*/
/* The names of the functions have been changed from regcomp and
- * regexec to pregcomp and pregexec in order to avoid conflicts
+ * regexec to pregcomp and pregexec in order to avoid conflicts
* with the POSIX routines of the same names.
*/
various inplace (keyhole style) optimisations. In addition study_chunk
and scan_commit populate this data structure with information about
what strings MUST appear in the pattern. We look for the longest
- string that must appear for at a fixed location, and we look for the
+ string that must appear at a fixed location, and we look for the
longest string that may appear at a floating location. So for instance
in the pattern:
- offset or min_offset
This is the position the string must appear at, or not before.
It also implicitly (when combined with minlenp) tells us how many
- character must match before the string we are searching.
- Likewise when combined with minlenp and the length of the string
+ characters must match before the string we are searching for.
+ Likewise when combined with minlenp and the length of the string it
tells us how many characters must appear after the string we have
found.
- max_offset
Only used for floating strings. This is the rightmost point that
- the string can appear at. Ifset to I32 max it indicates that the
+ the string can appear at. If set to I32 max it indicates that the
string can occur infinitely far to the right.
- minlenp
minimum length for the F is 1. This is important as the minimum length
is used to determine offsets in front of and behind the string being
looked for. Since strings can be composites this is the length of the
- pattern at the time it was commited with a scan_commit. Note that
+ pattern at the time it was committed with a scan_commit. Note that
the length is calculated by study_chunk, so that the minimum lengths
are not known until the full pattern has been compiled, thus the
pointer to the value.
SV **longest; /* Either &l_fixed, or &l_float. */
SV *longest_fixed; /* longest fixed string found in pattern */
I32 offset_fixed; /* offset where it starts */
- I32 *minlen_fixed; /* pointer to the minlen relevent to the string */
+ I32 *minlen_fixed; /* pointer to the minlen relevant to the string */
I32 lookbehind_fixed; /* is the position of the string modfied by LB */
SV *longest_float; /* longest floating string found in pattern */
I32 offset_float_min; /* earliest point in string it can appear */
I32 offset_float_max; /* latest point in string it can appear */
- I32 *minlen_float; /* pointer to the minlen relevent to the string */
+ I32 *minlen_float; /* pointer to the minlen relevant to the string */
I32 lookbehind_float; /* is the position of the string modified by LB */
I32 flags;
I32 whilem_c;
#define SCF_SEEN_ACCEPT 0x8000
#define UTF cBOOL(RExC_utf8)
-#define LOC cBOOL(RExC_flags & RXf_PMf_LOCALE)
-#define UNI_SEMANTICS cBOOL(RExC_flags & RXf_PMf_UNICODE)
+#define LOC (get_regex_charset(RExC_flags) == REGEX_LOCALE_CHARSET)
+#define UNI_SEMANTICS (get_regex_charset(RExC_flags) == REGEX_UNICODE_CHARSET)
+
#define FOLD cBOOL(RExC_flags & RXf_PMf_FOLD)
#define OOB_UNICODE 12345678
ANYOF_CLASS_ZERO(cl);
ANYOF_BITMAP_SETALL(cl);
- cl->flags = ANYOF_EOS|ANYOF_UNICODE_ALL;
+ cl->flags = ANYOF_EOS|ANYOF_UNICODE_ALL|ANYOF_LOC_NONBITMAP_FOLD|ANYOF_NON_UTF8_LATIN1_ALL;
if (LOC)
cl->flags |= ANYOF_LOCALE;
}
PERL_ARGS_ASSERT_CL_AND;
assert(and_with->type == ANYOF);
- if (!(and_with->flags & ANYOF_CLASS)
- && !(cl->flags & ANYOF_CLASS)
+
+ if (!(ANYOF_CLASS_TEST_ANY_SET(and_with))
+ && !(ANYOF_CLASS_TEST_ANY_SET(cl))
&& (and_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
- && !(and_with->flags & ANYOF_FOLD)
- && !(cl->flags & ANYOF_FOLD)) {
+ && !(and_with->flags & ANYOF_LOC_NONBITMAP_FOLD)
+ && !(cl->flags & ANYOF_LOC_NONBITMAP_FOLD)) {
int i;
if (and_with->flags & ANYOF_INVERT)
if (!(and_with->flags & ANYOF_EOS))
cl->flags &= ~ANYOF_EOS;
- if (cl->flags & ANYOF_UNICODE_ALL && and_with->flags & ANYOF_UNICODE &&
+ if (!(and_with->flags & ANYOF_LOC_NONBITMAP_FOLD))
+ cl->flags &= ~ANYOF_LOC_NONBITMAP_FOLD;
+ if (!(and_with->flags & ANYOF_NON_UTF8_LATIN1_ALL))
+ cl->flags &= ~ANYOF_NON_UTF8_LATIN1_ALL;
+
+ if (cl->flags & ANYOF_UNICODE_ALL && and_with->flags & ANYOF_NONBITMAP &&
!(and_with->flags & ANYOF_INVERT)) {
cl->flags &= ~ANYOF_UNICODE_ALL;
- cl->flags |= ANYOF_UNICODE;
+ cl->flags |= and_with->flags & ANYOF_NONBITMAP; /* field is 2 bits; use
+ only the one(s)
+ actually set */
ARG_SET(cl, ARG(and_with));
}
if (!(and_with->flags & ANYOF_UNICODE_ALL) &&
!(and_with->flags & ANYOF_INVERT))
cl->flags &= ~ANYOF_UNICODE_ALL;
- if (!(and_with->flags & (ANYOF_UNICODE|ANYOF_UNICODE_ALL)) &&
+ if (!(and_with->flags & (ANYOF_NONBITMAP|ANYOF_UNICODE_ALL)) &&
!(and_with->flags & ANYOF_INVERT))
- cl->flags &= ~ANYOF_UNICODE;
+ cl->flags &= ~ANYOF_NONBITMAP;
}
/* 'OR' a given class with another one. Can create false positives */
* (OK1(i) | OK1(i')) | (!OK1(i) & !OK1(i'))
*/
if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
- && !(or_with->flags & ANYOF_FOLD)
- && !(cl->flags & ANYOF_FOLD) ) {
+ && !(or_with->flags & ANYOF_LOC_NONBITMAP_FOLD)
+ && !(cl->flags & ANYOF_LOC_NONBITMAP_FOLD) ) {
int i;
for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
} else {
/* (B1 | CL1) | (B2 | CL2) = (B1 | B2) | (CL1 | CL2)) */
if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
- && (!(or_with->flags & ANYOF_FOLD)
- || (cl->flags & ANYOF_FOLD)) ) {
+ && (!(or_with->flags & ANYOF_LOC_NONBITMAP_FOLD)
+ || (cl->flags & ANYOF_LOC_NONBITMAP_FOLD)) ) {
int i;
/* OR char bitmap and class bitmap separately */
for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
cl->bitmap[i] |= or_with->bitmap[i];
- if (or_with->flags & ANYOF_CLASS) {
+ if (ANYOF_CLASS_TEST_ANY_SET(or_with)) {
for (i = 0; i < ANYOF_CLASSBITMAP_SIZE; i++)
cl->classflags[i] |= or_with->classflags[i];
cl->flags |= ANYOF_CLASS;
}
if (or_with->flags & ANYOF_EOS)
cl->flags |= ANYOF_EOS;
+ if (!(or_with->flags & ANYOF_NON_UTF8_LATIN1_ALL))
+ cl->flags |= ANYOF_NON_UTF8_LATIN1_ALL;
- if (cl->flags & ANYOF_UNICODE && or_with->flags & ANYOF_UNICODE &&
+ if (or_with->flags & ANYOF_LOC_NONBITMAP_FOLD)
+ cl->flags |= ANYOF_LOC_NONBITMAP_FOLD;
+
+ /* If both nodes match something outside the bitmap, but what they match
+ * outside is not the same pointer, and hence not easily compared, give up
+ * and allow the start class to match everything outside the bitmap */
+ if (cl->flags & ANYOF_NONBITMAP && or_with->flags & ANYOF_NONBITMAP &&
ARG(cl) != ARG(or_with)) {
cl->flags |= ANYOF_UNICODE_ALL;
- cl->flags &= ~ANYOF_UNICODE;
}
+
if (or_with->flags & ANYOF_UNICODE_ALL) {
cl->flags |= ANYOF_UNICODE_ALL;
- cl->flags &= ~ANYOF_UNICODE;
}
}
tables that are used to generate the final compressed
representation which is what dump_trie expects.
- Part of the reason for their existance is to provide a form
+ Part of the reason for their existence is to provide a form
of documentation as to how the different representations function.
