within pattern */
int num_code_blocks; /* size of code_blocks[] */
int code_index; /* next code_blocks[] slot */
-#if ADD_TO_REGEXEC
+ SSize_t maxlen; /* mininum possible number of chars in string to match */
+#ifdef ADD_TO_REGEXEC
char *starttry; /* -Dr: where regtry was called. */
#define RExC_starttry (pRExC_state->starttry)
#endif
#define RExC_sawback (pRExC_state->sawback)
#define RExC_seen (pRExC_state->seen)
#define RExC_size (pRExC_state->size)
+#define RExC_maxlen (pRExC_state->maxlen)
#define RExC_npar (pRExC_state->npar)
#define RExC_nestroot (pRExC_state->nestroot)
#define RExC_extralen (pRExC_state->extralen)
DEBUG_OPTIMISE_MORE_r({ \
PerlIO_printf(Perl_debug_log,"RExC_seen: "); \
\
- if (RExC_seen & REG_SEEN_ZERO_LEN) \
- PerlIO_printf(Perl_debug_log,"REG_SEEN_ZERO_LEN "); \
+ if (RExC_seen & REG_ZERO_LEN_SEEN) \
+ PerlIO_printf(Perl_debug_log,"REG_ZERO_LEN_SEEN "); \
\
- if (RExC_seen & REG_SEEN_LOOKBEHIND) \
- PerlIO_printf(Perl_debug_log,"REG_SEEN_LOOKBEHIND "); \
+ if (RExC_seen & REG_LOOKBEHIND_SEEN) \
+ PerlIO_printf(Perl_debug_log,"REG_LOOKBEHIND_SEEN "); \
\
- if (RExC_seen & REG_SEEN_GPOS) \
- PerlIO_printf(Perl_debug_log,"REG_SEEN_GPOS "); \
+ if (RExC_seen & REG_GPOS_SEEN) \
+ PerlIO_printf(Perl_debug_log,"REG_GPOS_SEEN "); \
\
- if (RExC_seen & REG_SEEN_CANY) \
- PerlIO_printf(Perl_debug_log,"REG_SEEN_CANY "); \
+ if (RExC_seen & REG_CANY_SEEN) \
+ PerlIO_printf(Perl_debug_log,"REG_CANY_SEEN "); \
\
- if (RExC_seen & REG_SEEN_RECURSE) \
- PerlIO_printf(Perl_debug_log,"REG_SEEN_RECURSE "); \
+ if (RExC_seen & REG_RECURSE_SEEN) \
+ PerlIO_printf(Perl_debug_log,"REG_RECURSE_SEEN "); \
\
- if (RExC_seen & REG_TOP_LEVEL_BRANCHES) \
- PerlIO_printf(Perl_debug_log,"REG_TOP_LEVEL_BRANCHES "); \
+ if (RExC_seen & REG_TOP_LEVEL_BRANCHES_SEEN) \
+ PerlIO_printf(Perl_debug_log,"REG_TOP_LEVEL_BRANCHES_SEEN "); \
\
- if (RExC_seen & REG_SEEN_VERBARG) \
- PerlIO_printf(Perl_debug_log,"REG_SEEN_VERBARG "); \
+ if (RExC_seen & REG_VERBARG_SEEN) \
+ PerlIO_printf(Perl_debug_log,"REG_VERBARG_SEEN "); \
\
- if (RExC_seen & REG_SEEN_CUTGROUP) \
- PerlIO_printf(Perl_debug_log,"REG_SEEN_CUTGROUP "); \
+ if (RExC_seen & REG_CUTGROUP_SEEN) \
+ PerlIO_printf(Perl_debug_log,"REG_CUTGROUP_SEEN "); \
\
- if (RExC_seen & REG_SEEN_RUN_ON_COMMENT) \
- PerlIO_printf(Perl_debug_log,"REG_SEEN_RUN_ON_COMMENT "); \
+ if (RExC_seen & REG_RUN_ON_COMMENT_SEEN) \
+ PerlIO_printf(Perl_debug_log,"REG_RUN_ON_COMMENT_SEEN "); \
\
- if (RExC_seen & REG_SEEN_EXACTF_SHARP_S) \
- PerlIO_printf(Perl_debug_log,"REG_SEEN_EXACTF_SHARP_S "); \
+ if (RExC_seen & REG_UNFOLDED_MULTI_SEEN) \
+ PerlIO_printf(Perl_debug_log,"REG_UNFOLDED_MULTI_SEEN "); \
\
- if (RExC_seen & REG_SEEN_GOSTART) \
- PerlIO_printf(Perl_debug_log,"REG_SEEN_GOSTART "); \
+ if (RExC_seen & REG_GOSTART_SEEN) \
+ PerlIO_printf(Perl_debug_log,"REG_GOSTART_SEEN "); \
+ \
+ if (RExC_seen & REG_UNBOUNDED_QUANTIFIER_SEEN) \
+ PerlIO_printf(Perl_debug_log,"REG_UNBOUNDED_QUANTIFIER_SEEN "); \
\
PerlIO_printf(Perl_debug_log,"\n"); \
});
if (RExC_contains_locale) {
ANYOF_POSIXL_SETALL(ssc);
ANYOF_FLAGS(ssc) |= ANYOF_LOCALE|ANYOF_POSIXL;
- if (RExC_contains_i) {
- ANYOF_FLAGS(ssc) |= ANYOF_LOC_FOLD;
- }
}
else {
ANYOF_POSIXL_ZERO(ssc);
return FALSE;
}
- if (RExC_contains_locale) {
- if (! (ANYOF_FLAGS(ssc) & ANYOF_LOCALE)
- || ! (ANYOF_FLAGS(ssc) & ANYOF_POSIXL)
- || ! ANYOF_POSIXL_TEST_ALL_SET(ssc))
- {
- return FALSE;
- }
- if (RExC_contains_i && ! (ANYOF_FLAGS(ssc) & ANYOF_LOC_FOLD)) {
- return FALSE;
- }
+ if (RExC_contains_locale
+ && ! ((ANYOF_FLAGS(ssc) & ANYOF_LOCALE)
+ || ! (ANYOF_FLAGS(ssc) & ANYOF_POSIXL)
+ || ! ANYOF_POSIXL_TEST_ALL_SET(ssc)))
+ {
+ return FALSE;
}
return TRUE;
STATIC SV*
S_get_ANYOF_cp_list_for_ssc(pTHX_ const RExC_state_t *pRExC_state,
- const regnode_charclass_posixl* const node)
+ const regnode_charclass_posixl_fold* const node)
{
/* Returns a mortal inversion list defining which code points are matched
* by 'node', which is of type ANYOF. Handles complementing the result if
* appropriate. If some code points aren't knowable at this time, the
- * returned list must, and will, contain every possible code point. */
+ * returned list must, and will, contain every code point that is a
+ * possibility. */
SV* invlist = sv_2mortal(_new_invlist(0));
unsigned int i;
const U32 n = ARG(node);
+ bool new_node_has_latin1 = FALSE;
PERL_ARGS_ASSERT_GET_ANYOF_CP_LIST_FOR_SSC;
for (i = 0; i < 256; i++) {
if (ANYOF_BITMAP_TEST(node, i)) {
invlist = add_cp_to_invlist(invlist, i);
+ new_node_has_latin1 = TRUE;
}
}
if (ANYOF_FLAGS(node) & ANYOF_INVERT) {
_invlist_invert(invlist);
}
+ else if (new_node_has_latin1 && ANYOF_FLAGS(node) & ANYOF_LOC_FOLD) {
+
+ /* Under /li, any 0-255 could fold to any other 0-255, depending on the
+ * locale. We can skip this if there are no 0-255 at all. */
+ _invlist_union(invlist, PL_Latin1, &invlist);
+ }
+
+ /* Similarly add the UTF-8 locale possible matches */
+ if (ANYOF_FLAGS(node) & ANYOF_LOC_FOLD && ANYOF_UTF8_LOCALE_INVLIST(node))
+ {
+ _invlist_union_maybe_complement_2nd(invlist,
+ ANYOF_UTF8_LOCALE_INVLIST(node),
+ ANYOF_FLAGS(node) & ANYOF_INVERT,
+ &invlist);
+ }
return invlist;
}
}
else {
anded_cp_list = get_ANYOF_cp_list_for_ssc(pRExC_state,
- (regnode_charclass_posixl*) and_with);
+ (regnode_charclass_posixl_fold*) and_with);
anded_flags = ANYOF_FLAGS(and_with) & ANYOF_COMMON_FLAGS;
}
* standard, in particular almost everything by Microsoft.
* The loop below just changes e.g., \w into \W and vice versa */
- regnode_charclass_posixl temp;
+ regnode_charclass_posixl_fold temp;
int add = 1; /* To calculate the index of the complement */
ANYOF_POSIXL_ZERO(&temp);
}
else {
ored_cp_list = get_ANYOF_cp_list_for_ssc(pRExC_state,
- (regnode_charclass_posixl*) or_with);
+ (regnode_charclass_posixl_fold*) or_with);
ored_flags = ANYOF_FLAGS(or_with) & ANYOF_COMMON_FLAGS;
}
set_ANYOF_arg(pRExC_state, (regnode *) ssc, invlist, NULL, NULL, FALSE);
+ /* The code points that could match under /li are already incorporated into
+ * the inversion list and bit map */
+ ANYOF_FLAGS(ssc) &= ~ANYOF_LOC_FOLD;
+
assert(! (ANYOF_FLAGS(ssc) & ANYOF_LOCALE) || RExC_contains_locale);
}
foldlen -= UTF8SKIP(uc);
}
else {
- foldlen = is_MULTI_CHAR_FOLD_utf8_safe(uc, e);
+ foldlen = is_MULTI_CHAR_FOLD_utf8(uc);
minbytes++;
}
}
foldlen--;
}
else {
- foldlen = is_MULTI_CHAR_FOLD_latin1_safe(uc, e);
+ foldlen = is_MULTI_CHAR_FOLD_latin1(uc);
minbytes++;
}
}
PerlIO_printf(Perl_debug_log, "\n");
});
Safefree(q);
- /*RExC_seen |= REG_SEEN_TRIEDFA;*/
+ /*RExC_seen |= REG_TRIEDFA_SEEN;*/
}
*
* If a node is to match under /i (folded), the number of characters it matches
* can be different than its character length if it contains a multi-character
- * fold. *min_subtract is set to the total delta of the input nodes.
+ * fold. *min_subtract is set to the total delta number of characters of the
+ * input nodes.
