3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
4 * 2000, 2001, 2002, 2003, 2005, 2006, 2007, by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
10 #include "regcharclass.h"
12 /* Convert branch sequences to more efficient trie ops? */
13 #define PERL_ENABLE_TRIE_OPTIMISATION 1
15 /* Be really aggressive about optimising patterns with trie sequences? */
16 #define PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION 1
18 /* Should the optimiser take positive assertions into account? */
19 #define PERL_ENABLE_POSITIVE_ASSERTION_STUDY 0
21 /* Not for production use: */
22 #define PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS 0
24 /* Activate offsets code - set to if 1 to enable */
26 #define RE_TRACK_PATTERN_OFFSETS
30 * The "internal use only" fields in regexp.h are present to pass info from
31 * compile to execute that permits the execute phase to run lots faster on
32 * simple cases. They are:
34 * regstart sv that must begin a match; NULL if none obvious
35 * reganch is the match anchored (at beginning-of-line only)?
36 * regmust string (pointer into program) that match must include, or NULL
37 * [regmust changed to SV* for bminstr()--law]
38 * regmlen length of regmust string
39 * [regmlen not used currently]
41 * Regstart and reganch permit very fast decisions on suitable starting points
42 * for a match, cutting down the work a lot. Regmust permits fast rejection
43 * of lines that cannot possibly match. The regmust tests are costly enough
44 * that pregcomp() supplies a regmust only if the r.e. contains something
45 * potentially expensive (at present, the only such thing detected is * or +
46 * at the start of the r.e., which can involve a lot of backup). Regmlen is
47 * supplied because the test in pregexec() needs it and pregcomp() is computing
49 * [regmust is now supplied always. The tests that use regmust have a
50 * heuristic that disables the test if it usually matches.]
52 * [In fact, we now use regmust in many cases to locate where the search
53 * starts in the string, so if regback is >= 0, the regmust search is never
54 * wasted effort. The regback variable says how many characters back from
55 * where regmust matched is the earliest possible start of the match.
56 * For instance, /[a-z].foo/ has a regmust of 'foo' and a regback of 2.]
60 * Structure for regexp "program". This is essentially a linear encoding
61 * of a nondeterministic finite-state machine (aka syntax charts or
62 * "railroad normal form" in parsing technology). Each node is an opcode
63 * plus a "next" pointer, possibly plus an operand. "Next" pointers of
64 * all nodes except BRANCH implement concatenation; a "next" pointer with
65 * a BRANCH on both ends of it is connecting two alternatives. (Here we
66 * have one of the subtle syntax dependencies: an individual BRANCH (as
67 * opposed to a collection of them) is never concatenated with anything
68 * because of operator precedence.) The operand of some types of node is
69 * a literal string; for others, it is a node leading into a sub-FSM. In
70 * particular, the operand of a BRANCH node is the first node of the branch.
71 * (NB this is *not* a tree structure: the tail of the branch connects
72 * to the thing following the set of BRANCHes.) The opcodes are defined
73 * in regnodes.h which is generated from regcomp.sym by regcomp.pl.
77 * A node is one char of opcode followed by two chars of "next" pointer.
78 * "Next" pointers are stored as two 8-bit pieces, high order first. The
79 * value is a positive offset from the opcode of the node containing it.
80 * An operand, if any, simply follows the node. (Note that much of the
81 * code generation knows about this implicit relationship.)
83 * Using two bytes for the "next" pointer is vast overkill for most things,
84 * but allows patterns to get big without disasters.
86 * [The "next" pointer is always aligned on an even
87 * boundary, and reads the offset directly as a short.]
90 /* This is the stuff that used to live in regexp.h that was truly
91 private to the engine itself. It now lives here. */
93 typedef struct regexp_internal {
94 int name_list_idx; /* Optional data index of an array of paren names */
96 U32 *offsets; /* offset annotations 20001228 MJD
97 data about mapping the program to the
99 offsets[0] is proglen when this is used
104 regnode *regstclass; /* Optional startclass as identified or constructed
106 struct reg_data *data; /* Additional miscellaneous data used by the program.