*/
/ (DUPE|DUPE) X? (?{ ... }) Y /x
-Thus EVAL blocks follwing a trie may be called a different number of times with
+Thus EVAL blocks following a trie may be called a different number of times with
and without the optimisation. With the optimisations dupes will be silently
-ignored. This inconsistant behaviour of EVAL type nodes is well established as
+ignored. This inconsistent behaviour of EVAL type nodes is well established as
the following demonstrates:
'words'=~/(word|word|word)(?{ print $1 })[xyz]/
Example of what happens on a structural level:
-The regexp /(ac|ad|ab)+/ will produce the folowing debug output:
+The regexp /(ac|ad|ab)+/ will produce the following debug output:
1: CURLYM[1] {1,32767}(18)
5: BRANCH(8)
regnode *convert = NULL;
U32 *prev_states; /* temp array mapping each state to previous one */
/* we just use folder as a flag in utf8 */
- const U8 * const folder = ( flags == EXACTF
- ? PL_fold
- : ( flags == EXACTFL
- ? PL_fold_locale
- : NULL
- )
- );
+ const U8 * folder = NULL;
#ifdef DEBUGGING
const U32 data_slot = add_data( pRExC_state, 4, "tuuu" );
PERL_UNUSED_ARG(depth);
#endif
+ switch (flags) {
+ case EXACTFU: folder = PL_fold_latin1; break;
+ case EXACTF: folder = PL_fold; break;
+ case EXACTFL: folder = PL_fold_locale; break;
+ }
+
trie = (reg_trie_data *) PerlMemShared_calloc( 1, sizeof(reg_trie_data) );
trie->refcount = 1;
trie->startstate = 1;
*TODO* If we keep track of how many times each character is used we can
remap the columns so that the table compression later on is more
- efficient in terms of memory by ensuring most common value is in the
+ efficient in terms of memory by ensuring the most common value is in the
middle and the least common are on the outside. IMO this would be better
than a most to least common mapping as theres a decent chance the most
common letter will share a node with the least common, meaning the node
- will not be compressable. With a middle is most common approach the worst
+ will not be compressible. With a middle is most common approach the worst
case is when we have the least common nodes twice.
*/
TRIE_STORE_REVCHAR;
}
if ( set_bit ) {
- /* store the codepoint in the bitmap, and if its ascii
- also store its folded equivelent. */
+ /* store the codepoint in the bitmap, and its folded
+ * equivalent. */
TRIE_BITMAP_SET(trie,uvc);
/* store the folded codepoint */
if ( !UTF ) {
/* store first byte of utf8 representation of
- codepoints in the 127 < uvc < 256 range */
- if (127 < uvc && uvc < 192) {
- TRIE_BITMAP_SET(trie,194);
- } else if (191 < uvc ) {
- TRIE_BITMAP_SET(trie,195);
- /* && uvc < 256 -- we know uvc is < 256 already */
+ variant codepoints */
+ if (! UNI_IS_INVARIANT(uvc)) {
+ TRIE_BITMAP_SET(trie, UTF8_TWO_BYTE_HI(uvc));
}
}
set_bit = 0; /* We've done our bit :-) */
We then construct the trie using only the .next slots of the entry
structs.
- We use the .check field of the first entry of the node temporarily to
+ We use the .check field of the first entry of the node temporarily to
make compression both faster and easier by keeping track of how many non
zero fields are in the node.
- Each states[] entry contains a .base field which indicates the
index in the state[] array wheres its transition data is stored.
- - If .base is 0 there are no valid transitions from that node.
+ - If .base is 0 there are no valid transitions from that node.
- If .base is nonzero then charid is added to it to find an entry in
the trans array.
XXX - wrong maybe?
The following process inplace converts the table to the compressed
- table: We first do not compress the root node 1,and mark its all its
+ table: We first do not compress the root node 1,and mark all its
.check pointers as 1 and set its .base pointer as 1 as well. This
- allows to do a DFA construction from the compressed table later, and
- ensures that any .base pointers we calculate later are greater than
- 0.
+ allows us to do a DFA construction from the compressed table later,
+ and ensures that any .base pointers we calculate later are greater
+ than 0.
- We set 'pos' to indicate the first entry of the second node.
PerlMemShared_realloc( trie->trans, trie->lasttrans
* sizeof(reg_trie_trans) );
- { /* Modify the program and insert the new TRIE node*/
+ { /* Modify the program and insert the new TRIE node */
U8 nodetype =(U8)(flags & 0xFF);
char *str=NULL;
depending on whether the thing following (in 'last') is a branch
or not and whther first is the startbranch (ie is it a sub part of
the alternation or is it the whole thing.)
- Assuming its a sub part we conver the EXACT otherwise we convert
+ Assuming its a sub part we convert the EXACT otherwise we convert
the whole branch sequence, including the first.
*/
/* Find the node we are going to overwrite */
* so, point the first word's .prev field at the second word. If the
* second already has a .prev field set, stop now. This will be the
* case either if we've already processed that word's accept state,
- * or that that state had multiple words, and the overspill words
- * were already linked up earlier.
+ * or that state had multiple words, and the overspill words were
+ * already linked up earlier.
*/
{
U16 word;
STATIC void
S_make_trie_failtable(pTHX_ RExC_state_t *pRExC_state, regnode *source, regnode *stclass, U32 depth)
{
-/* The Trie is constructed and compressed now so we can build a fail array now if its needed
+/* The Trie is constructed and compressed now so we can build a fail array if it's needed
This is basically the Aho-Corasick algorithm. Its from exercise 3.31 and 3.32 in the
"Red Dragon" -- Compilers, principles, techniques, and tools. Aho, Sethi, Ullman 1985/88
ISBN 0-201-10088-6
We find the fail state for each state in the trie, this state is the longest proper
- suffix of the current states 'word' that is also a proper prefix of another word in our
- trie. State 1 represents the word '' and is the thus the default fail state. This allows
+ suffix of the current state's 'word' that is also a proper prefix of another word in our
+ trie. State 1 represents the word '' and is thus the default fail state. This allows
the DFA not to have to restart after its tried and failed a word at a given point, it
simply continues as though it had been matching the other word in the first place.
Consider
'abcdgu'=~/abcdefg|cdgu/
When we get to 'd' we are still matching the first word, we would encounter 'g' which would
- fail, which would bring use to the state representing 'd' in the second word where we would
- try 'g' and succeed, prodceding to match 'cdgu'.
+ fail, which would bring us to the state representing 'd' in the second word where we would
+ try 'g' and succeed, proceeding to match 'cdgu'.
*/
/* add a fail transition */
const U32 trie_offset = ARG(source);
}
#endif
}
-
- if (UTF && ( OP(scan) == EXACTF ) && ( STR_LEN(scan) >= 6 ) ) {
+#define GREEK_SMALL_LETTER_IOTA_WITH_DIALYTIKA_AND_TONOS 0x0390
+#define IOTA_D_T GREEK_SMALL_LETTER_IOTA_WITH_DIALYTIKA_AND_TONOS
+#define GREEK_SMALL_LETTER_UPSILON_WITH_DIALYTIKA_AND_TONOS 0x03B0
+#define UPSILON_D_T GREEK_SMALL_LETTER_UPSILON_WITH_DIALYTIKA_AND_TONOS
+
+ if (UTF
+ && ( OP(scan) == EXACTF || OP(scan) == EXACTFU)
+ && ( STR_LEN(scan) >= 6 ) )
+ {
/*
Two problematic code points in Unicode casefolding of EXACT nodes:
return stopnow;
}
-/* REx optimizer. Converts nodes into quickier variants "in place".
+/* REx optimizer. Converts nodes into quicker variants "in place".
Finds fixed substrings. */
/* Stops at toplevel WHILEM as well as at "last". At end *scanp is set
SAVEFREEPV(and_withp)
/* this is a chain of data about sub patterns we are processing that
- need to be handled seperately/specially in study_chunk. Its so
+ need to be handled separately/specially in study_chunk. Its so
we can simulate recursion without losing state. */
struct scan_frame;
typedef struct scan_frame {
which would be constructed from a pattern like /A|LIST|OF|WORDS/
- If we can find such a subseqence we need to turn the first
+ If we can find such a subsequence we need to turn the first
element into a trie and then add the subsequent branch exact
strings to the trie.
We have two cases
- 1. patterns where the whole set of branch can be converted.
+ 1. patterns where the whole set of branches can be converted.
2. patterns where only a subset can be converted.
In case 1 we can replace the whole set with a single regop
for the trie. In case 2 we need to keep the start and end
- branchs so
+ branches so
'BRANCH EXACT; BRANCH EXACT; BRANCH X'
becomes BRANCH TRIE; BRANCH X;
If x(1..n)==tail then we can do a simple trie, if not we make
a "jump" trie, such that when we match the appropriate word
- we "jump" to the appopriate tail node. Essentailly we turn
+ we "jump" to the appropriate tail node. Essentially we turn
a nested if into a case structure of sorts.
*/
and noper_next is the same as scan (our current
position in the regex) then the EXACT branch is
a possible optimization target. Once we have
- two or more consequetive such branches we can
+ two or more consecutive such branches we can
create a trie of the EXACT's contents and stich
it in place. If the sequence represents all of
the branches we eliminate the whole thing and
/* Check whether it is compatible with what we know already! */
int compat = 1;
+
+ /* If compatible, we or it in below. It is compatible if is
+ * in the bitmp and either 1) its bit or its fold is set, or 2)
+ * it's for a locale. Even if there isn't unicode semantics
+ * here, at runtime there may be because of matching against a
+ * utf8 string, so accept a possible false positive for
+ * latin1-range folds */
if (uc >= 0x100 ||
(!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
&& !ANYOF_BITMAP_TEST(data->start_class, uc)
- && (!(data->start_class->flags & ANYOF_FOLD)
- || !ANYOF_BITMAP_TEST(data->start_class, PL_fold[uc])))
+ && (!(data->start_class->flags & ANYOF_LOC_NONBITMAP_FOLD)
+ || !ANYOF_BITMAP_TEST(data->start_class, PL_fold_latin1[uc])))
)
compat = 0;
ANYOF_CLASS_ZERO(data->start_class);
if (flags & SCF_DO_STCLASS_AND) {
/* Check whether it is compatible with what we know already! */
int compat = 1;
-
if (uc >= 0x100 ||
- (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
- && !ANYOF_BITMAP_TEST(data->start_class, uc)
- && !ANYOF_BITMAP_TEST(data->start_class, PL_fold[uc])))
+ (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
+ && !ANYOF_BITMAP_TEST(data->start_class, uc)
+ && !ANYOF_BITMAP_TEST(data->start_class, PL_fold_latin1[uc])))
+ {
compat = 0;
+ }
ANYOF_CLASS_ZERO(data->start_class);
ANYOF_BITMAP_ZERO(data->start_class);
if (compat) {
ANYOF_BITMAP_SET(data->start_class, uc);
data->start_class->flags &= ~ANYOF_EOS;
- data->start_class->flags |= ANYOF_FOLD;
- if (OP(scan) == EXACTFL)
+ data->start_class->flags |= ANYOF_LOC_NONBITMAP_FOLD;
+ if (OP(scan) == EXACTFL) {
data->start_class->flags |= ANYOF_LOCALE;
+ }
+ else {
+
+ /* Also set the other member of the fold pair. In case
+ * that unicode semantics is called for at runtime, use
+ * the full latin1 fold. (Can't do this for locale,
+ * because not known until runtime */
+ ANYOF_BITMAP_SET(data->start_class, PL_fold_latin1[uc]);
+ }
}
}
else if (flags & SCF_DO_STCLASS_OR) {
- if (data->start_class->flags & ANYOF_FOLD) {
+ if (data->start_class->flags & ANYOF_LOC_NONBITMAP_FOLD) {
/* false positive possible if the class is case-folded.
