glib/gregex.c

   1 /* GRegex -- regular expression API wrapper around PCRE.
   2  *
   3  * Copyright (C) 1999, 2000 Scott Wimer
   4  * Copyright (C) 2004, Matthias Clasen <mclasen@redhat.com>
   5  * Copyright (C) 2005 - 2007, Marco Barisione <marco@barisione.org>
   6  *
   7  * This library is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2.1 of the License, or (at your option) any later version.
  11  *
  12  * This library is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU Lesser General Public License
  18  * along with this library; if not, see <http://www.gnu.org/licenses/>.
  19  */
  20
  21 #include "config.h"
  22
  23 #include <string.h>
  24
  25 #ifdef USE_SYSTEM_PCRE
  26 #include <pcre.h>
  27 #else
  28 #include "pcre/pcre.h"
  29 #endif
  30
  31 #include "gtypes.h"
  32 #include "gregex.h"
  33 #include "glibintl.h"
  34 #include "glist.h"
  35 #include "gmessages.h"
  36 #include "gstrfuncs.h"
  37 #include "gatomic.h"
  38 #include "gthread.h"
  39
  40 /**
  41  * SECTION:gregex
  42  * @title: Perl-compatible regular expressions
  43  * @short_description: matches strings against regular expressions
  44  * @see_also: [Regular expression syntax][glib-regex-syntax]
  45  *
  46  * The g_regex_*() functions implement regular
  47  * expression pattern matching using syntax and semantics similar to
  48  * Perl regular expression.
  49  *
  50  * Some functions accept a @start_position argument, setting it differs
  51  * from just passing over a shortened string and setting #G_REGEX_MATCH_NOTBOL
  52  * in the case of a pattern that begins with any kind of lookbehind assertion.
  53  * For example, consider the pattern "\Biss\B" which finds occurrences of "iss"
  54  * in the middle of words. ("\B" matches only if the current position in the
  55  * subject is not a word boundary.) When applied to the string "Mississipi"
  56  * from the fourth byte, namely "issipi", it does not match, because "\B" is
  57  * always false at the start of the subject, which is deemed to be a word
  58  * boundary. However, if the entire string is passed , but with
  59  * @start_position set to 4, it finds the second occurrence of "iss" because
  60  * it is able to look behind the starting point to discover that it is
  61  * preceded by a letter.
  62  *
  63  * Note that, unless you set the #G_REGEX_RAW flag, all the strings passed
  64  * to these functions must be encoded in UTF-8. The lengths and the positions
  65  * inside the strings are in bytes and not in characters, so, for instance,
  66  * "\xc3\xa0" (i.e. "à") is two bytes long but it is treated as a
  67  * single character. If you set #G_REGEX_RAW the strings can be non-valid
  68  * UTF-8 strings and a byte is treated as a character, so "\xc3\xa0" is two
  69  * bytes and two characters long.
  70  *
  71  * When matching a pattern, "\n" matches only against a "\n" character in
  72  * the string, and "\r" matches only a "\r" character. To match any newline
  73  * sequence use "\R". This particular group matches either the two-character
  74  * sequence CR + LF ("\r\n"), or one of the single characters LF (linefeed,
  75  * U+000A, "\n"), VT vertical tab, U+000B, "\v"), FF (formfeed, U+000C, "\f"),
  76  * CR (carriage return, U+000D, "\r"), NEL (next line, U+0085), LS (line
  77  * separator, U+2028), or PS (paragraph separator, U+2029).
  78  *
  79  * The behaviour of the dot, circumflex, and dollar metacharacters are
  80  * affected by newline characters, the default is to recognize any newline
  81  * character (the same characters recognized by "\R"). This can be changed
  82  * with #G_REGEX_NEWLINE_CR, #G_REGEX_NEWLINE_LF and #G_REGEX_NEWLINE_CRLF
  83  * compile options, and with #G_REGEX_MATCH_NEWLINE_ANY,
  84  * #G_REGEX_MATCH_NEWLINE_CR, #G_REGEX_MATCH_NEWLINE_LF and
  85  * #G_REGEX_MATCH_NEWLINE_CRLF match options. These settings are also
  86  * relevant when compiling a pattern if #G_REGEX_EXTENDED is set, and an
  87  * unescaped "#" outside a character class is encountered. This indicates
  88  * a comment that lasts until after the next newline.
  89  *
  90  * When setting the %G_REGEX_JAVASCRIPT_COMPAT flag, pattern syntax and pattern
  91  * matching is changed to be compatible with the way that regular expressions
  92  * work in JavaScript. More precisely, a lonely ']' character in the pattern
  93  * is a syntax error; the '\x' escape only allows 0 to 2 hexadecimal digits, and
  94  * you must use the '\u' escape sequence with 4 hex digits to specify a unicode
  95  * codepoint instead of '\x' or 'x{....}'. If '\x' or '\u' are not followed by
  96  * the specified number of hex digits, they match 'x' and 'u' literally; also
  97  * '\U' always matches 'U' instead of being an error in the pattern. Finally,
  98  * pattern matching is modified so that back references to an unset subpattern
  99  * group produces a match with the empty string instead of an error. See
 100  * pcreapi(3) for more information.
 101  *
 102  * Creating and manipulating the same #GRegex structure from different
 103  * threads is not a problem as #GRegex does not modify its internal
 104  * state between creation and destruction, on the other hand #GMatchInfo
 105  * is not threadsafe.
 106  *
 107  * The regular expressions low-level functionalities are obtained through
 108  * the excellent
 109  * [PCRE](http://www.pcre.org/)
 110  * library written by Philip Hazel.
 111  */
 112
 113 /* Mask of all the possible values for GRegexCompileFlags. */
 114 #define G_REGEX_COMPILE_MASK (G_REGEX_CASELESS          | \
 115                               G_REGEX_MULTILINE         | \
 116                               G_REGEX_DOTALL            | \
 117                               G_REGEX_EXTENDED          | \
 118                               G_REGEX_ANCHORED          | \
 119                               G_REGEX_DOLLAR_ENDONLY    | \
 120                               G_REGEX_UNGREEDY          | \
 121                               G_REGEX_RAW               | \
 122                               G_REGEX_NO_AUTO_CAPTURE   | \
 123                               G_REGEX_OPTIMIZE          | \
 124                               G_REGEX_FIRSTLINE         | \
 125                               G_REGEX_DUPNAMES          | \
 126                               G_REGEX_NEWLINE_CR        | \
 127                               G_REGEX_NEWLINE_LF        | \
 128                               G_REGEX_NEWLINE_CRLF      | \
 129                               G_REGEX_NEWLINE_ANYCRLF   | \
 130                               G_REGEX_BSR_ANYCRLF       | \
 131                               G_REGEX_JAVASCRIPT_COMPAT)
 132
 133 /* Mask of all GRegexCompileFlags values that are (not) passed trough to PCRE */
 134 #define G_REGEX_COMPILE_PCRE_MASK (G_REGEX_COMPILE_MASK & ~G_REGEX_COMPILE_NONPCRE_MASK)
 135 #define G_REGEX_COMPILE_NONPCRE_MASK (G_REGEX_RAW              | \
 136                                       G_REGEX_OPTIMIZE)
 137
 138 /* Mask of all the possible values for GRegexMatchFlags. */
 139 #define G_REGEX_MATCH_MASK (G_REGEX_MATCH_ANCHORED         | \
 140                             G_REGEX_MATCH_NOTBOL           | \
 141                             G_REGEX_MATCH_NOTEOL           | \
 142                             G_REGEX_MATCH_NOTEMPTY         | \
 143                             G_REGEX_MATCH_PARTIAL          | \
 144                             G_REGEX_MATCH_NEWLINE_CR       | \
 145                             G_REGEX_MATCH_NEWLINE_LF       | \
 146                             G_REGEX_MATCH_NEWLINE_CRLF     | \
 147                             G_REGEX_MATCH_NEWLINE_ANY      | \
 148                             G_REGEX_MATCH_NEWLINE_ANYCRLF  | \
 149                             G_REGEX_MATCH_BSR_ANYCRLF      | \
 150                             G_REGEX_MATCH_BSR_ANY          | \
 151                             G_REGEX_MATCH_PARTIAL_SOFT     | \
 152                             G_REGEX_MATCH_PARTIAL_HARD     | \
 153                             G_REGEX_MATCH_NOTEMPTY_ATSTART)
 154
 155 /* we rely on these flags having the same values */
 156 G_STATIC_ASSERT (G_REGEX_CASELESS          == PCRE_CASELESS);
 157 G_STATIC_ASSERT (G_REGEX_MULTILINE         == PCRE_MULTILINE);
 158 G_STATIC_ASSERT (G_REGEX_DOTALL            == PCRE_DOTALL);
 159 G_STATIC_ASSERT (G_REGEX_EXTENDED          == PCRE_EXTENDED);
 160 G_STATIC_ASSERT (G_REGEX_ANCHORED          == PCRE_ANCHORED);
 161 G_STATIC_ASSERT (G_REGEX_DOLLAR_ENDONLY    == PCRE_DOLLAR_ENDONLY);
 162 G_STATIC_ASSERT (G_REGEX_UNGREEDY          == PCRE_UNGREEDY);
 163 G_STATIC_ASSERT (G_REGEX_NO_AUTO_CAPTURE   == PCRE_NO_AUTO_CAPTURE);
 164 G_STATIC_ASSERT (G_REGEX_FIRSTLINE         == PCRE_FIRSTLINE);
 165 G_STATIC_ASSERT (G_REGEX_DUPNAMES          == PCRE_DUPNAMES);
 166 G_STATIC_ASSERT (G_REGEX_NEWLINE_CR        == PCRE_NEWLINE_CR);
 167 G_STATIC_ASSERT (G_REGEX_NEWLINE_LF        == PCRE_NEWLINE_LF);
 168 G_STATIC_ASSERT (G_REGEX_NEWLINE_CRLF      == PCRE_NEWLINE_CRLF);
 169 G_STATIC_ASSERT (G_REGEX_NEWLINE_ANYCRLF   == PCRE_NEWLINE_ANYCRLF);
 170 G_STATIC_ASSERT (G_REGEX_BSR_ANYCRLF       == PCRE_BSR_ANYCRLF);
 171 G_STATIC_ASSERT (G_REGEX_JAVASCRIPT_COMPAT == PCRE_JAVASCRIPT_COMPAT);
 172
 173 G_STATIC_ASSERT (G_REGEX_MATCH_ANCHORED         == PCRE_ANCHORED);
 174 G_STATIC_ASSERT (G_REGEX_MATCH_NOTBOL           == PCRE_NOTBOL);
 175 G_STATIC_ASSERT (G_REGEX_MATCH_NOTEOL           == PCRE_NOTEOL);
 176 G_STATIC_ASSERT (G_REGEX_MATCH_NOTEMPTY         == PCRE_NOTEMPTY);
 177 G_STATIC_ASSERT (G_REGEX_MATCH_PARTIAL          == PCRE_PARTIAL);
 178 G_STATIC_ASSERT (G_REGEX_MATCH_NEWLINE_CR       == PCRE_NEWLINE_CR);
 179 G_STATIC_ASSERT (G_REGEX_MATCH_NEWLINE_LF       == PCRE_NEWLINE_LF);
 180 G_STATIC_ASSERT (G_REGEX_MATCH_NEWLINE_CRLF     == PCRE_NEWLINE_CRLF);
 181 G_STATIC_ASSERT (G_REGEX_MATCH_NEWLINE_ANY      == PCRE_NEWLINE_ANY);
 182 G_STATIC_ASSERT (G_REGEX_MATCH_NEWLINE_ANYCRLF  == PCRE_NEWLINE_ANYCRLF);
 183 G_STATIC_ASSERT (G_REGEX_MATCH_BSR_ANYCRLF      == PCRE_BSR_ANYCRLF);
 184 G_STATIC_ASSERT (G_REGEX_MATCH_BSR_ANY          == PCRE_BSR_UNICODE);
 185 G_STATIC_ASSERT (G_REGEX_MATCH_PARTIAL_SOFT     == PCRE_PARTIAL_SOFT);
 186 G_STATIC_ASSERT (G_REGEX_MATCH_PARTIAL_HARD     == PCRE_PARTIAL_HARD);
 187 G_STATIC_ASSERT (G_REGEX_MATCH_NOTEMPTY_ATSTART == PCRE_NOTEMPTY_ATSTART);
 188
 189 /* These PCRE flags are unused or not exposed publically in GRegexFlags, so
 190  * it should be ok to reuse them for different things.
 191  */
 192 G_STATIC_ASSERT (G_REGEX_OPTIMIZE          == PCRE_NO_UTF8_CHECK);
 193 G_STATIC_ASSERT (G_REGEX_RAW               == PCRE_UTF8);
 194
 195 /* if the string is in UTF-8 use g_utf8_ functions, else use
 196  * use just +/- 1. */
 197 #define NEXT_CHAR(re, s) (((re)->compile_opts & G_REGEX_RAW) ? \
 198                                 ((s) + 1) : \
 199                                 g_utf8_next_char (s))
 200 #define PREV_CHAR(re, s) (((re)->compile_opts & G_REGEX_RAW) ? \
 201                                 ((s) - 1) : \
 202                                 g_utf8_prev_char (s))
 203
 204 struct _GMatchInfo
 205 {
 206   volatile gint ref_count;      /* the ref count */
 207   GRegex *regex;                /* the regex */
 208   GRegexMatchFlags match_opts;  /* options used at match time on the regex */
 209   gint matches;                 /* number of matching sub patterns */
 210   gint pos;                     /* position in the string where last match left off */
 211   gint  n_offsets;              /* number of offsets */
 212   gint *offsets;                /* array of offsets paired 0,1 ; 2,3 ; 3,4 etc */
 213   gint *workspace;              /* workspace for pcre_dfa_exec() */
 214   gint n_workspace;             /* number of workspace elements */
 215   const gchar *string;          /* string passed to the match function */
 216   gssize string_len;            /* length of string */
 217 };
 218
 219 struct _GRegex
 220 {
 221   volatile gint ref_count;      /* the ref count for the immutable part */
 222   gchar *pattern;               /* the pattern */
 223   pcre *pcre_re;                /* compiled form of the pattern */
 224   GRegexCompileFlags compile_opts;      /* options used at compile time on the pattern */
 225   GRegexMatchFlags match_opts;  /* options used at match time on the regex */
 226   pcre_extra *extra;            /* data stored when G_REGEX_OPTIMIZE is used */
 227 };
 228
 229 /* TRUE if ret is an error code, FALSE otherwise. */
 230 #define IS_PCRE_ERROR(ret) ((ret) < PCRE_ERROR_NOMATCH && (ret) != PCRE_ERROR_PARTIAL)
 231
 232 typedef struct _InterpolationData InterpolationData;
 233 static gboolean  interpolation_list_needs_match (GList *list);
 234 static gboolean  interpolate_replacement        (const GMatchInfo *match_info,
 235                                                  GString *result,
 236                                                  gpointer data);
 237 static GList    *split_replacement              (const gchar *replacement,
 238                                                  GError **error);
 239 static void      free_interpolation_data        (InterpolationData *data);
 240
 241
 242 static const gchar *
 243 match_error (gint errcode)
 244 {
 245   switch (errcode)
 246     {
 247     case PCRE_ERROR_NOMATCH:
 248       /* not an error */
 249       break;
 250     case PCRE_ERROR_NULL:
 251       /* NULL argument, this should not happen in GRegex */
 252       g_warning ("A NULL argument was passed to PCRE");
 253       break;
 254     case PCRE_ERROR_BADOPTION:
 255       return "bad options";
 256     case PCRE_ERROR_BADMAGIC:
 257       return _("corrupted object");
 258     case PCRE_ERROR_UNKNOWN_OPCODE:
 259       return N_("internal error or corrupted object");
 260     case PCRE_ERROR_NOMEMORY:
 261       return _("out of memory");
 262     case PCRE_ERROR_NOSUBSTRING:
 263       /* not used by pcre_exec() */
 264       break;
 265     case PCRE_ERROR_MATCHLIMIT:
 266       return _("backtracking limit reached");
 267     case PCRE_ERROR_CALLOUT:
 268       /* callouts are not implemented */
 269       break;
 270     case PCRE_ERROR_BADUTF8:
 271     case PCRE_ERROR_BADUTF8_OFFSET:
 272       /* we do not check if strings are valid */
 273       break;
 274     case PCRE_ERROR_PARTIAL:
 275       /* not an error */
 276       break;
 277     case PCRE_ERROR_BADPARTIAL:
 278       return _("the pattern contains items not supported for partial matching");
 279     case PCRE_ERROR_INTERNAL:
 280       return _("internal error");
 281     case PCRE_ERROR_BADCOUNT:
 282       /* negative ovecsize, this should not happen in GRegex */
 283       g_warning ("A negative ovecsize was passed to PCRE");
 284       break;
 285     case PCRE_ERROR_DFA_UITEM:
 286       return _("the pattern contains items not supported for partial matching");
 287     case PCRE_ERROR_DFA_UCOND:
 288       return _("back references as conditions are not supported for partial matching");
 289     case PCRE_ERROR_DFA_UMLIMIT:
 290       /* the match_field field is not used in GRegex */
 291       break;
 292     case PCRE_ERROR_DFA_WSSIZE:
 293       /* handled expanding the workspace */
 294       break;
 295     case PCRE_ERROR_DFA_RECURSE:
 296     case PCRE_ERROR_RECURSIONLIMIT:
 297       return _("recursion limit reached");
 298     case PCRE_ERROR_BADNEWLINE:
 299       return _("invalid combination of newline flags");
 300     case PCRE_ERROR_BADOFFSET:
 301       return _("bad offset");
 302     case PCRE_ERROR_SHORTUTF8:
 303       return _("short utf8");
 304     case PCRE_ERROR_RECURSELOOP:
 305       return _("recursion loop");
 306     default:
 307       break;
 308     }
 309   return _("unknown error");
 310 }
 311
 312 static void
 313 translate_compile_error (gint *errcode, const gchar **errmsg)
 314 {
 315   /* Compile errors are created adding 100 to the error code returned
 316    * by PCRE.