107 Used to make it easier to clone and free arbitrary
108 data that the regops need. Often the ARG field of
109 a regop is an index into this structure */
110 struct reg_code_blocks *code_blocks;/* positions of literal (?{}) */
111 regnode program[1]; /* Unwarranted chumminess with compiler. */
114 #define RXi_SET(x,y) (x)->pprivate = (void*)(y)
115 #define RXi_GET(x) ((regexp_internal *)((x)->pprivate))
116 #define RXi_GET_DECL(r,ri) regexp_internal *ri = RXi_GET(r)
118 * Flags stored in regexp->intflags
119 * These are used only internally to the regexp engine
121 * See regexp.h for flags used externally to the regexp engine
123 #define RXp_INTFLAGS(rx) ((rx)->intflags)
124 #define RX_INTFLAGS(prog) RXp_INTFLAGS(ReANY(prog))
126 #define PREGf_SKIP 0x00000001
127 #define PREGf_IMPLICIT 0x00000002 /* Converted .* to ^.* */
128 #define PREGf_NAUGHTY 0x00000004 /* how exponential is this pattern? */
129 #define PREGf_VERBARG_SEEN 0x00000008
130 #define PREGf_CUTGROUP_SEEN 0x00000010
131 #define PREGf_USE_RE_EVAL 0x00000020 /* compiled with "use re 'eval'" */
132 /* these used to be extflags, but are now intflags */
133 #define PREGf_NOSCAN 0x00000040
135 #define PREGf_GPOS_SEEN 0x00000100
136 #define PREGf_GPOS_FLOAT 0x00000200
138 #define PREGf_ANCH_MBOL 0x00000400
139 #define PREGf_ANCH_SBOL 0x00000800
140 #define PREGf_ANCH_GPOS 0x00001000
141 #define PREGf_RECURSE_SEEN 0x00002000
144 ( PREGf_ANCH_SBOL | PREGf_ANCH_GPOS | PREGf_ANCH_MBOL )
146 /* this is where the old regcomp.h started */
148 struct regnode_string {
155 /* Argument bearing node - workhorse,
156 arg1 is often for the data field */
164 /* Similar to a regnode_1 but with an extra signed argument */
173 /* 'Two field' -- Two 16 bit unsigned args */
182 /* This give the number of code points that can be in the bitmap of an ANYOF
183 * node. The shift number must currently be one of: 8..12. It can't be less
184 * than 8 (256) because some code relies on it being at least that. Above 12
185 * (4096), and you start running into warnings that some data structure widths
186 * have been exceeded, though the test suite as of this writing still passes
187 * for up through 16, which is as high as anyone would ever want to go,
188 * encompassing all of the Unicode BMP, and thus including all the economically
189 * important world scripts. At 12 most of them are: including Arabic,
190 * Cyrillic, Greek, Hebrew, Indian subcontinent, Latin, and Thai; but not Han,
191 * Japanese, nor Korean. (The regarglen structure in regnodes.h is a U8, and
192 * the trie types TRIEC and AHOCORASICKC are larger than U8 for shift values
193 * below above 12.) Be sure to benchmark before changing, as larger sizes do
194 * significantly slow down the test suite */
195 #define NUM_ANYOF_CODE_POINTS (1 << 8)
197 #define ANYOF_BITMAP_SIZE (NUM_ANYOF_CODE_POINTS / 8) /* 8 bits/Byte */
199 /* Note that these form structs which are supersets of the next smaller one, by
200 * appending fields. Alignment problems can occur if one of those optional
201 * fields requires stricter alignment than the base struct. And formal
202 * parameters that can really be two or more of the structs should be
203 * declared as the smallest one it could be. See commit message for
204 * 7dcac5f6a5195002b55c935ee1d67f67e1df280b. Regnode allocation is done
205 * without regard to alignment, and changing it to would also require changing
206 * the code that inserts and deletes regnodes. The basic single-argument
207 * regnode has a U32, which is what reganode() allocates as a unit. Therefore
208 * no field can require stricter alignment than U32. */
210 /* also used by trie */
211 struct regnode_charclass {
