GDestroyNotify) free_interpolation_data);
2551               free_interpolation_data (data);
2552
2553               return NULL;
2554             }
2555           list = g_list_prepend (list, data);
2556         }
2557       else
2558         {
2559           p++;
2560           if (*p == '\\' || *p == '\0')
2561             {
2562               if (p - start > 0)
2563                 {
2564                   data = g_new0 (InterpolationData, 1);
2565                   data->text = g_strndup (start, p - start);
2566                   data->type = REPL_TYPE_STRING;
2567                   list = g_list_prepend (list, data);
2568                 }
2569             }
2570         }
2571     }
2572
2573   return g_list_reverse (list);
2574 }
2575
2576 /* Change the case of c based on change_case. */
2577 #define CHANGE_CASE(c, change_case) \
2578         (((change_case) & CHANGE_CASE_LOWER_MASK) ? \
2579                 g_unichar_tolower (c) : \
2580                 g_unichar_toupper (c))
2581
2582 static void
2583 string_append (GString     *string,
2584                const gchar *text,
2585                ChangeCase  *change_case)
2586 {
2587   gunichar c;
2588
2589   if (text[0] == '\0')
2590     return;
2591
2592   if (*change_case == CHANGE_CASE_NONE)
2593     {
2594       g_string_append (string, text);
2595     }
2596   else if (*change_case & CHANGE_CASE_SINGLE_MASK)
2597     {
2598       c = g_utf8_get_char (text);
2599       g_string_append_unichar (string, CHANGE_CASE (c, *change_case));
2600       g_string_append (string, g_utf8_next_char (text));
2601       *change_case = CHANGE_CASE_NONE;
2602     }
2603   else
2604     {
2605       while (*text != '\0')
2606         {
2607           c = g_utf8_get_char (text);
2608           g_string_append_unichar (string, CHANGE_CASE (c, *change_case));
2609           text = g_utf8_next_char (text);
2610         }
2611     }
2612 }
2613
2614 static gboolean
2615 interpolate_replacement (const GMatchInfo *match_info,
2616                          GString          *result,
2617                          gpointer          data)
2618 {
2619   GList *list;
2620   InterpolationData *idata;
2621   gchar *match;
2622   ChangeCase change_case = CHANGE_CASE_NONE;
2623
2624   for (list = data; list; list = list->next)
2625     {
2626       idata = list->data;
2627       switch (idata->type)
2628         {
2629         case REPL_TYPE_STRING:
2630           string_append (result, idata->text, &change_case);
2631           break;
2632         case REPL_TYPE_CHARACTER:
2633           g_string_append_c (result, CHANGE_CASE (idata->c, change_case));
2634           if (change_case & CHANGE_CASE_SINGLE_MASK)
2635             change_case = CHANGE_CASE_NONE;
2636           break;
2637         case REPL_TYPE_NUMERIC_REFERENCE:
2638           match = g_match_info_fetch (match_info, idata->num);
2639           if (match)
2640             {
2641               string_append (result, match, &change_case);
2642               g_free (match);
2643             }
2644           break;
2645         case REPL_TYPE_SYMBOLIC_REFERENCE:
2646           match = g_match_info_fetch_named (match_info, idata->text);
2647           if (match)
2648             {
2649               string_append (result, match, &change_case);
2650               g_free (match);
2651             }
2652           break;
2653         case REPL_TYPE_CHANGE_CASE:
2654           change_case = idata->change_case;
2655           break;
2656         }
2657     }
2658
2659   return FALSE;
2660 }
2661
2662 /* whether actual match_info is needed for replacement, i.e.