 317    * If errcode is known we put the translatable error message in
 318    * erromsg. If errcode is unknown we put the generic
 319    * G_REGEX_ERROR_COMPILE error code in errcode and keep the
 320    * untranslated error message returned by PCRE.
 321    * Note that there can be more PCRE errors with the same GRegexError
 322    * and that some PCRE errors are useless for us.
 323    */
 324   *errcode += 100;
 325
 326   switch (*errcode)
 327     {
 328     case G_REGEX_ERROR_STRAY_BACKSLASH:
 329       *errmsg = _("\\ at end of pattern");
 330       break;
 331     case G_REGEX_ERROR_MISSING_CONTROL_CHAR:
 332       *errmsg = _("\\c at end of pattern");
 333       break;
 334     case G_REGEX_ERROR_UNRECOGNIZED_ESCAPE:
 335       *errmsg = _("unrecognized character following \\");
 336       break;
 337     case G_REGEX_ERROR_QUANTIFIERS_OUT_OF_ORDER:
 338       *errmsg = _("numbers out of order in {} quantifier");
 339       break;
 340     case G_REGEX_ERROR_QUANTIFIER_TOO_BIG:
 341       *errmsg = _("number too big in {} quantifier");
 342       break;
 343     case G_REGEX_ERROR_UNTERMINATED_CHARACTER_CLASS:
 344       *errmsg = _("missing terminating ] for character class");
 345       break;
 346     case G_REGEX_ERROR_INVALID_ESCAPE_IN_CHARACTER_CLASS:
 347       *errmsg = _("invalid escape sequence in character class");
 348       break;
 349     case G_REGEX_ERROR_RANGE_OUT_OF_ORDER:
 350       *errmsg = _("range out of order in character class");
 351       break;
 352     case G_REGEX_ERROR_NOTHING_TO_REPEAT:
 353       *errmsg = _("nothing to repeat");
 354       break;
 355     case 111: /* internal error: unexpected repeat */
 356       *errcode = G_REGEX_ERROR_INTERNAL;
 357       *errmsg = _("unexpected repeat");
 358       break;
 359     case G_REGEX_ERROR_UNRECOGNIZED_CHARACTER:
 360       *errmsg = _("unrecognized character after (? or (?-");
 361       break;
 362     case G_REGEX_ERROR_POSIX_NAMED_CLASS_OUTSIDE_CLASS:
 363       *errmsg = _("POSIX named classes are supported only within a class");
 364       break;
 365     case G_REGEX_ERROR_UNMATCHED_PARENTHESIS:
 366       *errmsg = _("missing terminating )");
 367       break;
 368     case G_REGEX_ERROR_INEXISTENT_SUBPATTERN_REFERENCE:
 369       *errmsg = _("reference to non-existent subpattern");
 370       break;
 371     case G_REGEX_ERROR_UNTERMINATED_COMMENT:
 372       *errmsg = _("missing ) after comment");
 373       break;
 374     case G_REGEX_ERROR_EXPRESSION_TOO_LARGE:
 375       *errmsg = _("regular expression is too large");
 376       break;
 377     case G_REGEX_ERROR_MEMORY_ERROR:
 378       *errmsg = _("failed to get memory");
 379       break;
 380     case 122: /* unmatched parentheses */
 381       *errcode = G_REGEX_ERROR_UNMATCHED_PARENTHESIS;
 382       *errmsg = _(") without opening (");
 383       break;
 384     case 123: /* internal error: code overflow */
 385       *errcode = G_REGEX_ERROR_INTERNAL;
 386       *errmsg = _("code overflow");
 387       break;
 388     case 124: /* "unrecognized character after (?<\0 */
 389       *errcode = G_REGEX_ERROR_UNRECOGNIZED_CHARACTER;
 390       *errmsg = _("unrecognized character after (?<");
 391       break;
 392     case G_REGEX_ERROR_VARIABLE_LENGTH_LOOKBEHIND:
 393       *errmsg = _("lookbehind assertion is not fixed length");
 394       break;
 395     case G_REGEX_ERROR_MALFORMED_CONDITION:
 396       *errmsg = _("malformed number or name after (?(");
 397       break;
 398     case G_REGEX_ERROR_TOO_MANY_CONDITIONAL_BRANCHES:
 399       *errmsg = _("conditional group contains more than two branches");
 400       break;
 401     case G_REGEX_ERROR_ASSERTION_EXPECTED:
 402       *errmsg = _("assertion expected after (?(");
 403       break;
 404     case 129:
 405       *errcode = G_REGEX_ERROR_UNMATCHED_PARENTHESIS;
 406       /* translators: '(?R' and '(?[+-]digits' are both meant as (groups of)
 407        * sequences here, '(?-54' would be an example for the second group.
 408        */
 409       *errmsg = _("(?R or (?[+-]digits must be followed by )");
 410       break;
 411     case G_REGEX_ERROR_UNKNOWN_POSIX_CLASS_NAME:
 412       *errmsg = _("unknown POSIX class name");
 413       break;
 414     case G_REGEX_ERROR_POSIX_COLLATING_ELEMENTS_NOT_SUPPORTED:
 415       *errmsg = _("POSIX collating elements are not supported");
 416       break;
 417     case G_REGEX_ERROR_HEX_CODE_TOO_LARGE:
 418       *errmsg = _("character value in \\x{...} sequence is too large");
 419       break;
 420     case G_REGEX_ERROR_INVALID_CONDITION:
 421       *errmsg = _("invalid condition (?(0)");
 422       break;
 423     case G_REGEX_ERROR_SINGLE_BYTE_MATCH_IN_LOOKBEHIND:
 424       *errmsg = _("\\C not allowed in lookbehind assertion");
 425       break;
 426     case 137: /* PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0 */
 427       /* A number of Perl escapes are not handled by PCRE.
 428        * Therefore it explicitly raises ERR37.
 429        */
 430       *errcode = G_REGEX_ERROR_UNRECOGNIZED_ESCAPE;
 431       *errmsg = _("escapes \\L, \\l, \\N{name}, \\U, and \\u are not supported");
 432       break;
 433     case G_REGEX_ERROR_INFINITE_LOOP:
 434       *errmsg = _("recursive call could loop indefinitely");
 435       break;
 436     case 141: /* unrecognized character after (?P\0 */
 437       *errcode = G_REGEX_ERROR_UNRECOGNIZED_CHARACTER;
 438       *errmsg = _("unrecognized character after (?P");
 439       break;
 440     case G_REGEX_ERROR_MISSING_SUBPATTERN_NAME_TERMINATOR:
 441       *errmsg = _("missing terminator in subpattern name");
 442       break;
 443     case G_REGEX_ERROR_DUPLICATE_SUBPATTERN_NAME:
 444       *errmsg = _("two named subpatterns have the same name");
 445       break;
 446     case G_REGEX_ERROR_MALFORMED_PROPERTY:
 447       *errmsg = _("malformed \\P or \\p sequence");
 448       break;
 449     case G_REGEX_ERROR_UNKNOWN_PROPERTY:
 450       *errmsg = _("unknown property name after \\P or \\p");
 451       break;
 452     case G_REGEX_ERROR_SUBPATTERN_NAME_TOO_LONG:
 453       *errmsg = _("subpattern name is too long (maximum 32 characters)");
 454       break;
 455     case G_REGEX_ERROR_TOO_MANY_SUBPATTERNS:
 456       *errmsg = _("too many named subpatterns (maximum 10,000)");
 457       break;
 458     case G_REGEX_ERROR_INVALID_OCTAL_VALUE:
 459       *errmsg = _("octal value is greater than \\377");
 460       break;
 461     case 152: /* internal error: overran compiling workspace */
 462       *errcode = G_REGEX_ERROR_INTERNAL;
 463       *errmsg = _("overran compiling workspace");
 464       break;
 465     case 153: /* internal error: previously-checked referenced subpattern not found */
 466       *errcode = G_REGEX_ERROR_INTERNAL;
 467       *errmsg = _("previously-checked referenced subpattern not found");
 468       break;
 469     case G_REGEX_ERROR_TOO_MANY_BRANCHES_IN_DEFINE:
 470       *errmsg = _("DEFINE group contains more than one branch");
 471       break;
 472     case G_REGEX_ERROR_INCONSISTENT_NEWLINE_OPTIONS:
 473       *errmsg = _("inconsistent NEWLINE options");
 474       break;
 475     case G_REGEX_ERROR_MISSING_BACK_REFERENCE:
 476       *errmsg = _("\\g is not followed by a braced, angle-bracketed, or quoted name or "
 477                   "number, or by a plain number");
 478       break;
 479     case G_REGEX_ERROR_INVALID_RELATIVE_REFERENCE:
 480       *errmsg = _("a numbered reference must not be zero");
 481       break;
 482     case G_REGEX_ERROR_BACKTRACKING_CONTROL_VERB_ARGUMENT_FORBIDDEN:
 483       *errmsg = _("an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)");
 484       break;
 485     case G_REGEX_ERROR_UNKNOWN_BACKTRACKING_CONTROL_VERB:
 486       *errmsg = _("(*VERB) not recognized");
 487       break;
 488     case G_REGEX_ERROR_NUMBER_TOO_BIG:
 489       *errmsg = _("number is too big");
 490       break;
 491     case G_REGEX_ERROR_MISSING_SUBPATTERN_NAME:
 492       *errmsg = _("missing subpattern name after (?&");
 493       break;
 494     case G_REGEX_ERROR_MISSING_DIGIT:
 495       *errmsg = _("digit expected after (?+");
 496       break;
 497     case G_REGEX_ERROR_INVALID_DATA_CHARACTER:
 498       *errmsg = _("] is an invalid data character in JavaScript compatibility mode");
 499       break;
 500     case G_REGEX_ERROR_EXTRA_SUBPATTERN_NAME:
 501       *errmsg = _("different names for subpatterns of the same number are not allowed");
 502       break;
 503     case G_REGEX_ERROR_BACKTRACKING_CONTROL_VERB_ARGUMENT_REQUIRED:
 504       *errmsg = _("(*MARK) must have an argument");
 505       break;
 506     case G_REGEX_ERROR_INVALID_CONTROL_CHAR:
 507       *errmsg = _( "\\c must be followed by an ASCII character");
 508       break;
 509     case G_REGEX_ERROR_MISSING_NAME:
 510       *errmsg = _("\\k is not followed by a braced, angle-bracketed, or quoted name");
 511       break;
 512     case G_REGEX_ERROR_NOT_SUPPORTED_IN_CLASS:
 513       *errmsg = _("\\N is not supported in a class");
 514       break;
 515     case G_REGEX_ERROR_TOO_MANY_FORWARD_REFERENCES:
 516       *errmsg = _("too many forward references");
 517       break;
 518     case G_REGEX_ERROR_NAME_TOO_LONG:
 519       *errmsg = _("name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)");
 520       break;
 521     case G_REGEX_ERROR_CHARACTER_VALUE_TOO_LARGE:
 522       *errmsg = _("character value in \\u.... sequence is too large");
 523       break;
 524
 525     case 116: /* erroffset passed as NULL */
 526       /* This should not happen as we never pass a NULL erroffset */
 527       g_warning ("erroffset passed as NULL");
 528       *errcode = G_REGEX_ERROR_COMPILE;
 529       break;
 530     case 117: /* unknown option bit(s) set */
 531       /* This should not happen as we check options before passing them
 532        * to pcre_compile2() */
 533       g_warning ("unknown option bit(s) set");
 534       *errcode = G_REGEX_ERROR_COMPILE;
 535       break;
 536     case 132: /* this version of PCRE is compiled without UTF support */
 537     case 144: /* invalid UTF-8 string */
 538     case 145: /* support for \\P, \\p, and \\X has not been compiled */
 539     case 167: /* this version of PCRE is not compiled with Unicode property support */
 540     case 173: /* disallowed Unicode code point (>= 0xd800 && <= 0xdfff) */
 541     case 174: /* invalid UTF-16 string */
 542       /* These errors should not happen as we are using an UTF-8 and UCP-enabled PCRE
 543        * and we do not check if strings are valid */
 544     case 170: /* internal error: unknown opcode in find_fixedlength() */
 545       *errcode = G_REGEX_ERROR_INTERNAL;
 546       break;
 547
 548     default:
 549       *errcode = G_REGEX_ERROR_COMPILE;
 550     }
 551 }
 552
 553 /* GMatchInfo */
 554
 555 static GMatchInfo *
 556 match_info_new (const GRegex *regex,
 557                 const gchar  *string,
 558                 gint          string_len,
 559                 gint          start_position,
 560                 gint          match_options,
 561                 gboolean      is_dfa)
 562 {
 563   GMatchInfo *match_info;
 564
 565   if (string_len < 0)
 566     string_len = strlen (string);
 567
 568   match_info = g_new0 (GMatchInfo, 1);
 569   match_info->ref_count = 1;
 570   match_info->regex = g_regex_ref ((GRegex *)regex);
 571   match_info->string = string;
 572   match_info->string_len = string_len;
 573   match_info->matches = PCRE_ERROR_NOMATCH;
 574   match_info->pos = start_position;
 575   match_info->match_opts = match_options;
 576
 577   if (is_dfa)
 578     {
 579       /* These values should be enough for most cases, if they are not
 580        * enough g_regex_match_all_full() will expand them. */
 581       match_info->n_offsets = 24;
 582       match_info->n_workspace = 100;
 583       match_info->workspace = g_new (gint, match_info->n_workspace);
 584     }
 585   else
 586     {
 587       gint capture_count;
 588       pcre_fullinfo (regex->pcre_re, regex->extra,
 589                      PCRE_INFO_CAPTURECOUNT, &capture_count);
 590       match_info->n_offsets = (capture_count + 1) * 3;
 591     }
 592
 593   match_info->offsets = g_new0 (gint, match_info->n_offsets);
 594   /* Set an invalid position for the previous match. */
 595   match_info->offsets[0] = -1;
 596   match_info->offsets[1] = -1;
 597
 598   return match_info;
 599 }
 600
 601 /**
 602  * g_match_info_get_regex:
 603  * @match_info: a #GMatchInfo
 604  *
 605  * Returns #GRegex object used in @match_info. It belongs to Glib
 606  * and must not be freed. Use g_regex_ref() if you need to keep it
 607  * after you free @match_info object.