215 U32 arg1; /* set by set_ANYOF_arg() */
216 char bitmap[ANYOF_BITMAP_SIZE]; /* only compile-time */
219 /* has runtime (locale) \d, \w, ..., [:posix:] classes */
220 struct regnode_charclass_class {
221 U8 flags; /* ANYOF_MATCHES_POSIXL bit must go here */
225 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time ... */
226 U32 classflags; /* and run-time */
229 /* A synthetic start class (SSC); is a regnode_charclass_posixl_fold, plus an
230 * extra SV*, used only during its construction and which is not used by
231 * regexec.c. Note that the 'next_off' field is unused, as the SSC stands
232 * alone, so there is never a next node. Also, there is no alignment issue,
233 * because these are declared or allocated as a complete unit so the compiler
234 * takes care of alignment. This is unlike the other regnodes which are
235 * allocated in terms of multiples of a single-argument regnode. SSC nodes can
236 * have a pointer field because there is no alignment issue, and because it is
237 * set to NULL after construction, before any cloning of the pattern */
239 U8 flags; /* ANYOF_MATCHES_POSIXL bit must go here */
243 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time ... */
244 U32 classflags; /* ... and run-time */
246 /* Auxiliary, only used during construction; NULL afterwards: list of code
251 /* We take advantage of 'next_off' not otherwise being used in the SSC by
252 * actually using it: by setting it to 1. This allows us to test and
253 * distinguish between an SSC and other ANYOF node types, as 'next_off' cannot
254 * otherwise be 1, because it is the offset to the next regnode expressed in
255 * units of regnodes. Since an ANYOF node contains extra fields, it adds up
256 * to 12 regnode units on 32-bit systems, (hence the minimum this can be (if
257 * not 0) is 11 there. Even if things get tightly packed on a 64-bit system,
258 * it still would be more than 1. */
259 #define set_ANYOF_SYNTHETIC(n) STMT_START{ OP(n) = ANYOF; \
262 #define is_ANYOF_SYNTHETIC(n) (PL_regkind[OP(n)] == ANYOF && NEXT_OFF(n) == 1)
264 /* XXX fix this description.
265 Impose a limit of REG_INFTY on various pattern matching operations
266 to limit stack growth and to avoid "infinite" recursions.
268 /* The default size for REG_INFTY is I16_MAX, which is the same as
269 SHORT_MAX (see perl.h). Unfortunately I16 isn't necessarily 16 bits
270 (see handy.h). On the Cray C90, sizeof(short)==4 and hence I16_MAX is
271 ((1<<31)-1), while on the Cray T90, sizeof(short)==8 and I16_MAX is
272 ((1<<63)-1). To limit stack growth to reasonable sizes, supply a
274 --Andy Dougherty 11 June 1998
278 # define REG_INFTY ((1<<15)-1)
283 # define REG_INFTY I16_MAX
286 #define ARG_VALUE(arg) (arg)
287 #define ARG__SET(arg,val) ((arg) = (val))
293 #define ARG(p) ARG_VALUE(ARG_LOC(p))
294 #define ARG1(p) ARG_VALUE(ARG1_LOC(p))
295 #define ARG2(p) ARG_VALUE(ARG2_LOC(p))
296 #define ARG2L(p) ARG_VALUE(ARG2L_LOC(p))
298 #define ARG_SET(p, val) ARG__SET(ARG_LOC(p), (val))
299 #define ARG1_SET(p, val) ARG__SET(ARG1_LOC(p), (val))
300 #define ARG2_SET(p, val) ARG__SET(ARG2_LOC(p), (val))
301 #define ARG2L_SET(p, val) ARG__SET(ARG2L_LOC(p), (val))
306 #define NEXT_OFF(p) ((p)->next_off)
307 #define NODE_ALIGN(node)
308 /* the following define was set to 0xde in 075abff3
309 * as part of some linting logic. I have set it to 0
310 * as otherwise in every place where we /might/ set flags
311 * we have to set it 0 explicitly, which duplicates
312 * assignments and IMO adds an unacceptable level of
313 * surprise to working in the regex engine. If this
314 * is changed from 0 then at the very least make sure
315 * that SBOL for /^/ sets the flags to 0 explicitly.
