STATIC CHECKPOINT
S_regcppush(pTHX_ const regexp *rex, I32 parenfloor, U32 maxopenparen)
{
- dVAR;
const int retval = PL_savestack_ix;
const int paren_elems_to_push =
(maxopenparen - parenfloor) * REGCP_PAREN_ELEMS;
PERL_ARGS_ASSERT_REGCPPUSH;
if (paren_elems_to_push < 0)
- Perl_croak(aTHX_ "panic: paren_elems_to_push, %i < 0, maxopenparen: %i parenfloor: %i REGCP_PAREN_ELEMS: %i",
- paren_elems_to_push, maxopenparen, parenfloor, REGCP_PAREN_ELEMS);
+ Perl_croak(aTHX_ "panic: paren_elems_to_push, %i < 0, maxopenparen: %i parenfloor: %i REGCP_PAREN_ELEMS: %u",
+ (int)paren_elems_to_push, (int)maxopenparen,
+ (int)parenfloor, (unsigned)REGCP_PAREN_ELEMS);
if ((elems_shifted >> SAVE_TIGHT_SHIFT) != total_elems)
Perl_croak(aTHX_ "panic: paren_elems_to_push offset %"UVuf
STATIC void
S_regcppop(pTHX_ regexp *rex, U32 *maxopenparen_p)
{
- dVAR;
UV i;
U32 paren;
GET_RE_DEBUG_FLAGS_DECL;
}
#endif
-/*
- * Need to implement the following flags for reg_anch:
- *
- * USE_INTUIT_NOML - Useful to call re_intuit_start() first
- * USE_INTUIT_ML
- * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
- * INTUIT_AUTORITATIVE_ML
- * INTUIT_ONCE_NOML - Intuit can match in one location only.
- * INTUIT_ONCE_ML
- *
- * Another flag for this function: SECOND_TIME (so that float substrs
- * with giant delta may be not rechecked).
- */
-
-/* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend.
- Otherwise, only SvCUR(sv) is used to get strbeg. */
-
-/* XXXX Some places assume that there is a fixed substring.
- An update may be needed if optimizer marks as "INTUITable"
- RExen without fixed substrings. Similarly, it is assumed that
- lengths of all the strings are no more than minlen, thus they
- cannot come from lookahead.
- (Or minlen should take into account lookahead.)
- NOTE: Some of this comment is not correct. minlen does now take account
- of lookahead/behind. Further research is required. -- demerphq
-*/
-
-/* A failure to find a constant substring means that there is no need to make
- an expensive call to REx engine, thus we celebrate a failure. Similarly,
- finding a substring too deep into the string means that fewer calls to
- regtry() should be needed.
-
- REx compiler's optimizer found 4 possible hints:
- a) Anchored substring;
- b) Fixed substring;
- c) Whether we are anchored (beginning-of-line or \G);
- d) First node (of those at offset 0) which may distinguish positions;
- We use a)b)d) and multiline-part of c), and try to find a position in the
- string which does not contradict any of them.
- */
-/* Most of decisions we do here should have been done at compile time.
- The nodes of the REx which we used for the search should have been
- deleted from the finite automaton. */
-
-/* args:
- * rx: the regex to match against
- * sv: the SV being matched: only used for utf8 flag; the string
- * itself is accessed via the pointers below. Note that on
- * something like an overloaded SV, SvPOK(sv) may be false
- * and the string pointers may point to something unrelated to
- * the SV itself.
- * strbeg: real beginning of string
- * strpos: the point in the string at which to begin matching
- * strend: pointer to the byte following the last char of the string
- * flags currently unused; set to 0
- * data: currently unused; set to NULL
+/* re_intuit_start():
+ *
+ * Based on some optimiser hints, try to find the earliest position in the
+ * string where the regex could match.
+ *
+ * rx: the regex to match against
+ * sv: the SV being matched: only used for utf8 flag; the string
+ * itself is accessed via the pointers below. Note that on
+ * something like an overloaded SV, SvPOK(sv) may be false
+ * and the string pointers may point to something unrelated to
+ * the SV itself.
+ * strbeg: real beginning of string
+ * strpos: the point in the string at which to begin matching
+ * strend: pointer to the byte following the last char of the string
+ * flags currently unused; set to 0
+ * data: currently unused; set to NULL
+ *
+ * The basic idea of re_intuit_start() is to use some known information
+ * about the pattern, namely:
+ *
+ * a) the longest known anchored substring (i.e. one that's at a
+ * constant offset from the beginning of the pattern; but not
+ * necessarily at a fixed offset from the beginning of the
+ * string);
+ * b) the longest floating substring (i.e. one that's not at a constant
+ * offset from the beginning of the pattern);
+ * c) Whether the pattern is anchored to the string; either
+ * an absolute anchor: /^../, or anchored to \n: /^.../m,
+ * or anchored to pos(): /\G/;
+ * d) A start class: a real or synthetic character class which
+ * represents which characters are legal at the start of the pattern;
+ *
+ * to either quickly reject the match, or to find the earliest position
+ * within the string at which the pattern might match, thus avoiding
+ * running the full NFA engine at those earlier locations, only to
+ * eventually fail and retry further along.
+ *
+ * Returns NULL if the pattern can't match, or returns the address within
+ * the string which is the earliest place the match could occur.