2663  * whether there are references
2664  */
2665 static gboolean
2666 interpolation_list_needs_match (GList *list)
2667 {
2668   while (list != NULL)
2669     {
2670       InterpolationData *data = list->data;
2671
2672       if (data->type == REPL_TYPE_SYMBOLIC_REFERENCE ||
2673           data->type == REPL_TYPE_NUMERIC_REFERENCE)
2674         {
2675           return TRUE;
2676         }
2677
2678       list = list->next;
2679     }
2680
2681   return FALSE;
2682 }
2683
2684 /**
2685  * g_regex_replace:
2686  * @regex: a #GRegex structure
2687  * @string: (array length=string_len): the string to perform matches against
2688  * @string_len: the length of @string, or -1 if @string is nul-terminated
2689  * @start_position: starting index of the string to match
2690  * @replacement: text to replace each match with
2691  * @match_options: options for the match
2692  * @error: location to store the error occurring, or %NULL to ignore errors
2693  *
2694  * Replaces all occurrences of the pattern in @regex with the
2695  * replacement text. Backreferences of the form '\number' or
2696  * '\g&lt;number&gt;' in the replacement text are interpolated by the
2697  * number-th captured subexpression of the match, '\g&lt;name&gt;' refers
2698  * to the captured subexpression with the given name. '\0' refers to the
2699  * complete match, but '\0' followed by a number is the octal representation
2700  * of a character. To include a literal '\' in the replacement, write '\\'.
2701  * There are also escapes that changes the case of the following text:
2702  *
2703  * <variablelist>
2704  * <varlistentry><term>\l</term>
2705  * <listitem>
2706  * <para>Convert to lower case the next character</para>
2707  * </listitem>
2708  * </varlistentry>
2709  * <varlistentry><term>\u</term>
2710  * <listitem>
2711  * <para>Convert to upper case the next character</para>
2712  * </listitem>
2713  * </varlistentry>
2714  * <varlistentry><term>\L</term>
2715  * <listitem>
2716  * <para>Convert to lower case till \E</para>
2717  * </listitem>
2718  * </varlistentry>
2719  * <varlistentry><term>\U</term>
2720  * <listitem>
2721  * <para>Convert to upper case till \E</para>
2722  * </listitem>
2723  * </varlistentry>
2724  * <varlistentry><term>\E</term>
2725  * <listitem>
2726  * <para>End case modification</para>
2727  * </listitem>
2728  * </varlistentry>
2729  * </variablelist>
2730  *
2731  * If you do not need to use backreferences use g_regex_replace_literal().
2732  *
2733  * The @replacement string must be UTF-8 encoded even if #G_REGEX_RAW was
2734  * passed to g_regex_new(). If you want to use not UTF-8 encoded stings
2735  * you can use g_regex_replace_literal().
2736  *
2737  * Setting @start_position differs from just passing over a shortened
2738  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern that
2739  * begins with any kind of lookbehind assertion, such as "\b".
2740  *
2741  * Returns: a newly allocated string containing the replacements
2742  *
2743  * Since: 2.14
2744  */
2745 gchar *
2746 g_regex_replace (const GRegex      *regex,
2747                  const gchar       *string,
2748                  gssize             string_len,
2749                  gint               start_position,
2750                  const gchar       *replacement,
2751                  GRegexMatchFlags   match_options,
2752                  GError           **error)
2753 {
2754   gchar *result;
2755   GList *list;
2756   GError *tmp_error = NULL;
2757
2758   g_return_val_if_fail (regex != NULL, NULL);
2759   g_return_val_if_fail (string != NULL, NULL);
2760   g_return_val_if_fail (start_position >= 0, NULL);
2761   g_return_val_if_fail (replacement != NULL, NULL);
2762   g_return_val_if_fail (error == NULL || *error == NULL, NULL);
2763   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
2764
2765   list = split_replacement (replacement, &tmp_error);
2766   if (tmp_error != NULL)
2767     {
2768       g_propagate_error (error, tmp_error);
2769       return NULL;
2770     }
2771
2772   result = g_regex_replace_eval (regex,
2773                                  string, string_len, start_position,
2774                                  match_options,
2775                                  interpolate_replacement,
2776                                  (gpointer)list,
2777                                  &tmp_error);
2778   if (tmp_error != NULL)
2779     g_propagate_error (error, tmp_error);
2780
2781   g_list_free_full (list, (GDestroyNotify) free_interpolation_data);
2782
2783   return result;
2784 }
2785
2786 static gboolean
2787 literal_replacement (const GMatchInfo *match_info,
2788                      GString          *result,
2789                      gpointer          data)
2790 {
2791   g_string_append (result, data);
2792   return FALSE;
2793 }
2794
2795 /**
2796  * g_regex_replace_literal:
2797  * @regex: a #GRegex structure
2798  * @string: (array length=string_len): the string to perform matches against
2799  * @string_len: the length of @string, or -1 if @string is nul-terminated
2800  * @start_position: starting index of the string to match
2801  * @replacement: text to replace each match with
2802  * @match_options: options for the match
2803  * @error: location to store the error occurring, or %NULL to ignore errors
2804  *
2805  * Replaces all occurrences of the pattern in @regex with the
2806  * replacement text. @replacement is replaced literally, to
2807  * include backreferences use g_regex_replace().