317 #define NODE_ALIGN_FILL(node) ((node)->flags = 0)
319 #define SIZE_ALIGN NODE_ALIGN
326 #define OP(p) ((p)->type)
327 #define FLAGS(p) ((p)->flags) /* Caution: Doesn't apply to all \
328 regnode types. For some, it's the \
329 character set of the regnode */
330 #define OPERAND(p) (((struct regnode_string *)p)->string)
331 #define MASK(p) ((char*)OPERAND(p))
332 #define STR_LEN(p) (((struct regnode_string *)p)->str_len)
333 #define STRING(p) (((struct regnode_string *)p)->string)
334 #define STR_SZ(l) ((l + sizeof(regnode) - 1) / sizeof(regnode))
335 #define NODE_SZ_STR(p) (STR_SZ(STR_LEN(p))+1)
342 #define NODE_ALIGN(node)
343 #define ARG_LOC(p) (((struct regnode_1 *)p)->arg1)
344 #define ARG1_LOC(p) (((struct regnode_2 *)p)->arg1)
345 #define ARG2_LOC(p) (((struct regnode_2 *)p)->arg2)
346 #define ARG2L_LOC(p) (((struct regnode_2L *)p)->arg2)
348 #define NODE_STEP_REGNODE 1 /* sizeof(regnode)/sizeof(regnode) */
349 #define EXTRA_STEP_2ARGS EXTRA_SIZE(struct regnode_2)
351 #define NODE_STEP_B 4
353 #define NEXTOPER(p) ((p) + NODE_STEP_REGNODE)
354 #define PREVOPER(p) ((p) - NODE_STEP_REGNODE)
356 #define FILL_ADVANCE_NODE(ptr, op) STMT_START { \
357 (ptr)->type = op; (ptr)->next_off = 0; (ptr)++; } STMT_END
358 #define FILL_ADVANCE_NODE_ARG(ptr, op, arg) STMT_START { \
359 ARG_SET(ptr, arg); FILL_ADVANCE_NODE(ptr, op); (ptr) += 1; } STMT_END
360 #define FILL_ADVANCE_NODE_2L_ARG(ptr, op, arg1, arg2) \
362 ARG_SET(ptr, arg1); \
363 ARG2L_SET(ptr, arg2); \
364 FILL_ADVANCE_NODE(ptr, op); \
368 #define REG_MAGIC 0234
370 #define SIZE_ONLY cBOOL(RExC_emit == (regnode *) & RExC_emit_dummy)
371 #define PASS1 SIZE_ONLY
372 #define PASS2 (! SIZE_ONLY)
374 /* An ANYOF node is basically a bitmap with the index being a code point. If
375 * the bit for that code point is 1, the code point matches; if 0, it doesn't
376 * match (complemented if inverted). There is an additional mechanism to deal
377 * with cases where the bitmap is insufficient in and of itself. This #define
378 * indicates if the bitmap does fully represent what this ANYOF node can match.
379 * The ARG is set to this special value (since 0, 1, ... are legal, but will
380 * never reach this high). */
381 #define ANYOF_ONLY_HAS_BITMAP ((U32) -1)
383 /* When the bimap isn't completely sufficient for handling the ANYOF node,
384 * flags (in node->flags of the ANYOF node) get set to indicate this. These
385 * are perennially in short supply. Beyond several cases where warnings need
386 * to be raised under certain circumstances, currently, there are six cases
387 * where the bitmap alone isn't sufficient. We could use six flags to
388 * represent the 6 cases, but to save flags bits, we play some games. The
391 * 1) The bitmap has a compiled-in very finite size. So something else needs
392 * to be used to specify if a code point that is too large for the bitmap
393 * actually matches. The mechanism currently is a swash or inversion
394 * list. ANYOF_ONLY_HAS_BITMAP, described above, being TRUE indicates
395 * there are no matches of too-large code points. But if it is FALSE,
396 * then almost certainly there are matches too large for the bitmap. (The
397 * other cases, described below, either imply this one or are extremely
398 * rare in practice.) So we can just assume that a too-large code point
399 * will need something beyond the bitmap if ANYOF_ONLY_HAS_BITMAP is
400 * FALSE, instead of having a separate flag for this.
401 * 2) A subset of item 1) is if all possible code points outside the bitmap
402 * match. This is a common occurrence when the class is complemented,
403 * like /[^ij]/. Therefore a bit is reserved to indicate this,
404 * rather than having an expensive swash created,
405 * ANYOF_MATCHES_ALL_ABOVE_BITMAP.
406 * 3) Under /d rules, it can happen that code points that are in the upper
407 * latin1 range (\x80-\xFF or their equivalents on EBCDIC platforms) match
408 * only if the runtime target string being matched against is UTF-8. For
409 * example /[\w[:punct:]]/d. This happens only for posix classes (with a
410 * couple of exceptions, like \d where it doesn't happen), and all such
411 * ones also have above-bitmap matches. Thus, 3) implies 1) as well.