*
- * And *has_exactf_sharp_s is set to indicate whether or not the node is EXACTF
- * and contains LATIN SMALL LETTER SHARP S
+ * And *unfolded_multi_char is set to indicate whether or not the node contains
+ * an unfolded multi-char fold. This happens when whether the fold is valid or
+ * not won't be known until runtime; namely for EXACTF nodes that contain LATIN
+ * SMALL LETTER SHARP S, as only if the target string being matched against
+ * turns out to be UTF-8 is that fold valid; and also for EXACTFL nodes whose
+ * folding rules depend on the locale in force at runtime. (Multi-char folds
+ * whose components are all above the Latin1 range are not run-time locale
+ * dependent, and have already been folded by the time this function is
+ * called.)
*
* This is as good a place as any to discuss the design of handling these
* multi-character fold sequences. It's been wrong in Perl for a very long
* character folded sequences. Since a single character can fold into
* such a sequence, the minimum match length for this node is less than
* the number of characters in the node. This routine returns in
- * *min_subtract how much to subtract from the the actual length of the
- * string to get a real minimum match length; it is 0 if there are no
- * multi-char foldeds. This delta is used by the caller to adjust the min
- * length of the match, and the delta between min and max, so that the
- * optimizer doesn't reject these possibilities based on size constraints.
+ * *min_subtract how many characters to subtract from the the actual
+ * length of the string to get a real minimum match length; it is 0 if
+ * there are no multi-char foldeds. This delta is used by the caller to
+ * adjust the min length of the match, and the delta between min and max,
+ * so that the optimizer doesn't reject these possibilities based on size
+ * constraints.
* 2) For the sequence involving the Sharp s (\xDF), the node type EXACTFU_SS
* is used for an EXACTFU node that contains at least one "ss" sequence in
* it. For non-UTF-8 patterns and strings, this is the only case where
* EXACTFU node without UTF-8 involvement doesn't have to concern itself
* with length changes, and so can be processed faster. regexec.c takes
* advantage of this. Generally, an EXACTFish node that is in UTF-8 is
- * pre-folded by regcomp.c. This saves effort in regex matching.
- * However, the pre-folding isn't done for non-UTF8 patterns because the
- * fold of the MICRO SIGN requires UTF-8, and we don't want to slow things
- * down by forcing the pattern into UTF8 unless necessary. Also what
- * EXACTF and EXACTFL nodes fold to isn't known until runtime. The fold
+ * pre-folded by regcomp.c (except EXACTFL, some of whose folds aren't
+ * known until runtime). This saves effort in regex matching. However,
+ * the pre-folding isn't done for non-UTF8 patterns because the fold of
+ * the MICRO SIGN requires UTF-8, and we don't want to slow things down by
+ * forcing the pattern into UTF8 unless necessary. Also what EXACTF (and,
+ * again, EXACTFL) nodes fold to isn't known until runtime. The fold
* possibilities for the non-UTF8 patterns are quite simple, except for
* the sharp s. All the ones that don't involve a UTF-8 target string are
* members of a fold-pair, and arrays are set up for all of them so that
* this file makes sure that in EXACTFU nodes, the sharp s gets folded to
* 'ss', even if the pattern isn't UTF-8. This avoids the issues
* described in the next item.
- * 3) A problem remains for the sharp s in EXACTF and EXACTFA nodes when the
- * pattern isn't in UTF-8. (BTW, there cannot be an EXACTF node with a
- * UTF-8 pattern.) An assumption that the optimizer part of regexec.c
- * (probably unwittingly, in Perl_regexec_flags()) makes is that a
- * character in the pattern corresponds to at most a single character in
- * the target string. (And I do mean character, and not byte here, unlike
- * other parts of the documentation that have never been updated to
- * account for multibyte Unicode.) sharp s in EXACTF nodes can match the
- * two character string 'ss'; in EXACTFA nodes it can match
- * "\x{17F}\x{17F}". These violate the assumption, and they are the only
- * instances where it is violated. I'm reluctant to try to change the
- * assumption, as the code involved is impenetrable to me (khw), so
- * instead the code here punts. This routine examines (when the pattern
- * isn't UTF-8) EXACTF and EXACTFA nodes for the sharp s, and returns a
- * boolean indicating whether or not the node contains a sharp s. When it
- * is true, the caller sets a flag that later causes the optimizer in this
- * file to not set values for the floating and fixed string lengths, and
- * thus avoids the optimizer code in regexec.c that makes the invalid
+ * 3) A problem remains for unfolded multi-char folds. (These occur when the
+ * validity of the fold won't be known until runtime, and so must remain
+ * unfolded for now. This happens for the sharp s in EXACTF and EXACTFA
+ * nodes when the pattern isn't in UTF-8. (Note, BTW, that there cannot
+ * be an EXACTF node with a UTF-8 pattern.) They also occur for various
+ * folds in EXACTFL nodes, regardless of the UTF-ness of the pattern.)
+ * The reason this is a problem is that the optimizer part of regexec.c
+ * (probably unwittingly, in Perl_regexec_flags()) makes an assumption
+ * that a character in the pattern corresponds to at most a single
+ * character in the target string. (And I do mean character, and not byte
+ * here, unlike other parts of the documentation that have never been
+ * updated to account for multibyte Unicode.) sharp s in EXACTF and
+ * EXACTFL nodes can match the two character string 'ss'; in EXACTFA nodes
+ * it can match "\x{17F}\x{17F}". These, along with other ones in EXACTFL
+ * nodes, violate the assumption, and they are the only instances where it
+ * is violated. I'm reluctant to try to change the assumption, as the
+ * code involved is impenetrable to me (khw), so instead the code here
+ * punts. This routine examines EXACTFL nodes, and (when the pattern
+ * isn't UTF-8) EXACTF and EXACTFA for such unfolded folds, and returns a
+ * boolean indicating whether or not the node contains such a fold. When
+ * it is true, the caller sets a flag that later causes the optimizer in
+ * this file to not set values for the floating and fixed string lengths,
+ * and thus avoids the optimizer code in regexec.c that makes the invalid
* assumption. Thus, there is no optimization based on string lengths for
- * non-UTF8-pattern EXACTF and EXACTFA nodes that contain the sharp s.
- * (The reason the assumption is wrong only in these two cases is that all
- * other non-UTF-8 folds are 1-1; and, for UTF-8 patterns, we pre-fold all
- * other folds to their expanded versions. We can't prefold sharp s to
- * 'ss' in EXACTF nodes because we don't know at compile time if it
- * actually matches 'ss' or not. It will match iff the target string is
- * in UTF-8, unlike the EXACTFU nodes, where it always matches; and
- * EXACTFA and EXACTFL where it never does. In an EXACTFA node in a UTF-8
- * pattern, sharp s is folded to "\x{17F}\x{17F}, avoiding the problem;
- * but in a non-UTF8 pattern, folding it to that above-Latin1 string would
- * require the pattern to be forced into UTF-8, the overhead of which we
- * want to avoid.)
+ * EXACTFL nodes that contain these few folds, nor for non-UTF8-pattern
+ * EXACTF and EXACTFA nodes that contain the sharp s. (The reason the
+ * assumption is wrong only in these cases is that all other non-UTF-8
+ * folds are 1-1; and, for UTF-8 patterns, we pre-fold all other folds to
+ * their expanded versions. (Again, we can't prefold sharp s to 'ss' in
+ * EXACTF nodes because we don't know at compile time if it actually
+ * matches 'ss' or not. For EXACTF nodes it will match iff the target
+ * string is in UTF-8. This is in contrast to EXACTFU nodes, where it
+ * always matches; and EXACTFA where it never does. In an EXACTFA node in
+ * a UTF-8 pattern, sharp s is folded to "\x{17F}\x{17F}, avoiding the
+ * problem; but in a non-UTF8 pattern, folding it to that above-Latin1
+ * string would require the pattern to be forced into UTF-8, the overhead
+ * of which we want to avoid. Similarly the unfolded multi-char folds in
+ * EXACTFL nodes will match iff the locale at the time of match is a UTF-8
+ * locale.)
*
* Similarly, the code that generates tries doesn't currently handle
* not-already-folded multi-char folds, and it looks like a pain to change
* using /iaa matching will be doing so almost entirely with ASCII
* strings, so this should rarely be encountered in practice */
-#define JOIN_EXACT(scan,min_subtract,has_exactf_sharp_s, flags) \
+#define JOIN_EXACT(scan,min_subtract,unfolded_multi_char, flags) \
if (PL_regkind[OP(scan)] == EXACT) \
- join_exact(pRExC_state,(scan),(min_subtract),has_exactf_sharp_s, (flags),NULL,depth+1)
+ join_exact(pRExC_state,(scan),(min_subtract),unfolded_multi_char, (flags),NULL,depth+1)
STATIC U32
S_join_exact(pTHX_ RExC_state_t *pRExC_state, regnode *scan,
- UV *min_subtract, bool *has_exactf_sharp_s,
+ UV *min_subtract, bool *unfolded_multi_char,
U32 flags,regnode *val, U32 depth)
{
/* Merge several consecutive EXACTish nodes into one. */
const unsigned int oldl = STR_LEN(scan);
regnode * const nnext = regnext(n);
- /* XXX I (khw) kind of doubt that this works on platforms where
- * U8_MAX is above 255 because of lots of other assumptions */
+ /* XXX I (khw) kind of doubt that this works on platforms (should
+ * Perl ever run on one) where U8_MAX is above 255 because of lots
+ * of other assumptions */
/* Don't join if the sum can't fit into a single node */
if (oldl + STR_LEN(n) > U8_MAX)
break;
}
*min_subtract = 0;
- *has_exactf_sharp_s = FALSE;
+ *unfolded_multi_char = FALSE;
/* Here, all the adjacent mergeable EXACTish nodes have been merged. We
* can now analyze for sequences of problematic code points. (Prior to
* hence missed). The sequences only happen in folding, hence for any
* non-EXACT EXACTish node */
if (OP(scan) != EXACT) {
- const U8 * const s0 = (U8*) STRING(scan);
- const U8 * s = s0;
- const U8 * const s_end = s0 + STR_LEN(scan);
+ U8* s0 = (U8*) STRING(scan);
+ U8* s = s0;
+ U8* s_end = s0 + STR_LEN(scan);
+
+ int total_count_delta = 0; /* Total delta number of characters that
+ multi-char folds expand to */
/* One pass is made over the node's string looking for all the
- * possibilities. to avoid some tests in the loop, there are two main
+ * possibilities. To avoid some tests in the loop, there are two main
* cases, for UTF-8 patterns (which can't have EXACTF nodes) and
* non-UTF-8 */
if (UTF) {
+ U8* folded = NULL;
+
+ if (OP(scan) == EXACTFL) {
+ U8 *d;
+
+ /* An EXACTFL node would already have been changed to another
+ * node type unless there is at least one character in it that
+ * is problematic; likely a character whose fold definition
+ * won't be known until runtime, and so has yet to be folded.