+ *
+ * The longest of the anchored and floating substrings is called 'check'
+ * and is checked first. The other is called 'other' and is checked
+ * second. The 'other' substring may not be present. For example,
+ *
+ * /(abc|xyz)ABC\d{0,3}DEFG/
+ *
+ * will have
+ *
+ * check substr (float) = "DEFG", offset 6..9 chars
+ * other substr (anchored) = "ABC", offset 3..3 chars
+ * stclass = [ax]
+ *
+ * Be aware that during the course of this function, sometimes 'anchored'
+ * refers to a substring being anchored relative to the start of the
+ * pattern, and sometimes to the pattern itself being anchored relative to
+ * the string. For example:
+ *
+ * /\dabc/: "abc" is anchored to the pattern;
+ * /^\dabc/: "abc" is anchored to the pattern and the string;
+ * /\d+abc/: "abc" is anchored to neither the pattern nor the string;
+ * /^\d+abc/: "abc" is anchored to neither the pattern nor the string,
+ * but the pattern is anchored to the string.
*/
char *
const U32 flags,
re_scream_pos_data *data)
{
- dVAR;
struct regexp *const prog = ReANY(rx);
SSize_t start_shift = prog->check_offset_min;
/* Should be nonnegative! */
&& (prog->float_utf8 || prog->float_substr))
|| (prog->float_min_offset >= prog->anchored_offset));
- /* CHR_DIST() would be more correct here but it makes things slow. */
+ /* byte rather than char calculation for efficiency. It fails
+ * to quickly reject some cases that can't match, but will reject
+ * them later after doing full char arithmetic */
if (prog->minlen > strend - strpos) {
DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
" String too short...\n"));
});
if (prog->intflags & PREGf_ANCH) { /* Match at \G, beg-of-str or after \n */
- /* Check after \n? */
- ml_anch = ( (prog->intflags & PREGf_ANCH_MBOL)
- || ((prog->intflags & PREGf_ANCH_BOL) && multiline));
- if (!ml_anch) {
+ /* ml_anch: check after \n?
+ *
+ * A note about IMPLICIT: on an un-anchored pattern beginning
+ * with /.*.../, these flags will have been added by the
+ * compiler:
+ * /.*abc/, /.*abc/m: PREGf_IMPLICIT | PREGf_ANCH_MBOL
+ * /.*abc/s: PREGf_IMPLICIT | PREGf_ANCH_SBOL
+ */
+ ml_anch = (prog->intflags & PREGf_ANCH_MBOL)
+ && !(prog->intflags & PREGf_IMPLICIT);
+
+ if (!ml_anch && !(prog->intflags & PREGf_IMPLICIT)) {
/* we are only allowed to match at BOS or \G */
/* trivially reject if there's a BOS anchor and we're not at BOS.
- * In the case of \G, we hope(!) that the caller has already
- * set strpos to pos()-gofs, and will already have checked
- * that this anchor position is legal. So we can skip it here.
+ *
+ * Note that we don't try to do a similar quick reject for
+ * \G, since generally the caller will have calculated strpos
+ * based on pos() and gofs, so the string is already correctly
+ * anchored by definition; and handling the exceptions would
+ * be too fiddly (e.g. REXEC_IGNOREPOS).
*/
- if ( !(prog->intflags & PREGf_ANCH_GPOS)
- && !(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */
- && (strpos != strbeg))
+ if ( strpos != strbeg
+ && (prog->intflags & (PREGf_ANCH_BOL|PREGf_ANCH_SBOL)))
{
DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
" Not at start...\n"));
/* in the presence of an anchor, the anchored (relative to the
* start of the regex) substr must also be anchored relative
- * to strpos. So quickly reject if substr isn't found there */
+ * to strpos. So quickly reject if substr isn't found there.
+ * This works for \G too, because the caller will already have
+ * subtracted gofs from pos, and gofs is the offset from the
+ * \G to the start of the regex. For example, in /.abc\Gdef/,
+ * where substr="abcdef", pos()=3, gofs=4, offset_min=1:
+ * caller will have set strpos=pos()-4; we look for the substr
+ * at position pos()-4+1, which lines up with the "a" */
if (prog->check_offset_min == prog->check_offset_max
- && !(prog->intflags & PREGf_CANY_SEEN)
- && ! multiline) /* /m can cause \n's to match that aren't
- accounted for in the string max length.
- See [perl #115242] */
+ && !(prog->intflags & PREGf_CANY_SEEN))
{
/* Substring at constant offset from beg-of-str... */
SSize_t slen = SvCUR(check);
- char *s;
-
- s = HOP3c(strpos, prog->check_offset_min, strend);
+ char *s = HOP3c(strpos, prog->check_offset_min, strend);
DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
" Looking for check substr at fixed offset %"IVdf"...\n",
(IV)prog->check_offset_min));
if (SvTAIL(check)) {
- /* In this case, the regex is anchored at the end too,
- * so the lengths must match exactly, give or take a \n.
- * NB: slen >= 1 since the last char of check is \n */
- if ( strend - s > slen || strend - s < slen - 1
- || (strend - s == slen && strend[-1] != '\n'))
+ /* In this case, the regex is anchored at the end too.
+ * Unless it's a multiline match, the lengths must match
+ * exactly, give or take a \n. NB: slen >= 1 since
+ * the last char of check is \n */
+ if (!multiline
+ && ( strend - s > slen
+ || strend - s < slen - 1
+ || (strend - s == slen && strend[-1] != '\n')))
{
DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
" String too long...\n"));
#endif
restart:
- /* Find a candidate regex origin in the region rx_origin..strend
- * by looking for the "check" substring in that region, corrected by
- * start/end_shift.
- */
+ /* This is the (re)entry point of the main loop in this function.
+ * The goal of this loop is to:
+ * 1) find the "check" substring in the region rx_origin..strend
+ * (adjusted by start_shift / end_shift). If not found, reject
+ * immediately.
+ * 2) If it exists, look for the "other" substr too if defined; for
+ * example, if the check substr maps to the anchored substr, then
+ * check the floating substr, and vice-versa. If not found, go
+ * back to (1) with rx_origin suitably incremented.