2808  *
2809  * Setting @start_position differs from just passing over a
2810  * shortened string and setting #G_REGEX_MATCH_NOTBOL in the
2811  * case of a pattern that begins with any kind of lookbehind
2812  * assertion, such as "\b".
2813  *
2814  * Returns: a newly allocated string containing the replacements
2815  *
2816  * Since: 2.14
2817  */
2818 gchar *
2819 g_regex_replace_literal (const GRegex      *regex,
2820                          const gchar       *string,
2821                          gssize             string_len,
2822                          gint               start_position,
2823                          const gchar       *replacement,
2824                          GRegexMatchFlags   match_options,
2825                          GError           **error)
2826 {
2827   g_return_val_if_fail (replacement != NULL, NULL);
2828   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
2829
2830   return g_regex_replace_eval (regex,
2831                                string, string_len, start_position,
2832                                match_options,
2833                                literal_replacement,
2834                                (gpointer)replacement,
2835                                error);
2836 }
2837
2838 /**
2839  * g_regex_replace_eval:
2840  * @regex: a #GRegex structure from g_regex_new()
2841  * @string: (array length=string_len): string to perform matches against
2842  * @string_len: the length of @string, or -1 if @string is nul-terminated
2843  * @start_position: starting index of the string to match
2844  * @match_options: options for the match
2845  * @eval: a function to call for each match
2846  * @user_data: user data to pass to the function
2847  * @error: location to store the error occurring, or %NULL to ignore errors
2848  *
2849  * Replaces occurrences of the pattern in regex with the output of
2850  * @eval for that occurrence.
2851  *
2852  * Setting @start_position differs from just passing over a shortened
2853  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern
2854  * that begins with any kind of lookbehind assertion, such as "\b".