Assume that the locale settings are the same... */
- if (uc < 0x100)
+ if (uc < 0x100) {
ANYOF_BITMAP_SET(data->start_class, uc);
+ if (OP(scan) != EXACTFL) {
+
+ /* And set the other member of the fold pair, but
+ * can't do that in locale because not known until
+ * run-time */
+ ANYOF_BITMAP_SET(data->start_class,
+ PL_fold_latin1[uc]);
+ }
+ }
data->start_class->flags &= ~ANYOF_EOS;
}
cl_and(data->start_class, and_withp);
f |= SCF_DO_STCLASS_AND;
f &= ~SCF_DO_STCLASS_OR;
}
- /* These are the cases when once a subexpression
- fails at a particular position, it cannot succeed
- even after backtracking at the enclosing scope.
-
- XXXX what if minimal match and we are at the
- initial run of {n,m}? */
- if ((mincount != maxcount - 1) && (maxcount != REG_INFTY))
+ /* Exclude from super-linear cache processing any {n,m}
+ regops for which the combination of input pos and regex
+ pos is not enough information to determine if a match
+ will be possible.
+
+ For example, in the regex /foo(bar\s*){4,8}baz/ with the
+ regex pos at the \s*, the prospects for a match depend not
+ only on the input position but also on how many (bar\s*)
+ repeats into the {4,8} we are. */
+ if ((mincount > 1) || (maxcount > 1 && maxcount != REG_INFTY))
f &= ~SCF_WHILEM_VISITED_POS;
/* This will finish on WHILEM, setting scan, or on NULL: */
#ifdef DEBUGGING
OP(nxt1 + 1) = OPTIMIZED; /* was count. */
OP(nxt + 1) = OPTIMIZED; /* was count. */
- NEXT_OFF(nxt1 + 1) = 0; /* just for consistancy. */
- NEXT_OFF(nxt + 1) = 0; /* just for consistancy. */
+ NEXT_OFF(nxt1 + 1) = 0; /* just for consistency. */
+ NEXT_OFF(nxt + 1) = 0; /* just for consistency. */
#endif
#if 0
while ( nxt1 && (OP(nxt1) != WHILEM)) {
else if ((OP(oscan) == CURLYX)
&& (flags & SCF_WHILEM_VISITED_POS)
/* See the comment on a similar expression above.
- However, this time it not a subexpression
+ However, this time it's not a subexpression
we care about, but the expression itself. */
&& (maxcount == REG_INFTY)
&& data && ++data->whilem_c < 16) {
NEXT_OFF(oscan) += NEXT_OFF(next);
}
continue;
- default: /* REF and CLUMP only? */
+ default: /* REF, ANYOFV, and CLUMP only? */
if (flags & SCF_DO_SUBSTR) {
SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */
data->longest = &(data->longest_float);
}
}
else if (OP(scan) == FOLDCHAR) {
- int d = ARG(scan)==0xDF ? 1 : 2;
+ int d = ARG(scan) == LATIN_SMALL_LETTER_SHARP_S ? 1 : 2;
flags &= ~SCF_DO_STCLASS;
min += 1;
delta += d;
goto do_default;
if (flags & SCF_DO_STCLASS_OR) { /* Everything but \n */
value = (ANYOF_BITMAP_TEST(data->start_class,'\n')
- || (data->start_class->flags & ANYOF_CLASS));
+ || ANYOF_CLASS_TEST_ANY_SET(data->start_class));
cl_anything(pRExC_state, data->start_class);
}
if (flags & SCF_DO_STCLASS_AND || !value)
if (flags & SCF_DO_STCLASS_AND) {
if (!(data->start_class->flags & ANYOF_LOCALE)) {
ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
- if (FLAGS(scan) & USE_UNI) {
+ if (FLAGS(scan) == REGEX_UNICODE_CHARSET) {
for (value = 0; value < 256; value++) {
if (!isWORDCHAR_L1(value)) {
ANYOF_BITMAP_CLEAR(data->start_class, value);
else {
if (data->start_class->flags & ANYOF_LOCALE)
ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
- else if (FLAGS(scan) & USE_UNI) {
+ else if (FLAGS(scan) == REGEX_UNICODE_CHARSET) {
for (value = 0; value < 256; value++) {
if (isWORDCHAR_L1(value)) {
ANYOF_BITMAP_SET(data->start_class, value);
}
}
break;
- case ALNUML:
- if (flags & SCF_DO_STCLASS_AND) {
- if (data->start_class->flags & ANYOF_LOCALE)
- ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
- }
- else {
- ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
- data->start_class->flags |= ANYOF_LOCALE;
- }
- break;
case NALNUM:
if (flags & SCF_DO_STCLASS_AND) {
if (!(data->start_class->flags & ANYOF_LOCALE)) {
ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
- if (FLAGS(scan) & USE_UNI) {
+ if (FLAGS(scan) == REGEX_UNICODE_CHARSET) {
for (value = 0; value < 256; value++) {
if (isWORDCHAR_L1(value)) {
ANYOF_BITMAP_CLEAR(data->start_class, value);
if (data->start_class->flags & ANYOF_LOCALE)
ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
else {
- for (value = 0; value < 256; value++)
- if (!isALNUM(value))
- ANYOF_BITMAP_SET(data->start_class, value);
+ if (FLAGS(scan) == REGEX_UNICODE_CHARSET) {
+ for (value = 0; value < 256; value++) {
+ if (! isWORDCHAR_L1(value)) {
+ ANYOF_BITMAP_SET(data->start_class, value);
+ }
+ }
+ } else {
+ for (value = 0; value < 256; value++) {
+ if (! isALNUM(value)) {
+ ANYOF_BITMAP_SET(data->start_class, value);
+ }
+ }
+ }
}
}
break;
- case NALNUML:
- if (flags & SCF_DO_STCLASS_AND) {
- if (data->start_class->flags & ANYOF_LOCALE)
- ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
- }
- else {
- data->start_class->flags |= ANYOF_LOCALE;
- ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
- }
- break;
case SPACE:
if (flags & SCF_DO_STCLASS_AND) {
if (!(data->start_class->flags & ANYOF_LOCALE)) {
ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
- if (FLAGS(scan) & USE_UNI) {
+ if (FLAGS(scan) == REGEX_UNICODE_CHARSET) {
for (value = 0; value < 256; value++) {
if (!isSPACE_L1(value)) {
ANYOF_BITMAP_CLEAR(data->start_class, value);
if (data->start_class->flags & ANYOF_LOCALE) {
ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
}
- else if (FLAGS(scan) & USE_UNI) {
+ else if (FLAGS(scan) == REGEX_UNICODE_CHARSET) {
for (value = 0; value < 256; value++) {
if (isSPACE_L1(value)) {
ANYOF_BITMAP_SET(data->start_class, value);
}
}
break;
- case SPACEL:
- if (flags & SCF_DO_STCLASS_AND) {
- if (data->start_class->flags & ANYOF_LOCALE)
- ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
- }
- else {
- data->start_class->flags |= ANYOF_LOCALE;
- ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
- }
- break;
case NSPACE:
if (flags & SCF_DO_STCLASS_AND) {
if (!(data->start_class->flags & ANYOF_LOCALE)) {
ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
- if (FLAGS(scan) & USE_UNI) {
+ if (FLAGS(scan) == REGEX_UNICODE_CHARSET) {
for (value = 0; value < 256; value++) {
if (isSPACE_L1(value)) {
ANYOF_BITMAP_CLEAR(data->start_class, value);
else {
if (data->start_class->flags & ANYOF_LOCALE)
ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
- else if (FLAGS(scan) & USE_UNI) {
+ else if (FLAGS(scan) == REGEX_UNICODE_CHARSET) {
for (value = 0; value < 256; value++) {
if (!isSPACE_L1(value)) {
ANYOF_BITMAP_SET(data->start_class, value);
}
}
break;
- case NSPACEL:
- if (flags & SCF_DO_STCLASS_AND) {
- if (data->start_class->flags & ANYOF_LOCALE) {
- ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
- for (value = 0; value < 256; value++)
- if (!isSPACE(value))
- ANYOF_BITMAP_CLEAR(data->start_class, value);
- }
- }
- else {
- data->start_class->flags |= ANYOF_LOCALE;
- ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
- }
- break;
case DIGIT:
if (flags & SCF_DO_STCLASS_AND) {
ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NDIGIT);
int f = 0;
/* We use SAVEFREEPV so that when the full compile
is finished perl will clean up the allocated
- minlens when its all done. This was we don't
+ minlens when it's all done. This way we don't
have to worry about freeing them when we know
they wont be used, which would be a pain.