 608  *
 609  * Returns: #GRegex object used in @match_info
 610  *
 611  * Since: 2.14
 612  */
 613 GRegex *
 614 g_match_info_get_regex (const GMatchInfo *match_info)
 615 {
 616   g_return_val_if_fail (match_info != NULL, NULL);
 617   return match_info->regex;
 618 }
 619
 620 /**
 621  * g_match_info_get_string:
 622  * @match_info: a #GMatchInfo
 623  *
 624  * Returns the string searched with @match_info. This is the
 625  * string passed to g_regex_match() or g_regex_replace() so
 626  * you may not free it before calling this function.
 627  *
 628  * Returns: the string searched with @match_info
 629  *
 630  * Since: 2.14
 631  */
 632 const gchar *
 633 g_match_info_get_string (const GMatchInfo *match_info)
 634 {
 635   g_return_val_if_fail (match_info != NULL, NULL);
 636   return match_info->string;
 637 }
 638
 639 /**
 640  * g_match_info_ref:
 641  * @match_info: a #GMatchInfo
 642  *
 643  * Increases reference count of @match_info by 1.
 644  *
 645  * Returns: @match_info
 646  *
 647  * Since: 2.30
 648  */
 649 GMatchInfo       *
 650 g_match_info_ref (GMatchInfo *match_info)
 651 {
 652   g_return_val_if_fail (match_info != NULL, NULL);
 653   g_atomic_int_inc (&match_info->ref_count);
 654   return match_info;
 655 }
 656
 657 /**
 658  * g_match_info_unref:
 659  * @match_info: a #GMatchInfo
 660  *
 661  * Decreases reference count of @match_info by 1. When reference count drops
 662  * to zero, it frees all the memory associated with the match_info structure.
 663  *
 664  * Since: 2.30
 665  */
 666 void
 667 g_match_info_unref (GMatchInfo *match_info)
 668 {
 669   if (g_atomic_int_dec_and_test (&match_info->ref_count))
 670     {
 671       g_regex_unref (match_info->regex);
 672       g_free (match_info->offsets);
 673       g_free (match_info->workspace);
 674       g_free (match_info);
 675     }
 676 }
 677
 678 /**
 679  * g_match_info_free:
 680  * @match_info: (nullable): a #GMatchInfo, or %NULL
 681  *
 682  * If @match_info is not %NULL, calls g_match_info_unref(); otherwise does
 683  * nothing.
 684  *
 685  * Since: 2.14
 686  */
 687 void
 688 g_match_info_free (GMatchInfo *match_info)
 689 {
 690   if (match_info == NULL)
 691     return;
 692
 693   g_match_info_unref (match_info);
 694 }
 695
 696 /**
 697  * g_match_info_next:
 698  * @match_info: a #GMatchInfo structure
 699  * @error: location to store the error occurring, or %NULL to ignore errors
 700  *
 701  * Scans for the next match using the same parameters of the previous
 702  * call to g_regex_match_full() or g_regex_match() that returned
 703  * @match_info.
 704  *
 705  * The match is done on the string passed to the match function, so you
 706  * cannot free it before calling this function.
 707  *
 708  * Returns: %TRUE is the string matched, %FALSE otherwise
 709  *
 710  * Since: 2.14
 711  */
 712 gboolean
 713 g_match_info_next (GMatchInfo  *match_info,
 714                    GError     **error)
 715 {
 716   gint prev_match_start;
 717   gint prev_match_end;
 718
 719   g_return_val_if_fail (match_info != NULL, FALSE);
 720   g_return_val_if_fail (error == NULL || *error == NULL, FALSE);
 721   g_return_val_if_fail (match_info->pos >= 0, FALSE);
 722
 723   prev_match_start = match_info->offsets[0];
 724   prev_match_end = match_info->offsets[1];
 725
 726   if (match_info->pos > match_info->string_len)
 727     {
 728       /* we have reached the end of the string */
 729       match_info->pos = -1;
 730       match_info->matches = PCRE_ERROR_NOMATCH;
 731       return FALSE;
 732     }
 733
 734   match_info->matches = pcre_exec (match_info->regex->pcre_re,
 735                                    match_info->regex->extra,
 736                                    match_info->string,
 737                                    match_info->string_len,
 738                                    match_info->pos,
 739                                    match_info->regex->match_opts | match_info->match_opts,
 740                                    match_info->offsets,
 741                                    match_info->n_offsets);
 742   if (IS_PCRE_ERROR (match_info->matches))
 743     {
 744       g_set_error (error, G_REGEX_ERROR, G_REGEX_ERROR_MATCH,
 745                    _("Error while matching regular expression %s: %s"),
 746                    match_info->regex->pattern, match_error (match_info->matches));
 747       return FALSE;
 748     }
 749
 750   /* avoid infinite loops if the pattern is an empty string or something
 751    * equivalent */
 752   if (match_info->pos == match_info->offsets[1])
 753     {
 754       if (match_info->pos > match_info->string_len)
 755         {
 756           /* we have reached the end of the string */
 757           match_info->pos = -1;
 758           match_info->matches = PCRE_ERROR_NOMATCH;
 759           return FALSE;
 760         }
 761
 762       match_info->pos = NEXT_CHAR (match_info->regex,
 763                                    &match_info->string[match_info->pos]) -
 764                                    match_info->string;
 765     }
 766   else
 767     {
 768       match_info->pos = match_info->offsets[1];
 769     }
 770
 771   /* it's possible to get two identical matches when we are matching
 772    * empty strings, for instance if the pattern is "(?=[A-Z0-9])" and
 773    * the string is "RegExTest" we have:
 774    *  - search at position 0: match from 0 to 0
 775    *  - search at position 1: match from 3 to 3
 776    *  - search at position 3: match from 3 to 3 (duplicate)
 777    *  - search at position 4: match from 5 to 5
 778    *  - search at position 5: match from 5 to 5 (duplicate)
 779    *  - search at position 6: no match -> stop
 780    * so we have to ignore the duplicates.
 781    * see bug #515944: http://bugzilla.gnome.org/show_bug.cgi?id=515944 */
 782   if (match_info->matches >= 0 &&
 783       prev_match_start == match_info->offsets[0] &&
 784       prev_match_end == match_info->offsets[1])
 785     {
 786       /* ignore this match and search the next one */
 787       return g_match_info_next (match_info, error);
 788     }
 789
 790   return match_info->matches >= 0;
 791 }
 792
 793 /**
 794  * g_match_info_matches:
 795  * @match_info: a #GMatchInfo structure
 796  *
 797  * Returns whether the previous match operation succeeded.
 798  *
 799  * Returns: %TRUE if the previous match operation succeeded,
 800  *   %FALSE otherwise
 801  *
 802  * Since: 2.14
 803  */
 804 gboolean
 805 g_match_info_matches (const GMatchInfo *match_info)
 806 {
 807   g_return_val_if_fail (match_info != NULL, FALSE);
 808
 809   return match_info->matches >= 0;
 810 }
 811
 812 /**
 813  * g_match_info_get_match_count:
 814  * @match_info: a #GMatchInfo structure
 815  *
 816  * Retrieves the number of matched substrings (including substring 0,
 817  * that is the whole matched text), so 1 is returned if the pattern
 818  * has no substrings in it and 0 is returned if the match failed.
 819  *
 820  * If the last match was obtained using the DFA algorithm, that is
 821  * using g_regex_match_all() or g_regex_match_all_full(), the retrieved
 822  * count is not that of the number of capturing parentheses but that of
 823  * the number of matched substrings.
 824  *
 825  * Returns: Number of matched substrings, or -1 if an error occurred
 826  *
 827  * Since: 2.14
 828  */
 829 gint
 830 g_match_info_get_match_count (const GMatchInfo *match_info)
 831 {
 832   g_return_val_if_fail (match_info, -1);
 833
 834   if (match_info->matches == PCRE_ERROR_NOMATCH)
 835     /* no match */
 836     return 0;
 837   else if (match_info->matches < PCRE_ERROR_NOMATCH)
 838     /* error */
 839     return -1;
 840   else
 841     /* match */
 842     return match_info->matches;
 843 }
 844
 845 /**
 846  * g_match_info_is_partial_match:
 847  * @match_info: a #GMatchInfo structure
 848  *
 849  * Usually if the string passed to g_regex_match*() matches as far as
 850  * it goes, but is too short to match the entire pattern, %FALSE is
 851  * returned. There are circumstances where it might be helpful to
 852  * distinguish this case from other cases in which there is no match.
 853  *
 854  * Consider, for example, an application where a human is required to
 855  * type in data for a field with specific formatting requirements. An
 856  * example might be a date in the form ddmmmyy, defined by the pattern
 857  * "^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$".
 858  * If the application sees the user’s keystrokes one by one, and can
 859  * check that what has been typed so far is potentially valid, it is
 860  * able to raise an error as soon as a mistake is made.
 861  *
 862  * GRegex supports the concept of partial matching by means of the
 863  * #G_REGEX_MATCH_PARTIAL_SOFT and #G_REGEX_MATCH_PARTIAL_HARD flags.
 864  * When they are used, the return code for
 865  * g_regex_match() or g_regex_match_full() is, as usual, %TRUE
 866  * for a complete match, %FALSE otherwise. But, when these functions
 867  * return %FALSE, you can check if the match was partial calling
 868  * g_match_info_is_partial_match().
 869  *
 870  * The difference between #G_REGEX_MATCH_PARTIAL_SOFT and
 871  * #G_REGEX_MATCH_PARTIAL_HARD is that when a partial match is encountered
 872  * with #G_REGEX_MATCH_PARTIAL_SOFT, matching continues to search for a
 873  * possible complete match, while with #G_REGEX_MATCH_PARTIAL_HARD matching
 874  * stops at the partial match.
 875  * When both #G_REGEX_MATCH_PARTIAL_SOFT and #G_REGEX_MATCH_PARTIAL_HARD
 876  * are set, the latter takes precedence.
 877  *
 878  * There were formerly some restrictions on the pattern for partial matching.
 879  * The restrictions no longer apply.
 880  *
 881  * See pcrepartial(3) for more information on partial matching.
 882  *
 883  * Returns: %TRUE if the match was partial, %FALSE otherwise
 884  *
 885  * Since: 2.14
 886  */
 887 gboolean
 888 g_match_info_is_partial_match (const GMatchInfo *match_info)
 889 {
 890   g_return_val_if_fail (match_info != NULL, FALSE);
 891
 892   return match_info->matches == PCRE_ERROR_PARTIAL;
 893 }
 894
 895 /**
 896  * g_match_info_expand_references:
 897  * @match_info: (nullable): a #GMatchInfo or %NULL
 898  * @string_to_expand: the string to expand
 899  * @error: location to store the error occurring, or %NULL to ignore errors
 900  *
 901  * Returns a new string containing the text in @string_to_expand with
 902  * references and escape sequences expanded. References refer to the last
 903  * match done with @string against @regex and have the same syntax used by
 904  * g_regex_replace().
 905  *
 906  * The @string_to_expand must be UTF-8 encoded even if #G_REGEX_RAW was
 907  * passed to g_regex_new().
 908  *
 909  * The backreferences are extracted from the string passed to the match
 910  * function, so you cannot call this function after freeing the string.
 911  *
 912  * @match_info may be %NULL in which case @string_to_expand must not
 913  * contain references. For instance "foo\n" does not refer to an actual
 914  * pattern and '\n' merely will be replaced with \n character,
 915  * while to expand "\0" (whole match) one needs the result of a match.
 916  * Use g_regex_check_replacement() to find out whether @string_to_expand
 917  * contains references.
 918  *
 919  * Returns: (nullable): the expanded string, or %NULL if an error occurred
 920  *
 921  * Since: 2.14
 922  */
 923 gchar *
 924 g_match_info_expand_references (const GMatchInfo  *match_info,
 925                                 const gchar       *string_to_expand,
 926                                 GError           **error)
 927 {
 928   GString *result;
 929   GList *list;
 930   GError *tmp_error = NULL;
 931
 932   g_return_val_if_fail (string_to_expand != NULL, NULL);
 933   g_return_val_if_fail (error == NULL || *error == NULL, NULL);
 934
 935   list = split_replacement (string_to_expand, &tmp_error);
 936   if (tmp_error != NULL)
 937     {
 938       g_propagate_error (error, tmp_error);
 939       return NULL;
 940     }
 941
 942   if (!match_info && interpolation_list_needs_match (list))
 943     {
 944       g_critical ("String '%s' contains references to the match, can't "
 945                   "expand references without GMatchInfo object",
 946                   string_to_expand);
 947       return NULL;
 948     }
 949
 950   result = g_string_sized_new (strlen (string_to_expand));
 951   interpolate_replacement (match_info, result, list);
 952
 953   g_list_free_full (list, (GDestroyNotify) free_interpolation_data);
 954
 955   return g_string_free (result, FALSE);
 956 }
 957
 958 /**
 959  * g_match_info_fetch:
 960  * @match_info: #GMatchInfo structure
 961  * @match_num: number of the sub expression
 962  *
 963  * Retrieves the text matching the @match_num'th capturing
 964  * parentheses. 0 is the full text of the match, 1 is the first paren
 965  * set, 2 the second, and so on.
 966  *
 967  * If @match_num is a valid sub pattern but it didn't match anything
 968  * (e.g. sub pattern 1, matching "b" against "(a)?b") then an empty
 969  * string is returned.
 970  *
 971  * If the match was obtained using the DFA algorithm, that is using
 972  * g_regex_match_all() or g_regex_match_all_full(), the retrieved
 973  * string is not that of a set of parentheses but that of a matched
 974  * substring. Substrings are matched in reverse order of length, so
 975  * 0 is the longest match.
 976  *
 977  * The string is fetched from the string passed to the match function,
 978  * so you cannot call this function after freeing the string.