2855  *
2856  * The following example uses g_regex_replace_eval() to replace multiple
2857  * strings at once:
2858  * |[
2859  * static gboolean
2860  * eval_cb (const GMatchInfo *info,
2861  *          GString          *res,
2862  *          gpointer          data)
2863  * {
2864  *   gchar *match;
2865  *   gchar *r;
2866  *
2867  *    match = g_match_info_fetch (info, 0);
2868  *    r = g_hash_table_lookup ((GHashTable *)data, match);
2869  *    g_string_append (res, r);
2870  *    g_free (match);
2871  *
2872  *    return FALSE;
2873  * }
2874  *
2875  * /&ast; ... &ast;/
2876  *
2877  * GRegex *reg;
2878  * GHashTable *h;
2879  * gchar *res;
2880  *
2881  * h = g_hash_table_new (g_str_hash, g_str_equal);
2882  *
2883  * g_hash_table_insert (h, "1", "ONE");
2884  * g_hash_table_insert (h, "2", "TWO");
2885  * g_hash_table_insert (h, "3", "THREE");
2886  * g_hash_table_insert (h, "4", "FOUR");
2887  *
2888  * reg = g_regex_new ("1|2|3|4", 0, 0, NULL);
2889  * res = g_regex_replace_eval (reg, text, -1, 0, 0, eval_cb, h, NULL);
2890  * g_hash_table_destroy (h);
2891  *
2892  * /&ast; ... &ast;/
2893  * ]|
2894  *
2895  * Returns: a newly allocated string containing the replacements
2896  *
2897  * Since: 2.14
2898  */
2899 gchar *
2900 g_regex_replace_eval (const GRegex        *regex,
2901                       const gchar         *string,
2902                       gssize               string_len,
2903                       gint                 start_position,
2904                       GRegexMatchFlags     match_options,
2905                       GRegexEvalCallback   eval,
2906                       gpointer             user_data,
2907                       GError             **error)
2908 {
2909   GMatchInfo *match_info;
2910   GString *result;
2911   gint str_pos = 0;
2912   gboolean done = FALSE;
2913   GError *tmp_error = NULL;
2914
2915   g_return_val_if_fail (regex != NULL, NULL);
2916   g_return_val_if_fail (string != NULL, NULL);
2917   g_return_val_if_fail (start_position >= 0, NULL);
2918   g_return_val_if_fail (eval != NULL, NULL);
2919   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
2920
2921   if (string_len < 0)
2922     string_len = strlen (string);
2923
2924   result = g_string_sized_new (string_len);
2925
2926   /* run down the string making matches. */
2927   g_regex_match_full (regex, string, string_len, start_position,
2928                       match_options, &match_info, &tmp_error);
2929   while (!done && g_match_info_matches (match_info))
2930     {
2931       g_string_append_len (result,
2932                            string + str_pos,
2933                            match_info->offsets[0] - str_pos);
2934       done = (*eval) (match_info, result, user_data);
2935       str_pos = match_info->offsets[1];
2936       g_match_info_next (match_info, &tmp_error);
2937     }
2938   g_match_info_free (match_info);
2939   if (tmp_error != NULL)
2940     {
2941       g_propagate_error (error, tmp_error);
2942       g_string_free (result, TRUE);
2943       return NULL;
2944     }
2945
2946   g_string_append_len (result, string + str_pos, string_len - str_pos);
2947   return g_string_free (result, FALSE);
2948 }
2949
2950 /**
2951  * g_regex_check_replacement:
2952  * @replacement: the replacement string
2953  * @has_references: (out) (allow-none): location to store information about
2954  *   references in @replacement or %NULL
2955  * @error: location to store error
2956  *
2957  * Checks whether @replacement is a valid replacement string
2958  * (see g_regex_replace()), i.e. that all escape sequences in
2959  * it are valid.
2960  *
2961  * If @has_references is not %NULL then @replacement is checked
2962  * for pattern references. For instance, replacement text 'foo\n'
2963  * does not contain references and may be evaluated without information
2964  * about actual match, but '\0\1' (whole match followed by first
2965  * subpattern) requires valid #GMatchInfo object.
2966  *
2967  * Returns: whether @replacement is a valid replacement string
2968  *
2969  * Since: 2.14
2970  */
2971 gboolean
2972 g_regex_check_replacement (const gchar  *replacement,
2973                            gboolean     *has_references,
2974                            GError      **error)
2975 {
2976   GList *list;
2977   GError *tmp = NULL;
2978
2979   list = split_replacement (replacement, &tmp);
2980
2981   if (tmp)
2982   {
2983     g_propagate_error (error, tmp);
2984     return FALSE;
2985   }
2986
2987   if (has_references)
2988     *has_references = interpolation_list_needs_match (list);
2989
2990   g_list_free_full (list, (GDestroyNotify) free_interpolation_data);
2991
2992   return TRUE;
2993 }
2994
2995 /**
2996  * g_regex_escape_nul:
2997  * @string: the string to escape
2998  * @length: the length of @string
2999  *
3000  * Escapes the nul characters in @string to "\x00".  It can be used
3001  * to compile a regex with embedded nul characters.