*/
#endif
REGEXP *
-Perl_re_compile(pTHX_ SV * const pattern, U32 pm_flags)
+Perl_re_compile(pTHX_ SV * const pattern, U32 orig_pm_flags)
{
dVAR;
REGEXP *rx;
regnode *scan;
I32 flags;
I32 minlen = 0;
+ U32 pm_flags;
/* these are all flags - maybe they should be turned
* into a single int with different bit masks */
I32 sawlookahead = 0;
I32 sawplus = 0;
I32 sawopen = 0;
+ bool used_setjump = FALSE;
U8 jump_ret = 0;
dJMPENV;
RExC_utf8 = RExC_orig_utf8 = SvUTF8(pattern);
-
+ /****************** LONG JUMP TARGET HERE***********************/
/* Longjmp back to here if have to switch in midstream to utf8 */
if (! RExC_orig_utf8) {
JMPENV_PUSH(jump_ret);
+ used_setjump = TRUE;
}
if (jump_ret == 0) { /* First time through */
- exp = SvPV(pattern, plen);
- xend = exp + plen;
+ exp = SvPV(pattern, plen);
+ xend = exp + plen;
+ /* ignore the utf8ness if the pattern is 0 length */
+ if (plen == 0) {
+ RExC_utf8 = RExC_orig_utf8 = 0;
+ }
DEBUG_COMPILE_r({
SV *dsv= sv_newmortal();
restudied = 0;
#endif
+ /* Set to use unicode semantics if the pattern is in utf8 and has the
+ * 'depends' charset specified, as it means unicode when utf8 */
+ pm_flags = orig_pm_flags;
+
+ if (RExC_utf8 && get_regex_charset(pm_flags) == REGEX_DEPENDS_CHARSET) {
+ set_regex_charset(&pm_flags, REGEX_UNICODE_CHARSET);
+ }
+
RExC_precomp = exp;
RExC_flags = pm_flags;
RExC_sawback = 0;
return(NULL);
}
- /* Here, finished first pass. Get rid of our setjmp, which we added for
- * efficiency only if the passed-in string wasn't in utf8, as shown by
- * RExC_orig_utf8. But if the first pass was redone, that variable will be
- * 1 here even though the original string wasn't utf8, but in this case
- * there will have been a long jump */
- if (jump_ret == UTF8_LONGJMP || ! RExC_orig_utf8) {
+ /* Here, finished first pass. Get rid of any added setjmp */
+ if (used_setjump) {
JMPENV_POP;
}
DEBUG_PARSE_r({
r->extflags = pm_flags;
{
bool has_p = ((r->extflags & RXf_PMf_KEEPCOPY) == RXf_PMf_KEEPCOPY);
- bool has_charset = cBOOL(r->extflags & (RXf_PMf_LOCALE|RXf_PMf_UNICODE));
+ bool has_charset = (get_regex_charset(r->extflags) != REGEX_DEPENDS_CHARSET);
/* The caret is output if there are any defaults: if not all the STD
* flags are set, or if no character set specifier is needed */
* covered by the caret */
const STRLEN wraplen = plen + has_p + has_runon
+ has_default /* If needs a caret */
- + has_charset /* If needs a character set specifier */
+
+ /* If needs a character set specifier */
+ + ((has_charset) ? MAX_CHARSET_NAME_LENGTH : 0)
+ (sizeof(STD_PAT_MODS) - 1)
+ (sizeof("(?:)") - 1);
*p++= DEFAULT_PAT_MOD;
}
if (has_charset) {
- if (r->extflags & RXf_PMf_LOCALE) {
- *p++ = LOCALE_PAT_MOD;
- } else {
- *p++ = UNICODE_PAT_MOD;
- }
+ STRLEN len;
+ const char* const name = get_regex_charset_name(r->extflags, &len);
+ Copy(name, p, len, char);
+ p += len;
}
if (has_p)
*p++ = KEEPCOPY_PAT_MOD; /*'p'*/
if (PL_regkind[OP(first)] == EXACT) {
if (OP(first) == EXACT)
NOOP; /* Empty, get anchored substr later. */
- else if ((OP(first) == EXACTF || OP(first) == EXACTFL))
+ else
ri->regstclass = first;
}
#ifdef TRIE_STCLASS
if (ri->regstclass
&& (OP(ri->regstclass) == REG_ANY || OP(ri->regstclass) == SANY))
ri->regstclass = NULL;
+
+ /* If the synthetic start class were to ever be used when EOS is set,
+ * that bit would have to be cleared, as it is shared with another */
if ((!(r->anchored_substr || r->anchored_utf8) || r->anchored_offset)
&& stclass_flag
&& !(data.start_class->flags & ANYOF_EOS)
r->check_substr = r->check_utf8 = r->anchored_substr = r->anchored_utf8
= r->float_substr = r->float_utf8 = NULL;
+
+ /* If the synthetic start class were to ever be used when EOS is set,
+ * that bit would have to be cleared, as it is shared with another */
if (!(data.start_class->flags & ANYOF_EOS)
&& !cl_is_anything(data.start_class))
{
SvREFCNT_inc_simple_void(sv_dat);
}
RExC_sawback = 1;
- ret = reganode(pRExC_state,
- (U8)(FOLD ? (LOC ? NREFFL : NREFF) : NREF),
- num);
+ ret = reganode(pRExC_state,
+ ((! FOLD)
+ ? NREF
+ : (UNI_SEMANTICS)
+ ? NREFFU
+ : (LOC)
+ ? NREFFL
+ : NREFF),
+ num);
*flagp |= HASWIDTH;
Set_Node_Offset(ret, parse_start+1);
if (SIZE_ONLY) {
HE *he_str;
SV *sv_dat = NULL;
- if (!svname) /* shouldnt happen */
+ if (!svname) /* shouldn't happen */
Perl_croak(aTHX_
"panic: reg_scan_name returned NULL");
if (!RExC_paren_names) {
/*
Diagram of capture buffer numbering.
Top line is the normal capture buffer numbers
- Botton line is the negative indexing as from
+ Bottom line is the negative indexing as from
the X (the (?-2))
+ 1 2 3 4 5 X 6 7
ENTER;
Perl_save_re_context(aTHX);
- rop = sv_compile_2op(sv, &sop, "re", &pad);
+ rop = Perl_sv_compile_2op_is_broken(aTHX_ sv, &sop, "re", &pad);
sop->op_private |= OPpREFCOUNTED;
/* re_dup will OpREFCNT_inc */
OpREFCNT_set(sop, 1);
that follow */
has_use_defaults = TRUE;
STD_PMMOD_FLAGS_CLEAR(&RExC_flags);
- RExC_flags &= ~(RXf_PMf_LOCALE|RXf_PMf_UNICODE);
+ if (RExC_utf8) { /* But the default for a utf8 pattern is
+ unicode semantics */
+ set_regex_charset(&RExC_flags, REGEX_UNICODE_CHARSET);
+ }
goto parse_flags;
default:
--RExC_parse;
U32 posflags = 0, negflags = 0;
U32 *flagsp = &posflags;
bool has_charset_modifier = 0;
+ regex_charset cs = REGEX_DEPENDS_CHARSET;
while (*RExC_parse) {
/* && strchr("iogcmsx", *RExC_parse) */
if (has_charset_modifier || flagsp == &negflags) {
goto fail_modifiers;
}
- posflags |= RXf_PMf_LOCALE;
- negflags |= RXf_PMf_UNICODE;
+ cs = REGEX_LOCALE_CHARSET;
has_charset_modifier = 1;
break;
case UNICODE_PAT_MOD:
if (has_charset_modifier || flagsp == &negflags) {
goto fail_modifiers;
}
- posflags |= RXf_PMf_UNICODE;
- negflags |= RXf_PMf_LOCALE;
+ cs = REGEX_UNICODE_CHARSET;
has_charset_modifier = 1;
break;
- case DUAL_PAT_MOD:
+ case DEPENDS_PAT_MOD:
if (has_use_defaults
|| has_charset_modifier
|| flagsp == &negflags)
{
goto fail_modifiers;
}
- negflags |= (RXf_PMf_LOCALE|RXf_PMf_UNICODE);
+
+ /* The dual charset means unicode semantics if the
+ * pattern (or target, not known until runtime) are
+ * utf8 */
+ cs = (RExC_utf8)
+ ? REGEX_UNICODE_CHARSET
+ : REGEX_DEPENDS_CHARSET;
has_charset_modifier = 1;
break;
case ONCE_PAT_MOD: /* 'o' */
}
break;
case '-':
- /* A flag is a default iff it is following a minus, so
+ /* A flag is a default iff it is following a minus, so
* if there is a minus, it means will be trying to
* re-specify a default which is an error */
if (has_use_defaults || flagsp == &negflags) {
case ')':
RExC_flags |= posflags;
RExC_flags &= ~negflags;
+ set_regex_charset(&RExC_flags, cs);
if (paren != ':') {
oregflags |= posflags;
oregflags &= ~negflags;
+ set_regex_charset(&oregflags, cs);
}
nextchar(pRExC_state);
if (paren != ':') {
char *endchar; /* Points to '.' or '}' ending cur char in the input
stream */
- ret = reg_node(pRExC_state,
- (U8)(FOLD ? (LOC ? EXACTFL : EXACTF) : EXACT));
+ ret = reg_node(pRExC_state, (U8) ((! FOLD) ? EXACT
+ : (LOC)
+ ? EXACTFL
+ : UNI_SEMANTICS
+ ? EXACTFU
+ : EXACTF));
s= STRING(ret);
/* Exact nodes can hold only a U8 length's of text = 255. Loop through
Note: we have to be careful with escapes, as they can be both literal
and special, and in the case of \10 and friends can either, depending
- on context. Specifically there are two seperate switches for handling
+ on context. Specifically there are two separate switches for handling
escape sequences, with the one for handling literal escapes requiring
a dummy entry for all of the special escapes that are actually handled
by the other.