 979  *
 980  * Returns: (nullable): The matched substring, or %NULL if an error
 981  *     occurred. You have to free the string yourself
 982  *
 983  * Since: 2.14
 984  */
 985 gchar *
 986 g_match_info_fetch (const GMatchInfo *match_info,
 987                     gint              match_num)
 988 {
 989   /* we cannot use pcre_get_substring() because it allocates the
 990    * string using pcre_malloc(). */
 991   gchar *match = NULL;
 992   gint start, end;
 993
 994   g_return_val_if_fail (match_info != NULL, NULL);
 995   g_return_val_if_fail (match_num >= 0, NULL);
 996
 997   /* match_num does not exist or it didn't matched, i.e. matching "b"
 998    * against "(a)?b" then group 0 is empty. */
 999   if (!g_match_info_fetch_pos (match_info, match_num, &start, &end))
1000     match = NULL;
1001   else if (start == -1)
1002     match = g_strdup ("");
1003   else
1004     match = g_strndup (&match_info->string[start], end - start);
1005
1006   return match;
1007 }
1008
1009 /**
1010  * g_match_info_fetch_pos:
1011  * @match_info: #GMatchInfo structure
1012  * @match_num: number of the sub expression
1013  * @start_pos: (out) (optional): pointer to location where to store
1014  *     the start position, or %NULL
1015  * @end_pos: (out) (optional): pointer to location where to store
1016  *     the end position, or %NULL
1017  *
1018  * Retrieves the position in bytes of the @match_num'th capturing
1019  * parentheses. 0 is the full text of the match, 1 is the first
1020  * paren set, 2 the second, and so on.
1021  *
1022  * If @match_num is a valid sub pattern but it didn't match anything
1023  * (e.g. sub pattern 1, matching "b" against "(a)?b") then @start_pos
1024  * and @end_pos are set to -1 and %TRUE is returned.
1025  *
1026  * If the match was obtained using the DFA algorithm, that is using
1027  * g_regex_match_all() or g_regex_match_all_full(), the retrieved
1028  * position is not that of a set of parentheses but that of a matched
1029  * substring. Substrings are matched in reverse order of length, so
1030  * 0 is the longest match.
1031  *
1032  * Returns: %TRUE if the position was fetched, %FALSE otherwise. If
1033  *   the position cannot be fetched, @start_pos and @end_pos are left
1034  *   unchanged
1035  *
1036  * Since: 2.14
1037  */
1038 gboolean
1039 g_match_info_fetch_pos (const GMatchInfo *match_info,
1040                         gint              match_num,
1041                         gint             *start_pos,
1042                         gint             *end_pos)
1043 {
1044   g_return_val_if_fail (match_info != NULL, FALSE);
1045   g_return_val_if_fail (match_num >= 0, FALSE);
1046
1047   /* make sure the sub expression number they're requesting is less than
1048    * the total number of sub expressions that were matched. */
1049   if (match_num >= match_info->matches)
1050     return FALSE;
1051
1052   if (start_pos != NULL)
1053     *start_pos = match_info->offsets[2 * match_num];
1054
1055   if (end_pos != NULL)
1056     *end_pos = match_info->offsets[2 * match_num + 1];
1057
1058   return TRUE;
1059 }
1060
1061 /*
1062  * Returns number of first matched subpattern with name @name.
1063  * There may be more than one in case when DUPNAMES is used,
1064  * and not all subpatterns with that name match;
1065  * pcre_get_stringnumber() does not work in that case.
1066  */
1067 static gint
1068 get_matched_substring_number (const GMatchInfo *match_info,
1069                               const gchar      *name)
1070 {
1071   gint entrysize;
1072   gchar *first, *last;
1073   guchar *entry;
1074
1075   if (!(match_info->regex->compile_opts & G_REGEX_DUPNAMES))
1076     return pcre_get_stringnumber (match_info->regex->pcre_re, name);
1077
1078   /* This code is copied from pcre_get.c: get_first_set() */
1079   entrysize = pcre_get_stringtable_entries (match_info->regex->pcre_re,
1080                                             name,
1081                                             &first,
1082                                             &last);
1083
1084   if (entrysize <= 0)
1085     return entrysize;
1086
1087   for (entry = (guchar*) first; entry <= (guchar*) last; entry += entrysize)
1088     {
1089       gint n = (entry[0] << 8) + entry[1];
1090       if (match_info->offsets[n*2] >= 0)
1091         return n;
1092     }
1093
1094   return (first[0] << 8) + first[1];
1095 }
1096
1097 /**
1098  * g_match_info_fetch_named:
1099  * @match_info: #GMatchInfo structure
1100  * @name: name of the subexpression
1101  *
1102  * Retrieves the text matching the capturing parentheses named @name.
1103  *
1104  * If @name is a valid sub pattern name but it didn't match anything
1105  * (e.g. sub pattern "X", matching "b" against "(?P<X>a)?b")
1106  * then an empty string is returned.
1107  *
1108  * The string is fetched from the string passed to the match function,
1109  * so you cannot call this function after freeing the string.
1110  *
1111  * Returns: (nullable): The matched substring, or %NULL if an error
1112  *     occurred. You have to free the string yourself
1113  *
1114  * Since: 2.14
1115  */
1116 gchar *
1117 g_match_info_fetch_named (const GMatchInfo *match_info,
1118                           const gchar      *name)
1119 {
1120   /* we cannot use pcre_get_named_substring() because it allocates the
1121    * string using pcre_malloc(). */
1122   gint num;
1123
1124   g_return_val_if_fail (match_info != NULL, NULL);
1125   g_return_val_if_fail (name != NULL, NULL);
1126
1127   num = get_matched_substring_number (match_info, name);
1128   if (num < 0)
1129     return NULL;
1130   else
1131     return g_match_info_fetch (match_info, num);
1132 }
1133
1134 /**
1135  * g_match_info_fetch_named_pos:
1136  * @match_info: #GMatchInfo structure
1137  * @name: name of the subexpression
1138  * @start_pos: (out) (optional): pointer to location where to store
1139  *     the start position, or %NULL
1140  * @end_pos: (out) (optional): pointer to location where to store
1141  *     the end position, or %NULL
1142  *
1143  * Retrieves the position in bytes of the capturing parentheses named @name.
1144  *
1145  * If @name is a valid sub pattern name but it didn't match anything
1146  * (e.g. sub pattern "X", matching "b" against "(?P<X>a)?b")
1147  * then @start_pos and @end_pos are set to -1 and %TRUE is returned.
1148  *
1149  * Returns: %TRUE if the position was fetched, %FALSE otherwise.
1150  *     If the position cannot be fetched, @start_pos and @end_pos
1151  *     are left unchanged.
1152  *
1153  * Since: 2.14
1154  */
1155 gboolean
1156 g_match_info_fetch_named_pos (const GMatchInfo *match_info,
1157                               const gchar      *name,
1158                               gint             *start_pos,
1159                               gint             *end_pos)
1160 {
1161   gint num;
1162
1163   g_return_val_if_fail (match_info != NULL, FALSE);
1164   g_return_val_if_fail (name != NULL, FALSE);
1165
1166   num = get_matched_substring_number (match_info, name);
1167   if (num < 0)
1168     return FALSE;
1169
1170   return g_match_info_fetch_pos (match_info, num, start_pos, end_pos);
1171 }
1172
1173 /**
1174  * g_match_info_fetch_all:
1175  * @match_info: a #GMatchInfo structure
1176  *
1177  * Bundles up pointers to each of the matching substrings from a match
1178  * and stores them in an array of gchar pointers. The first element in
1179  * the returned array is the match number 0, i.e. the entire matched
1180  * text.
1181  *
1182  * If a sub pattern didn't match anything (e.g. sub pattern 1, matching
1183  * "b" against "(a)?b") then an empty string is inserted.
1184  *
1185  * If the last match was obtained using the DFA algorithm, that is using
1186  * g_regex_match_all() or g_regex_match_all_full(), the retrieved
1187  * strings are not that matched by sets of parentheses but that of the
1188  * matched substring. Substrings are matched in reverse order of length,
1189  * so the first one is the longest match.
1190  *
1191  * The strings are fetched from the string passed to the match function,
1192  * so you cannot call this function after freeing the string.
1193  *
1194  * Returns: (transfer full): a %NULL-terminated array of gchar *
1195  *     pointers.  It must be freed using g_strfreev(). If the previous
1196  *     match failed %NULL is returned
1197  *
1198  * Since: 2.14
1199  */
1200 gchar **
1201 g_match_info_fetch_all (const GMatchInfo *match_info)
1202 {
1203   /* we cannot use pcre_get_substring_list() because the returned value
1204    * isn't suitable for g_strfreev(). */
1205   gchar **result;
1206   gint i;
1207
1208   g_return_val_if_fail (match_info != NULL, NULL);
1209
1210   if (match_info->matches < 0)
1211     return NULL;
1212
1213   result = g_new (gchar *, match_info->matches + 1);
1214   for (i = 0; i < match_info->matches; i++)
1215     result[i] = g_match_info_fetch (match_info, i);
1216   result[i] = NULL;
1217
1218   return result;
1219 }
1220
1221
1222 /* GRegex */
1223
1224 G_DEFINE_QUARK (g-regex-error-quark, g_regex_error)
1225
1226 /**
1227  * g_regex_ref:
1228  * @regex: a #GRegex
1229  *
1230  * Increases reference count of @regex by 1.
1231  *
1232  * Returns: @regex
1233  *
1234  * Since: 2.14
1235  */
1236 GRegex *
1237 g_regex_ref (GRegex *regex)
1238 {
1239   g_return_val_if_fail (regex != NULL, NULL);
1240   g_atomic_int_inc (&regex->ref_count);
1241   return regex;
1242 }
1243
1244 /**
1245  * g_regex_unref:
1246  * @regex: a #GRegex
1247  *
1248  * Decreases reference count of @regex by 1. When reference count drops
1249  * to zero, it frees all the memory associated with the regex structure.
1250  *
1251  * Since: 2.14
1252  */
1253 void
1254 g_regex_unref (GRegex *regex)
1255 {
1256   g_return_if_fail (regex != NULL);
1257
1258   if (g_atomic_int_dec_and_test (&regex->ref_count))
1259     {
1260       g_free (regex->pattern);
1261       if (regex->pcre_re != NULL)
1262         pcre_free (regex->pcre_re);
1263       if (regex->extra != NULL)
1264         pcre_free (regex->extra);
1265       g_free (regex);
1266     }
1267 }
1268
1269 /*
1270  * @match_options: (inout) (optional):
1271  */
1272 static pcre *regex_compile (const gchar         *pattern,
1273                             GRegexCompileFlags   compile_options,
1274                             GRegexCompileFlags  *compile_options_out,
1275                             GRegexMatchFlags    *match_options,
1276                             GError             **error);
1277
1278 /**
1279  * g_regex_new:
1280  * @pattern: the regular expression
1281  * @compile_options: compile options for the regular expression, or 0
1282  * @match_options: match options for the regular expression, or 0
1283  * @error: return location for a #GError
1284  *
1285  * Compiles the regular expression to an internal form, and does
1286  * the initial setup of the #GRegex structure.
1287  *
1288  * Returns: (nullable): a #GRegex structure or %NULL if an error occured. Call
1289  *   g_regex_unref() when you are done with it
1290  *
1291  * Since: 2.14
1292  */
1293 GRegex *
1294 g_regex_new (const gchar         *pattern,
1295              GRegexCompileFlags   compile_options,
1296              GRegexMatchFlags     match_options,
1297              GError             **error)
1298 {
1299   GRegex *regex;
1300   pcre *re;
1301   const gchar *errmsg;
1302   gboolean optimize = FALSE;
1303   static volatile gsize initialised = 0;
1304
1305   g_return_val_if_fail (pattern != NULL, NULL);
1306   g_return_val_if_fail (error == NULL || *error == NULL, NULL);
1307   g_return_val_if_fail ((compile_options & ~G_REGEX_COMPILE_MASK) == 0, NULL);
1308   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
1309
1310   if (g_once_init_enter (&initialised))
1311     {
1312       int supports_utf8, supports_ucp;
1313
1314       pcre_config (PCRE_CONFIG_UTF8, &supports_utf8);
1315       if (!supports_utf8)
1316         g_critical (_("PCRE library is compiled without UTF8 support"));
1317
1318       pcre_config (PCRE_CONFIG_UNICODE_PROPERTIES, &supports_ucp);
1319       if (!supports_ucp)
1320         g_critical (_("PCRE library is compiled without UTF8 properties support"));
1321
1322       g_once_init_leave (&initialised, supports_utf8 && supports_ucp ? 1 : 2);
1323     }
1324
1325   if (G_UNLIKELY (initialised != 1))
1326     {
1327       g_set_error_literal (error, G_REGEX_ERROR, G_REGEX_ERROR_COMPILE,
1328                            _("PCRE library is compiled with incompatible options"));
1329       return NULL;
1330     }
1331
1332   /* G_REGEX_OPTIMIZE has the same numeric value of PCRE_NO_UTF8_CHECK,
1333    * as we do not need to wrap PCRE_NO_UTF8_CHECK. */
1334   if (compile_options & G_REGEX_OPTIMIZE)
1335     optimize = TRUE;
1336
1337   re = regex_compile (pattern, compile_options, &compile_options,
1338                       &match_options, error);
1339
1340   if (re == NULL)
1341     return NULL;
1342
1343   regex = g_new0 (GRegex, 1);
1344   regex->ref_count = 1;
1345   regex->pattern = g_strdup (pattern);
1346   regex->pcre_re = re;
1347   regex->compile_opts = compile_options;
1348   regex->match_opts = match_options;
1349
1350   if (optimize)
1351     {
1352       regex->extra = pcre_study (regex->pcre_re, 0, &errmsg);
1353       if (errmsg != NULL)
1354         {
1355           GError *tmp_error = g_error_new (G_REGEX_ERROR,
1356                                            G_REGEX_ERROR_OPTIMIZE,
1357                                            _("Error while optimizing "
1358                                              "regular expression %s: %s"),
1359                                            regex->pattern,
1360                                            errmsg);
1361           g_propagate_error (error, tmp_error);
1362
1363           g_regex_unref (regex);
1364           return NULL;
1365         }
1366     }
1367
1368   return regex;
1369 }
1370
1371 static pcre *
1372 regex_compile (const gchar         *pattern,
1373                GRegexCompileFlags   compile_options,
1374                GRegexCompileFlags  *compile_options_out,
1375                GRegexMatchFlags    *match_options,
1376                GError             **error)
1377 {
1378   pcre *re;
1379   const gchar *errmsg;
1380   gint erroffset;
1381   gint errcode;
1382   GRegexCompileFlags nonpcre_compile_options;
1383   unsigned long int pcre_compile_options;
1384
1385   nonpcre_compile_options = compile_options & G_REGEX_COMPILE_NONPCRE_MASK;
1386
1387   /* In GRegex the string are, by default, UTF-8 encoded. PCRE
1388    * instead uses UTF-8 only if required with PCRE_UTF8. */
1389   if (compile_options & G_REGEX_RAW)
1390     {
1391       /* disable utf-8 */
1392       compile_options &= ~G_REGEX_RAW;
1393     }
1394   else
1395     {
1396       /* enable utf-8 */
1397       compile_options |= PCRE_UTF8 | PCRE_NO_UTF8_CHECK;
1398
1399       if (match_options != NULL)
1400         *match_options |= PCRE_NO_UTF8_CHECK;
1401     }
1402
1403   /* PCRE_NEWLINE_ANY is the default for the internal PCRE but
1404    * not for the system one. */
1405   if (!(compile_options & G_REGEX_NEWLINE_CR) &&
1406       !(compile_options & G_REGEX_NEWLINE_LF))
1407     {
1408       compile_options |= PCRE_NEWLINE_ANY;
1409     }
1410
1411   compile_options |= PCRE_UCP;
1412
1413   /* PCRE_BSR_UNICODE is the default for the internal PCRE but
1414    * possibly not for the system one.