+ * For all but the UTF-8 locale, folds are 1-1 in length, but
+ * to handle the UTF-8 case, we need to create a temporary
+ * folded copy using UTF-8 locale rules in order to analyze it.
+ * This is because our macros that look to see if a sequence is
+ * a multi-char fold assume everything is folded (otherwise the
+ * tests in those macros would be too complicated and slow).
+ * Note that here, the non-problematic folds will have already
+ * been done, so we can just copy such characters. We actually
+ * don't completely fold the EXACTFL string. We skip the
+ * unfolded multi-char folds, as that would just create work
+ * below to figure out the size they already are */
+
+ Newx(folded, UTF8_MAX_FOLD_CHAR_EXPAND * STR_LEN(scan) + 1, U8);
+ d = folded;
+ while (s < s_end) {
+ STRLEN s_len = UTF8SKIP(s);
+ if (! is_PROBLEMATIC_LOCALE_FOLD_utf8(s)) {
+ Copy(s, d, s_len, U8);
+ d += s_len;
+ }
+ else if (is_FOLDS_TO_MULTI_utf8(s)) {
+ *unfolded_multi_char = TRUE;
+ Copy(s, d, s_len, U8);
+ d += s_len;
+ }
+ else if (isASCII(*s)) {
+ *(d++) = toFOLD(*s);
+ }
+ else {
+ STRLEN len;
+ _to_utf8_fold_flags(s, d, &len, FOLD_FLAGS_FULL);
+ d += len;
+ }
+ s += s_len;
+ }
+
+ /* Point the remainder of the routine to look at our temporary
+ * folded copy */
+ s = folded;
+ s_end = d;
+ } /* End of creating folded copy of EXACTFL string */
/* Examine the string for a multi-character fold sequence. UTF-8
* patterns have all characters pre-folded by the time this code is
length sequence we are looking for is 2 */
{
int count = 0; /* How many characters in a multi-char fold */
- int len = is_MULTI_CHAR_FOLD_utf8_safe(s, s_end);
+ int len = is_MULTI_CHAR_FOLD_utf8(s);
if (! len) { /* Not a multi-char fold: get next char */
s += UTF8SKIP(s);
continue;
}
- /* Nodes with 'ss' require special handling, except for EXACTFL
- * and EXACTFA-ish for which there is no multi-char fold to
- * this */
+ /* Nodes with 'ss' require special handling, except for
+ * EXACTFA-ish for which there is no multi-char fold to this */
if (len == 2 && *s == 's' && *(s+1) == 's'
- && OP(scan) != EXACTFL
&& OP(scan) != EXACTFA
&& OP(scan) != EXACTFA_NO_TRIE)
{
count = 2;
- OP(scan) = EXACTFU_SS;
+ if (OP(scan) != EXACTFL) {
+ OP(scan) = EXACTFU_SS;
+ }
s += 2;
}
else { /* Here is a generic multi-char fold. */
- const U8* multi_end = s + len;
-
- /* Count how many characters in it. In the case of /l and
- * /aa, no folds which contain ASCII code points are
- * allowed, so check for those, and skip if found. (In
- * EXACTFL, no folds are allowed to any Latin1 code point,
- * not just ASCII. But there aren't any of these
- * currently, nor ever likely, so don't take the time to
- * test for them. The code that generates the
- * is_MULTI_foo() macros croaks should one actually get put
- * into Unicode .) */
- if (OP(scan) != EXACTFL
- && OP(scan) != EXACTFA
- && OP(scan) != EXACTFA_NO_TRIE)
- {
+ U8* multi_end = s + len;
+
+ /* Count how many characters in it. In the case of /aa, no
+ * folds which contain ASCII code points are allowed, so
+ * check for those, and skip if found. */
+ if (OP(scan) != EXACTFA && OP(scan) != EXACTFA_NO_TRIE) {
count = utf8_length(s, multi_end);
s = multi_end;
}
/* The delta is how long the sequence is minus 1 (1 is how long
* the character that folds to the sequence is) */
- *min_subtract += count - 1;
- next_iteration: ;
+ total_count_delta += count - 1;
+ next_iteration: ;
}
+
+ /* We created a temporary folded copy of the string in EXACTFL
+ * nodes. Therefore we need to be sure it doesn't go below zero,
+ * as the real string could be shorter */
+ if (OP(scan) == EXACTFL) {
+ int total_chars = utf8_length((U8*) STRING(scan),
+ (U8*) STRING(scan) + STR_LEN(scan));
+ if (total_count_delta > total_chars) {
+ total_count_delta = total_chars;
+ }
+ }
+
+ *min_subtract += total_count_delta;
+ Safefree(folded);
}
else if (OP(scan) == EXACTFA) {
while (s < s_end) {
if (*s == LATIN_SMALL_LETTER_SHARP_S) {
OP(scan) = EXACTFA_NO_TRIE;
- *has_exactf_sharp_s = TRUE;
+ *unfolded_multi_char = TRUE;
break;
}
s++;
continue;
}
}
- else if (OP(scan) != EXACTFL) {
-
- /* Non-UTF-8 pattern, not EXACTFA nor EXACTFL node. Look for the
- * multi-char folds that are all Latin1. (This code knows that
- * there are no current multi-char folds possible with EXACTFL,
- * relying on fold_grind.t to catch any errors if the very unlikely
- * event happens that some get added in future Unicode versions.)
- * As explained in the comments preceding this function, we look
- * also for the sharp s in EXACTF nodes; it can be in the final
- * position. Otherwise we can stop looking 1 byte earlier because
- * have to find at least two characters for a multi-fold */
- const U8* upper = (OP(scan) == EXACTF) ? s_end : s_end -1;
+ else {
+
+ /* Non-UTF-8 pattern, not EXACTFA node. Look for the multi-char
+ * folds that are all Latin1. As explained in the comments
+ * preceding this function, we look also for the sharp s in EXACTF
+ * and EXACTFL nodes; it can be in the final position. Otherwise
+ * we can stop looking 1 byte earlier because have to find at least
+ * two characters for a multi-fold */
+ const U8* upper = (OP(scan) == EXACTF || OP(scan) == EXACTFL)
+ ? s_end
+ : s_end -1;
while (s < upper) {
- int len = is_MULTI_CHAR_FOLD_latin1_safe(s, s_end);
+ int len = is_MULTI_CHAR_FOLD_latin1(s);
if (! len) { /* Not a multi-char fold. */
- if (*s == LATIN_SMALL_LETTER_SHARP_S && OP(scan) == EXACTF)
+ if (*s == LATIN_SMALL_LETTER_SHARP_S
+ && (OP(scan) == EXACTF || OP(scan) == EXACTFL))
{
- *has_exactf_sharp_s = TRUE;
+ *unfolded_multi_char = TRUE;
}
s++;
continue;
* changed so that a sharp s in the string can match this
* ss in the pattern, but they remain EXACTF nodes, as they
* won't match this unless the target string is is UTF-8,
- * which we don't know until runtime */
- if (OP(scan) != EXACTF) {
+ * which we don't know until runtime. EXACTFL nodes can't
+ * transform into EXACTFU nodes */
+ if (OP(scan) != EXACTF && OP(scan) != EXACTFL) {
OP(scan) = EXACTFU_SS;
}
}
} scan_frame;
-#define SCAN_COMMIT(s, data, m) scan_commit(s, data, m, is_inf)
-
STATIC SSize_t
S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp,
SSize_t *minlenp, SSize_t *deltap,
UV min_subtract = 0; /* How mmany chars to subtract from the minimum
node length to get a real minimum (because
the folded version may be shorter) */
- bool has_exactf_sharp_s = FALSE;
+ bool unfolded_multi_char = FALSE;
/* Peephole optimizer: */
DEBUG_OPTIMISE_MORE_r(
{
/* Its not clear to khw or hv why this is done here, and not in the
* clauses that deal with EXACT nodes. khw's guess is that it's
* because of a previous design */
- JOIN_EXACT(scan,&min_subtract, &has_exactf_sharp_s, 0);
+ JOIN_EXACT(scan,&min_subtract, &unfolded_multi_char, 0);
/* Follow the next-chain of the current node and optimize
away all the NOTHINGs from it. */
regnode_ssc accum;
regnode * const startbranch=scan;
- if (flags & SCF_DO_SUBSTR)
- SCAN_COMMIT(pRExC_state, data, minlenp); /* Cannot merge
- strings after
- this. */
- if (flags & SCF_DO_STCLASS)
+ if (flags & SCF_DO_SUBSTR) {
+ /* Cannot merge strings after this. */
+ scan_commit(pRExC_state, data, minlenp, is_inf);
+ }
+
+ if (flags & SCF_DO_STCLASS)
ssc_init_zero(pRExC_state, &accum);
while (OP(scan) == code) {
if ( startbranch == first
&& scan == tail )
{
- RExC_seen &=~REG_TOP_LEVEL_BRANCHES;
+ RExC_seen &=~REG_TOP_LEVEL_BRANCHES_SEEN;
}
}
#endif
/* some form of infinite recursion, assume infinite length
* */
if (flags & SCF_DO_SUBSTR) {
- SCAN_COMMIT(pRExC_state,data,minlenp);
+ scan_commit(pRExC_state, data, minlenp, is_inf);
data->longest = &(data->longest_float);
}
is_inf = is_inf_internal = 1;
/* Search for fixed substrings supports EXACT only. */
if (flags & SCF_DO_SUBSTR) {
assert(data);
- SCAN_COMMIT(pRExC_state, data, minlenp);
+ scan_commit(pRExC_state, data, minlenp, is_inf);
}
if (UTF) {
const U8 * const s = (U8 *)STRING(scan);
uc = utf8_to_uvchr_buf(s, s + l, NULL);
l = utf8_length(s, s + l);
}
- if (has_exactf_sharp_s) {
- RExC_seen |= REG_SEEN_EXACTF_SHARP_S;
+ if (unfolded_multi_char) {
+ RExC_seen |= REG_UNFOLDED_MULTI_SEEN;
}
min += l - min_subtract;
assert (min >= 0);
}
if (OP(scan) == EXACTFL) {
if (flags & SCF_DO_STCLASS_AND) {
- ssc_flags_and(data->start_class,
- ANYOF_LOCALE|ANYOF_LOC_FOLD);
+ ssc_flags_and(data->start_class, ANYOF_LOCALE);
}
else if (flags & SCF_DO_STCLASS_OR) {
- ANYOF_FLAGS(data->start_class)
- |= ANYOF_LOCALE|ANYOF_LOC_FOLD;
+ ANYOF_FLAGS(data->start_class) |= ANYOF_LOCALE;
}
/* We don't know what the folds are; it could be anything. XXX
scan = NEXTOPER(scan);
goto do_curly;
}
- is_inf = is_inf_internal = 1;
- scan = regnext(scan);
if (flags & SCF_DO_SUBSTR) {
- SCAN_COMMIT(pRExC_state, data, minlenp);
+ scan_commit(pRExC_state, data, minlenp, is_inf);
/* Cannot extend fixed substrings */
data->longest = &(data->longest_float);
}
+ is_inf = is_inf_internal = 1;
+ scan = regnext(scan);
goto optimize_curly_tail;
case CURLY:
if (stopparen>0 && (OP(scan)==CURLYN || OP(scan)==CURLYM)
next_is_eval = (OP(scan) == EVAL);
do_curly:
if (flags & SCF_DO_SUBSTR) {
- if (mincount == 0) SCAN_COMMIT(pRExC_state,data,minlenp);
+ if (mincount == 0)
+ scan_commit(pRExC_state, data, minlenp, is_inf);
/* Cannot extend fixed substrings */
pos_before = data->pos_min;
}
is_inf_internal |= deltanext == SSize_t_MAX
|| (maxcount == REG_INFTY && minnext + deltanext > 0);
is_inf |= is_inf_internal;
- if (is_inf)
+ if (is_inf) {
delta = SSize_t_MAX;
- else
+ } else {
delta += (minnext + deltanext) * maxcount
- minnext * mincount;
-
+ }
/* Try powerful optimization CURLYX => CURLYN. */
if ( OP(oscan) == CURLYX && data
&& data->flags & SF_IN_PAR
&& !(data->flags & SF_HAS_EVAL)
&& !deltanext /* atom is fixed width */
&& minnext != 0 /* CURLYM can't handle zero width */
- && ! (RExC_seen & REG_SEEN_EXACTF_SHARP_S) /* Nor \xDF */
+
+ /* Nor characters whose fold at run-time may be
+ * multi-character */
+ && ! (RExC_seen & REG_UNFOLDED_MULTI_SEEN)
) {
/* XXXX How to optimize if data == 0? */
/* Optimize to a simpler form. */
pars++;
if (flags & SCF_DO_SUBSTR) {
SV *last_str = NULL;
+ STRLEN last_chrs = 0;
int counted = mincount != 0;
if (data->last_end > 0 && mincount != 0) { /* Ends with a
l -= old;
/* Get the added string: */
last_str = newSVpvn_utf8(s + old, l, UTF);
+ last_chrs = UTF ? utf8_length((U8*)(s + old),
+ (U8*)(s + old + l)) : l;
if (deltanext == 0 && pos_before == b) {
/* What was added is a constant string */
if (mincount > 1) {
+
SvGROW(last_str, (mincount * l) + 1);
repeatcpy(SvPVX(last_str) + l,
SvPVX_const(last_str), l,
SvUTF8(sv) && SvMAGICAL(sv) ?