+ * 3) If we find an rx_origin position that doesn't contradict
+ * either of the substrings, then check the possible additional
+ * constraints on rx_origin of /^.../m or a known start class.
+ * If these fail, then depending on which constraints fail, jump
+ * back to here, or to various other re-entry points further along
+ * that skip some of the first steps.
+ * 4) If we pass all those tests, update the BmUSEFUL() count on the
+ * substring. If the start position was determined to be at the
+ * beginning of the string - so, not rejected, but not optimised,
+ * since we have to run regmatch from position 0 - decrement the
+ * BmUSEFUL() count. Otherwise increment it.
+ */
+
+
+ /* first, look for the 'check' substring */
+
{
U8* start_point;
U8* end_point;
}
- /* if the regex is absolutely anchored to the start of the string,
- * then check_offset_max represents an upper bound on the string
- * where the substr could start */
+ /* If the regex is absolutely anchored to either the start of the
+ * string (BOL,SBOL) or to pos() (ANCH_GPOS), then
+ * check_offset_max represents an upper bound on the string where
+ * the substr could start. For the ANCH_GPOS case, we assume that
+ * the caller of intuit will have already set strpos to
+ * pos()-gofs, so in this case strpos + offset_max will still be
+ * an upper bound on the substr.
+ */
if (!ml_anch
&& prog->intflags & PREGf_ANCH
- && prog->check_offset_max != SSize_t_MAX
- && start_shift < prog->check_offset_max)
+ && prog->check_offset_max != SSize_t_MAX)
{
SSize_t len = SvCUR(check) - !!SvTAIL(check);
- end_point = HOP3lim(start_point,
- prog->check_offset_max - start_shift,
- end_point -len)
- + len;
+ const char * const anchor =
+ (prog->intflags & PREGf_ANCH_GPOS ? strpos : strbeg);
+
+ /* do a bytes rather than chars comparison. It's conservative;
+ * so it skips doing the HOP if the result can't possibly end
+ * up earlier than the old value of end_point.
+ */
+ if ((char*)end_point - anchor > prog->check_offset_max) {
+ end_point = HOP3lim((U8*)anchor,
+ prog->check_offset_max,
+ end_point -len)
+ + len;
+ }
}
DEBUG_OPTIMISE_MORE_r({
check_at = fbm_instr( start_point, end_point,
check, multiline ? FBMrf_MULTILINE : 0);
- }
- /* Update the count-of-usability, remove useless subpatterns,
- unshift s. */
-
- DEBUG_EXECUTE_r({
- RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
- SvPVX_const(check), RE_SV_DUMPLEN(check), 30);
- PerlIO_printf(Perl_debug_log, " %s %s substr %s%s%s",
- (check_at ? "Found" : "Did not find"),
- (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr)
- ? "anchored" : "floating"),
- quoted,
- RE_SV_TAIL(check),
- (check_at ? " at offset " : "...\n") );
- });
+ /* Update the count-of-usability, remove useless subpatterns,
+ unshift s. */
- if (!check_at)
- goto fail_finish;
- /* Finish the diagnostic message */
- DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(check_at - strpos)) );
+ DEBUG_EXECUTE_r({
+ RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
+ SvPVX_const(check), RE_SV_DUMPLEN(check), 30);
+ PerlIO_printf(Perl_debug_log, " %s %s substr %s%s%s",
+ (check_at ? "Found" : "Did not find"),
+ (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr)
+ ? "anchored" : "floating"),
+ quoted,
+ RE_SV_TAIL(check),
+ (check_at ? " at offset " : "...\n") );
+ });
- /* set rx_origin to the minimum position where the regex could start
- * matching, given the constraint of the just-matched check substring.
- * But don't set it lower than previously.
- */
+ if (!check_at)
+ goto fail_finish;
+ /* Finish the diagnostic message */
+ DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(check_at - strpos)) );
- if (check_at - rx_origin > prog->check_offset_max)
- rx_origin = HOP3c(check_at, -prog->check_offset_max, rx_origin);
+ /* set rx_origin to the minimum position where the regex could start
+ * matching, given the constraint of the just-matched check substring.
+ * But don't set it lower than previously.
+ */
+ if (check_at - rx_origin > prog->check_offset_max)
+ rx_origin = HOP3c(check_at, -prog->check_offset_max, rx_origin);
+ }
- /* XXX dmq: first branch is for positive lookbehind...
- Our check string is offset from the beginning of the pattern.
- So we need to do any stclass tests offset forward from that
- point. I think. :-(
- */
-
- /* Got a candidate. Check MBOL anchoring, and the *other* substr.
- Start with the other substr.
- XXXX no SCREAM optimization yet - and a very coarse implementation
- XXXX /ttx+/ results in anchored="ttx", floating="x". floating will
- *always* match. Probably should be marked during compile...
- Probably it is right to do no SCREAM here...