1415    */
1416   if (~compile_options & G_REGEX_BSR_ANYCRLF)
1417     compile_options |= PCRE_BSR_UNICODE;
1418
1419   /* compile the pattern */
1420   re = pcre_compile2 (pattern, compile_options, &errcode,
1421                       &errmsg, &erroffset, NULL);
1422
1423   /* if the compilation failed, set the error member and return
1424    * immediately */
1425   if (re == NULL)
1426     {
1427       GError *tmp_error;
1428
1429       /* Translate the PCRE error code to GRegexError and use a translated
1430        * error message if possible */
1431       translate_compile_error (&errcode, &errmsg);
1432
1433       /* PCRE uses byte offsets but we want to show character offsets */
1434       erroffset = g_utf8_pointer_to_offset (pattern, &pattern[erroffset]);
1435
1436       tmp_error = g_error_new (G_REGEX_ERROR, errcode,
1437                                _("Error while compiling regular "
1438                                  "expression %s at char %d: %s"),
1439                                pattern, erroffset, errmsg);
1440       g_propagate_error (error, tmp_error);
1441
1442       return NULL;
1443     }
1444
1445   /* For options set at the beginning of the pattern, pcre puts them into
1446    * compile options, e.g. "(?i)foo" will make the pcre structure store
1447    * PCRE_CASELESS even though it wasn't explicitly given for compilation. */
1448   pcre_fullinfo (re, NULL, PCRE_INFO_OPTIONS, &pcre_compile_options);
1449   compile_options = pcre_compile_options & G_REGEX_COMPILE_PCRE_MASK;
1450
1451   /* Don't leak PCRE_NEWLINE_ANY, which is part of PCRE_NEWLINE_ANYCRLF */
1452   if ((pcre_compile_options & PCRE_NEWLINE_ANYCRLF) != PCRE_NEWLINE_ANYCRLF)
1453     compile_options &= ~PCRE_NEWLINE_ANY;
1454
1455   compile_options |= nonpcre_compile_options;
1456
1457   if (!(compile_options & G_REGEX_DUPNAMES))
1458     {
1459       gboolean jchanged = FALSE;
1460       pcre_fullinfo (re, NULL, PCRE_INFO_JCHANGED, &jchanged);
1461       if (jchanged)
1462         compile_options |= G_REGEX_DUPNAMES;
1463     }
1464
1465   if (compile_options_out != 0)
1466     *compile_options_out = compile_options;
1467
1468   return re;
1469 }
1470
1471 /**
1472  * g_regex_get_pattern:
1473  * @regex: a #GRegex structure
1474  *
1475  * Gets the pattern string associated with @regex, i.e. a copy of
1476  * the string passed to g_regex_new().
1477  *
1478  * Returns: the pattern of @regex
1479  *
1480  * Since: 2.14
1481  */
1482 const gchar *
1483 g_regex_get_pattern (const GRegex *regex)
1484 {
1485   g_return_val_if_fail (regex != NULL, NULL);
1486
1487   return regex->pattern;
1488 }
1489
1490 /**
1491  * g_regex_get_max_backref:
1492  * @regex: a #GRegex
1493  *
1494  * Returns the number of the highest back reference
1495  * in the pattern, or 0 if the pattern does not contain
1496  * back references.
1497  *
1498  * Returns: the number of the highest back reference
1499  *
1500  * Since: 2.14
1501  */
1502 gint
1503 g_regex_get_max_backref (const GRegex *regex)
1504 {
1505   gint value;
1506
1507   pcre_fullinfo (regex->pcre_re, regex->extra,
1508                  PCRE_INFO_BACKREFMAX, &value);
1509
1510   return value;
1511 }
1512
1513 /**
1514  * g_regex_get_capture_count:
1515  * @regex: a #GRegex
1516  *
1517  * Returns the number of capturing subpatterns in the pattern.
1518  *
1519  * Returns: the number of capturing subpatterns
1520  *
1521  * Since: 2.14
1522  */
1523 gint
1524 g_regex_get_capture_count (const GRegex *regex)
1525 {
1526   gint value;
1527
1528   pcre_fullinfo (regex->pcre_re, regex->extra,
1529                  PCRE_INFO_CAPTURECOUNT, &value);
1530
1531   return value;
1532 }
1533
1534 /**
1535  * g_regex_get_has_cr_or_lf:
1536  * @regex: a #GRegex structure
1537  *
1538  * Checks whether the pattern contains explicit CR or LF references.
1539  *
1540  * Returns: %TRUE if the pattern contains explicit CR or LF references
1541  *
1542  * Since: 2.34
1543  */
1544 gboolean
1545 g_regex_get_has_cr_or_lf (const GRegex *regex)
1546 {
1547   gint value;
1548
1549   pcre_fullinfo (regex->pcre_re, regex->extra,
1550                  PCRE_INFO_HASCRORLF, &value);
1551
1552   return !!value;
1553 }
1554
1555 /**
1556  * g_regex_get_max_lookbehind:
1557  * @regex: a #GRegex structure
1558  *
1559  * Gets the number of characters in the longest lookbehind assertion in the
1560  * pattern. This information is useful when doing multi-segment matching using
1561  * the partial matching facilities.
1562  *
1563  * Returns: the number of characters in the longest lookbehind assertion.
1564  *
1565  * Since: 2.38
1566  */
1567 gint
1568 g_regex_get_max_lookbehind (const GRegex *regex)
1569 {
1570   gint max_lookbehind;
1571
1572   pcre_fullinfo (regex->pcre_re, regex->extra,
1573                  PCRE_INFO_MAXLOOKBEHIND, &max_lookbehind);
1574
1575   return max_lookbehind;
1576 }
1577
1578 /**
1579  * g_regex_get_compile_flags:
1580  * @regex: a #GRegex
1581  *
1582  * Returns the compile options that @regex was created with.
1583  *
1584  * Depending on the version of PCRE that is used, this may or may not
1585  * include flags set by option expressions such as `(?i)` found at the
1586  * top-level within the compiled pattern.
1587  *
1588  * Returns: flags from #GRegexCompileFlags
1589  *
1590  * Since: 2.26
1591  */
1592 GRegexCompileFlags
1593 g_regex_get_compile_flags (const GRegex *regex)
1594 {
1595   g_return_val_if_fail (regex != NULL, 0);
1596
1597   return regex->compile_opts;
1598 }
1599
1600 /**
1601  * g_regex_get_match_flags:
1602  * @regex: a #GRegex
1603  *
1604  * Returns the match options that @regex was created with.
1605  *
1606  * Returns: flags from #GRegexMatchFlags
1607  *
1608  * Since: 2.26
1609  */
1610 GRegexMatchFlags
1611 g_regex_get_match_flags (const GRegex *regex)
1612 {
1613   g_return_val_if_fail (regex != NULL, 0);
1614
1615   return regex->match_opts & G_REGEX_MATCH_MASK;
1616 }
1617
1618 /**
1619  * g_regex_match_simple:
1620  * @pattern: the regular expression
1621  * @string: the string to scan for matches
1622  * @compile_options: compile options for the regular expression, or 0
1623  * @match_options: match options, or 0
1624  *
1625  * Scans for a match in @string for @pattern.
1626  *
1627  * This function is equivalent to g_regex_match() but it does not
1628  * require to compile the pattern with g_regex_new(), avoiding some
1629  * lines of code when you need just to do a match without extracting
1630  * substrings, capture counts, and so on.
1631  *
1632  * If this function is to be called on the same @pattern more than
1633  * once, it's more efficient to compile the pattern once with
1634  * g_regex_new() and then use g_regex_match().
1635  *
1636  * Returns: %TRUE if the string matched, %FALSE otherwise
1637  *
1638  * Since: 2.14
1639  */
1640 gboolean
1641 g_regex_match_simple (const gchar        *pattern,
1642                       const gchar        *string,
1643                       GRegexCompileFlags  compile_options,
1644                       GRegexMatchFlags    match_options)
1645 {
1646   GRegex *regex;
1647   gboolean result;
1648
1649   regex = g_regex_new (pattern, compile_options, 0, NULL);
1650   if (!regex)
1651     return FALSE;
1652   result = g_regex_match_full (regex, string, -1, 0, match_options, NULL, NULL);
1653   g_regex_unref (regex);
1654   return result;
1655 }
1656
1657 /**
1658  * g_regex_match:
1659  * @regex: a #GRegex structure from g_regex_new()
1660  * @string: the string to scan for matches
1661  * @match_options: match options
1662  * @match_info: (out) (optional): pointer to location where to store
1663  *     the #GMatchInfo, or %NULL if you do not need it
1664  *
1665  * Scans for a match in string for the pattern in @regex.
1666  * The @match_options are combined with the match options specified
1667  * when the @regex structure was created, letting you have more
1668  * flexibility in reusing #GRegex structures.
1669  *
1670  * A #GMatchInfo structure, used to get information on the match,
1671  * is stored in @match_info if not %NULL. Note that if @match_info
1672  * is not %NULL then it is created even if the function returns %FALSE,
1673  * i.e. you must free it regardless if regular expression actually matched.
1674  *
1675  * To retrieve all the non-overlapping matches of the pattern in
1676  * string you can use g_match_info_next().
1677  *
1678  * |[<!-- language="C" -->
1679  * static void
1680  * print_uppercase_words (const gchar *string)
1681  * {
1682  *   // Print all uppercase-only words.
1683  *   GRegex *regex;
1684  *   GMatchInfo *match_info;
1685  *
1686  *   regex = g_regex_new ("[A-Z]+", 0, 0, NULL);
1687  *   g_regex_match (regex, string, 0, &match_info);
1688  *   while (g_match_info_matches (match_info))
1689  *     {
1690  *       gchar *word = g_match_info_fetch (match_info, 0);
1691  *       g_print ("Found: %s\n", word);
1692  *       g_free (word);
1693  *       g_match_info_next (match_info, NULL);
1694  *     }
1695  *   g_match_info_free (match_info);
1696  *   g_regex_unref (regex);
1697  * }
1698  * ]|
1699  *
1700  * @string is not copied and is used in #GMatchInfo internally. If
1701  * you use any #GMatchInfo method (except g_match_info_free()) after
1702  * freeing or modifying @string then the behaviour is undefined.
1703  *
1704  * Returns: %TRUE is the string matched, %FALSE otherwise
1705  *
1706  * Since: 2.14
1707  */
1708 gboolean
1709 g_regex_match (const GRegex      *regex,
1710                const gchar       *string,
1711                GRegexMatchFlags   match_options,
1712                GMatchInfo       **match_info)
1713 {
1714   return g_regex_match_full (regex, string, -1, 0, match_options,
1715                              match_info, NULL);
1716 }
1717
1718 /**
1719  * g_regex_match_full:
1720  * @regex: a #GRegex structure from g_regex_new()
1721  * @string: (array length=string_len): the string to scan for matches
1722  * @string_len: the length of @string, or -1 if @string is nul-terminated
1723  * @start_position: starting index of the string to match, in bytes
1724  * @match_options: match options
1725  * @match_info: (out) (optional): pointer to location where to store
1726  *     the #GMatchInfo, or %NULL if you do not need it
1727  * @error: location to store the error occurring, or %NULL to ignore errors
1728  *
1729  * Scans for a match in string for the pattern in @regex.
1730  * The @match_options are combined with the match options specified
1731  * when the @regex structure was created, letting you have more
1732  * flexibility in reusing #GRegex structures.
1733  *
1734  * Setting @start_position differs from just passing over a shortened
1735  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern
1736  * that begins with any kind of lookbehind assertion, such as "\b".
1737  *
1738  * A #GMatchInfo structure, used to get information on the match, is
1739  * stored in @match_info if not %NULL. Note that if @match_info is
1740  * not %NULL then it is created even if the function returns %FALSE,
1741  * i.e. you must free it regardless if regular expression actually
1742  * matched.
1743  *
1744  * @string is not copied and is used in #GMatchInfo internally. If
1745  * you use any #GMatchInfo method (except g_match_info_free()) after
1746  * freeing or modifying @string then the behaviour is undefined.
1747  *
1748  * To retrieve all the non-overlapping matches of the pattern in
1749  * string you can use g_match_info_next().
1750  *
1751  * |[<!-- language="C" -->
1752  * static void
1753  * print_uppercase_words (const gchar *string)
1754  * {
1755  *   // Print all uppercase-only words.
1756  *   GRegex *regex;
1757  *   GMatchInfo *match_info;
1758  *   GError *error = NULL;
1759  *
1760  *   regex = g_regex_new ("[A-Z]+", 0, 0, NULL);
1761  *   g_regex_match_full (regex, string, -1, 0, 0, &match_info, &error);
1762  *   while (g_match_info_matches (match_info))
1763  *     {
1764  *       gchar *word = g_match_info_fetch (match_info, 0);
1765  *       g_print ("Found: %s\n", word);
1766  *       g_free (word);
1767  *       g_match_info_next (match_info, &error);
1768  *     }
1769  *   g_match_info_free (match_info);
1770  *   g_regex_unref (regex);
1771  *   if (error != NULL)
1772  *     {
1773  *       g_printerr ("Error while matching: %s\n", error->message);
1774  *       g_error_free (error);
1775  *     }
1776  * }
1777  * ]|
1778  *
1779  * Returns: %TRUE is the string matched, %FALSE otherwise
1780  *
1781  * Since: 2.14
1782  */
1783 gboolean
1784 g_regex_match_full (const GRegex      *regex,
1785                     const gchar       *string,
1786                     gssize             string_len,
1787                     gint               start_position,
1788                     GRegexMatchFlags   match_options,
1789                     GMatchInfo       **match_info,
1790                     GError           **error)
1791 {
1792   GMatchInfo *info;
1793   gboolean match_ok;
1794
1795   g_return_val_if_fail (regex != NULL, FALSE);
1796   g_return_val_if_fail (string != NULL, FALSE);
1797   g_return_val_if_fail (start_position >= 0, FALSE);
1798   g_return_val_if_fail (error == NULL || *error == NULL, FALSE);
1799   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, FALSE);
1800
1801   info = match_info_new (regex, string, string_len, start_position,
1802                          match_options, FALSE);
1803   match_ok = g_match_info_next (info, error);
1804   if (match_info != NULL)
1805     *match_info = info;
1806   else
1807     g_match_info_free (info);
1808
1809   return match_ok;
1810 }
1811
1812 /**
1813  * g_regex_match_all:
1814  * @regex: a #GRegex structure from g_regex_new()
1815  * @string: the string to scan for matches
1816  * @match_options: match options
1817  * @match_info: (out) (optional): pointer to location where to store
1818  *     the #GMatchInfo, or %NULL if you do not need it
1819  *
1820  * Using the standard algorithm for regular expression matching only
1821  * the longest match in the string is retrieved. This function uses
1822  * a different algorithm so it can retrieve all the possible matches.