mg_find(sv, PERL_MAGIC_utf8) : NULL;
if (mg && mg->mg_len >= 0)
- mg->mg_len += CHR_SVLEN(last_str) - l;
+ mg->mg_len += last_chrs * (mincount-1);
}
+ last_chrs *= mincount;
data->last_end += l * (mincount - 1);
}
} else {
if (mincount != maxcount) {
/* Cannot extend fixed substrings found inside
the group. */
- SCAN_COMMIT(pRExC_state,data,minlenp);
+ scan_commit(pRExC_state, data, minlenp, is_inf);
if (mincount && last_str) {
SV * const sv = data->last_found;
MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ?
mg->mg_len = -1;
sv_setsv(sv, last_str);
data->last_end = data->pos_min;
- data->last_start_min =
- data->pos_min - CHR_SVLEN(last_str);
+ data->last_start_min = data->pos_min - last_chrs;
data->last_start_max = is_inf
? SSize_t_MAX
- : data->pos_min + data->pos_delta
- - CHR_SVLEN(last_str);
+ : data->pos_min + data->pos_delta - last_chrs;
}
data->longest = &(data->longest_float);
}
case REF:
case CLUMP:
if (flags & SCF_DO_SUBSTR) {
- SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect
- anything... */
+ /* Cannot expect anything... */
+ scan_commit(pRExC_state, data, minlenp, is_inf);
data->longest = &(data->longest_float);
}
is_inf = is_inf_internal = 1;
min++;
delta++; /* Because of the 2 char string cr-lf */
if (flags & SCF_DO_SUBSTR) {
- SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect
- anything... */
+ /* Cannot expect anything... */
+ scan_commit(pRExC_state, data, minlenp, is_inf);
data->pos_min += 1;
data->pos_delta += 1;
data->longest = &(data->longest_float);
else if (REGNODE_SIMPLE(OP(scan))) {
if (flags & SCF_DO_SUBSTR) {
- SCAN_COMMIT(pRExC_state,data,minlenp);
+ scan_commit(pRExC_state, data, minlenp, is_inf);
data->pos_min++;
}
min++;
/* NPOSIXD matches all upper Latin1 code points unless the
* target string being matched is UTF-8, which is
- * unknowable until match time */
- if (PL_regkind[OP(scan)] == NPOSIXD) {
- _invlist_union_complement_2nd(my_invlist,
- PL_XPosix_ptrs[_CC_ASCII], &my_invlist);
+ * unknowable until match time. Since we are going to
+ * invert, we want to get rid of all of them so that the
+ * inversion will match all */
+ if (OP(scan) == NPOSIXD) {
+ _invlist_subtract(my_invlist, PL_UpperLatin1,
+ &my_invlist);
}
join_posix:
data->flags |= (OP(scan) == MEOL
? SF_BEFORE_MEOL
: SF_BEFORE_SEOL);
- SCAN_COMMIT(pRExC_state, data, minlenp);
+ scan_commit(pRExC_state, data, minlenp, is_inf);
}
else if ( PL_regkind[OP(scan)] == BRANCHJ
if ((flags & SCF_DO_SUBSTR) && data->last_found) {
f |= SCF_DO_SUBSTR;
if (scan->flags)
- SCAN_COMMIT(pRExC_state, &data_fake,minlenp);
+ scan_commit(pRExC_state, &data_fake, minlenp, is_inf);
data_fake.last_found=newSVsv(data->last_found);
}
}
if ((flags & SCF_DO_SUBSTR) && data_fake.last_found) {
if (RExC_rx->minlen<*minnextp)
RExC_rx->minlen=*minnextp;
- SCAN_COMMIT(pRExC_state, &data_fake, minnextp);
+ scan_commit(pRExC_state, &data_fake, minnextp, is_inf);
SvREFCNT_dec_NN(data_fake.last_found);
if ( data_fake.minlen_fixed != minlenp )
}
else if ( PL_regkind[OP(scan)] == ENDLIKE ) {
if (flags & SCF_DO_SUBSTR) {
- SCAN_COMMIT(pRExC_state,data,minlenp);
+ scan_commit(pRExC_state, data, minlenp, is_inf);
flags &= ~SCF_DO_SUBSTR;
}
if (data && OP(scan)==ACCEPT) {
else if (OP(scan) == LOGICAL && scan->flags == 2) /* Embedded follows */
{
if (flags & SCF_DO_SUBSTR) {
- SCAN_COMMIT(pRExC_state,data,minlenp);
+ scan_commit(pRExC_state, data, minlenp, is_inf);
data->longest = &(data->longest_float);
}
is_inf = is_inf_internal = 1;
flags &= ~SCF_DO_STCLASS;
}
else if (OP(scan) == GPOS) {
- if (!(RExC_rx->extflags & RXf_GPOS_FLOAT) &&
+ if (!(RExC_rx->intflags & PREGf_GPOS_FLOAT) &&
!(delta || is_inf || (data && data->pos_delta)))
{
- if (!(RExC_rx->extflags & RXf_ANCH) && (flags & SCF_DO_SUBSTR))
- RExC_rx->extflags |= RXf_ANCH_GPOS;
+ if (!(RExC_rx->intflags & PREGf_ANCH) && (flags & SCF_DO_SUBSTR))
+ RExC_rx->intflags |= PREGf_ANCH_GPOS;
if (RExC_rx->gofs < (STRLEN)min)
RExC_rx->gofs = min;
} else {
- RExC_rx->extflags |= RXf_GPOS_FLOAT;
+ RExC_rx->intflags |= PREGf_GPOS_FLOAT;
RExC_rx->gofs = 0;
}
}
SSize_t max1 = 0, min1 = SSize_t_MAX;
regnode_ssc accum;
- if (flags & SCF_DO_SUBSTR) /* XXXX Add !SUSPEND? */
- SCAN_COMMIT(pRExC_state, data,minlenp); /* Cannot merge strings
- after this. */
+ if (flags & SCF_DO_SUBSTR) { /* XXXX Add !SUSPEND? */
+ /* Cannot merge strings after this. */
+ scan_commit(pRExC_state, data, minlenp, is_inf);
+ }
if (flags & SCF_DO_STCLASS)
ssc_init_zero(pRExC_state, &accum);
delta += (trie->maxlen - trie->minlen);
flags &= ~SCF_DO_STCLASS; /* xxx */
if (flags & SCF_DO_SUBSTR) {
- SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect
- anything... */
+ /* Cannot expect anything... */
+ scan_commit(pRExC_state, data, minlenp, is_inf);
data->pos_min += trie->minlen;
data->pos_delta += (trie->maxlen - trie->minlen);
if (trie->maxlen != trie->minlen)
*scanp = scan;
*deltap = is_inf_internal ? SSize_t_MAX : delta;
+
if (flags & SCF_DO_SUBSTR && is_inf)
data->pos_delta = SSize_t_MAX - data->pos_min;
if (is_par > (I32)U8_MAX)
DEBUG_STUDYDATA("post-fin:",data,depth);
- return min < stopmin ? min : stopmin;
+ {
+ SSize_t final_minlen= min < stopmin ? min : stopmin;
+
+ if (!(RExC_seen & REG_UNBOUNDED_QUANTIFIER_SEEN) && (RExC_maxlen < final_minlen + delta)) {
+ RExC_maxlen = final_minlen + delta;
+ }
+ return final_minlen;
+ }
+ /* not-reached */
}
STATIC U32
|| (eol /* Can't have SEOL and MULTI */
&& (! meol || (RExC_flags & RXf_PMf_MULTILINE)))
)
- /* See comments for join_exact for why REG_SEEN_EXACTF_SHARP_S */
- || (RExC_seen & REG_SEEN_EXACTF_SHARP_S))
+ /* See comments for join_exact for why REG_UNFOLDED_MULTI_SEEN */
+ || (RExC_seen & REG_UNFOLDED_MULTI_SEEN))
{
return FALSE;
}
RExC_sawback = 0;
RExC_seen = 0;
+ RExC_maxlen = 0;
RExC_in_lookbehind = 0;
RExC_seen_zerolen = *exp == '^' ? -1 : 0;
RExC_extralen = 0;
bool has_default =
(((r->extflags & RXf_PMf_STD_PMMOD) != RXf_PMf_STD_PMMOD)
|| ! has_charset);
- bool has_runon = ((RExC_seen & REG_SEEN_RUN_ON_COMMENT)
- == REG_SEEN_RUN_ON_COMMENT);
+ bool has_runon = ((RExC_seen & REG_RUN_ON_COMMENT_SEEN)
+ == REG_RUN_ON_COMMENT_SEEN);
U16 reganch = (U16)((r->extflags & RXf_PMf_STD_PMMOD)
>> RXf_PMf_STD_PMMOD_SHIFT);
const char *fptr = STD_PAT_MODS; /*"msix"*/
/* setup various meta data about recursion, this all requires
* RExC_npar to be correctly set, and a bit later on we clear it */
- if (RExC_seen & REG_SEEN_RECURSE) {
+ if (RExC_seen & REG_RECURSE_SEEN) {
Newxz(RExC_open_parens, RExC_npar,regnode *);
SAVEFREEPV(RExC_open_parens);
Newxz(RExC_close_parens,RExC_npar,regnode *);
SAVEFREEPV(RExC_close_parens);
}
- if (RExC_seen & (REG_SEEN_RECURSE | REG_SEEN_GOSTART)) {
+ if (RExC_seen & (REG_RECURSE_SEEN | REG_GOSTART_SEEN)) {
/* Note, RExC_npar is 1 + the number of parens in a pattern.