- */
+
+ /* now look for the 'other' substring if defined */
if (utf8_target ? prog->substrs->data[other_ix].utf8_substr
: prog->substrs->data[other_ix].substr)
postprocess_substr_matches:
- /* handle the extra constraint of /^/m */
+ /* handle the extra constraint of /^.../m if present */
- if (ml_anch && rx_origin != strbeg && rx_origin[-1] != '\n'
- /* May be due to an implicit anchor of m{.*foo} */
- && !(prog->intflags & PREGf_IMPLICIT))
- {
+ if (ml_anch && rx_origin != strbeg && rx_origin[-1] != '\n') {
char *s;
DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
}
else {
DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
- " Starting position does not contradict /%s^%s/m...\n",
- PL_colors[0], PL_colors[1]));
+ " (multiline anchor test skipped)\n"));
}
success_at_start:
}
DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
" This position contradicts STCLASS...\n") );
- if ((prog->intflags & PREGf_ANCH) && !ml_anch)
+ if ((prog->intflags & PREGf_ANCH) && !ml_anch
+ && !(prog->intflags & PREGf_IMPLICIT))
goto fail;
/* Contradict one of substrings */
prog->check_substr = prog->check_utf8 = NULL; /* disable */
prog->float_substr = prog->float_utf8 = NULL; /* clear */
check = NULL; /* abort */
- /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevant flag
- see http://bugs.activestate.com/show_bug.cgi?id=87173 */
- if (prog->intflags & PREGf_IMPLICIT) {
- prog->intflags &= ~PREGf_ANCH_MBOL;
- /* maybe we have no anchors left after this... */
- if (!(prog->intflags & PREGf_ANCH))
- prog->extflags &= ~RXf_IS_ANCHORED;
- }
/* XXXX This is a remnant of the old implementation. It
looks wasteful, since now INTUIT can use many
other heuristics. */
prog->extflags &= ~RXf_USE_INTUIT;
- /* XXXX What other flags might need to be cleared in this branch? */
}
}
switch (trie_type) { \
case trie_utf8_exactfa_fold: \
flags |= FOLD_FLAGS_NOMIX_ASCII; \
- /* FALL THROUGH */ \
+ /* FALLTHROUGH */ \
case trie_utf8_fold: \
if ( foldlen>0 ) { \
uvc = utf8n_to_uvchr( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \
break; \
case trie_latin_utf8_exactfa_fold: \
flags |= FOLD_FLAGS_NOMIX_ASCII; \
- /* FALL THROUGH */ \
+ /* FALLTHROUGH */ \
case trie_latin_utf8_fold: \
if ( foldlen>0 ) { \
uvc = utf8n_to_uvchr( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \
} \
} STMT_END
-#define REXEC_FBC_EXACTISH_SCAN(CoNd) \
+#define REXEC_FBC_EXACTISH_SCAN(COND) \
STMT_START { \
while (s <= e) { \
- if ( (CoNd) \
+ if ( (COND) \
&& (ln == 1 || folder(s, pat_string, ln)) \
&& (reginfo->intuit || regtry(reginfo, &s)) )\
goto got_it; \
} \
} STMT_END
-#define REXEC_FBC_UTF8_SCAN(CoDe) \
+#define REXEC_FBC_UTF8_SCAN(CODE) \
STMT_START { \
while (s < strend) { \
- CoDe \
+ CODE \
s += UTF8SKIP(s); \
} \
} STMT_END
-#define REXEC_FBC_SCAN(CoDe) \
+#define REXEC_FBC_SCAN(CODE) \
STMT_START { \
while (s < strend) { \
- CoDe \
+ CODE \
s++; \
} \
} STMT_END
-#define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \
-REXEC_FBC_UTF8_SCAN( \
- if (CoNd) { \
- if (tmp && (reginfo->intuit || regtry(reginfo, &s))) \
- goto got_it; \
- else \
- tmp = doevery; \
- } \
- else \
- tmp = 1; \
+#define REXEC_FBC_UTF8_CLASS_SCAN(COND) \
+REXEC_FBC_UTF8_SCAN( /* Loops while (s < strend) */ \
+ if (COND) { \
+ if (tmp && (reginfo->intuit || regtry(reginfo, &s))) \
+ goto got_it; \
+ else \
+ tmp = doevery; \
+ } \
+ else \
+ tmp = 1; \
)
-#define REXEC_FBC_CLASS_SCAN(CoNd) \
-REXEC_FBC_SCAN( \
- if (CoNd) { \
- if (tmp && (reginfo->intuit || regtry(reginfo, &s))) \
- goto got_it; \
- else \
- tmp = doevery; \
- } \
- else \
- tmp = 1; \
+#define REXEC_FBC_CLASS_SCAN(COND) \
+REXEC_FBC_SCAN( /* Loops while (s < strend) */ \
+ if (COND) { \
+ if (tmp && (reginfo->intuit || regtry(reginfo, &s))) \
+ goto got_it; \
+ else \
+ tmp = doevery; \
+ } \
+ else \
+ tmp = 1; \
)
-#define REXEC_FBC_TRYIT \
-if ((reginfo->intuit || regtry(reginfo, &s))) \
- goto got_it
+/* This is the macro to use when we want to see if something that looks like it
+ * could match, actually does, and if so exits the loop */
+#define REXEC_FBC_TRYIT \
+ if ((reginfo->intuit || regtry(reginfo, &s))) \
+ goto got_it
-#define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \
+#define REXEC_FBC_CSCAN(CONDUTF8,COND) \
if (utf8_target) { \
- REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
+ REXEC_FBC_UTF8_CLASS_SCAN(CONDUTF8); \
} \
else { \
- REXEC_FBC_CLASS_SCAN(CoNd); \
+ REXEC_FBC_CLASS_SCAN(COND); \
}
-
+
#define DUMP_EXEC_POS(li,s,doutf8) \
dump_exec_pos(li,s,(reginfo->strend),(reginfo->strbeg), \
startpos, doutf8)
-
-#define UTF8_NOLOAD(TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \
- tmp = (s != reginfo->strbeg) ? UCHARAT(s - 1) : '\n'; \
- tmp = TEST_NON_UTF8(tmp); \
- REXEC_FBC_UTF8_SCAN( \
- if (tmp == ! TEST_NON_UTF8((U8) *s)) { \
- tmp = !tmp; \
- IF_SUCCESS; \
- } \
- else { \
- IF_FAIL; \
- } \
- ); \
-
-#define UTF8_LOAD(TeSt1_UtF8, TeSt2_UtF8, IF_SUCCESS, IF_FAIL) \
- if (s == reginfo->strbeg) { \
- tmp = '\n'; \
- } \
- else { \
- U8 * const r = reghop3((U8*)s, -1, (U8*)reginfo->strbeg); \
- tmp = utf8n_to_uvchr(r, (U8*) reginfo->strend - r, \
+/* The three macros below are slightly different versions of the same logic.