RExC_parse++;
vFAIL("Quantifier follows nothing");
break;
- case 0xDF:
- case 0xC3:
- case 0xCE:
+ case LATIN_SMALL_LETTER_SHARP_S:
+ case UTF8_TWO_BYTE_HI_nocast(LATIN_SMALL_LETTER_SHARP_S):
+ case UTF8_TWO_BYTE_HI_nocast(IOTA_D_T):
+#if UTF8_TWO_BYTE_HI_nocast(UPSILON_D_T) != UTF8_TWO_BYTE_HI_nocast(IOTA_D_T)
+#error The beginning utf8 byte of IOTA_D_T and UPSILON_D_T unexpectedly differ. Other instances in this code should have the case statement below.
+ case UTF8_TWO_BYTE_HI_nocast(UPSILON_D_T):
+#endif
do_foldchar:
if (!LOC && FOLD) {
U32 len,cp;
literal text handling code.
*/
switch ((U8)*++RExC_parse) {
- case 0xDF:
- case 0xC3:
- case 0xCE:
+ case LATIN_SMALL_LETTER_SHARP_S:
+ case UTF8_TWO_BYTE_HI_nocast(LATIN_SMALL_LETTER_SHARP_S):
+ case UTF8_TWO_BYTE_HI_nocast(IOTA_D_T):
goto do_foldchar;
/* Special Escapes */
case 'A':
ret = reg_node(pRExC_state, (U8)(ALNUML));
} else {
ret = reg_node(pRExC_state, (U8)(ALNUM));
- FLAGS(ret) = (UNI_SEMANTICS) ? USE_UNI : 0;
}
+ FLAGS(ret) = get_regex_charset(RExC_flags);
*flagp |= HASWIDTH|SIMPLE;
goto finish_meta_pat;
case 'W':
ret = reg_node(pRExC_state, (U8)(NALNUML));
} else {
ret = reg_node(pRExC_state, (U8)(NALNUM));
- FLAGS(ret) = (UNI_SEMANTICS) ? USE_UNI : 0;
}
+ FLAGS(ret) = get_regex_charset(RExC_flags);
*flagp |= HASWIDTH|SIMPLE;
goto finish_meta_pat;
case 'b':
ret = reg_node(pRExC_state, (U8)(BOUNDL));
} else {
ret = reg_node(pRExC_state, (U8)(BOUND));
- FLAGS(ret) = (UNI_SEMANTICS) ? USE_UNI : 0;
}
+ FLAGS(ret) = get_regex_charset(RExC_flags);
*flagp |= SIMPLE;
goto finish_meta_pat;
case 'B':
ret = reg_node(pRExC_state, (U8)(NBOUNDL));
} else {
ret = reg_node(pRExC_state, (U8)(NBOUND));
- FLAGS(ret) = (UNI_SEMANTICS) ? USE_UNI : 0;
}
+ FLAGS(ret) = get_regex_charset(RExC_flags);
*flagp |= SIMPLE;
goto finish_meta_pat;
case 's':
ret = reg_node(pRExC_state, (U8)(SPACEL));
} else {
ret = reg_node(pRExC_state, (U8)(SPACE));
- FLAGS(ret) = (UNI_SEMANTICS) ? USE_UNI : 0;
}
+ FLAGS(ret) = get_regex_charset(RExC_flags);
*flagp |= HASWIDTH|SIMPLE;
goto finish_meta_pat;
case 'S':
ret = reg_node(pRExC_state, (U8)(NSPACEL));
} else {
ret = reg_node(pRExC_state, (U8)(NSPACE));
- FLAGS(ret) = (UNI_SEMANTICS) ? USE_UNI : 0;
}
+ FLAGS(ret) = get_regex_charset(RExC_flags);
*flagp |= HASWIDTH|SIMPLE;
goto finish_meta_pat;
case 'd':
- ret = reg_node(pRExC_state, DIGIT);
+ if (LOC) {
+ ret = reg_node(pRExC_state, (U8)(DIGITL));
+ } else {
+ ret = reg_node(pRExC_state, (U8)(DIGIT));
+ }
*flagp |= HASWIDTH|SIMPLE;
goto finish_meta_pat;
case 'D':
- ret = reg_node(pRExC_state, NDIGIT);
+ if (LOC) {
+ ret = reg_node(pRExC_state, (U8)(NDIGITL));
+ } else {
+ ret = reg_node(pRExC_state, (U8)(NDIGIT));
+ }
*flagp |= HASWIDTH|SIMPLE;
goto finish_meta_pat;
case 'R':
RExC_sawback = 1;
ret = reganode(pRExC_state,
- (U8)(FOLD ? (LOC ? NREFFL : NREFF) : NREF),
- num);
+ ((! FOLD)
+ ? NREF
+ : (UNI_SEMANTICS)
+ ? NREFFU
+ : (LOC)
+ ? NREFFL
+ : NREFF),
+ num);
*flagp |= HASWIDTH;
/* override incorrect value set in reganode MJD */
}
RExC_sawback = 1;
ret = reganode(pRExC_state,
- (U8)(FOLD ? (LOC ? REFFL : REFF) : REF),
- num);
+ ((! FOLD)
+ ? REF
+ : (UNI_SEMANTICS)
+ ? REFFU
+ : (LOC)
+ ? REFFL
+ : REFF),
+ num);
*flagp |= HASWIDTH;
/* override incorrect value set in reganode MJD */
defchar:
ender = 0;
ret = reg_node(pRExC_state,
- (U8)(FOLD ? (LOC ? EXACTFL : EXACTF) : EXACT));
+ (U8) ((! FOLD) ? EXACT
+ : (LOC)
+ ? EXACTFL
+ : (UNI_SEMANTICS)
+ ? EXACTFU
+ : EXACTF)
+ );
s = STRING(ret);
for (len = 0, p = RExC_parse - 1;
len < 127 && p < RExC_end;
if (RExC_flags & RXf_PMf_EXTENDED)
p = regwhite( pRExC_state, p );
switch ((U8)*p) {
- case 0xDF:
- case 0xC3:
- case 0xCE:
+ case LATIN_SMALL_LETTER_SHARP_S:
+ case UTF8_TWO_BYTE_HI_nocast(LATIN_SMALL_LETTER_SHARP_S):
+ case UTF8_TWO_BYTE_HI_nocast(IOTA_D_T):
if (LOC || !FOLD || !is_TRICKYFOLD_safe(p,RExC_end,UTF))
goto normal_default;
case '^':
switch ((U8)*++p) {
/* These are all the special escapes. */
- case 0xDF:
- case 0xC3:
- case 0xCE:
+ case LATIN_SMALL_LETTER_SHARP_S:
+ case UTF8_TWO_BYTE_HI_nocast(LATIN_SMALL_LETTER_SHARP_S):
+ case UTF8_TWO_BYTE_HI_nocast(IOTA_D_T):
if (LOC || !FOLD || !is_TRICKYFOLD_safe(p,RExC_end,UTF))
goto normal_default;
case 'A': /* Start assertion */
}
}
-/* No locale test */
-#define _C_C_T_NOLOC_(NAME,TEST,WORD) \
-ANYOF_##NAME: \
- for (value = 0; value < 256; value++) \
- if (TEST) \
- ANYOF_BITMAP_SET(ret, value); \
- yesno = '+'; \
- what = WORD; \
- break; \
-case ANYOF_N##NAME: \
- for (value = 0; value < 256; value++) \
- if (!TEST) \
- ANYOF_BITMAP_SET(ret, value); \
- yesno = '!'; \
- what = WORD; \
+/* No locale test, and always Unicode semantics */
+#define _C_C_T_NOLOC_(NAME,TEST,WORD) \
+ANYOF_##NAME: \
+ for (value = 0; value < 256; value++) \
+ if (TEST) \
+ stored += S_set_regclass_bit(aTHX_ pRExC_state, ret, (U8) value); \
+ yesno = '+'; \
+ what = WORD; \
+ break; \
+case ANYOF_N##NAME: \
+ for (value = 0; value < 256; value++) \
+ if (!TEST) \
+ stored += S_set_regclass_bit(aTHX_ pRExC_state, ret, (U8) value); \
+ yesno = '!'; \
+ what = WORD; \
break
/* Like the above, but there are differences if we are in uni-8-bit or not, so
* there are two tests passed in, to use depending on that. There aren't any
* cases where the label is different from the name, so no need for that
* parameter */
-#define _C_C_T_(NAME,TEST_8,TEST_7,WORD) \
-ANYOF_##NAME: \
- if (LOC) ANYOF_CLASS_SET(ret, ANYOF_##NAME); \
- else if (UNI_SEMANTICS) { \
- for (value = 0; value < 256; value++) { \
- if (TEST_8) ANYOF_BITMAP_SET(ret, value); \
- } \
- } \
- else { \
- for (value = 0; value < 256; value++) { \
- if (TEST_7) ANYOF_BITMAP_SET(ret, value); \
- } \
- } \
- yesno = '+'; \
- what = WORD; \
- break; \
-case ANYOF_N##NAME: \
- if (LOC) ANYOF_CLASS_SET(ret, ANYOF_N##NAME); \
- else if (UNI_SEMANTICS) { \
- for (value = 0; value < 256; value++) { \
- if (! TEST_8) ANYOF_BITMAP_SET(ret, value); \
- } \
- } \
- else { \
- for (value = 0; value < 256; value++) { \
- if (! TEST_7) ANYOF_BITMAP_SET(ret, value); \
- } \
- } \
- yesno = '!'; \
- what = WORD; \
+#define _C_C_T_(NAME,TEST_8,TEST_7,WORD) \
+ANYOF_##NAME: \
+ if (LOC) ANYOF_CLASS_SET(ret, ANYOF_##NAME); \
+ else if (UNI_SEMANTICS) { \
+ for (value = 0; value < 256; value++) { \
+ if (TEST_8) stored += \
+ S_set_regclass_bit(aTHX_ pRExC_state, ret, (U8) value); \
+ } \
+ } \
+ else { \
+ for (value = 0; value < 128; value++) { \
+ if (TEST_7) stored += \
+ S_set_regclass_bit(aTHX_ pRExC_state, ret, \
+ (U8) UNI_TO_NATIVE(value)); \
+ } \
+ } \
+ yesno = '+'; \
+ what = WORD; \
+ break; \
+case ANYOF_N##NAME: \
+ if (LOC) ANYOF_CLASS_SET(ret, ANYOF_N##NAME); \
+ else if (UNI_SEMANTICS) { \
+ for (value = 0; value < 256; value++) { \
+ if (! TEST_8) stored += \
+ S_set_regclass_bit(aTHX_ pRExC_state, ret, (U8) value); \
+ } \
+ } \
+ else { \
+ for (value = 0; value < 128; value++) { \
+ if (! TEST_7) stored += \
+ S_set_regclass_bit(aTHX_ pRExC_state, ret, (U8) value); \
+ } \
+ /* For a non-ut8 target string with DEPENDS semantics, all above ASCII \
+ * Latin1 code points match the complement of any of the classes. But \
+ * in utf8, they have their Unicode semantics, so can't just set them \
+ * in the bitmap, or else regexec.c will think they matched when they \
+ * shouldn't. */ \
+ ANYOF_FLAGS(ret) |= ANYOF_NON_UTF8_LATIN1_ALL|ANYOF_UTF8; \
+ } \
+ yesno = '!'; \
+ what = WORD; \
break
/*
#define POSIX_CC_UNI_NAME(CCNAME) "Posix" CCNAME
#endif
+STATIC U8
+S_set_regclass_bit_fold(pTHX_ RExC_state_t *pRExC_state, regnode* node, const U8 value)
+{
+
+ /* Handle the setting of folds in the bitmap for non-locale ANYOF nodes.