412 * Note that /d rules are no longer encouraged; 'use 5.14' or higher
413 * deselects them. But a flag is required so that they can be properly
414 * handled. But it can be a shared flag: see 5) below.
415 * 4) Also under /d rules, something like /[\Wfoo]/ will match everything in
416 * the \x80-\xFF range, unless the string being matched against is UTF-8.
417 * A swash could be created for this case, but this is relatively common,
418 * and it turns out that it's all or nothing: if any one of these code
419 * points matches, they all do. Hence a single bit suffices. We use a
420 * shared flag that doesn't take up space by itself:
421 * ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER.
422 * This also implies 1), with one exception: [:^cntrl:].
423 * 5) A user-defined \p{} property may not have been defined by the time the
424 * regex is compiled. In this case, we don't know until runtime what it
425 * will match, so we have to assume it could match anything, including
426 * code points that ordinarily would be in the bitmap. A flag bit is
427 * necessary to indicate this , though it can be shared with the item 3)
428 * flag, as that only occurs under /d, and this only occurs under non-d.
429 * This case is quite uncommon in the field, and the /(?[ ...])/ construct
430 * is a better way to accomplish what this feature does. This case also
432 * ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP
433 * is the shared flag.
434 * 6) /[foo]/il may have folds that are only valid if the runtime locale is a
435 * UTF-8 one. These are quite rare, so it would be good to avoid the
436 * expense of looking for them. But /l matching is slow anyway, and we've
437 * traditionally not worried too much about its performance. And this
438 * condition requires the ANYOFL_FOLD flag to be set, so testing for
439 * that flag would be sufficient to rule out most cases of this. So it is
440 * unclear if this should have a flag or not. But, this flag can be
441 * shared with another, so it doesn't occupy extra space.
443 * At the moment, there is one spare bit, but this could be increased by
446 * If just one more bit is needed, at this writing it seems to khw that the
447 * best choice would be to make ANYOF_MATCHES_ALL_ABOVE_BITMAP not a flag, but
450 * #define ANYOF_MATCHES_ALL_ABOVE_BITMAP ((U32) -2)
452 * and access it through the ARG like ANYOF_ONLY_HAS_BITMAP is. This flag is
453 * used by all ANYOF node types, and it could be used to avoid calling the
454 * handler function, as the macro REGINCLASS in regexec.c does now for other
457 * Another possibility is to instead (or additionally) rename the ANYOF_POSIXL
458 * flag to be ANYOFL_LARGE, to mean that the ANYOF node has an extra 32 bits
459 * beyond what a regular one does. That's what it effectively means now, with
460 * the extra space all for the POSIX class flags. But those classes actually
461 * only occupy 30 bits, so the ANYOFL_FOLD and
462 * ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD flags could be moved
463 * to that extra space. The 30 bits in the extra word would indicate if a
464 * posix class should be looked up or not. The downside of this is that ANYOFL
465 * nodes with folding would always have to have the extra space allocated, even
466 * if they didn't use the 30 posix bits. There isn't an SSC problem as all
467 * SSCs are this large anyway.
469 * One could completely remove ANYOFL_LARGE and make all ANYOFL nodes large.
470 * REGINCLASS would have to be modified so that if the node type were this, it
471 * would call reginclass(), as the flag bit that indicates to do this now would
474 * All told, 5 bits could be available for other uses if all of the above were
477 * Some flags are not used in synthetic start class (SSC) nodes, so could be
478 * shared should new flags be needed for SSCs, like SSC_MATCHES_EMPTY_STRING
481 /* If this is set, the result of the match should be complemented. regexec.c
482 * is expecting this to be in the low bit. Never in an SSC */
483 #define ANYOF_INVERT 0x01
485 /* For the SSC node only, which cannot be inverted, so is shared with that bit.