+ *
+ * The first is for /a and /aa when the target string is UTF-8. This can only
+ * match ascii, but it must advance based on UTF-8. The other two handle the
+ * non-UTF-8 and the more generic UTF-8 cases. In all three, we are looking
+ * for the boundary (or non-boundary) between a word and non-word character.
+ * The utf8 and non-utf8 cases have the same logic, but the details must be
+ * different. Find the "wordness" of the character just prior to this one, and
+ * compare it with the wordness of this one. If they differ, we have a
+ * boundary. At the beginning of the string, pretend that the previous
+ * character was a new-line.
+ *
+ * All these macros uncleanly have side-effects with each other and outside
+ * variables. So far it's been too much trouble to clean-up
+ *
+ * TEST_NON_UTF8 is the macro or function to call to test if its byte input is
+ * a word character or not.
+ * IF_SUCCESS is code to do if it finds that we are at a boundary between
+ * word/non-word
+ * IF_FAIL is code to do if we aren't at a boundary between word/non-word
+ *
+ * Exactly one of the two IF_FOO parameters is a no-op, depending on whether we
+ * are looking for a boundary or for a non-boundary. If we are looking for a
+ * boundary, we want IF_FAIL to be the no-op, and for IF_SUCCESS to go out and
+ * see if this tentative match actually works, and if so, to quit the loop
+ * here. And vice-versa if we are looking for a non-boundary.
+ *
+ * 'tmp' below in the next three macros in the REXEC_FBC_SCAN and
+ * REXEC_FBC_UTF8_SCAN loops is a loop invariant, a bool giving the return of
+ * TEST_NON_UTF8(s-1). To see this, note that that's what it is defined to be
+ * at entry to the loop, and to get to the IF_FAIL branch, tmp must equal
+ * TEST_NON_UTF8(s), and in the opposite branch, IF_SUCCESS, tmp is that
+ * complement. But in that branch we complement tmp, meaning that at the
+ * bottom of the loop tmp is always going to be equal to TEST_NON_UTF8(s),
+ * which means at the top of the loop in the next iteration, it is
+ * TEST_NON_UTF8(s-1) */
+#define FBC_UTF8_A(TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \
+ tmp = (s != reginfo->strbeg) ? UCHARAT(s - 1) : '\n'; \
+ tmp = TEST_NON_UTF8(tmp); \
+ REXEC_FBC_UTF8_SCAN( /* advances s while s < strend */ \
+ if (tmp == ! TEST_NON_UTF8((U8) *s)) { \
+ tmp = !tmp; \
+ IF_SUCCESS; /* Is a boundary if values for s-1 and s differ */ \
+ } \
+ else { \
+ IF_FAIL; \
+ } \
+ ); \
+
+/* Like FBC_UTF8_A, but TEST_UV is a macro which takes a UV as its input, and
+ * TEST_UTF8 is a macro that for the same input code points returns identically
+ * to TEST_UV, but takes a pointer to a UTF-8 encoded string instead */
+#define FBC_UTF8(TEST_UV, TEST_UTF8, IF_SUCCESS, IF_FAIL) \
+ if (s == reginfo->strbeg) { \
+ tmp = '\n'; \
+ } \
+ else { /* Back-up to the start of the previous character */ \
+ U8 * const r = reghop3((U8*)s, -1, (U8*)reginfo->strbeg); \
+ tmp = utf8n_to_uvchr(r, (U8*) reginfo->strend - r, \
0, UTF8_ALLOW_DEFAULT); \
- } \
- tmp = TeSt1_UtF8; \
- LOAD_UTF8_CHARCLASS_ALNUM(); \
- REXEC_FBC_UTF8_SCAN( \
- if (tmp == ! (TeSt2_UtF8)) { \
- tmp = !tmp; \
- IF_SUCCESS; \
- } \
- else { \
- IF_FAIL; \
- } \
- ); \
+ } \
+ tmp = TEST_UV(tmp); \
+ LOAD_UTF8_CHARCLASS_ALNUM(); \
+ REXEC_FBC_UTF8_SCAN( /* advances s while s < strend */ \
+ if (tmp == ! (TEST_UTF8((U8 *) s))) { \
+ tmp = !tmp; \
+ IF_SUCCESS; \
+ } \
+ else { \
+ IF_FAIL; \
+ } \
+ );
/* The only difference between the BOUND and NBOUND cases is that
* REXEC_FBC_TRYIT is called when matched in BOUND, and when non-matched in
- * NBOUND. This is accomplished by passing it in either the if or else clause,
- * with the other one being empty */
-#define FBC_BOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \
- FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER)
-
-#define FBC_BOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \
- FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER)
-
-#define FBC_NBOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \
- FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT)
-
-#define FBC_NBOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \
- FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT)
-
-
-/* Common to the BOUND and NBOUND cases. Unfortunately the UTF8 tests need to
- * be passed in completely with the variable name being tested, which isn't
- * such a clean interface, but this is easier to read than it was before. We
- * are looking for the boundary (or non-boundary between a word and non-word
- * character. The utf8 and non-utf8 cases have the same logic, but the details
- * must be different. Find the "wordness" of the character just prior to this
- * one, and compare it with the wordness of this one. If they differ, we have
- * a boundary. At the beginning of the string, pretend that the previous
- * character was a new-line */
+ * NBOUND. This is accomplished by passing it as either the if or else clause,
+ * with the other one being empty (PLACEHOLDER is defined as empty).