+ * Locale folding is done at run-time, so this function should not be
+ * called for nodes that are for locales.
+ *
+ * This function simply sets the bit corresponding to the fold of the input
+ * 'value', if not already set. The fold of 'f' is 'F', and the fold of
+ * 'F' is 'f'.
+ *
+ * It also sets any necessary flags, and returns the number of bits that
+ * actually changed from 0 to 1 */
+
+ U8 stored = 0;
+ U8 fold;
+
+ fold = (UNI_SEMANTICS) ? PL_fold_latin1[value]
+ : PL_fold[value];
+
+ /* It assumes the bit for 'value' has already been set */
+ if (fold != value && ! ANYOF_BITMAP_TEST(node, fold)) {
+ ANYOF_BITMAP_SET(node, fold);
+ stored++;
+ }
+ if (_HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(value)
+ || (! UNI_SEMANTICS
+ && ! isASCII(value)
+ && PL_fold_latin1[value] != value))
+ { /* A character that has a fold outside of Latin1 matches outside the
+ bitmap, but only when the target string is utf8. Similarly when we
+ don't have unicode semantics for the above ASCII Latin-1 characters,
+ and they have a fold, they should match if the target is utf8, and
+ not otherwise */
+ ANYOF_FLAGS(node) |= ANYOF_UTF8;
+ }
+
+ return stored;
+}
+
+
+PERL_STATIC_INLINE U8
+S_set_regclass_bit(pTHX_ RExC_state_t *pRExC_state, regnode* node, const U8 value)
+{
+ /* This inline function sets a bit in the bitmap if not already set, and if
+ * appropriate, its fold, returning the number of bits that actually
+ * changed from 0 to 1 */
+
+ U8 stored;
+
+ if (ANYOF_BITMAP_TEST(node, value)) { /* Already set */
+ return 0;
+ }
+
+ ANYOF_BITMAP_SET(node, value);
+ stored = 1;
+
+ if (FOLD && ! LOC) { /* Locale folds aren't known until runtime */
+ stored += S_set_regclass_bit_fold(aTHX_ pRExC_state, node, value);
+ }
+
+ return stored;
+}
+
/*
parse a class specification and produce either an ANYOF node that
matches the pattern or if the pattern matches a single char only and
bool need_class = 0;
SV *listsv = NULL;
UV n;
- bool optimize_invert = TRUE;
AV* unicode_alternate = NULL;
#ifdef EBCDIC
UV literal_endpoint = 0;
#endif
- UV stored = 0; /* 0, 1, or more than 1 chars stored in the class */
+ UV stored = 0; /* how many chars stored in the bitmap */
regnode * const orig_emit = RExC_emit; /* Save the original RExC_emit in
case we need to change the emitted regop to an EXACT. */
if (SIZE_ONLY) {
RExC_size += ANYOF_SKIP;
+#ifdef ANYOF_ADD_LOC_SKIP
+ if (LOC) {
+ RExC_size += ANYOF_ADD_LOC_SKIP;
+ }
+#endif
listsv = &PL_sv_undef; /* For code scanners: listsv always non-NULL. */
}
else {
RExC_emit += ANYOF_SKIP;
- if (FOLD)
- ANYOF_FLAGS(ret) |= ANYOF_FOLD;
- if (LOC)
+ if (LOC) {
ANYOF_FLAGS(ret) |= ANYOF_LOCALE;
+#ifdef ANYOF_ADD_LOC_SKIP
+ RExC_emit += ANYOF_ADD_LOC_SKIP;
+#endif
+ }
ANYOF_BITMAP_ZERO(ret);
listsv = newSVpvs("# comment\n");
}
(value=='p' ? '+' : '!'), (int)n, RExC_parse);
}
RExC_parse = e + 1;
- ANYOF_FLAGS(ret) |= ANYOF_UNICODE;
+
+ /* The \p could match something in the Latin1 range, hence
+ * something that isn't utf8 */
+ ANYOF_FLAGS(ret) |= ANYOF_NONBITMAP;
+ if (FOLD) { /* And one of these could have a multi-char fold */
+ OP(ret) = ANYOFV;
+ }
namedclass = ANYOF_MAX; /* no official name, but it's named */
}
break;
if (LOC && namedclass < ANYOF_MAX && ! need_class) {
need_class = 1;
if (SIZE_ONLY) {
+#ifdef ANYOF_CLASS_ADD_SKIP
RExC_size += ANYOF_CLASS_ADD_SKIP;
+#endif
}
else {
+#ifdef ANYOF_CLASS_ADD_SKIP
RExC_emit += ANYOF_CLASS_ADD_SKIP;
+#endif
ANYOF_CLASS_ZERO(ret);
}
- ANYOF_FLAGS(ret) |= ANYOF_CLASS|ANYOF_LARGE;
+ ANYOF_FLAGS(ret) |= ANYOF_CLASS;
}
- /* a bad range like a-\d, a-[:digit:] ? */
+ /* a bad range like a-\d, a-[:digit:]. The '-' is taken as a
+ * literal */
if (range) {
if (!SIZE_ONLY) {
const int w =
w, w, rangebegin);
if (prevvalue < 256) {
- ANYOF_BITMAP_SET(ret, prevvalue);
- ANYOF_BITMAP_SET(ret, '-');
+ stored +=
+ S_set_regclass_bit(aTHX_ pRExC_state, ret, (U8) prevvalue);
+ stored +=
+ S_set_regclass_bit(aTHX_ pRExC_state, ret, '-');
}
else {
- ANYOF_FLAGS(ret) |= ANYOF_UNICODE;
+ ANYOF_FLAGS(ret) |= ANYOF_UTF8;
Perl_sv_catpvf(aTHX_ listsv,
- "%04"UVxf"\n%04"UVxf"\n", (UV)prevvalue, (UV) '-');
+ "%04"UVxf"\n%04"UVxf"\n", (UV)prevvalue, (UV) '-');
}
}
const char *what = NULL;
char yesno = 0;
- if (namedclass > OOB_NAMEDCLASS)
- optimize_invert = FALSE;
/* Possible truncation here but in some 64-bit environments
* the compiler gets heartburn about switch on 64-bit values.