3002  *
3003  * For completeness, @length can be -1 for a nul-terminated string.
3004  * In this case the output string will be of course equal to @string.
3005  *
3006  * Returns: a newly-allocated escaped string
3007  *
3008  * Since: 2.30
3009  */
3010 gchar *
3011 g_regex_escape_nul (const gchar *string,
3012                     gint         length)
3013 {
3014   GString *escaped;
3015   const gchar *p, *piece_start, *end;
3016   gint backslashes;
3017
3018   g_return_val_if_fail (string != NULL, NULL);
3019
3020   if (length < 0)
3021     return g_strdup (string);
3022
3023   end = string + length;
3024   p = piece_start = string;
3025   escaped = g_string_sized_new (length + 1);
3026
3027   backslashes = 0;
3028   while (p < end)
3029     {
3030       switch (*p)
3031         {
3032         case '\0':
3033           if (p != piece_start)
3034             {
3035               /* copy the previous piece. */
3036               g_string_append_len (escaped, piece_start, p - piece_start);
3037             }
3038           if ((backslashes & 1) == 0)
3039             g_string_append_c (escaped, '\\');
3040           g_string_append_c (escaped, 'x');
3041           g_string_append_c (escaped, '0');
3042           g_string_append_c (escaped, '0');
3043           piece_start = ++p;
3044           backslashes = 0;
3045           break;
3046         case '\\':
3047           backslashes++;
3048           ++p;
3049           break;
3050         default:
3051           backslashes = 0;
3052           p = g_utf8_next_char (p);
3053           break;
3054         }
3055     }
3056
3057   if (piece_start < end)
3058     g_string_append_len (escaped, piece_start, end - piece_start);
3059
3060   return g_string_free (escaped, FALSE);
3061 }
3062
3063 /**
3064  * g_regex_escape_string:
3065  * @string: (array length=length): the string to escape
3066  * @length: the length of @string, or -1 if @string is nul-terminated
3067  *
3068  * Escapes the special characters used for regular expressions
3069  * in @string, for instance "a.b*c" becomes "a\.b\*c". This
3070  * function is useful to dynamically generate regular expressions.
3071  *
3072  * @string can contain nul characters that are replaced with "\0",
3073  * in this case remember to specify the correct length of @string
3074  * in @length.
3075  *
3076  * Returns: a newly-allocated escaped string
3077  *
3078  * Since: 2.14
3079  */
3080 gchar *
3081 g_regex_escape_string (const gchar *string,
3082                        gint         length)
3083 {
3084   GString *escaped;
3085   const char *p, *piece_start, *end;
3086
3087   g_return_val_if_fail (string != NULL, NULL);
3088
3089   if (length < 0)
3090     length = strlen (string);
3091
3092   end = string + length;
3093   p = piece_start = string;
3094   escaped = g_string_sized_new (length + 1);
3095
3096   while (p < end)
3097     {
3098       switch (*p)
3099         {
3100         case '\0':
3101         case '\\':
3102         case '|':
3103         case '(':
3104         case ')':
3105         case '[':
3106         case ']':
3107         case '{':
3108         case '}':
3109         case '^':
3110         case '$':
3111         case '*':
3112         case '+':
3113         case '?':
3114         case '.':
3115           if (p != piece_start)
3116             /* copy the previous piece. */
3117             g_string_append_len (escaped, piece_start, p - piece_start);
3118           g_string_append_c (escaped, '\\');
3119           if (*p == '\0')
3120             g_string_append_c (escaped, '0');
3121           else
3122             g_string_append_c (escaped, *p);
3123           piece_start = ++p;
3124           break;
3125         default:
3126           p = g_utf8_next_char (p);
3127           break;
3128         }
3129   }
3130
3131   if (piece_start < end)
3132     g_string_append_len (escaped, piece_start, end - piece_start);
3133
3134   return g_string_free (escaped, FALSE);
3135 }