486 * This is used only during regex compilation. */
487 #define SSC_MATCHES_EMPTY_STRING ANYOF_INVERT
489 /* Set if this is a regnode_charclass_posixl vs a regnode_charclass. This
490 * is used for runtime \d, \w, [:posix:], ..., which are used only in locale
491 * and the optimizer's synthetic start class. Non-locale \d, etc are resolved
492 * at compile-time. Only set under /l; can be in SSC */
493 #define ANYOF_MATCHES_POSIXL 0x02
495 /* The fold is calculated and stored in the bitmap where possible at compile
496 * time. However under locale, the actual folding varies depending on
497 * what the locale is at the time of execution, so it has to be deferred until
498 * then. Only set under /l; never in an SSC */
499 #define ANYOFL_FOLD 0x04
501 /* Shared bit set only with ANYOFL and SSC nodes:
502 * If ANYOFL_FOLD is set, this means there are potential matches valid
503 * only if the locale is a UTF-8 one.
504 * If ANYOFL_FOLD is NOT set, this means to warn if the runtime locale
505 * isn't a UTF-8 one (and the generated node assumes a UTF-8 locale).
506 * None of INVERT, POSIXL,
507 * ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP
509 #define ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD 0x08
511 /* Convenience macros for teasing apart the meanings when reading the above bit
513 #define ANYOFL_SOME_FOLDS_ONLY_IN_UTF8_LOCALE(flags) \
514 ((flags & ( ANYOFL_FOLD /* Both bits are set */ \
515 |ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD)) \
517 |ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD))
519 #define ANYOFL_UTF8_LOCALE_REQD(flags) \
520 ((flags & ( ANYOFL_FOLD /* Only REQD bit is set */ \
521 |ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD)) \
522 == ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD)
524 /* Spare: Be sure to change ANYOF_FLAGS_ALL if this gets used 0x10 */
526 /* If set, the node matches every code point NUM_ANYOF_CODE_POINTS and above.
527 * Can be in an SSC */
528 #define ANYOF_MATCHES_ALL_ABOVE_BITMAP 0x20
531 * Under /d it means the ANYOFD node matches more things if the target
532 * string is encoded in UTF-8; any such things will be non-ASCII,
533 * characters that are < 256, and can be accessed via the swash.
534 * When not under /d, it means the ANYOF node contains a user-defined
535 * property that wasn't yet defined at the time the regex was compiled,
536 * and so must be looked up at runtime, by creating a swash
537 * (These uses are mutually exclusive because a user-defined property is
538 * specified by \p{}, and \p{} implies /u which deselects /d). The long macro
539 * name is to make sure that you are cautioned about its shared nature. Only
540 * the non-/d meaning can be in an SSC */
541 #define ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP 0x40
544 * Under /d it means the ANYOFD node matches all non-ASCII Latin1
545 * characters when the target string is not in utf8.
546 * When not under /d, it means the ANYOF node should raise a warning if
547 * matching against an above-Unicode code point.
548 * (These uses are mutually exclusive because the warning requires a \p{}, and
549 * \p{} implies /u which deselects /d). An SSC node only has this bit set if
550 * what is meant is the warning. The long macro name is to make sure that you
551 * are cautioned about its shared nature */
552 #define ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER 0x80
554 #define ANYOF_FLAGS_ALL (0xff & ~0x10)
556 #define ANYOF_LOCALE_FLAGS (ANYOFL_FOLD | ANYOF_MATCHES_POSIXL)