* A similar issue a little earlier when switching on value.
if (LOC)
ANYOF_CLASS_SET(ret, ANYOF_ASCII);
else {
-#ifndef EBCDIC
for (value = 0; value < 128; value++)
- ANYOF_BITMAP_SET(ret, value);
-#else /* EBCDIC */
- for (value = 0; value < 256; value++) {
- if (isASCII(value))
- ANYOF_BITMAP_SET(ret, value);
- }
-#endif /* EBCDIC */
+ stored +=
+ S_set_regclass_bit(aTHX_ pRExC_state, ret, (U8) ASCII_TO_NATIVE(value));
}
yesno = '+';
- what = "ASCII";
+ what = NULL; /* Doesn't match outside ascii, so
+ don't want to add +utf8:: */
break;
case ANYOF_NASCII:
if (LOC)
ANYOF_CLASS_SET(ret, ANYOF_NASCII);
else {
-#ifndef EBCDIC
for (value = 128; value < 256; value++)
- ANYOF_BITMAP_SET(ret, value);
-#else /* EBCDIC */
- for (value = 0; value < 256; value++) {
- if (!isASCII(value))
- ANYOF_BITMAP_SET(ret, value);
- }
-#endif /* EBCDIC */
+ stored +=
+ S_set_regclass_bit(aTHX_ pRExC_state, ret, (U8) ASCII_TO_NATIVE(value));
}
yesno = '!';
what = "ASCII";
else {
/* consecutive digits assumed */
for (value = '0'; value <= '9'; value++)
- ANYOF_BITMAP_SET(ret, value);
+ stored +=
+ S_set_regclass_bit(aTHX_ pRExC_state, ret, (U8) value);
}
yesno = '+';
what = POSIX_CC_UNI_NAME("Digit");
else {
/* consecutive digits assumed */
for (value = 0; value < '0'; value++)
- ANYOF_BITMAP_SET(ret, value);
+ stored +=
+ S_set_regclass_bit(aTHX_ pRExC_state, ret, (U8) value);
for (value = '9' + 1; value < 256; value++)
- ANYOF_BITMAP_SET(ret, value);
+ stored +=
+ S_set_regclass_bit(aTHX_ pRExC_state, ret, (U8) value);
}
yesno = '!';
what = POSIX_CC_UNI_NAME("Digit");
if (what) {
/* Strings such as "+utf8::isWord\n" */
Perl_sv_catpvf(aTHX_ listsv, "%cutf8::Is%s\n", yesno, what);
+ ANYOF_FLAGS(ret) |= ANYOF_UTF8;
}
- stored+=2; /* can't optimize this class */
+
continue;
}
} /* end of namedclass \blah */
w, w, rangebegin);
}
if (!SIZE_ONLY)
- ANYOF_BITMAP_SET(ret, '-');
+ stored +=
+ S_set_regclass_bit(aTHX_ pRExC_state, ret, '-');
} else
range = 1; /* yeah, it's a range! */
continue; /* but do it the next time */
}
/* now is the next time */
- /*stored += (value - prevvalue + 1);*/
if (!SIZE_ONLY) {
if (prevvalue < 256) {
const IV ceilvalue = value < 256 ? value : 255;
if (isLOWER(prevvalue)) {
for (i = prevvalue; i <= ceilvalue; i++)
if (isLOWER(i) && !ANYOF_BITMAP_TEST(ret,i)) {
- stored++;
- ANYOF_BITMAP_SET(ret, i);
+ stored +=
+ S_set_regclass_bit(aTHX_ pRExC_state, ret, (U8) i);
}
} else {
for (i = prevvalue; i <= ceilvalue; i++)
if (isUPPER(i) && !ANYOF_BITMAP_TEST(ret,i)) {
- stored++;
- ANYOF_BITMAP_SET(ret, i);
+ stored +=
+ S_set_regclass_bit(aTHX_ pRExC_state, ret, (U8) i);
}
}
}
else
#endif
for (i = prevvalue; i <= ceilvalue; i++) {
- if (!ANYOF_BITMAP_TEST(ret,i)) {
- stored++;
- ANYOF_BITMAP_SET(ret, i);
- }
+ stored += S_set_regclass_bit(aTHX_ pRExC_state, ret, (U8) i);
}
}
if (value > 255 || UTF) {
const UV prevnatvalue = NATIVE_TO_UNI(prevvalue);
const UV natvalue = NATIVE_TO_UNI(value);
- stored+=2; /* can't optimize this class */
- ANYOF_FLAGS(ret) |= ANYOF_UNICODE;
- if (prevnatvalue < natvalue) { /* what about > ? */
+
+ /* If the code point requires utf8 to represent, and we are not
+ * folding, it can't match unless the target is in utf8. Only
+ * a few code points above 255 fold to below it, so XXX an
+ * optimization would be to know which ones and set the flag
+ * appropriately. */
+ ANYOF_FLAGS(ret) |= (FOLD || value < 256)
+ ? ANYOF_NONBITMAP
+ : ANYOF_UTF8;
+ if (prevnatvalue < natvalue) { /* '>' case is fatal error above */
+
+ /* The \t sets the whole range */
Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\t%04"UVxf"\n",
prevnatvalue, natvalue);
+
+ /* Currently, we don't look at every value in the range.
+ * Therefore we have to assume the worst case: that if
+ * folding, it will match more than one character */
+ if (FOLD) {
+ OP(ret) = ANYOFV;
+ }
}
else if (prevnatvalue == natvalue) {
Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n", natvalue);
sv = newSVpvn_utf8((char*)foldbuf, foldlen,
TRUE);
av_push(unicode_alternate, sv);
+ OP(ret) = ANYOFV;
}
}
return ret;
/****** !SIZE_ONLY AFTER HERE *********/
- if( stored == 1 && (value < 128 || (value < 256 && !UTF))
- && !( ANYOF_FLAGS(ret) & ( ANYOF_FLAGS_ALL ^ ANYOF_FOLD ) )
- ) {
- /* optimize single char class to an EXACT node
- but *only* when its not a UTF/high char */
+ /* Optimize inverted simple patterns (e.g. [^a-z]). Note that we haven't
+ * set the FOLD flag yet, so this this does optimize those. It doesn't
+ * optimize locale. Doing so perhaps could be done as long as there is
+ * nothing like \w in it; some thought also would have to be given to the
+ * interaction with above 0x100 chars */
+ if (! LOC && (ANYOF_FLAGS(ret) & ANYOF_FLAGS_ALL) == ANYOF_INVERT) {
+ for (value = 0; value < ANYOF_BITMAP_SIZE; ++value)
+ ANYOF_BITMAP(ret)[value] ^= 0xFF;
+ stored = 256 - stored;
+
+ /* The inversion means that everything above 255 is matched; and at the
+ * same time we clear the invert flag */
+ ANYOF_FLAGS(ret) = ANYOF_UTF8|ANYOF_UNICODE_ALL;
+ }
+
+ if (FOLD) {
+ SV *sv;
+
+ /* This is the one character in the bitmap that needs special handling
+ * under non-locale folding, as it folds to two characters 'ss'. This
+ * happens if it is set and not inverting, or isn't set and are
+ * inverting */
+ if (! LOC
+ && (cBOOL(ANYOF_BITMAP_TEST(ret, LATIN_SMALL_LETTER_SHARP_S))
+ ^ cBOOL(ANYOF_FLAGS(ret) & ANYOF_INVERT)))
+ {
+ OP(ret) = ANYOFV; /* Can match more than a single char */
+
+ /* Under Unicode semantics), it can do this when the target string
+ * isn't in utf8 */
+ if (UNI_SEMANTICS) {
+ ANYOF_FLAGS(ret) |= ANYOF_NONBITMAP_NON_UTF8;
+ }
+
+ if (!unicode_alternate) {
+ unicode_alternate = newAV();
+ }
+ sv = newSVpvn_utf8("ss", 2, TRUE);
+ av_push(unicode_alternate, sv);
+ }
+
+ /* Folding in the bitmap is taken care of above, but not for locale
+ * (for which we have to wait to see what folding is in effect at
+ * runtime), and for things not in the bitmap. Set run-time fold flag
+ * for these */
+ if ((LOC || (ANYOF_FLAGS(ret) & ANYOF_NONBITMAP))) {
+ ANYOF_FLAGS(ret) |= ANYOF_LOC_NONBITMAP_FOLD;
+ }
+ }
+
+ /* A single character class can be "optimized" into an EXACTish node.
+ * Note that since we don't currently count how many characters there are
+ * outside the bitmap, we are XXX missing optimization possibilities for
+ * them. This optimization can't happen unless this is a truly single
+ * character class, which means that it can't be an inversion into a
+ * many-character class, and there must be no possibility of there being
+ * things outside the bitmap. 'stored' (only) for locales doesn't include
+ * \w, etc, so have to make a special test that they aren't present
+ *
+ * Similarly A 2-character class of the very special form like [bB] can be
+ * optimized into an EXACTFish node, but only for non-locales, and for
+ * characters which only have the two folds; so things like 'fF' and 'Ii'
+ * wouldn't work because they are part of the fold of 'LATIN SMALL LIGATURE
+ * FI'. */
+ if (! (ANYOF_FLAGS(ret) & (ANYOF_NONBITMAP|ANYOF_INVERT|ANYOF_UNICODE_ALL))
+ && (((stored == 1 && ((! (ANYOF_FLAGS(ret) & ANYOF_LOCALE))
+ || (! ANYOF_CLASS_TEST_ANY_SET(ret)))))
+ || (stored == 2 && ((! (ANYOF_FLAGS(ret) & ANYOF_LOCALE))
+ && (! _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(value))
+ /* If the latest code point has a fold whose
+ * bit is set, it must be the only other one */
+ && ((prevvalue = PL_fold_latin1[value]) != (IV)value)
+ && ANYOF_BITMAP_TEST(ret, prevvalue)))))
+ {
+ /* Note that the information needed to decide to do this optimization
+ * is not currently available until the 2nd pass, and that the actually
+ * used EXACTish node takes less space than the calculated ANYOF node,
+ * and hence the amount of space calculated in the first pass is larger
+ * than actually used, so this optimization doesn't gain us any space.