* So its 1 if there are no parens. */
RExC_study_chunk_recursed_bytes= (RExC_npar >> 3) +
DEBUG_OPTIMISE_r(PerlIO_printf(Perl_debug_log,"Restudying\n"));
RExC_state = copyRExC_state;
- if (seen & REG_TOP_LEVEL_BRANCHES)
- RExC_seen |= REG_TOP_LEVEL_BRANCHES;
+ if (seen & REG_TOP_LEVEL_BRANCHES_SEEN)
+ RExC_seen |= REG_TOP_LEVEL_BRANCHES_SEEN;
else
- RExC_seen &= ~REG_TOP_LEVEL_BRANCHES;
+ RExC_seen &= ~REG_TOP_LEVEL_BRANCHES_SEEN;
StructCopy(&zero_scan_data, &data, scan_data_t);
}
#else
/* testing for BRANCH here tells us whether there is "must appear"
data in the pattern. If there is then we can use it for optimisations */
- if (!(RExC_seen & REG_TOP_LEVEL_BRANCHES)) { /* Only one top-level choice.
+ if (!(RExC_seen & REG_TOP_LEVEL_BRANCHES_SEEN)) { /* Only one top-level choice.
*/
SSize_t fake;
STRLEN longest_float_length, longest_fixed_length;
PL_regkind[OP(first)] == NBOUND)
ri->regstclass = first;
else if (PL_regkind[OP(first)] == BOL) {
- r->extflags |= (OP(first) == MBOL
- ? RXf_ANCH_MBOL
+ r->intflags |= (OP(first) == MBOL
+ ? PREGf_ANCH_MBOL
: (OP(first) == SBOL
- ? RXf_ANCH_SBOL
- : RXf_ANCH_BOL));
+ ? PREGf_ANCH_SBOL
+ : PREGf_ANCH_BOL));
first = NEXTOPER(first);
goto again;
}
else if (OP(first) == GPOS) {
- r->extflags |= RXf_ANCH_GPOS;
+ r->intflags |= PREGf_ANCH_GPOS;
first = NEXTOPER(first);
goto again;
}
else if ((!sawopen || !RExC_sawback) &&
(OP(first) == STAR &&
PL_regkind[OP(NEXTOPER(first))] == REG_ANY) &&
- !(r->extflags & RXf_ANCH) && !pRExC_state->num_code_blocks)
+ !(r->intflags & PREGf_ANCH) && !pRExC_state->num_code_blocks)
{
/* turn .* into ^.* with an implied $*=1 */
const int type =
(OP(NEXTOPER(first)) == REG_ANY)
- ? RXf_ANCH_MBOL
- : RXf_ANCH_SBOL;
- r->extflags |= type;
- r->intflags |= PREGf_IMPLICIT;
+ ? PREGf_ANCH_MBOL
+ : PREGf_ANCH_SBOL;
+ r->intflags |= (type | PREGf_IMPLICIT);
first = NEXTOPER(first);
goto again;
}
if ( RExC_npar == 1 && data.longest == &(data.longest_fixed)
&& data.last_start_min == 0 && data.last_end > 0
&& !RExC_seen_zerolen
- && !(RExC_seen & REG_SEEN_VERBARG)
- && !((RExC_seen & REG_SEEN_GPOS) || (r->extflags & RXf_ANCH_GPOS)))
+ && !(RExC_seen & REG_VERBARG_SEEN)
+ && !(RExC_seen & REG_GPOS_SEEN)
+ ){
r->extflags |= RXf_CHECK_ALL;
+ }
scan_commit(pRExC_state, &data,&minlen,0);
longest_float_length = CHR_SVLEN(data.longest_float);
/* A temporary algorithm prefers floated substr to fixed one to dig
* more info. */
if (longest_fixed_length > longest_float_length) {
+ r->substrs->check_ix = 0;
r->check_end_shift = r->anchored_end_shift;
r->check_substr = r->anchored_substr;
r->check_utf8 = r->anchored_utf8;
r->check_offset_min = r->check_offset_max = r->anchored_offset;
- if (r->extflags & RXf_ANCH_SINGLE)
- r->extflags |= RXf_NOSCAN;
+ if (r->intflags & (PREGf_ANCH_SBOL|PREGf_ANCH_GPOS))
+ r->intflags |= PREGf_NOSCAN;
}
else {
+ r->substrs->check_ix = 1;
r->check_end_shift = r->float_end_shift;
r->check_substr = r->float_substr;
r->check_utf8 = r->float_utf8;
if (SvTAIL(r->check_substr ? r->check_substr : r->check_utf8))
r->extflags |= RXf_INTUIT_TAIL;
}
+ r->substrs->data[0].max_offset = r->substrs->data[0].min_offset;
+
/* XXX Unneeded? dmq (shouldn't as this is handled elsewhere)
if ( (STRLEN)minlen < longest_float_length )
minlen= longest_float_length;
}
}
+ if (RExC_seen & REG_UNBOUNDED_QUANTIFIER_SEEN) {
+ r->extflags |= RXf_UNBOUNDED_QUANTIFIER_SEEN;
+ r->maxlen = REG_INFTY;
+ }
+ else {
+ r->maxlen = RExC_maxlen;
+ }
+
/* Guard against an embedded (?=) or (?<=) with a longer minlen than
the "real" pattern. */
DEBUG_OPTIMISE_r({
- PerlIO_printf(Perl_debug_log,"minlen: %"IVdf" r->minlen:%"IVdf"\n",
- (IV)minlen, (IV)r->minlen);
+ PerlIO_printf(Perl_debug_log,"minlen: %"IVdf" r->minlen:%"IVdf" maxlen:%ld\n",
+ (IV)minlen, (IV)r->minlen, RExC_maxlen);
});
r->minlenret = minlen;
if (r->minlen < minlen)
r->minlen = minlen;
- if (RExC_seen & REG_SEEN_GPOS)
- r->extflags |= RXf_GPOS_SEEN;
- if (RExC_seen & REG_SEEN_LOOKBEHIND)
+ if (RExC_seen & REG_GPOS_SEEN)
+ r->intflags |= PREGf_GPOS_SEEN;
+ if (RExC_seen & REG_LOOKBEHIND_SEEN)
r->extflags |= RXf_NO_INPLACE_SUBST; /* inplace might break the
lookbehind */
if (pRExC_state->num_code_blocks)
r->extflags |= RXf_EVAL_SEEN;
- if (RExC_seen & REG_SEEN_CANY)
- r->extflags |= RXf_CANY_SEEN;
- if (RExC_seen & REG_SEEN_VERBARG)
+ if (RExC_seen & REG_CANY_SEEN)
+ r->intflags |= PREGf_CANY_SEEN;
+ if (RExC_seen & REG_VERBARG_SEEN)
{
r->intflags |= PREGf_VERBARG_SEEN;
r->extflags |= RXf_NO_INPLACE_SUBST; /* don't understand this! Yves */
}
- if (RExC_seen & REG_SEEN_CUTGROUP)
+ if (RExC_seen & REG_CUTGROUP_SEEN)
r->intflags |= PREGf_CUTGROUP_SEEN;
if (pm_flags & PMf_USE_RE_EVAL)
r->intflags |= PREGf_USE_RE_EVAL;
else
RXp_PAREN_NAMES(r) = NULL;
+ /* If we have seen an anchor in our pattern then we set the extflag RXf_IS_ANCHORED
+ * so it can be used in pp.c */
+ if (r->intflags & PREGf_ANCH)
+ r->extflags |= RXf_IS_ANCHORED;
+
{
regnode *first = ri->program + 1;
U8 fop = OP(first);
assert(s >= rx->subbeg);
assert((STRLEN)rx->sublen >= (STRLEN)((s - rx->subbeg) + i) );
if (i >= 0) {
-#if NO_TAINT_SUPPORT
+#ifdef NO_TAINT_SUPPORT
sv_setpvn(sv, s, i);
#else
const int oldtainted = TAINT_get;
sv_setpvn(sv, s, i);
TAINT_set(oldtainted);
#endif
- if ( (rx->extflags & RXf_CANY_SEEN)
+ if ( (rx->intflags & PREGf_CANY_SEEN)
? (RXp_MATCH_UTF8(rx)
&& (!i || is_utf8_string((U8*)s, i)))
: (RXp_MATCH_UTF8(rx)) )
/* [19:06] <TimToady> :: is then */
if ( memEQs(start_verb,verb_len,"THEN") ) {
op = CUTGROUP;
- RExC_seen |= REG_SEEN_CUTGROUP;
+ RExC_seen |= REG_CUTGROUP_SEEN;
}
break;
}
}
}
if (!internal_argval)
- RExC_seen |= REG_SEEN_VERBARG;
+ RExC_seen |= REG_VERBARG_SEEN;
} else if ( start_arg ) {
vFAIL3("Verb pattern '%.*s' may not have an argument",
verb_len, start_verb);
paren = 1;
goto capturing_parens;
}
- RExC_seen |= REG_SEEN_LOOKBEHIND;
+ RExC_seen |= REG_LOOKBEHIND_SEEN;
RExC_in_lookbehind++;
RExC_parse++;
case '=': /* (?=...) */
if (*RExC_parse != ')')
FAIL("Sequence (?R) not terminated");
ret = reg_node(pRExC_state, GOSTART);
- RExC_seen |= REG_SEEN_GOSTART;
+ RExC_seen |= REG_GOSTART_SEEN;
*flagp |= POSTPONED;
nextchar(pRExC_state);
return ret;
} else {
RExC_size++;
}
- RExC_seen |= REG_SEEN_RECURSE;
+ RExC_seen |= REG_RECURSE_SEEN;
Set_Node_Length(ret, 1 + regarglen[OP(ret)]); /* MJD */
Set_Node_Offset(ret, parse_start); /* MJD */
if (!SIZE_ONLY ){
if (!RExC_nestroot)
RExC_nestroot = parno;
- if (RExC_seen & REG_SEEN_RECURSE
+ if (RExC_seen & REG_RECURSE_SEEN
&& !RExC_open_parens[parno-1])
{
DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
break;
case 1: case 2:
ender = reganode(pRExC_state, CLOSE, parno);
- if (!SIZE_ONLY && RExC_seen & REG_SEEN_RECURSE) {
+ if (!SIZE_ONLY && RExC_seen & REG_RECURSE_SEEN) {
DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
"Setting close paren #%"IVdf" to %d\n",
(IV)parno, REG_NODE_NUM(ender)));
if (have_branch && !SIZE_ONLY) {
char is_nothing= 1;
if (depth==1)
- RExC_seen |= REG_TOP_LEVEL_BRANCHES;
+ RExC_seen |= REG_TOP_LEVEL_BRANCHES_SEEN;
/* Hook the tails of the branches to the closing node. */
for (br = ret; br; br = regnext(br)) {
ARG1_SET(ret, (U16)min);
ARG2_SET(ret, (U16)max);
}
+ if (max == REG_INFTY)
+ RExC_seen |= REG_UNBOUNDED_QUANTIFIER_SEEN;
goto nest_check;
}
reginsert(pRExC_state, STAR, ret, depth+1);
ret->flags = 0;
RExC_naughty += 4;
+ RExC_seen |= REG_UNBOUNDED_QUANTIFIER_SEEN;
}
else if (op == '*') {
min = 0;
reginsert(pRExC_state, PLUS, ret, depth+1);
ret->flags = 0;
RExC_naughty += 3;
+ RExC_seen |= REG_UNBOUNDED_QUANTIFIER_SEEN;
}
else if (op == '+') {
min = 1;
* If <len> is zero, the function assumes that the node is to contain only
* the single character given by <code_point> and calculates what <len>
* should be. In pass 1, it sizes the node appropriately. In pass 2, it
- * additionally will populate the node's STRING with <code_point>, if <len>
- * is 0. In both cases <*flagp> is appropriately set
+ * additionally will populate the node's STRING with <code_point> or its
+ * fold if folding.