1823  * For more documentation see g_regex_match_all_full().
1824  *
1825  * A #GMatchInfo structure, used to get information on the match, is
1826  * stored in @match_info if not %NULL. Note that if @match_info is
1827  * not %NULL then it is created even if the function returns %FALSE,
1828  * i.e. you must free it regardless if regular expression actually
1829  * matched.
1830  *
1831  * @string is not copied and is used in #GMatchInfo internally. If
1832  * you use any #GMatchInfo method (except g_match_info_free()) after
1833  * freeing or modifying @string then the behaviour is undefined.
1834  *
1835  * Returns: %TRUE is the string matched, %FALSE otherwise
1836  *
1837  * Since: 2.14
1838  */
1839 gboolean
1840 g_regex_match_all (const GRegex      *regex,
1841                    const gchar       *string,
1842                    GRegexMatchFlags   match_options,
1843                    GMatchInfo       **match_info)
1844 {
1845   return g_regex_match_all_full (regex, string, -1, 0, match_options,
1846                                  match_info, NULL);
1847 }
1848
1849 /**
1850  * g_regex_match_all_full:
1851  * @regex: a #GRegex structure from g_regex_new()
1852  * @string: (array length=string_len): the string to scan for matches
1853  * @string_len: the length of @string, or -1 if @string is nul-terminated
1854  * @start_position: starting index of the string to match, in bytes
1855  * @match_options: match options
1856  * @match_info: (out) (optional): pointer to location where to store
1857  *     the #GMatchInfo, or %NULL if you do not need it
1858  * @error: location to store the error occurring, or %NULL to ignore errors
1859  *
1860  * Using the standard algorithm for regular expression matching only
1861  * the longest match in the string is retrieved, it is not possible
1862  * to obtain all the available matches. For instance matching
1863  * "<a> <b> <c>" against the pattern "<.*>"
1864  * you get "<a> <b> <c>".
1865  *
1866  * This function uses a different algorithm (called DFA, i.e. deterministic
1867  * finite automaton), so it can retrieve all the possible matches, all
1868  * starting at the same point in the string. For instance matching
1869  * "<a> <b> <c>" against the pattern "<.*>;"
1870  * you would obtain three matches: "<a> <b> <c>",
1871  * "<a> <b>" and "<a>".
1872  *
1873  * The number of matched strings is retrieved using
1874  * g_match_info_get_match_count(). To obtain the matched strings and
1875  * their position you can use, respectively, g_match_info_fetch() and
1876  * g_match_info_fetch_pos(). Note that the strings are returned in
1877  * reverse order of length; that is, the longest matching string is
1878  * given first.
1879  *
1880  * Note that the DFA algorithm is slower than the standard one and it
1881  * is not able to capture substrings, so backreferences do not work.
1882  *
1883  * Setting @start_position differs from just passing over a shortened
1884  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern
1885  * that begins with any kind of lookbehind assertion, such as "\b".
1886  *
1887  * A #GMatchInfo structure, used to get information on the match, is
1888  * stored in @match_info if not %NULL. Note that if @match_info is
1889  * not %NULL then it is created even if the function returns %FALSE,
1890  * i.e. you must free it regardless if regular expression actually
1891  * matched.
1892  *
1893  * @string is not copied and is used in #GMatchInfo internally. If
1894  * you use any #GMatchInfo method (except g_match_info_free()) after
1895  * freeing or modifying @string then the behaviour is undefined.
1896  *
1897  * Returns: %TRUE is the string matched, %FALSE otherwise
1898  *
1899  * Since: 2.14
1900  */
1901 gboolean
1902 g_regex_match_all_full (const GRegex      *regex,
1903                         const gchar       *string,
1904                         gssize             string_len,
1905                         gint               start_position,
1906                         GRegexMatchFlags   match_options,
1907                         GMatchInfo       **match_info,
1908                         GError           **error)
1909 {
1910   GMatchInfo *info;
1911   gboolean done;
1912   pcre *pcre_re;
1913   pcre_extra *extra;
1914   gboolean retval;
1915
1916   g_return_val_if_fail (regex != NULL, FALSE);
1917   g_return_val_if_fail (string != NULL, FALSE);
1918   g_return_val_if_fail (start_position >= 0, FALSE);
1919   g_return_val_if_fail (error == NULL || *error == NULL, FALSE);
1920   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, FALSE);
1921
1922 #ifdef PCRE_NO_AUTO_POSSESS
1923   /* For PCRE >= 8.34 we need to turn off PCRE_NO_AUTO_POSSESS, which
1924    * is an optimization for normal regex matching, but results in omitting
1925    * some shorter matches here, and an observable behaviour change.
1926    *
1927    * DFA matching is rather niche, and very rarely used according to
1928    * codesearch.debian.net, so don't bother caching the recompiled RE. */
1929   pcre_re = regex_compile (regex->pattern,
1930                            regex->compile_opts | PCRE_NO_AUTO_POSSESS,
1931                            NULL, NULL, error);
1932
1933   if (pcre_re == NULL)
1934     return FALSE;
1935
1936   /* Not bothering to cache the optimization data either, with similar
1937    * reasoning */
1938   extra = NULL;
1939 #else
1940   /* For PCRE < 8.33 the precompiled regex is fine. */
1941   pcre_re = regex->pcre_re;
1942   extra = regex->extra;
1943 #endif
1944
1945   info = match_info_new (regex, string, string_len, start_position,
1946                          match_options, TRUE);
1947
1948   done = FALSE;
1949   while (!done)
1950     {
1951       done = TRUE;
1952       info->matches = pcre_dfa_exec (pcre_re, extra,
1953                                      info->string, info->string_len,
1954                                      info->pos,
1955                                      regex->match_opts | match_options,
1956                                      info->offsets, info->n_offsets,
1957                                      info->workspace, info->n_workspace);
1958       if (info->matches == PCRE_ERROR_DFA_WSSIZE)
1959         {
1960           /* info->workspace is too small. */
1961           info->n_workspace *= 2;
1962           info->workspace = g_realloc (info->workspace,
1963                                        info->n_workspace * sizeof (gint));
1964           done = FALSE;
1965         }
1966       else if (info->matches == 0)
1967         {
1968           /* info->offsets is too small. */
1969           info->n_offsets *= 2;
1970           info->offsets = g_realloc (info->offsets,
1971                                      info->n_offsets * sizeof (gint));
1972           done = FALSE;
1973         }
1974       else if (IS_PCRE_ERROR (info->matches))
1975         {
1976           g_set_error (error, G_REGEX_ERROR, G_REGEX_ERROR_MATCH,
1977                        _("Error while matching regular expression %s: %s"),
1978                        regex->pattern, match_error (info->matches));
1979         }
1980     }
1981
1982 #ifdef PCRE_NO_AUTO_POSSESS
1983   pcre_free (pcre_re);
1984 #endif
1985
1986   /* set info->pos to -1 so that a call to g_match_info_next() fails. */
1987   info->pos = -1;
1988   retval = info->matches >= 0;
1989
1990   if (match_info != NULL)
1991     *match_info = info;
1992   else
1993     g_match_info_free (info);
1994
1995   return retval;
1996 }
1997
1998 /**
1999  * g_regex_get_string_number:
2000  * @regex: #GRegex structure
2001  * @name: name of the subexpression
2002  *
2003  * Retrieves the number of the subexpression named @name.
2004  *
2005  * Returns: The number of the subexpression or -1 if @name
2006  *   does not exists
2007  *
2008  * Since: 2.14
2009  */
2010 gint
2011 g_regex_get_string_number (const GRegex *regex,
2012                            const gchar  *name)
2013 {
2014   gint num;
2015
2016   g_return_val_if_fail (regex != NULL, -1);
2017   g_return_val_if_fail (name != NULL, -1);
2018
2019   num = pcre_get_stringnumber (regex->pcre_re, name);
2020   if (num == PCRE_ERROR_NOSUBSTRING)
2021     num = -1;
2022
2023   return num;
2024 }
2025
2026 /**
2027  * g_regex_split_simple:
2028  * @pattern: the regular expression
2029  * @string: the string to scan for matches
2030  * @compile_options: compile options for the regular expression, or 0
2031  * @match_options: match options, or 0
2032  *
2033  * Breaks the string on the pattern, and returns an array of
2034  * the tokens. If the pattern contains capturing parentheses,
2035  * then the text for each of the substrings will also be returned.
2036  * If the pattern does not match anywhere in the string, then the
2037  * whole string is returned as the first token.
2038  *
2039  * This function is equivalent to g_regex_split() but it does
2040  * not require to compile the pattern with g_regex_new(), avoiding
2041  * some lines of code when you need just to do a split without
2042  * extracting substrings, capture counts, and so on.
2043  *
2044  * If this function is to be called on the same @pattern more than
2045  * once, it's more efficient to compile the pattern once with
2046  * g_regex_new() and then use g_regex_split().
2047  *
2048  * As a special case, the result of splitting the empty string ""
2049  * is an empty vector, not a vector containing a single string.
2050  * The reason for this special case is that being able to represent
2051  * a empty vector is typically more useful than consistent handling
2052  * of empty elements. If you do need to represent empty elements,
2053  * you'll need to check for the empty string before calling this
2054  * function.
2055  *
2056  * A pattern that can match empty strings splits @string into
2057  * separate characters wherever it matches the empty string between
2058  * characters. For example splitting "ab c" using as a separator
2059  * "\s*", you will get "a", "b" and "c".
2060  *
2061  * Returns: (transfer full): a %NULL-terminated array of strings. Free
2062  * it using g_strfreev()
2063  *
2064  * Since: 2.14
2065  **/
2066 gchar **
2067 g_regex_split_simple (const gchar        *pattern,
2068                       const gchar        *string,
2069                       GRegexCompileFlags  compile_options,
2070                       GRegexMatchFlags    match_options)
2071 {
2072   GRegex *regex;
2073   gchar **result;
2074
2075   regex = g_regex_new (pattern, compile_options, 0, NULL);
2076   if (!regex)
2077     return NULL;
2078
2079   result = g_regex_split_full (regex, string, -1, 0, match_options, 0, NULL);
2080   g_regex_unref (regex);
2081   return result;
2082 }
2083
2084 /**
2085  * g_regex_split:
2086  * @regex: a #GRegex structure
2087  * @string: the string to split with the pattern
2088  * @match_options: match time option flags
2089  *
2090  * Breaks the string on the pattern, and returns an array of the tokens.
2091  * If the pattern contains capturing parentheses, then the text for each
2092  * of the substrings will also be returned. If the pattern does not match
2093  * anywhere in the string, then the whole string is returned as the first
2094  * token.
2095  *
2096  * As a special case, the result of splitting the empty string "" is an
2097  * empty vector, not a vector containing a single string. The reason for
2098  * this special case is that being able to represent a empty vector is
2099  * typically more useful than consistent handling of empty elements. If
2100  * you do need to represent empty elements, you'll need to check for the
2101  * empty string before calling this function.
2102  *
2103  * A pattern that can match empty strings splits @string into separate
2104  * characters wherever it matches the empty string between characters.
2105  * For example splitting "ab c" using as a separator "\s*", you will get
2106  * "a", "b" and "c".
2107  *
2108  * Returns: (transfer full): a %NULL-terminated gchar ** array. Free
2109  * it using g_strfreev()
2110  *
2111  * Since: 2.14
2112  **/
2113 gchar **
2114 g_regex_split (const GRegex     *regex,
2115                const gchar      *string,
2116                GRegexMatchFlags  match_options)
2117 {
2118   return g_regex_split_full (regex, string, -1, 0,
2119                              match_options, 0, NULL);
2120 }
2121
2122 /**
2123  * g_regex_split_full:
2124  * @regex: a #GRegex structure
2125  * @string: (array length=string_len): the string to split with the pattern
2126  * @string_len: the length of @string, or -1 if @string is nul-terminated
2127  * @start_position: starting index of the string to match, in bytes
2128  * @match_options: match time option flags
2129  * @max_tokens: the maximum number of tokens to split @string into.
2130  *   If this is less than 1, the string is split completely
2131  * @error: return location for a #GError
2132  *
2133  * Breaks the string on the pattern, and returns an array of the tokens.
2134  * If the pattern contains capturing parentheses, then the text for each
2135  * of the substrings will also be returned. If the pattern does not match
2136  * anywhere in the string, then the whole string is returned as the first
2137  * token.
2138  *
2139  * As a special case, the result of splitting the empty string "" is an
2140  * empty vector, not a vector containing a single string. The reason for
2141  * this special case is that being able to represent a empty vector is
2142  * typically more useful than consistent handling of empty elements. If
2143  * you do need to represent empty elements, you'll need to check for the
2144  * empty string before calling this function.
2145  *
2146  * A pattern that can match empty strings splits @string into separate
2147  * characters wherever it matches the empty string between characters.
2148  * For example splitting "ab c" using as a separator "\s*", you will get
2149  * "a", "b" and "c".
2150  *
2151  * Setting @start_position differs from just passing over a shortened
2152  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern
2153  * that begins with any kind of lookbehind assertion, such as "\b".