+ *
+ * The TEST_FOO parameters are for operating on different forms of input, but
+ * all should be ones that return identically for the same underlying code
+ * points */
+#define FBC_BOUND(TEST_NON_UTF8, TEST_UV, TEST_UTF8) \
+ FBC_BOUND_COMMON( \
+ FBC_UTF8(TEST_UV, TEST_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), \
+ TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER)
+
+#define FBC_BOUND_A(TEST_NON_UTF8, TEST_UV, TEST_UTF8) \
+ FBC_BOUND_COMMON( \
+ FBC_UTF8_A(TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), \
+ TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER)
+
+#define FBC_NBOUND(TEST_NON_UTF8, TEST_UV, TEST_UTF8) \
+ FBC_BOUND_COMMON( \
+ FBC_UTF8(TEST_UV, TEST_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), \
+ TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT)
+
+#define FBC_NBOUND_A(TEST_NON_UTF8, TEST_UV, TEST_UTF8) \
+ FBC_BOUND_COMMON( \
+ FBC_UTF8_A(TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), \
+ TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT)
+
+/* Like the above two macros. UTF8_CODE is the complete code for handling
+ * UTF-8. Common to the BOUND and NBOUND cases, set-up by the FBC_BOUND, etc
+ * macros below */
#define FBC_BOUND_COMMON(UTF8_CODE, TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \
if (utf8_target) { \
- UTF8_CODE \
+ UTF8_CODE \
} \
else { /* Not utf8 */ \
tmp = (s != reginfo->strbeg) ? UCHARAT(s - 1) : '\n'; \
tmp = TEST_NON_UTF8(tmp); \
- REXEC_FBC_SCAN( \
+ REXEC_FBC_SCAN( /* advances s while s < strend */ \
if (tmp == ! TEST_NON_UTF8((U8) *s)) { \
- tmp = !tmp; \
IF_SUCCESS; \
+ tmp = !tmp; \
} \
else { \
IF_FAIL; \
if ((!prog->minlen && tmp) && (reginfo->intuit || regtry(reginfo, &s))) \
goto got_it;
+
/* We know what class REx starts with. Try to find this position... */
/* if reginfo->intuit, its a dryrun */
/* annoyingly all the vars in this routine have different names from their counterparts
in regmatch. /grrr */
-
STATIC char *
S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s,
const char *strend, regmatch_info *reginfo)
case EXACTFA_NO_TRIE: /* This node only generated for non-utf8 patterns */
assert(! is_utf8_pat);
- /* FALL THROUGH */
+ /* FALLTHROUGH */
case EXACTFA:
if (is_utf8_pat || utf8_target) {
utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII;
fold_array = PL_fold_latin1;
folder = foldEQ_latin1;
- /* FALL THROUGH */
+ /* FALLTHROUGH */
do_exactf_non_utf8: /* Neither pattern nor string are UTF8, and there
are no glitches with fold-length differences
}
break;
}
+
case BOUNDL:
- FBC_BOUND(isWORDCHAR_LC,
- isWORDCHAR_LC_uvchr(tmp),
- isWORDCHAR_LC_utf8((U8*)s));
+ FBC_BOUND(isWORDCHAR_LC, isWORDCHAR_LC_uvchr, isWORDCHAR_LC_utf8);
break;
case NBOUNDL:
- FBC_NBOUND(isWORDCHAR_LC,
- isWORDCHAR_LC_uvchr(tmp),
- isWORDCHAR_LC_utf8((U8*)s));
+ FBC_NBOUND(isWORDCHAR_LC, isWORDCHAR_LC_uvchr, isWORDCHAR_LC_utf8);
break;
case BOUND:
- FBC_BOUND(isWORDCHAR,
- isWORDCHAR_uni(tmp),
- cBOOL(swash_fetch(PL_utf8_swash_ptrs[_CC_WORDCHAR], (U8*)s, utf8_target)));
+ FBC_BOUND(isWORDCHAR, isWORDCHAR_uni, isWORDCHAR_utf8);
break;
case BOUNDA:
- FBC_BOUND_NOLOAD(isWORDCHAR_A,
- isWORDCHAR_A(tmp),
- isWORDCHAR_A((U8*)s));
+ FBC_BOUND_A(isWORDCHAR_A, isWORDCHAR_A, isWORDCHAR_A);
break;
case NBOUND:
- FBC_NBOUND(isWORDCHAR,
- isWORDCHAR_uni(tmp),
- cBOOL(swash_fetch(PL_utf8_swash_ptrs[_CC_WORDCHAR], (U8*)s, utf8_target)));
+ FBC_NBOUND(isWORDCHAR, isWORDCHAR_uni, isWORDCHAR_utf8);
break;
case NBOUNDA:
- FBC_NBOUND_NOLOAD(isWORDCHAR_A,
- isWORDCHAR_A(tmp),
- isWORDCHAR_A((U8*)s));
+ FBC_NBOUND_A(isWORDCHAR_A, isWORDCHAR_A, isWORDCHAR_A);
break;
case BOUNDU:
- FBC_BOUND(isWORDCHAR_L1,
- isWORDCHAR_uni(tmp),
- cBOOL(swash_fetch(PL_utf8_swash_ptrs[_CC_WORDCHAR], (U8*)s, utf8_target)));
+ FBC_BOUND(isWORDCHAR_L1, isWORDCHAR_uni, isWORDCHAR_utf8);
break;
case NBOUNDU:
- FBC_NBOUND(isWORDCHAR_L1,
- isWORDCHAR_uni(tmp),
- cBOOL(swash_fetch(PL_utf8_swash_ptrs[_CC_WORDCHAR], (U8*)s, utf8_target)));
+ FBC_NBOUND(isWORDCHAR_L1, isWORDCHAR_uni, isWORDCHAR_utf8);
break;
case LNBREAK:
REXEC_FBC_CSCAN(is_LNBREAK_utf8_safe(s, strend),
case NPOSIXA:
if (utf8_target) {