+ *
+ * In both cases <*flagp> is appropriately set
*
* It knows that under FOLD, the Latin Sharp S and UTF characters above
* 255, must be folded (the former only when the rules indicate it can
if (! len_passed_in) {
if (UTF) {
- if (FOLD && (! LOC || code_point > 255)) {
+ if (UNI_IS_INVARIANT(code_point)) {
+ if (LOC || ! FOLD) { /* /l defers folding until runtime */
+ *character = (U8) code_point;
+ }
+ else { /* Here is /i and not /l (toFOLD() is defined on just
+ ASCII, which isn't the same thing as INVARIANT on
+ EBCDIC, but it works there, as the extra invariants
+ fold to themselves) */
+ *character = toFOLD((U8) code_point);
+ }
+ len = 1;
+ }
+ else if (FOLD && (! LOC
+ || ! is_PROBLEMATIC_LOCALE_FOLD_cp(code_point)))
+ { /* Folding, and ok to do so now */
_to_uni_fold_flags(code_point,
character,
&len,
- FOLD_FLAGS_FULL | ((LOC)
- ? FOLD_FLAGS_LOCALE
- : (ASCII_FOLD_RESTRICTED)
- ? FOLD_FLAGS_NOMIX_ASCII
- : 0));
+ FOLD_FLAGS_FULL | ((ASCII_FOLD_RESTRICTED)
+ ? FOLD_FLAGS_NOMIX_ASCII
+ : 0));
+ }
+ else if (code_point <= MAX_UTF8_TWO_BYTE) {
+
+ /* Not folding this cp, and can output it directly */
+ *character = UTF8_TWO_BYTE_HI(code_point);
+ *(character + 1) = UTF8_TWO_BYTE_LO(code_point);
+ len = 2;
}
else {
uvchr_to_utf8( character, code_point);
len = UTF8SKIP(character);
}
- }
- else if (! FOLD
- || code_point != LATIN_SMALL_LETTER_SHARP_S
- || ASCII_FOLD_RESTRICTED
- || ! AT_LEAST_UNI_SEMANTICS)
+ } /* Else pattern isn't UTF8. We only fold the sharp s, when
+ appropriate */
+ else if (UNLIKELY(code_point == LATIN_SMALL_LETTER_SHARP_S)
+ && FOLD
+ && AT_LEAST_UNI_SEMANTICS
+ && ! ASCII_FOLD_RESTRICTED)
{
- *character = (U8) code_point;
- len = 1;
- }
- else {
*character = 's';
*(character + 1) = 's';
len = 2;
}
+ else {
+ *character = (U8) code_point;
+ len = 1;
+ }
}
if (SIZE_ONLY) {
goto finish_meta_pat;
case 'G':
ret = reg_node(pRExC_state, GPOS);
- RExC_seen |= REG_SEEN_GPOS;
+ RExC_seen |= REG_GPOS_SEEN;
*flagp |= SIMPLE;
goto finish_meta_pat;
case 'K':
* be necessary here to avoid cases of memory corruption, as
* with: C<$_="x" x 80; s/x\K/y/> -- rgs
*/
- RExC_seen |= REG_SEEN_LOOKBEHIND;
+ RExC_seen |= REG_LOOKBEHIND_SEEN;
goto finish_meta_pat;
case 'Z':
ret = reg_node(pRExC_state, SEOL);
goto finish_meta_pat;
case 'C':
ret = reg_node(pRExC_state, CANY);
- RExC_seen |= REG_SEEN_CANY;
+ RExC_seen |= REG_CANY_SEEN;
*flagp |= HASWIDTH|SIMPLE;
goto finish_meta_pat;
case 'X':
case 'b':
RExC_seen_zerolen++;
- RExC_seen |= REG_SEEN_LOOKBEHIND;
+ RExC_seen |= REG_LOOKBEHIND_SEEN;
op = BOUND + get_regex_charset(RExC_flags);
if (op > BOUNDA) { /* /aa is same as /a */
op = BOUNDA;
FLAGS(ret) = get_regex_charset(RExC_flags);
*flagp |= SIMPLE;
if (! SIZE_ONLY && (U8) *(RExC_parse + 1) == '{') {
- ckWARNdep(RExC_parse, "\"\\b{\" is deprecated; use \"\\b\\{\" or \"\\b[{]\" instead");
+ /* diag_listed_as: Use "%s" instead of "%s" */
+ vFAIL("Use \"\\b\\{\" instead of \"\\b{\"");
}
goto finish_meta_pat;
case 'B':
RExC_seen_zerolen++;
- RExC_seen |= REG_SEEN_LOOKBEHIND;
+ RExC_seen |= REG_LOOKBEHIND_SEEN;
op = NBOUND + get_regex_charset(RExC_flags);
if (op > NBOUNDA) { /* /aa is same as /a */
op = NBOUNDA;
FLAGS(ret) = get_regex_charset(RExC_flags);
*flagp |= SIMPLE;
if (! SIZE_ONLY && (U8) *(RExC_parse + 1) == '{') {
- ckWARNdep(RExC_parse, "\"\\B{\" is deprecated; use \"\\B\\{\" or \"\\B[{]\" instead");
+ /* diag_listed_as: Use "%s" instead of "%s" */
+ vFAIL("Use \"\\B\\{\" instead of \"\\B{\"");
}
goto finish_meta_pat;
char foldbuf[MAX_NODE_STRING_SIZE+UTF8_MAXBYTES_CASE];
char *s0;
U8 upper_parse = MAX_NODE_STRING_SIZE;
- STRLEN foldlen;
U8 node_type = compute_EXACTish(pRExC_state);
bool next_is_quantifier;
char * oldp = NULL;
/* We can convert EXACTF nodes to EXACTFU if they contain only
* characters that match identically regardless of the target
* string's UTF8ness. The reason to do this is that EXACTF is not
- * trie-able, EXACTFU is. (We don't need to figure this out until
- * pass 2) */
- bool maybe_exactfu = node_type == EXACTF && PASS2;
+ * trie-able, EXACTFU is.