+ * But an EXACT node is faster than an ANYOF node, and can be combined
+ * with any adjacent EXACT nodes later by the optimizer for further
+ * gains. The speed of executing an EXACTF is similar to an ANYOF
+ * node, so the optimization advantage comes from the ability to join
+ * it to adjacent EXACT nodes */
+
const char * cur_parse= RExC_parse;
+ U8 op;
RExC_emit = (regnode *)orig_emit;
RExC_parse = (char *)orig_parse;
- ret = reg_node(pRExC_state,
- (U8)((ANYOF_FLAGS(ret) & ANYOF_FOLD) ? EXACTF : EXACT));
- RExC_parse = (char *)cur_parse;
- *STRING(ret)= (char)value;
- STR_LEN(ret)= 1;
- RExC_emit += STR_SZ(1);
- SvREFCNT_dec(listsv);
- return ret;
- }
- /* optimize case-insensitive simple patterns (e.g. /[a-z]/i) */
- if ( /* If the only flag is folding (plus possibly inversion). */
- ((ANYOF_FLAGS(ret) & (ANYOF_FLAGS_ALL ^ ANYOF_INVERT)) == ANYOF_FOLD)
- ) {
- for (value = 0; value < 256; ++value) {
- if (ANYOF_BITMAP_TEST(ret, value)) {
- UV fold = PL_fold[value];
- if (fold != value)
- ANYOF_BITMAP_SET(ret, fold);
+ if (stored == 1) {
+
+ /* A locale node with one point can be folded; all the other cases
+ * with folding will have two points, since we calculate them above
+ */
+ if (ANYOF_FLAGS(ret) & ANYOF_LOC_NONBITMAP_FOLD) {
+ op = EXACTFL;
+ }
+ else {
+ op = EXACT;
}
+ } /* else 2 chars in the bit map: the folds of each other */
+ else if (UNI_SEMANTICS || !isASCII(value)) {
+
+ /* To join adjacent nodes, they must be the exact EXACTish type.
+ * Try to use the most likely type, by using EXACTFU if the regex
+ * calls for them, or is required because the character is
+ * non-ASCII */
+ op = EXACTFU;
+ }
+ else { /* Otherwise, more likely to be EXACTF type */
+ op = EXACTF;
}
- ANYOF_FLAGS(ret) &= ~ANYOF_FOLD;
- }
- /* optimize inverted simple patterns (e.g. [^a-z]) */
- if (optimize_invert &&
- /* If the only flag is inversion. */
- (ANYOF_FLAGS(ret) & ANYOF_FLAGS_ALL) == ANYOF_INVERT) {
- for (value = 0; value < ANYOF_BITMAP_SIZE; ++value)
- ANYOF_BITMAP(ret)[value] ^= ANYOF_FLAGS_ALL;
- ANYOF_FLAGS(ret) = ANYOF_UNICODE_ALL;
+ ret = reg_node(pRExC_state, op);
+ RExC_parse = (char *)cur_parse;
+ if (UTF && ! NATIVE_IS_INVARIANT(value)) {
+ *STRING(ret)= UTF8_EIGHT_BIT_HI((U8) value);
+ *(STRING(ret) + 1)= UTF8_EIGHT_BIT_LO((U8) value);
+ STR_LEN(ret)= 2;
+ RExC_emit += STR_SZ(2);
+ }
+ else {
+ *STRING(ret)= (char)value;
+ STR_LEN(ret)= 1;
+ RExC_emit += STR_SZ(1);
+ }
+ SvREFCNT_dec(listsv);
+ return ret;
}
+
{
AV * const av = newAV();
SV *rv;
/* nextchar()
- Advance that parse position, and optionally absorbs
+ Advances the parse position, and optionally absorbs
"whitespace" from the inputstream.
Without /x "whitespace" means (?#...) style comments only,
- Look for optimizable sequences at the same time.
- currently only looks for EXACT chains.
-This is expermental code. The idea is to use this routine to perform
+This is experimental code. The idea is to use this routine to perform
in place optimizations on branches and groups as they are constructed,
with the long term intention of removing optimization from study_chunk so
that it is purely analytical.
switch (OP(scan)) {
case EXACT:
case EXACTF:
+ case EXACTFU:
case EXACTFL:
if( exact == PSEUDO )
exact= OP(scan);
{
int bit;
int set=0;
+ regex_charset cs;
for (bit=0; bit<32; bit++) {
if (flags & (1<<bit)) {
+ if ((1<<bit) & RXf_PMf_CHARSET) { /* Output separately, below */
+ continue;
+ }
if (!set++ && lead)
PerlIO_printf(Perl_debug_log, "%s",lead);
PerlIO_printf(Perl_debug_log, "%s ",PL_reg_extflags_name[bit]);
}
}
+ if ((cs = get_regex_charset(flags)) != REGEX_DEPENDS_CHARSET) {
+ if (!set++ && lead) {
+ PerlIO_printf(Perl_debug_log, "%s",lead);
+ }
+ switch (cs) {
+ case REGEX_UNICODE_CHARSET:
+ PerlIO_printf(Perl_debug_log, "UNICODE");
+ break;
+ case REGEX_LOCALE_CHARSET:
+ PerlIO_printf(Perl_debug_log, "LOCALE");
+ break;
+ default:
+ PerlIO_printf(Perl_debug_log, "UNKNOWN CHARACTER SET");
+ break;
+ }
+ }
if (lead) {
if (set)
PerlIO_printf(Perl_debug_log, "\n");
* --jhi */
pv_pretty(sv, STRING(o), STR_LEN(o), 60, PL_colors[0], PL_colors[1],
PERL_PV_ESCAPE_UNI_DETECT |
+ PERL_PV_ESCAPE_NONASCII |
PERL_PV_PRETTY_ELLIPSES |
PERL_PV_PRETTY_LTGT |
PERL_PV_PRETTY_NOCLEAR
else if (k == REF || k == OPEN || k == CLOSE || k == GROUPP || OP(o)==ACCEPT) {
Perl_sv_catpvf(aTHX_ sv, "%d", (int)ARG(o)); /* Parenth number */
if ( RXp_PAREN_NAMES(prog) ) {
- if ( k != REF || OP(o) < NREF) {
+ if ( k != REF || (OP(o) < NREF)) {
AV *list= MUTABLE_AV(progi->data->data[progi->name_list_idx]);
SV **name= av_fetch(list, ARG(o), 0 );
if (name)
if (flags & ANYOF_LOCALE)
sv_catpvs(sv, "{loc}");
- if (flags & ANYOF_FOLD)
+ if (flags & ANYOF_LOC_NONBITMAP_FOLD)
sv_catpvs(sv, "{i}");
Perl_sv_catpvf(aTHX_ sv, "[%s", PL_colors[0]);
if (flags & ANYOF_INVERT)
}
EMIT_ANYOF_TEST_SEPARATOR(do_sep,sv,flags);
- /* output any special charclass tests (used mostly under use locale) */
- if (o->flags & ANYOF_CLASS)
+ /* output any special charclass tests (used entirely under use locale) */
+ if (ANYOF_CLASS_TEST_ANY_SET(o))
for (i = 0; i < (int)(sizeof(anyofs)/sizeof(char*)); i++)
if (ANYOF_CLASS_TEST(o,i)) {
sv_catpv(sv, anyofs[i]);
EMIT_ANYOF_TEST_SEPARATOR(do_sep,sv,flags);
+ if (flags & ANYOF_NON_UTF8_LATIN1_ALL) {
+ sv_catpvs(sv, "{non-utf8-latin1-all}");
+ }
+
/* output information about the unicode matching */
- if (flags & ANYOF_UNICODE)
- sv_catpvs(sv, "{unicode}");
- else if (flags & ANYOF_UNICODE_ALL)
+ if (flags & ANYOF_UNICODE_ALL)
sv_catpvs(sv, "{unicode_all}");
+ else if (flags & ANYOF_UTF8)
+ sv_catpvs(sv, "{unicode}");
+ if (flags & ANYOF_NONBITMAP_NON_UTF8)
+ sv_catpvs(sv, "{outside bitmap}");
{
SV *lv;
handles refcounting and freeing the perl core regexp structure. When
it is necessary to actually free the structure the first thing it
- does is call the 'free' method of the regexp_engine associated to to
+ does is call the 'free' method of the regexp_engine associated to
the regexp, allowing the handling of the void *pprivate; member
first. (This routine is not overridable by extensions, which is why
the extensions free is called first.)
The solution is to make a lightweight copy of the regexp structure
when a qr// is returned from the code executed by (??{$qr}) this
- lightweight copy doesnt actually own any of its data except for
+ lightweight copy doesn't actually own any of its data except for
the starp/end and the actual regexp structure itself.
*/
Free the private data in a regexp. This is overloadable by
extensions. Perl takes care of the regexp structure in pregfree(),
- this covers the *pprivate pointer which technically perldoesnt
+ this covers the *pprivate pointer which technically perl doesn't
know about, however of course we have to handle the
regexp_internal structure when no extension is in use.
ones (binary 1111 1111, hexadecimal FF). It is similar, but not
identical, to the ASCII delete (DEL) or rubout control character.
) So the old condition can be simplified to !isPRINT(c) */
- if (!isPRINT(c))
- Perl_sv_catpvf(aTHX_ sv, "\\%o", c);
+ if (!isPRINT(c)) {
+ if (c < 256) {
+ Perl_sv_catpvf(aTHX_ sv, "\\x%02x", c);
+ }
+ else {
+ Perl_sv_catpvf(aTHX_ sv, "\\x{%x}", c);
+ }
+ }
else {
const char string = c;
if (c == '-' || c == ']' || c == '\\' || c == '^')
else if ( op == PLUS || op == STAR) {
DUMPUNTIL(NEXTOPER(node), NEXTOPER(node) + 1);
}
- else if (op == ANYOF) {
+ else if (PL_regkind[(U8)op] == ANYOF) {
/* arglen 1 + class block */
- node += 1 + ((ANYOF_FLAGS(node) & ANYOF_LARGE)
+ node += 1 + ((ANYOF_FLAGS(node) & ANYOF_CLASS)
? ANYOF_CLASS_SKIP : ANYOF_SKIP);
node = NEXTOPER(node);
}