/* The complement of something that matches only ASCII matches all
- * UTF-8 variant code points, plus everything in ASCII that isn't
- * in the class */
- REXEC_FBC_UTF8_CLASS_SCAN(! UTF8_IS_INVARIANT(*s)
+ * non-ASCII, plus everything in ASCII that isn't in the class. */
+ REXEC_FBC_UTF8_CLASS_SCAN(! isASCII_utf8(s)
|| ! _generic_isCC_A(*s, FLAGS(c)));
break;
}
macros */
case _CC_ENUM_SPACE: /* XXX would require separate code if we
revert the change of \v matching this */
- /* FALL THROUGH */
+ /* FALLTHROUGH */
case _CC_ENUM_PSXSPC:
REXEC_FBC_UTF8_CLASS_SCAN(
break;
default:
Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
- break;
}
return 0;
got_it:
/* flags: For optimizations. See REXEC_* in regexp.h */
{
- dVAR;
struct regexp *const prog = ReANY(rx);
char *s;
regnode *c;
/* Be paranoid... */
if (prog == NULL || stringarg == NULL) {
Perl_croak(aTHX_ "NULL regexp parameter");
- return 0;
}
DEBUG_EXECUTE_r(
Perl_croak(aTHX_ "corrupted regexp program");
}
+ RX_MATCH_TAINTED_off(rx);
+
reginfo->prog = rx; /* Yes, sorry that this is confusing. */
reginfo->intuit = 0;
reginfo->is_utf8_target = cBOOL(utf8_target);
STATIC I32 /* 0 failure, 1 success */
S_regtry(pTHX_ regmatch_info *reginfo, char **startposp)
{
- dVAR;
CHECKPOINT lastcp;
REGEXP *const rx = reginfo->prog;
regexp *const prog = ReANY(rx);
* or 0 if non of the buffers matched.
*/
STATIC I32
-S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan)
+S_reg_check_named_buff_matched(const regexp *rex, const regnode *scan)
{
I32 n;
RXi_GET_DECL(rex,rexi);
}
else { /* an EXACTFish node which doesn't begin with a multi-char fold */
c1 = is_utf8_pat ? valid_utf8_to_uvchr(pat, NULL) : *pat;
- if (c1 > 256) {
+ if (c1 > 255) {
/* Load the folds hash, if not already done */
SV** listp;
if (! PL_utf8_foldclosures) {
- if (! PL_utf8_tofold) {
- U8 dummy[UTF8_MAXBYTES_CASE+1];
-
- /* Force loading this by folding an above-Latin1 char */
- to_utf8_fold((U8*) HYPHEN_UTF8, dummy, NULL);
- assert(PL_utf8_tofold); /* Verify that worked */
- }
- PL_utf8_foldclosures = _swash_inversion_hash(PL_utf8_tofold);
+ _load_PL_utf8_foldclosures();
}
/* The fold closures data structure is a hash with the keys
/* Folds that cross the 255/256 boundary are forbidden
* if EXACTFL (and isnt a UTF8 locale), or EXACTFA and
* one is ASCIII. Since the pattern character is above
- * 256, and its only other match is below 256, the only
+ * 255, and its only other match is below 256, the only
* legal match will be to itself. We have thrown away
* the original, so have to compute which is the one
- * above 255 */
+ * above 255. */
if ((c1 < 256) != (c2 < 256)) {
if ((OP(text_node) == EXACTFL
&& ! IN_UTF8_CTYPE_LOCALE)
}
}
}
- else /* Here, c1 is < 255 */
+ else /* Here, c1 is <= 255 */
if (utf8_target
&& HAS_NONLATIN1_FOLD_CLOSURE(c1)
&& ( ! (OP(text_node) == EXACTFL && ! IN_UTF8_CTYPE_LOCALE))
case EXACTFA_NO_TRIE: /* This node only generated for
non-utf8 patterns */
assert(! is_utf8_pat);
- /* FALL THROUGH */
+ /* FALLTHROUGH */
case EXACTFA:
case EXACTFU_SS:
case EXACTFU:
break;
case EOL: /* /..$/ */
- /* FALL THROUGH */
+ /* FALLTHROUGH */
case SEOL: /* /..$/s */
if (!NEXTCHR_IS_EOS && nextchr != '\n')
sayNO;
sayNO_SILENT;
assert(0); /* NOTREACHED */
}
- /* FALL THROUGH */
+ /* FALLTHROUGH */
case TRIE: /* (ab|cd) */
/* the basic plan of execution of the trie is:
* At the beginning, run though all the states, and
case EXACTFA_NO_TRIE: /* This node only generated for non-utf8
patterns */
assert(! is_utf8_pat);
- /* FALL THROUGH */
+ /* FALLTHROUGH */
case EXACTFA: /* /abc/iaa */
folder = foldEQ_latin1;
fold_array = PL_fold_latin1;
break;
default:
Perl_croak(aTHX_ "panic: Unexpected FLAGS %u in op %u", FLAGS(scan), OP(scan));
- break;
}
}
/* Note requires that all BOUNDs be lower than all NBOUNDs in
assert(0); /* NOTREACHED */
case WHILEM_A_min_fail: /* just failed to match A in a minimal match */
- /* FALL THROUGH */
+ /* FALLTHROUGH */
case WHILEM_A_pre_fail: /* just failed to match even minimal A */
REGCP_UNWIND(ST.lastcp);
regcppop(rex, &maxopenparen);
if (next == scan)
next = NULL;
scan = NEXTOPER(scan);
- /* FALL THROUGH */
+ /* FALLTHROUGH */
case BRANCH: /* /(...|A|...)/ */