+ *
+ * Similarly, we can convert EXACTFL nodes to EXACTFU if they
+ * contain only above-Latin1 characters (hence must be in UTF8),
+ * which don't participate in folds with Latin1-range characters,
+ * as the latter's folds aren't known until runtime. (We don't
+ * need to figure this out until pass 2) */
+ bool maybe_exactfu = PASS2
+ && (node_type == EXACTF || node_type == EXACTFL);
/* If a folding node contains only code points that don't
* participate in folds, it can be changed into an EXACT node,
reparse:
- /* We do the EXACTFish to EXACT node only if folding, and not if in
- * locale, as whether a character folds or not isn't known until
- * runtime. (And we don't need to figure this out until pass 2) */
- maybe_exact = FOLD && ! LOC && PASS2;
+ /* We do the EXACTFish to EXACT node only if folding. (And we
+ * don't need to figure this out until pass 2) */
+ maybe_exact = FOLD && PASS2;
/* XXX The node can hold up to 255 bytes, yet this only goes to
* 127. I (khw) do not know why. Keeping it somewhat less than
}
case 'c':
p++;
- ender = grok_bslash_c(*p++, UTF, SIZE_ONLY);
+ ender = grok_bslash_c(*p++, SIZE_ONLY);
break;
case '8': case '9': /* must be a backreference */
--p;
goto loopdone;
}
- if (! FOLD) { /* The simple case, just append the literal */
+ if (! FOLD /* The simple case, just append the literal */
+ || (LOC /* Also don't fold for tricky chars under /l */
+ && is_PROBLEMATIC_LOCALE_FOLD_cp(ender)))
+ {
if (UTF) {
const STRLEN unilen = reguni(pRExC_state, ender, s);
if (unilen > 0) {
else {
REGC((char)ender, s++);
}
+
+ /* Can get here if folding only if is one of the /l
+ * characters whose fold depends on the locale. The
+ * occurrence of any of these indicate that we can't
+ * simplify things */
+ if (FOLD) {
+ maybe_exact = FALSE;
+ maybe_exactfu = FALSE;
+ }
}
- else /* FOLD */ if (! ( UTF
+ else /* FOLD */
+ if (! ( UTF
/* See comments for join_exact() as to why we fold this
* non-UTF at compile time */
|| (node_type == EXACTFU
&& ender == LATIN_SMALL_LETTER_SHARP_S)))
{
/* Here, are folding and are not UTF-8 encoded; therefore
- * the character must be in the range 0-255. */
+ * the character must be in the range 0-255, and is not /l
+ * (Not /l because we already handled these under /l in
+ * is_PROBLEMATIC_LOCALE_FOLD_cp */
if (IS_IN_SOME_FOLD_L1(ender)) {
maybe_exact = FALSE;
* unfolded, and we have to calculate how many EXACTish
* nodes it will take; and we may run out of room in a node
* in the middle of a potential multi-char fold, and have
- * to back off accordingly. */
-
- /* Prime the casefolded buffer. Locale rules, which apply
- * only to code points < 256, aren't known until execution,
- * so for them, just output the original character using
- * utf8. If we start to fold non-UTF patterns, be sure to
- * update join_exact() */
- if (LOC && ender < 256) {
- if (UVCHR_IS_INVARIANT(ender)) {
- *s = (U8) ender;
- foldlen = 1;
- } else {
- *s = UTF8_TWO_BYTE_HI(ender);
- *(s + 1) = UTF8_TWO_BYTE_LO(ender);
- foldlen = 2;
- }
+ * to back off accordingly. (Hence we can't use REGC for
+ * the simple case just below.) */
+
+ UV folded;
+ if (isASCII(ender)) {
+ folded = toFOLD(ender);
+ *(s)++ = (U8) folded;
+ }
+ else {
+ STRLEN foldlen;
+
+ folded = _to_uni_fold_flags(
+ ender,
+ (U8 *) s,
+ &foldlen,
+ FOLD_FLAGS_FULL | ((ASCII_FOLD_RESTRICTED)
+ ? FOLD_FLAGS_NOMIX_ASCII
+ : 0));
+ s += foldlen;
+
+ /* The loop increments <len> each time, as all but this
+ * path (and one other) through it add a single byte to
+ * the EXACTish node. But this one has changed len to
+ * be the correct final value, so subtract one to
+ * cancel out the increment that follows */
+ len += foldlen - 1;
}
- else { /* ender >= 256 */
- UV folded = _to_uni_fold_flags(
- ender,
- (U8 *) s,
- &foldlen,
- FOLD_FLAGS_FULL
- | ((LOC) ? FOLD_FLAGS_LOCALE
- : (ASCII_FOLD_RESTRICTED)
- ? FOLD_FLAGS_NOMIX_ASCII
- : 0)
- );
-
- /* If this node only contains non-folding code points
- * so far, see if this new one is also non-folding */
- if (maybe_exact) {
- if (folded != ender) {
+ /* If this node only contains non-folding code points so
+ * far, see if this new one is also non-folding */
+ if (maybe_exact) {
+ if (folded != ender) {
+ maybe_exact = FALSE;
+ }
+ else {
+ /* Here the fold is the original; we have to check
+ * further to see if anything folds to it */
+ if (_invlist_contains_cp(PL_utf8_foldable,
+ ender))
+ {
maybe_exact = FALSE;
}
- else {
- /* Here the fold is the original; we have
- * to check further to see if anything
- * folds to it */
- if (_invlist_contains_cp(PL_utf8_foldable,
- ender))
- {
- maybe_exact = FALSE;
- }
- }
}
- ender = folded;
}
- s += foldlen;
-
- /* The loop increments <len> each time, as all but this
- * path (and one other) through it add a single byte to the
- * EXACTish node. But this one has changed len to be the
- * correct final value, so subtract one to cancel out the
- * increment that follows */
- len += foldlen - 1;
+ ender = folded;
}
if (next_is_quantifier) {
if (! UTF) {
- /* These two have no multi-char folds to non-UTF characters
- */
- if (ASCII_FOLD_RESTRICTED || LOC) {
+ /* This has no multi-char folds to non-UTF characters */
+ if (ASCII_FOLD_RESTRICTED) {
goto loopdone;
}
}
}
else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
-
- /* No Latin1 characters participate in multi-char
- * folds under /l */
- if (LOC
- || ! IS_NON_FINAL_FOLD(TWO_BYTE_UTF8_TO_NATIVE(
+ if (! IS_NON_FINAL_FOLD(TWO_BYTE_UTF8_TO_NATIVE(
*s, *(s+1))))
{
break;
* code points in the node that participate in folds;
* similarly for 'maybe_exactfu' and code points that match
* differently depending on UTF8ness of the target string
- * */
+ * (for /u), or depending on locale for /l */
if (maybe_exact) {
OP(ret) = EXACT;
}
}
} while (p < e);
if (!ended)
- RExC_seen |= REG_SEEN_RUN_ON_COMMENT;
+ RExC_seen |= REG_RUN_ON_COMMENT_SEEN;
}
else
break;
/* Returns the next non-pattern-white space, non-comment character (the
* latter only if 'recognize_comment is true) in the string p, which is
* ended by RExC_end. If there is no line break ending a comment,
- * RExC_seen has added the REG_SEEN_RUN_ON_COMMENT flag; */
+ * RExC_seen has added the REG_RUN_ON_COMMENT_SEEN flag; */
const char *e = RExC_end;
PERL_ARGS_ASSERT_REGPATWS;
}
} while (p < e);
if (!ended)
- RExC_seen |= REG_SEEN_RUN_ON_COMMENT;
+ RExC_seen |= REG_RUN_ON_COMMENT_SEEN;
}
else
break;
goto recode_encoding;
break;
case 'c':
- value = grok_bslash_c(*RExC_parse++, UTF, SIZE_ONLY);
+ value = grok_bslash_c(*RExC_parse++, SIZE_ONLY);
break;
case '0': case '1': case '2': case '3': case '4':
case '5': case '6': case '7':
* space will contain a bit for each named class that is to be matched
* against. This isn't needed for \p{} and pseudo-classes, as they are
* not affected by locale, and hence are dealt with separately */
- if (LOC
- && ! need_class
- && (ANYOF_LOCALE == ANYOF_POSIXL
- || (namedclass > OOB_NAMEDCLASS
- && namedclass < ANYOF_POSIXL_MAX)))
- {
- need_class = 1;
- if (SIZE_ONLY) {
- RExC_size += ANYOF_POSIXL_SKIP - ANYOF_SKIP;
+ if (LOC) {
+ if (FOLD && ! need_class) {
+ need_class = 1;
+ if (SIZE_ONLY) {
+ RExC_size += ANYOF_POSIXL_FOLD_SKIP - ANYOF_SKIP;
+ }
+ else {
+ RExC_emit += ANYOF_POSIXL_FOLD_SKIP - ANYOF_SKIP;
+ }
+
+ /* We need to initialize this here because this node type has
+ * this field, and will skip getting initialized when we get to
+ * a posix class since are doing it here */
+ ANYOF_POSIXL_ZERO(ret);
}
- else {
- RExC_emit += ANYOF_POSIXL_SKIP - ANYOF_SKIP;
+ if (ANYOF_LOCALE == ANYOF_POSIXL
+ || (namedclass > OOB_NAMEDCLASS
+ && namedclass < ANYOF_POSIXL_MAX))
+ {
+ if (! need_class) {
+ need_class = 1;
+ if (SIZE_ONLY) {
+ RExC_size += ANYOF_POSIXL_SKIP - ANYOF_SKIP;
+ }
+ else {
+ RExC_emit += ANYOF_POSIXL_SKIP - ANYOF_SKIP;
+ }
+ ANYOF_POSIXL_ZERO(ret);
+ }
+ ANYOF_FLAGS(ret) |= ANYOF_POSIXL;
}
- ANYOF_POSIXL_ZERO(ret);
- ANYOF_FLAGS(ret) |= ANYOF_POSIXL;
}
if (namedclass > OOB_NAMEDCLASS) { /* this is a named class \blah */
#ifndef HAS_ISASCII
&& classnum != _CC_ASCII
#endif
-#ifndef HAS_ISBLANK
- && classnum != _CC_BLANK
-#endif
) {
/* See if it already matches the complement of this POSIX
? &posixes
: &nposixes;
SV** source_ptr = &PL_XPosix_ptrs[classnum];
-#ifndef HAS_ISBLANK
- /* If the platform doesn't have isblank(), we handle locale
- * with the hardcoded ASII values. */
- if (LOC && classnum == _CC_BLANK) {
- _invlist_subtract(*source_ptr,
- PL_UpperLatin1,
- source_ptr);
- }
-#endif
-
_invlist_union_maybe_complement_2nd(
*posixes_ptr,
*source_ptr,
value,
foldbuf,
&foldlen,
- FOLD_FLAGS_FULL
- | ((LOC) ? FOLD_FLAGS_LOCALE
- : (ASCII_FOLD_RESTRICTED)
- ? FOLD_FLAGS_NOMIX_ASCII
- : 0)
+ FOLD_FLAGS_FULL | (ASCII_FOLD_RESTRICTED
+ ? FOLD_FLAGS_NOMIX_ASCII
+ : 0)
);
/* Here, <folded> should be the first character of the
if (op > POSIXA) { /* /aa is same as /a */
op = POSIXA;
}
-#ifndef HAS_ISBLANK
- if (op == POSIXL
- && (namedclass == ANYOF_BLANK
- || namedclass == ANYOF_NBLANK))
- {
- op = POSIXA;
- }
-#endif
join_posix:
/* The odd numbered ones are the complements of the
UV start, end; /* End points of code point ranges */
SV* fold_intersection = NULL;
+ SV** use_list;
+
+ /* Our calculated list will be for Unicode rules. For locale
+ * matching, we have to keep a separate list that is consulted at
+ * runtime only when the locale indicates Unicode rules. For
+ * non-locale, we just use to the general list */
+ if (LOC) {
+ use_list = &ANYOF_UTF8_LOCALE_INVLIST(ret);
+ *use_list = NULL;
+ }
+ else {
+ use_list = &cp_list;
+ }
/* Only the characters in this class that participate in folds need
* be checked. Get the intersection of this class and all the
while (invlist_iternext(fold_intersection, &start, &end)) {
UV j;
- /* Locale folding for Latin1 characters is deferred until
- * runtime */
- if (LOC && start < 256) {
- start = 256;
- }
-
/* Look at every character in the range */
for (j = start; j <= end; j++) {
-
U8 foldbuf[UTF8_MAXBYTES_CASE+1];
STRLEN foldlen;
SV** listp;
if (IS_IN_SOME_FOLD_L1(j)) {
/* ASCII is always matched; non-ASCII is matched
- * only under Unicode rules */
- if (isASCII(j) || AT_LEAST_UNI_SEMANTICS) {
- cp_list = add_cp_to_invlist(cp_list,
+ * only under Unicode rules (which could happen
+ * under /l if the locale is a UTF-8 one */
+ if (isASCII(j) || ! DEPENDS_SEMANTICS) {
+ *use_list = add_cp_to_invlist(*use_list,
PL_fold_latin1[j]);
}
else {
switch (j) {
case 'k':
case 'K':
- cp_list =
- add_cp_to_invlist(cp_list, KELVIN_SIGN);
+ *use_list =
+ add_cp_to_invlist(*use_list, KELVIN_SIGN);
break;
case 's':
case 'S':
- cp_list = add_cp_to_invlist(cp_list,
+ *use_list = add_cp_to_invlist(*use_list,
LATIN_SMALL_LETTER_LONG_S);
break;
case MICRO_SIGN:
- cp_list = add_cp_to_invlist(cp_list,
- GREEK_CAPITAL_LETTER_MU);
- cp_list = add_cp_to_invlist(cp_list,
- GREEK_SMALL_LETTER_MU);
+ *use_list = add_cp_to_invlist(*use_list,
+ GREEK_CAPITAL_LETTER_MU);
+ *use_list = add_cp_to_invlist(*use_list,
+ GREEK_SMALL_LETTER_MU);
break;
case LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE:
case LATIN_SMALL_LETTER_A_WITH_RING_ABOVE:
- cp_list =
- add_cp_to_invlist(cp_list,
- ANGSTROM_SIGN);
+ *use_list =
+ add_cp_to_invlist(*use_list, ANGSTROM_SIGN);
break;
case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS:
- cp_list = add_cp_to_invlist(cp_list,
+ *use_list = add_cp_to_invlist(*use_list,
LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS);
break;
case LATIN_SMALL_LETTER_SHARP_S:
- cp_list = add_cp_to_invlist(cp_list,
+ *use_list = add_cp_to_invlist(*use_list,
LATIN_CAPITAL_LETTER_SHARP_S);
break;
case 'F': case 'f':
* the simple fold, as the multi-character folds have been
* handled earlier and separated out */
_to_uni_fold_flags(j, foldbuf, &foldlen,
- ((LOC)
- ? FOLD_FLAGS_LOCALE
- : (ASCII_FOLD_RESTRICTED)
- ? FOLD_FLAGS_NOMIX_ASCII
- : 0));
+ (ASCII_FOLD_RESTRICTED)
+ ? FOLD_FLAGS_NOMIX_ASCII
+ : 0);
/* Single character fold of above Latin1. Add everything in
* its fold closure to the list that this node should match.