2154  *
2155  * Returns: (transfer full): a %NULL-terminated gchar ** array. Free
2156  * it using g_strfreev()
2157  *
2158  * Since: 2.14
2159  **/
2160 gchar **
2161 g_regex_split_full (const GRegex      *regex,
2162                     const gchar       *string,
2163                     gssize             string_len,
2164                     gint               start_position,
2165                     GRegexMatchFlags   match_options,
2166                     gint               max_tokens,
2167                     GError           **error)
2168 {
2169   GError *tmp_error = NULL;
2170   GMatchInfo *match_info;
2171   GList *list, *last;
2172   gint i;
2173   gint token_count;
2174   gboolean match_ok;
2175   /* position of the last separator. */
2176   gint last_separator_end;
2177   /* was the last match 0 bytes long? */
2178   gboolean last_match_is_empty;
2179   /* the returned array of char **s */
2180   gchar **string_list;
2181
2182   g_return_val_if_fail (regex != NULL, NULL);
2183   g_return_val_if_fail (string != NULL, NULL);
2184   g_return_val_if_fail (start_position >= 0, NULL);
2185   g_return_val_if_fail (error == NULL || *error == NULL, NULL);
2186   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
2187
2188   if (max_tokens <= 0)
2189     max_tokens = G_MAXINT;
2190
2191   if (string_len < 0)
2192     string_len = strlen (string);
2193
2194   /* zero-length string */
2195   if (string_len - start_position == 0)
2196     return g_new0 (gchar *, 1);
2197
2198   if (max_tokens == 1)
2199     {
2200       string_list = g_new0 (gchar *, 2);
2201       string_list[0] = g_strndup (&string[start_position],
2202                                   string_len - start_position);
2203       return string_list;
2204     }
2205
2206   list = NULL;
2207   token_count = 0;
2208   last_separator_end = start_position;
2209   last_match_is_empty = FALSE;
2210
2211   match_ok = g_regex_match_full (regex, string, string_len, start_position,
2212                                  match_options, &match_info, &tmp_error);
2213
2214   while (tmp_error == NULL)
2215     {
2216       if (match_ok)
2217         {
2218           last_match_is_empty =
2219                     (match_info->offsets[0] == match_info->offsets[1]);
2220
2221           /* we need to skip empty separators at the same position of the end
2222            * of another separator. e.g. the string is "a b" and the separator
2223            * is " *", so from 1 to 2 we have a match and at position 2 we have
2224            * an empty match. */
2225           if (last_separator_end != match_info->offsets[1])
2226             {
2227               gchar *token;
2228               gint match_count;
2229
2230               token = g_strndup (string + last_separator_end,
2231                                  match_info->offsets[0] - last_separator_end);
2232               list = g_list_prepend (list, token);
2233               token_count++;
2234
2235               /* if there were substrings, these need to be added to
2236                * the list. */
2237               match_count = g_match_info_get_match_count (match_info);
2238               if (match_count > 1)
2239                 {
2240                   for (i = 1; i < match_count; i++)
2241                     list = g_list_prepend (list, g_match_info_fetch (match_info, i));
2242                 }
2243             }
2244         }
2245       else
2246         {
2247           /* if there was no match, copy to end of string. */
2248           if (!last_match_is_empty)
2249             {
2250               gchar *token = g_strndup (string + last_separator_end,
2251                                         match_info->string_len - last_separator_end);
2252               list = g_list_prepend (list, token);
2253             }
2254           /* no more tokens, end the loop. */
2255           break;
2256         }
2257
2258       /* -1 to leave room for the last part. */
2259       if (token_count >= max_tokens - 1)
2260         {
2261           /* we have reached the maximum number of tokens, so we copy
2262            * the remaining part of the string. */
2263           if (last_match_is_empty)
2264             {
2265               /* the last match was empty, so we have moved one char
2266                * after the real position to avoid empty matches at the
2267                * same position. */
2268               match_info->pos = PREV_CHAR (regex, &string[match_info->pos]) - string;
2269             }
2270           /* the if is needed in the case we have terminated the available
2271            * tokens, but we are at the end of the string, so there are no
2272            * characters left to copy. */
2273           if (string_len > match_info->pos)
2274             {
2275               gchar *token = g_strndup (string + match_info->pos,
2276                                         string_len - match_info->pos);
2277               list = g_list_prepend (list, token);
2278             }
2279           /* end the loop. */
2280           break;
2281         }
2282
2283       last_separator_end = match_info->pos;
2284       if (last_match_is_empty)
2285         /* if the last match was empty, g_match_info_next() has moved
2286          * forward to avoid infinite loops, but we still need to copy that
2287          * character. */
2288         last_separator_end = PREV_CHAR (regex, &string[last_separator_end]) - string;
2289
2290       match_ok = g_match_info_next (match_info, &tmp_error);
2291     }
2292   g_match_info_free (match_info);
2293   if (tmp_error != NULL)
2294     {
2295       g_propagate_error (error, tmp_error);
2296       g_list_free_full (list, g_free);
2297       return NULL;
2298     }
2299
2300   string_list = g_new (gchar *, g_list_length (list) + 1);
2301   i = 0;
2302   for (last = g_list_last (list); last; last = g_list_previous (last))
2303     string_list[i++] = last->data;
2304   string_list[i] = NULL;
2305   g_list_free (list);
2306
2307   return string_list;
2308 }
2309
2310 enum
2311 {
2312   REPL_TYPE_STRING,
2313   REPL_TYPE_CHARACTER,
2314   REPL_TYPE_SYMBOLIC_REFERENCE,
2315   REPL_TYPE_NUMERIC_REFERENCE,
2316   REPL_TYPE_CHANGE_CASE
2317 };
2318
2319 typedef enum
2320 {
2321   CHANGE_CASE_NONE         = 1 << 0,
2322   CHANGE_CASE_UPPER        = 1 << 1,
2323   CHANGE_CASE_LOWER        = 1 << 2,
2324   CHANGE_CASE_UPPER_SINGLE = 1 << 3,
2325   CHANGE_CASE_LOWER_SINGLE = 1 << 4,
2326   CHANGE_CASE_SINGLE_MASK  = CHANGE_CASE_UPPER_SINGLE | CHANGE_CASE_LOWER_SINGLE,
2327   CHANGE_CASE_LOWER_MASK   = CHANGE_CASE_LOWER | CHANGE_CASE_LOWER_SINGLE,
2328   CHANGE_CASE_UPPER_MASK   = CHANGE_CASE_UPPER | CHANGE_CASE_UPPER_SINGLE
2329 } ChangeCase;
2330
2331 struct _InterpolationData
2332 {
2333   gchar     *text;
2334   gint       type;
2335   gint       num;
2336   gchar      c;
2337   ChangeCase change_case;
2338 };
2339
2340 static void
2341 free_interpolation_data (InterpolationData *data)
2342 {
2343   g_free (data->text);
2344   g_free (data);
2345 }
2346
2347 static const gchar *
2348 expand_escape (const gchar        *replacement,
2349                const gchar        *p,
2350                InterpolationData  *data,
2351                GError            **error)
2352 {
2353   const gchar *q, *r;
2354   gint x, d, h, i;
2355   const gchar *error_detail;
2356   gint base = 0;
2357   GError *tmp_error = NULL;
2358
2359   p++;
2360   switch (*p)
2361     {
2362     case 't':
2363       p++;
2364       data->c = '\t';
2365       data->type = REPL_TYPE_CHARACTER;
2366       break;
2367     case 'n':
2368       p++;
2369       data->c = '\n';
2370       data->type = REPL_TYPE_CHARACTER;
2371       break;
2372     case 'v':
2373       p++;
2374       data->c = '\v';
2375       data->type = REPL_TYPE_CHARACTER;
2376       break;
2377     case 'r':
2378       p++;
2379       data->c = '\r';
2380       data->type = REPL_TYPE_CHARACTER;
2381       break;
2382     case 'f':
2383       p++;
2384       data->c = '\f';
2385       data->type = REPL_TYPE_CHARACTER;
2386       break;
2387     case 'a':
2388       p++;
2389       data->c = '\a';
2390       data->type = REPL_TYPE_CHARACTER;
2391       break;
2392     case 'b':
2393       p++;
2394       data->c = '\b';
2395       data->type = REPL_TYPE_CHARACTER;
2396       break;
2397     case '\\':
2398       p++;
2399       data->c = '\\';
2400       data->type = REPL_TYPE_CHARACTER;
2401       break;
2402     case 'x':
2403       p++;
2404       x = 0;
2405       if (*p == '{')
2406         {
2407           p++;
2408           do
2409             {
2410               h = g_ascii_xdigit_value (*p);
2411               if (h < 0)
2412                 {
2413                   error_detail = _("hexadecimal digit or “}” expected");
2414                   goto error;
2415                 }
2416               x = x * 16 + h;
2417               p++;
2418             }
2419           while (*p != '}');
2420           p++;
2421         }
2422       else
2423         {
2424           for (i = 0; i < 2; i++)
2425             {
2426               h = g_ascii_xdigit_value (*p);
2427               if (h < 0)
2428                 {
2429                   error_detail = _("hexadecimal digit expected");
2430                   goto error;
2431                 }
2432               x = x * 16 + h;
2433               p++;
2434             }
2435         }
2436       data->type = REPL_TYPE_STRING;
2437       data->text = g_new0 (gchar, 8);
2438       g_unichar_to_utf8 (x, data->text);
2439       break;
2440     case 'l':
2441       p++;
2442       data->type = REPL_TYPE_CHANGE_CASE;
2443       data->change_case = CHANGE_CASE_LOWER_SINGLE;
2444       break;
2445     case 'u':
2446       p++;
2447       data->type = REPL_TYPE_CHANGE_CASE;
2448       data->change_case = CHANGE_CASE_UPPER_SINGLE;
2449       break;
2450     case 'L':
2451       p++;
2452       data->type = REPL_TYPE_CHANGE_CASE;
2453       data->change_case = CHANGE_CASE_LOWER;
2454       break;
2455     case 'U':
2456       p++;
2457       data->type = REPL_TYPE_CHANGE_CASE;
2458       data->change_case = CHANGE_CASE_UPPER;
2459       break;
2460     case 'E':
2461       p++;
2462       data->type = REPL_TYPE_CHANGE_CASE;
2463       data->change_case = CHANGE_CASE_NONE;
2464       break;
2465     case 'g':
2466       p++;
2467       if (*p != '<')
2468         {
2469           error_detail = _("missing “<” in symbolic reference");
2470           goto error;
2471         }
2472       q = p + 1;
2473       do
2474         {
2475           p++;
2476           if (!*p)
2477             {
2478               error_detail = _("unfinished symbolic reference");
2479               goto error;
2480             }
2481         }
2482       while (*p != '>');
2483       if (p - q == 0)
2484         {
2485           error_detail = _("zero-length symbolic reference");
2486           goto error;
2487         }
2488       if (g_ascii_isdigit (*q))
2489         {
2490           x = 0;
2491           do
2492             {
2493               h = g_ascii_digit_value (*q);
2494               if (h < 0)
2495                 {
2496                   error_detail = _("digit expected");
2497                   p = q;
2498                   goto error;
2499                 }
2500               x = x * 10 + h;
2501               q++;
2502             }
2503           while (q != p);
2504           data->num = x;
2505           data->type = REPL_TYPE_NUMERIC_REFERENCE;
2506         }
2507       else
2508         {
2509           r = q;
2510           do
2511             {
2512               if (!g_ascii_isalnum (*r))
2513                 {
2514                   error_detail = _("illegal symbolic reference");
2515                   p = r;
2516                   goto error;
2517                 }
2518               r++;
2519             }
2520           while (r != p);
2521           data->text = g_strndup (q, p - q);
2522           data->type = REPL_TYPE_SYMBOLIC_REFERENCE;
2523         }
2524       p++;
2525       break;
2526     case '0':
2527       /* if \0 is followed by a number is an octal number representing a
2528        * character, else it is a numeric reference. */
2529       if (g_ascii_digit_value (*g_utf8_next_char (p)) >= 0)
2530         {
2531           base = 8;
2532           p = g_utf8_next_char (p);
2533         }
2534     case '1':
2535     case '2':
2536     case '3':
2537     case '4':
2538     case '5':
2539     case '6':
2540     case '7':
2541     case '8':
2542     case '9':
2543       x = 0;
2544       d = 0;
2545       for (i = 0; i < 3; i++)
2546         {
2547           h = g_ascii_digit_value (*p);
2548           if (h < 0)
2549             break;
2550           if (h > 7)
2551             {
2552               if (base == 8)
2553                 break;
2554               else
2555                 base = 10;
2556             }
2557           if (i == 2 && base == 10)
2558             break;
2559           x = x * 8 + h;
2560           d = d * 10 + h;
2561           p++;
2562         }
2563       if (base == 8 || i == 3)
2564         {
2565           data->type = REPL_TYPE_STRING;
2566           data->text = g_new0 (gchar, 8);
2567           g_unichar_to_utf8 (x, data->text);
2568         }
2569       else
2570         {
2571           data->type = REPL_TYPE_NUMERIC_REFERENCE;
2572           data->num = d;
2573         }
2574       break;
2575     case 0:
2576       error_detail = _("stray final “\\”");
2577       goto error;
2578       break;
2579     default:
2580       error_detail = _("unknown escape sequence");
2581       goto error;
2582     }
2583
2584   return p;
2585
2586  error:
2587   /* G_GSSIZE_FORMAT doesn't work with gettext, so we use %lu */
2588   tmp_error = g_error_new (G_REGEX_ERROR,
2589                            G_REGEX_ERROR_REPLACE,
2590                            _("Error while parsing replacement "
2591                              "text “%s” at char %lu: %s"),
2592                            replacement,
2593                            (gulong)(p - replacement),
2594                            error_detail);
2595   g_propagate_error (error, tmp_error);
2596
2597   return NULL;
2598 }
2599
2600 static GList *
2601 split_replacement (const gchar  *replacement,
2602                    GError      **error)
2603 {
2604   GList *list = NULL;
2605   InterpolationData *data;
2606   const gchar *p, *start;
2607
2608   start = p = replacement;
2609   while (*p)
2610     {
2611       if (*p == '\\')
2612         {
2613           data = g_new0 (InterpolationData, 1);
2614           start = p = expand_escape (replacement, p, data, error);
2615           if (p == NULL)
2616             {
2617               g_list_free_full (list, (GDestroyNotify) free_interpolation_data);
2618               free_interpolation_data (data);
2619
2620               return NULL;
2621             }
2622           list = g_list_prepend (list, data);
2623         }
2624       else
2625         {
2626           p++;
2627           if (*p == '\\' || *p == '\0')
2628             {
2629               if (p - start > 0)
2630                 {
2631                   data = g_new0 (InterpolationData, 1);
2632                   data->text = g_strndup (start, p - start);
2633                   data->type = REPL_TYPE_STRING;
2634                   list = g_list_prepend (list, data);
2635                 }
2636             }
2637         }
2638     }
2639
2640   return g_list_reverse (list);
2641 }
2642
2643 /* Change the case of c based on change_case. */
2644 #define CHANGE_CASE(c, change_case) \
2645         (((change_case) & CHANGE_CASE_LOWER_MASK) ? \
2646                 g_unichar_tolower (c) : \
2647                 g_unichar_toupper (c))
2648
2649 static void
2650 string_append (GString     *string,
2651                const gchar *text,
2652                ChangeCase  *change_case)
2653 {
2654   gunichar c;
2655
2656   if (text[0] == '\0')
2657     return;
2658
2659   if (*change_case == CHANGE_CASE_NONE)
2660     {
2661       g_string_append (string, text);
2662     }
2663   else if (*change_case & CHANGE_CASE_SINGLE_MASK)
2664     {
2665       c = g_utf8_get_char (text);
2666       g_string_append_unichar (string, CHANGE_CASE (c, *change_case));
2667       g_string_append (string, g_utf8_next_char (text));
2668       *change_case = CHANGE_CASE_NONE;
2669     }
2670   else
2671     {
2672       while (*text != '\0')
2673         {
2674           c = g_utf8_get_char (text);
2675           g_string_append_unichar (string, CHANGE_CASE (c, *change_case));
2676           text = g_utf8_next_char (text);
2677         }
2678     }
2679 }
2680
2681 static gboolean
2682 interpolate_replacement (const GMatchInfo *match_info,
2683                          GString          *result,
2684                          gpointer          data)
2685 {
2686   GList *list;
2687   InterpolationData *idata;
2688   gchar *match;
2689   ChangeCase change_case = CHANGE_CASE_NONE;
2690
2691   for (list = data; list; list = list->next)
2692     {
2693       idata = list->data;
2694       switch (idata->type)
2695         {
2696         case REPL_TYPE_STRING:
2697           string_append (result, idata->text, &change_case);
2698           break;
2699         case REPL_TYPE_CHARACTER:
2700           g_string_append_c (result, CHANGE_CASE (idata->c, change_case));
2701           if (change_case & CHANGE_CASE_SINGLE_MASK)
2702             change_case = CHANGE_CASE_NONE;
2703           break;
2704         case REPL_TYPE_NUMERIC_REFERENCE:
2705           match = g_match_info_fetch (match_info, idata->num);
2706           if (match)
2707             {
2708               string_append (result, match, &change_case);
2709               g_free (match);
2710             }
2711           break;
2712         case REPL_TYPE_SYMBOLIC_REFERENCE:
2713           match = g_match_info_fetch_named (match_info, idata->text);
2714           if (match)
2715             {
2716               string_append (result, match, &change_case);
2717               g_free (match);
2718             }
2719           break;
2720         case REPL_TYPE_CHANGE_CASE:
2721           change_case = idata->change_case;
2722           break;
2723         }
2724     }
2725
2726   return FALSE;
2727 }
2728
2729 /* whether actual match_info is needed for replacement, i.e.