scan = NEXTOPER(scan); /* scan now points to inner node */
assert(0); /* NOTREACHED */
}
}
- /* FALL THROUGH */
+ /* FALLTHROUGH */
case CURLY_B_max_fail:
/* failed to find B in a greedy match */
case IFMATCH_A_fail: /* body of (?...A) failed */
ST.wanted = !ST.wanted;
- /* FALL THROUGH */
+ /* FALLTHROUGH */
case IFMATCH_A: /* body of (?...A) succeeded */
if (ST.logical) {
/* push a state that backtracks on success */
st->u.yes.prev_yes_state = yes_state;
yes_state = st;
- /* FALL THROUGH */
+ /* FALLTHROUGH */
push_state:
/* push a new regex state, then continue at scan */
{
sv_commit = &PL_sv_yes;
sv_yes_mark = &PL_sv_no;
}
+ assert(sv_err);
+ assert(sv_mrk);
sv_setsv(sv_err, sv_commit);
sv_setsv(sv_mrk, sv_yes_mark);
}
S_regrepeat(pTHX_ regexp *prog, char **startposp, const regnode *p,
regmatch_info *const reginfo, I32 max, int depth)
{
- dVAR;
char *scan; /* Pointer to current position in target string */
I32 c;
char *loceol = reginfo->strend; /* local version */
case EXACTFA_NO_TRIE: /* This node only generated for non-utf8 patterns */
assert(! reginfo->is_utf8_pat);
- /* FALL THROUGH */
+ /* FALLTHROUGH */
case EXACTFA:
utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII;
goto do_exactf;
to_complement = 1;
goto utf8_posix;
}
- /* FALL THROUGH */
+ /* FALLTHROUGH */
case NPOSIXA:
if (! utf8_target) {
else {
/* The complement of something that matches only ASCII matches all
- * UTF-8 variant code points, plus everything in ASCII that isn't
- * in the class. */
+ * non-ASCII, plus everything in ASCII that isn't in the class. */
while (hardcount < max && scan < loceol
- && (! UTF8_IS_INVARIANT(*scan)
+ && (! isASCII_utf8(scan)
|| ! _generic_isCC_A((U8) *scan, FLAGS(p))))
{
scan += UTF8SKIP(scan);
case _CC_ENUM_SPACE: /* XXX would require separate code
if we revert the change of \v
matching this */
- /* FALL THROUGH */
+ /* FALLTHROUGH */
case _CC_ENUM_PSXSPC:
while (hardcount < max
&& scan < loceol
* swash are returned (in a printable form).
* Tied intimately to how regcomp.c sets up the data structure */
- dVAR;
SV *sw = NULL;
SV *si = NULL; /* Input swash initialization string */
SV* invlist = NULL;
*only_utf8_locale_ptr = ary[2];
}
else {
+ assert(only_utf8_locale_ptr);
*only_utf8_locale_ptr = NULL;
}
}
else if (doinit && ((si && si != &PL_sv_undef)
|| (invlist && invlist != &PL_sv_undef))) {
-
+ assert(si);
sw = _core_swash_init("utf8", /* the utf8 package */
"", /* nameless */
si,
/* If requested, return a printable version of what this swash matches */
if (listsvp) {
- SV* matches_string = newSVpvn("", 0);
+ SV* matches_string = newSVpvs("");
/* The swash should be used, if possible, to get the data, as it
* contains the resolved data. But this function can be called at
SV * const sw = _get_regclass_nonbitmap_data(prog, n, TRUE, 0,
&only_utf8_locale);
if (sw) {
+ U8 utf8_buffer[2];
U8 * utf8_p;
if (utf8_target) {
utf8_p = (U8 *) p;
} else { /* Convert to utf8 */
- STRLEN len = 1;
- utf8_p = bytes_to_utf8(p, &len);
+ utf8_p = utf8_buffer;
+ append_utf8_from_native_byte(*p, &utf8_p);
+ utf8_p = utf8_buffer;
}
if (swash_fetch(sw, utf8_p, TRUE)) {
match = TRUE;
}
-
- /* If we allocated a string above, free it */
- if (! utf8_target) Safefree(utf8_p);
}
if (! match && only_utf8_locale && IN_UTF8_CTYPE_LOCALE) {
match = _invlist_contains_cp(only_utf8_locale, c);
* 'off' >= 0, backwards if negative. But don't go outside of position
* 'lim', which better be < s if off < 0 */
- dVAR;
-
PERL_ARGS_ASSERT_REGHOP3;
if (off >= 0) {
STATIC U8 *
S_reghop4(U8 *s, SSize_t off, const U8* llim, const U8* rlim)
{
- dVAR;
-
PERL_ARGS_ASSERT_REGHOP4;
if (off >= 0) {
STATIC U8 *
S_reghopmaybe3(U8* s, SSize_t off, const U8* lim)
{
- dVAR;
-
PERL_ARGS_ASSERT_REGHOPMAYBE3;
if (off >= 0) {
static void
S_cleanup_regmatch_info_aux(pTHX_ void *arg)
{
- dVAR;
regmatch_info_aux *aux = (regmatch_info_aux *) arg;
regmatch_info_aux_eval *eval_state = aux->info_aux_eval;
regmatch_slab *s;
/* Converts substr fields in prog from UTF-8 to bytes, calling fbm_compile
* on the converted value; returns FALSE if can't be converted. */
- dVAR;
int i = 1;
PERL_ARGS_ASSERT_TO_BYTE_SUBSTR;
https://perl5.git.perl.org / perl5.git / blobdiff