}
c = SvUV(*c_p);
- /* /aa doesn't allow folds between ASCII and non-;
- * /l doesn't allow them between above and below
- * 256 */
+ /* /aa doesn't allow folds between ASCII and non- */
if ((ASCII_FOLD_RESTRICTED
- && (isASCII(c) != isASCII(j)))
- || (LOC && c < 256)) {
+ && (isASCII(c) != isASCII(j))))
+ {
+ continue;
+ }
+
+ /* Folds under /l which cross the 255/256 boundary
+ * are added to a separate list. (These are valid
+ * only when the locale is UTF-8.) */
+ if (c < 256 && LOC) {
+ *use_list = add_cp_to_invlist(*use_list, c);
continue;
}
- /* Folds involving non-ascii Latin1 characters
- * under /d are added to a separate list */
if (isASCII(c) || c > 255 || AT_LEAST_UNI_SEMANTICS)
{
cp_list = add_cp_to_invlist(cp_list, c);
}
else {
- depends_list = add_cp_to_invlist(depends_list, c);
+ /* Similarly folds involving non-ascii Latin1
+ * characters under /d are added to their list */
+ depends_list = add_cp_to_invlist(depends_list,
+ c);
}
}
}
/* If we didn't do folding, it's because some information isn't available
* until runtime; set the run-time fold flag for these. (We don't have to
* worry about properties folding, as that is taken care of by the swash
- * fetching) */
- if (FOLD && LOC)
- {
- ANYOF_FLAGS(ret) |= ANYOF_LOC_FOLD;
+ * fetching). We know to set the flag if we have a non-NULL list for UTF-8
+ * locales, or the class matches at least one 0-255 range code point */
+ if (LOC && FOLD) {
+ if (ANYOF_UTF8_LOCALE_INVLIST(ret)) {
+ ANYOF_FLAGS(ret) |= ANYOF_LOC_FOLD;
+ }
+ else if (cp_list) { /* Look to see if there a 0-255 code point is in
+ the list */
+ UV start, end;
+ invlist_iterinit(cp_list);
+ if (invlist_iternext(cp_list, &start, &end) && start < 256) {
+ ANYOF_FLAGS(ret) |= ANYOF_LOC_FOLD;
+ }
+ invlist_iterfinish(cp_list);
+ }
}
/* Optimize inverted simple patterns (e.g. [^a-z]) when everything is known
if (cp_list
&& ! invert
&& ! depends_list
- && ! (ANYOF_FLAGS(ret) & ANYOF_POSIXL)
+ && ! (ANYOF_FLAGS(ret) & (ANYOF_LOC_FOLD|ANYOF_POSIXL))
&& ! HAS_NONLOCALE_RUNTIME_PROPERTY_DEFINITION
/* We don't optimize if we are supposed to make sure all non-Unicode
Absorbs an /x style # comments from the input stream.
Returns true if there is more text remaining in the stream.
- Will set the REG_SEEN_RUN_ON_COMMENT flag if the comment
+ Will set the REG_RUN_ON_COMMENT_SEEN flag if the comment
terminates the pattern without including a newline.
Note its the callers responsibility to ensure that we are
if (!ended) {
/* we ran off the end of the pattern without ending
the comment, so we have to add an \n when wrapping */
- RExC_seen |= REG_SEEN_RUN_ON_COMMENT;
+ RExC_seen |= REG_RUN_ON_COMMENT_SEEN;
return 0;
} else
return 1;
regnode * const temp = regnext(scan);
#ifdef EXPERIMENTAL_INPLACESCAN
if (PL_regkind[OP(scan)] == EXACT) {
- bool has_exactf_sharp_s; /* Unexamined in this routine */
+ bool unfolded_multi_char; /* Unexamined in this routine */
if (join_exact(pRExC_state, scan, &min,
- &has_exactf_sharp_s, 1, val, depth+1))
+ &unfolded_multi_char, 1, val, depth+1))
return EXACT;
}
#endif
(r->check_substr == r->float_substr
&& r->check_utf8 == r->float_utf8
? "(checking floating" : "(checking anchored"));
- if (r->extflags & RXf_NOSCAN)
+ if (r->intflags & PREGf_NOSCAN)
PerlIO_printf(Perl_debug_log, " noscan");
if (r->extflags & RXf_CHECK_ALL)
PerlIO_printf(Perl_debug_log, " isall");
regprop(r, sv, ri->regstclass);
PerlIO_printf(Perl_debug_log, "stclass %s ", SvPVX_const(sv));
}
- if (r->extflags & RXf_ANCH) {
+ if (r->intflags & PREGf_ANCH) {
PerlIO_printf(Perl_debug_log, "anchored");
- if (r->extflags & RXf_ANCH_BOL)
+ if (r->intflags & PREGf_ANCH_BOL)
PerlIO_printf(Perl_debug_log, "(BOL)");
- if (r->extflags & RXf_ANCH_MBOL)
+ if (r->intflags & PREGf_ANCH_MBOL)
PerlIO_printf(Perl_debug_log, "(MBOL)");
- if (r->extflags & RXf_ANCH_SBOL)
+ if (r->intflags & PREGf_ANCH_SBOL)
PerlIO_printf(Perl_debug_log, "(SBOL)");
- if (r->extflags & RXf_ANCH_GPOS)
+ if (r->intflags & PREGf_ANCH_GPOS)
PerlIO_printf(Perl_debug_log, "(GPOS)");
PerlIO_putc(Perl_debug_log, ' ');
}
- if (r->extflags & RXf_GPOS_SEEN)
+ if (r->intflags & PREGf_GPOS_SEEN)
PerlIO_printf(Perl_debug_log, "GPOS:%"UVuf" ", (UV)r->gofs);
if (r->intflags & PREGf_SKIP)
PerlIO_printf(Perl_debug_log, "plus ");
}
}
- if (flags & (ANYOF_ABOVE_LATIN1_ALL|ANYOF_ABOVE_LATIN1_ALL)
- || ANYOF_NONBITMAP(o))
+ if ((flags & ANYOF_ABOVE_LATIN1_ALL)
+ || ANYOF_UTF8_LOCALE_INVLIST(o) || ANYOF_NONBITMAP(o))
{
if (do_sep) {
Perl_sv_catpvf(aTHX_ sv,"%s][%s",PL_colors[1],PL_colors[0]);
SvREFCNT_dec_NN(lv);
}
}
+
+ /* Output any UTF-8 locale code points */
+ if (flags & ANYOF_LOC_FOLD && ANYOF_UTF8_LOCALE_INVLIST(o)) {
+ UV start, end;
+ int max_entries = 256;
+
+ sv_catpvs(sv, "{utf8 locale}");
+ invlist_iterinit(ANYOF_UTF8_LOCALE_INVLIST(o));
+ while (invlist_iternext(ANYOF_UTF8_LOCALE_INVLIST(o),
+ &start, &end)) {
+ put_range(sv, start, end);
+ max_entries --;
+ if (max_entries < 0) {
+ sv_catpvs(sv, "...");
+ break;
+ }
+ }
+ invlist_iterfinish(ANYOF_UTF8_LOCALE_INVLIST(o));
+ }
}
Perl_sv_catpvf(aTHX_ sv, "%s]", PL_colors[1]);
}
else if (PL_regkind[(U8)op] == ANYOF) {
/* arglen 1 + class block */
- node += 1 + ((ANYOF_FLAGS(node) & ANYOF_POSIXL)
- ? ANYOF_POSIXL_SKIP : ANYOF_SKIP);
+ node += 1 + ((ANYOF_FLAGS(node) & ANYOF_LOC_FOLD)
+ ? ANYOF_POSIXL_FOLD_SKIP
+ : (ANYOF_FLAGS(node) & ANYOF_POSIXL)
+ ? ANYOF_POSIXL_SKIP
+ : ANYOF_SKIP);
node = NEXTOPER(node);
}
else if (PL_regkind[(U8)op] == EXACT) {