2730  * whether there are references
2731  */
2732 static gboolean
2733 interpolation_list_needs_match (GList *list)
2734 {
2735   while (list != NULL)
2736     {
2737       InterpolationData *data = list->data;
2738
2739       if (data->type == REPL_TYPE_SYMBOLIC_REFERENCE ||
2740           data->type == REPL_TYPE_NUMERIC_REFERENCE)
2741         {
2742           return TRUE;
2743         }
2744
2745       list = list->next;
2746     }
2747
2748   return FALSE;
2749 }
2750
2751 /**
2752  * g_regex_replace:
2753  * @regex: a #GRegex structure
2754  * @string: (array length=string_len): the string to perform matches against
2755  * @string_len: the length of @string, or -1 if @string is nul-terminated
2756  * @start_position: starting index of the string to match, in bytes
2757  * @replacement: text to replace each match with
2758  * @match_options: options for the match
2759  * @error: location to store the error occurring, or %NULL to ignore errors
2760  *
2761  * Replaces all occurrences of the pattern in @regex with the
2762  * replacement text. Backreferences of the form '\number' or
2763  * '\g<number>' in the replacement text are interpolated by the
2764  * number-th captured subexpression of the match, '\g<name>' refers
2765  * to the captured subexpression with the given name. '\0' refers
2766  * to the complete match, but '\0' followed by a number is the octal
2767  * representation of a character. To include a literal '\' in the
2768  * replacement, write '\\'.
2769  *
2770  * There are also escapes that changes the case of the following text:
2771  *
2772  * - \l: Convert to lower case the next character
2773  * - \u: Convert to upper case the next character
2774  * - \L: Convert to lower case till \E
2775  * - \U: Convert to upper case till \E
2776  * - \E: End case modification
2777  *
2778  * If you do not need to use backreferences use g_regex_replace_literal().
2779  *
2780  * The @replacement string must be UTF-8 encoded even if #G_REGEX_RAW was
2781  * passed to g_regex_new(). If you want to use not UTF-8 encoded stings
2782  * you can use g_regex_replace_literal().
2783  *
2784  * Setting @start_position differs from just passing over a shortened
2785  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern that
2786  * begins with any kind of lookbehind assertion, such as "\b".
2787  *
2788  * Returns: a newly allocated string containing the replacements
2789  *
2790  * Since: 2.14
2791  */
2792 gchar *
2793 g_regex_replace (const GRegex      *regex,
2794                  const gchar       *string,
2795                  gssize             string_len,
2796                  gint               start_position,
2797                  const gchar       *replacement,
2798                  GRegexMatchFlags   match_options,
2799                  GError           **error)
2800 {
2801   gchar *result;
2802   GList *list;
2803   GError *tmp_error = NULL;
2804
2805   g_return_val_if_fail (regex != NULL, NULL);
2806   g_return_val_if_fail (string != NULL, NULL);
2807   g_return_val_if_fail (start_position >= 0, NULL);
2808   g_return_val_if_fail (replacement != NULL, NULL);
2809   g_return_val_if_fail (error == NULL || *error == NULL, NULL);
2810   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
2811
2812   list = split_replacement (replacement, &tmp_error);
2813   if (tmp_error != NULL)
2814     {
2815       g_propagate_error (error, tmp_error);
2816       return NULL;
2817     }
2818
2819   result = g_regex_replace_eval (regex,
2820                                  string, string_len, start_position,
2821                                  match_options,
2822                                  interpolate_replacement,
2823                                  (gpointer)list,
2824                                  &tmp_error);
2825   if (tmp_error != NULL)
2826     g_propagate_error (error, tmp_error);
2827
2828   g_list_free_full (list, (GDestroyNotify) free_interpolation_data);
2829
2830   return result;
2831 }
2832
2833 static gboolean
2834 literal_replacement (const GMatchInfo *match_info,
2835                      GString          *result,
2836                      gpointer          data)
2837 {
2838   g_string_append (result, data);
2839   return FALSE;
2840 }
2841
2842 /**
2843  * g_regex_replace_literal:
2844  * @regex: a #GRegex structure
2845  * @string: (array length=string_len): the string to perform matches against
2846  * @string_len: the length of @string, or -1 if @string is nul-terminated
2847  * @start_position: starting index of the string to match, in bytes
2848  * @replacement: text to replace each match with
2849  * @match_options: options for the match
2850  * @error: location to store the error occurring, or %NULL to ignore errors
2851  *
2852  * Replaces all occurrences of the pattern in @regex with the
2853  * replacement text. @replacement is replaced literally, to
2854  * include backreferences use g_regex_replace().
2855  *
2856  * Setting @start_position differs from just passing over a
2857  * shortened string and setting #G_REGEX_MATCH_NOTBOL in the
2858  * case of a pattern that begins with any kind of lookbehind
2859  * assertion, such as "\b".
2860  *
2861  * Returns: a newly allocated string containing the replacements
2862  *
2863  * Since: 2.14
2864  */
2865 gchar *
2866 g_regex_replace_literal (const GRegex      *regex,
2867                          const gchar       *string,
2868                          gssize             string_len,
2869                          gint               start_position,
2870                          const gchar       *replacement,
2871                          GRegexMatchFlags   match_options,
2872                          GError           **error)
2873 {
2874   g_return_val_if_fail (replacement != NULL, NULL);
2875   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
2876
2877   return g_regex_replace_eval (regex,
2878                                string, string_len, start_position,
2879                                match_options,
2880                                literal_replacement,
2881                                (gpointer)replacement,
2882                                error);
2883 }
2884
2885 /**
2886  * g_regex_replace_eval:
2887  * @regex: a #GRegex structure from g_regex_new()
2888  * @string: (array length=string_len): string to perform matches against
2889  * @string_len: the length of @string, or -1 if @string is nul-terminated
2890  * @start_position: starting index of the string to match, in bytes
2891  * @match_options: options for the match
2892  * @eval: a function to call for each match
2893  * @user_data: user data to pass to the function
2894  * @error: location to store the error occurring, or %NULL to ignore errors
2895  *
2896  * Replaces occurrences of the pattern in regex with the output of
2897  * @eval for that occurrence.
2898  *
2899  * Setting @start_position differs from just passing over a shortened
2900  * string and setting #G_REGEX_MATCH_NOTBOL in the case of a pattern
2901  * that begins with any kind of lookbehind assertion, such as "\b".
2902  *
2903  * The following example uses g_regex_replace_eval() to replace multiple
2904  * strings at once:
2905  * |[<!-- language="C" -->
2906  * static gboolean
2907  * eval_cb (const GMatchInfo *info,
2908  *          GString          *res,
2909  *          gpointer          data)
2910  * {
2911  *   gchar *match;
2912  *   gchar *r;
2913  *
2914  *    match = g_match_info_fetch (info, 0);
2915  *    r = g_hash_table_lookup ((GHashTable *)data, match);
2916  *    g_string_append (res, r);
2917  *    g_free (match);
2918  *
2919  *    return FALSE;
2920  * }
2921  *
2922  * ...
2923  *
2924  * GRegex *reg;
2925  * GHashTable *h;
2926  * gchar *res;
2927  *
2928  * h = g_hash_table_new (g_str_hash, g_str_equal);
2929  *
2930  * g_hash_table_insert (h, "1", "ONE");
2931  * g_hash_table_insert (h, "2", "TWO");
2932  * g_hash_table_insert (h, "3", "THREE");
2933  * g_hash_table_insert (h, "4", "FOUR");
2934  *
2935  * reg = g_regex_new ("1|2|3|4", 0, 0, NULL);
2936  * res = g_regex_replace_eval (reg, text, -1, 0, 0, eval_cb, h, NULL);
2937  * g_hash_table_destroy (h);
2938  *
2939  * ...
2940  * ]|
2941  *
2942  * Returns: a newly allocated string containing the replacements
2943  *
2944  * Since: 2.14
2945  */
2946 gchar *
2947 g_regex_replace_eval (const GRegex        *regex,
2948                       const gchar         *string,
2949                       gssize               string_len,
2950                       gint                 start_position,
2951                       GRegexMatchFlags     match_options,
2952                       GRegexEvalCallback   eval,
2953                       gpointer             user_data,
2954                       GError             **error)
2955 {
2956   GMatchInfo *match_info;
2957   GString *result;
2958   gint str_pos = 0;
2959   gboolean done = FALSE;
2960   GError *tmp_error = NULL;
2961
2962   g_return_val_if_fail (regex != NULL, NULL);
2963   g_return_val_if_fail (string != NULL, NULL);
2964   g_return_val_if_fail (start_position >= 0, NULL);
2965   g_return_val_if_fail (eval != NULL, NULL);
2966   g_return_val_if_fail ((match_options & ~G_REGEX_MATCH_MASK) == 0, NULL);
2967
2968   if (string_len < 0)
2969     string_len = strlen (string);
2970
2971   result = g_string_sized_new (string_len);
2972
2973   /* run down the string making matches. */
2974   g_regex_match_full (regex, string, string_len, start_position,
2975                       match_options, &match_info, &tmp_error);
2976   while (!done && g_match_info_matches (match_info))
2977     {
2978       g_string_append_len (result,
2979                            string + str_pos,
2980                            match_info->offsets[0] - str_pos);
2981       done = (*eval) (match_info, result, user_data);
2982       str_pos = match_info->offsets[1];
2983       g_match_info_next (match_info, &tmp_error);
2984     }
2985   g_match_info_free (match_info);
2986   if (tmp_error != NULL)
2987     {
2988       g_propagate_error (error, tmp_error);
2989       g_string_free (result, TRUE);
2990       return NULL;
2991     }
2992
2993   g_string_append_len (result, string + str_pos, string_len - str_pos);
2994   return g_string_free (result, FALSE);
2995 }
2996
2997 /**
2998  * g_regex_check_replacement:
2999  * @replacement: the replacement string
3000  * @has_references: (out) (optional): location to store information about
3001  *   references in @replacement or %NULL
3002  * @error: location to store error
3003  *
3004  * Checks whether @replacement is a valid replacement string
3005  * (see g_regex_replace()), i.e. that all escape sequences in
3006  * it are valid.
3007  *
3008  * If @has_references is not %NULL then @replacement is checked
3009  * for pattern references. For instance, replacement text 'foo\n'
3010  * does not contain references and may be evaluated without information
3011  * about actual match, but '\0\1' (whole match followed by first
3012  * subpattern) requires valid #GMatchInfo object.
3013  *
3014  * Returns: whether @replacement is a valid replacement string
3015  *
3016  * Since: 2.14
3017  */
3018 gboolean
3019 g_regex_check_replacement (const gchar  *replacement,
3020                            gboolean     *has_references,
3021                            GError      **error)
3022 {
3023   GList *list;
3024   GError *tmp = NULL;
3025
3026   list = split_replacement (replacement, &tmp);
3027
3028   if (tmp)
3029   {
3030     g_propagate_error (error, tmp);
3031     return FALSE;
3032   }
3033
3034   if (has_references)
3035     *has_references = interpolation_list_needs_match (list);
3036
3037   g_list_free_full (list, (GDestroyNotify) free_interpolation_data);
3038
3039   return TRUE;
3040 }
3041
3042 /**
3043  * g_regex_escape_nul:
3044  * @string: the string to escape
3045  * @length: the length of @string
3046  *
3047  * Escapes the nul characters in @string to "\x00".  It can be used
3048  * to compile a regex with embedded nul characters.
3049  *
3050  * For completeness, @length can be -1 for a nul-terminated string.
3051  * In this case the output string will be of course equal to @string.
3052  *
3053  * Returns: a newly-allocated escaped string
3054  *
3055  * Since: 2.30
3056  */
3057 gchar *
3058 g_regex_escape_nul (const gchar *string,
3059                     gint         length)
3060 {
3061   GString *escaped;
3062   const gchar *p, *piece_start, *end;
3063   gint backslashes;
3064
3065   g_return_val_if_fail (string != NULL, NULL);
3066
3067   if (length < 0)
3068     return g_strdup (string);
3069
3070   end = string + length;
3071   p = piece_start = string;
3072   escaped = g_string_sized_new (length + 1);
3073
3074   backslashes = 0;
3075   while (p < end)
3076     {
3077       switch (*p)
3078         {
3079         case '\0':
3080           if (p != piece_start)
3081             {
3082               /* copy the previous piece. */
3083               g_string_append_len (escaped, piece_start, p - piece_start);
3084             }
3085           if ((backslashes & 1) == 0)
3086             g_string_append_c (escaped, '\\');
3087           g_string_append_c (escaped, 'x');
3088           g_string_append_c (escaped, '0');
3089           g_string_append_c (escaped, '0');
3090           piece_start = ++p;
3091           backslashes = 0;
3092           break;
3093         case '\\':
3094           backslashes++;
3095           ++p;
3096           break;
3097         default:
3098           backslashes = 0;
3099           p = g_utf8_next_char (p);
3100           break;
3101         }
3102     }
3103
3104   if (piece_start < end)
3105     g_string_append_len (escaped, piece_start, end - piece_start);
3106
3107   return g_string_free (escaped, FALSE);
3108 }
3109
3110 /**
3111  * g_regex_escape_string:
3112  * @string: (array length=length): the string to escape
3113  * @length: the length of @string, or -1 if @string is nul-terminated
3114  *
3115  * Escapes the special characters used for regular expressions
3116  * in @string, for instance "a.b*c" becomes "a\.b\*c". This
3117  * function is useful to dynamically generate regular expressions.
3118  *
3119  * @string can contain nul characters that are replaced with "\0",
3120  * in this case remember to specify the correct length of @string
3121  * in @length.
3122  *
3123  * Returns: a newly-allocated escaped string
3124  *
3125  * Since: 2.14
3126  */
3127 gchar *
3128 g_regex_escape_string (const gchar *string,
3129                        gint         length)
3130 {
3131   GString *escaped;
3132   const char *p, *piece_start, *end;
3133
3134   g_return_val_if_fail (string != NULL, NULL);
3135
3136   if (length < 0)
3137     length = strlen (string);
3138
3139   end = string + length;
3140   p = piece_start = string;
3141   escaped = g_string_sized_new (length + 1);
3142
3143   while (p < end)
3144     {
3145       switch (*p)
3146         {
3147         case '\0':
3148         case '\\':
3149         case '|':
3150         case '(':
3151         case ')':
3152         case '[':
3153         case ']':
3154         case '{':
3155         case '}':
3156         case '^':
3157         case '$':
3158         case '*':
3159         case '+':
3160         case '?':
3161         case '.':
3162           if (p != piece_start)
3163             /* copy the previous piece. */
3164             g_string_append_len (escaped, piece_start, p - piece_start);
3165           g_string_append_c (escaped, '\\');
3166           if (*p == '\0')
3167             g_string_append_c (escaped, '0');
3168           else
3169             g_string_append_c (escaped, *p);
3170           piece_start = ++p;
3171           break;
3172         default:
3173           p = g_utf8_next_char (p);
3174           break;
3175         }
3176   }
3177
3178   if (piece_start < end)
3179     g_string_append_len (escaped, piece_start, end - piece_start);
3180
3181   return g_string_free (escaped, FALSE);
3182 }