558 /* These are the flags that apply to both regular ANYOF nodes and synthetic
559 * start class nodes during construction of the SSC. During finalization of
560 * the SSC, other of the flags may get added to it */
561 #define ANYOF_COMMON_FLAGS 0
563 /* Character classes for node->classflags of ANYOF */
564 /* Should be synchronized with a table in regprop() */
565 /* 2n should be the normal one, paired with its complement at 2n+1 */
567 #define ANYOF_ALPHA ((_CC_ALPHA) * 2)
568 #define ANYOF_NALPHA ((ANYOF_ALPHA) + 1)
569 #define ANYOF_ALPHANUMERIC ((_CC_ALPHANUMERIC) * 2) /* [[:alnum:]] isalnum(3), utf8::IsAlnum */
570 #define ANYOF_NALPHANUMERIC ((ANYOF_ALPHANUMERIC) + 1)
571 #define ANYOF_ASCII ((_CC_ASCII) * 2)
572 #define ANYOF_NASCII ((ANYOF_ASCII) + 1)
573 #define ANYOF_BLANK ((_CC_BLANK) * 2) /* GNU extension: space and tab: non-vertical space */
574 #define ANYOF_NBLANK ((ANYOF_BLANK) + 1)
575 #define ANYOF_CASED ((_CC_CASED) * 2) /* Pseudo class for [:lower:] or
576 [:upper:] under /i */
577 #define ANYOF_NCASED ((ANYOF_CASED) + 1)
578 #define ANYOF_CNTRL ((_CC_CNTRL) * 2)
579 #define ANYOF_NCNTRL ((ANYOF_CNTRL) + 1)
580 #define ANYOF_DIGIT ((_CC_DIGIT) * 2) /* \d */
581 #define ANYOF_NDIGIT ((ANYOF_DIGIT) + 1)
582 #define ANYOF_GRAPH ((_CC_GRAPH) * 2)
583 #define ANYOF_NGRAPH ((ANYOF_GRAPH) + 1)
584 #define ANYOF_LOWER ((_CC_LOWER) * 2)
585 #define ANYOF_NLOWER ((ANYOF_LOWER) + 1)
586 #define ANYOF_PRINT ((_CC_PRINT) * 2)
587 #define ANYOF_NPRINT ((ANYOF_PRINT) + 1)
588 #define ANYOF_PUNCT ((_CC_PUNCT) * 2)
589 #define ANYOF_NPUNCT ((ANYOF_PUNCT) + 1)
590 #define ANYOF_SPACE ((_CC_SPACE) * 2) /* \s */
591 #define ANYOF_NSPACE ((ANYOF_SPACE) + 1)
592 #define ANYOF_UPPER ((_CC_UPPER) * 2)
593 #define ANYOF_NUPPER ((ANYOF_UPPER) + 1)
594 #define ANYOF_WORDCHAR ((_CC_WORDCHAR) * 2) /* \w, PL_utf8_alnum, utf8::IsWord, ALNUM */
595 #define ANYOF_NWORDCHAR ((ANYOF_WORDCHAR) + 1)
596 #define ANYOF_XDIGIT ((_CC_XDIGIT) * 2)
597 #define ANYOF_NXDIGIT ((ANYOF_XDIGIT) + 1)
599 /* pseudo classes below this, not stored in the class bitmap, but used as flags
600 during compilation of char classes */
602 #define ANYOF_VERTWS ((_CC_VERTSPACE) * 2)
603 #define ANYOF_NVERTWS ((ANYOF_VERTWS)+1)
605 /* It is best if this is the last one, as all above it are stored as bits in a
606 * bitmap, and it isn't part of that bitmap */
607 #if _CC_VERTSPACE != _HIGHEST_REGCOMP_DOT_H_SYNC
608 # error Problem with handy.h _HIGHEST_REGCOMP_DOT_H_SYNC #define
611 #define ANYOF_POSIXL_MAX (ANYOF_VERTWS) /* So upper loop limit is written:
613 * Hence doesn't include VERTWS, as that
614 * is a pseudo class */
615 #define ANYOF_MAX ANYOF_POSIXL_MAX
617 #if (ANYOF_POSIXL_MAX > 32) /* Must fit in 32-bit word */
618 # error Problem with handy.h _CC_foo #defines
621 #define ANYOF_HORIZWS ((ANYOF_POSIXL_MAX)+2) /* = (ANYOF_NVERTWS + 1) */
622 #define ANYOF_NHORIZWS ((ANYOF_POSIXL_MAX)+3)
624 #define ANYOF_UNIPROP ((ANYOF_POSIXL_MAX)+4) /* Used to indicate a Unicode
625 property: \p{} or \P{} */
627 /* Backward source code compatibility. */
629 #define ANYOF_ALNUML ANYOF_ALNUM
630 #define ANYOF_NALNUML ANYOF_NALNUM
631 #define ANYOF_SPACEL ANYOF_SPACE
632 #define ANYOF_NSPACEL ANYOF_NSPACE
633 #define ANYOF_ALNUM ANYOF_WORDCHAR
634 #define ANYOF_NALNUM ANYOF_NWORDCHAR
636 /* Utility macros for the bitmap and classes of ANYOF */
638 #define ANYOF_FLAGS(p) ((p)->flags)
640 #define ANYOF_BIT(c) (1U << ((c) & 7))
642 #define ANYOF_POSIXL_SET(p, c) (((regnode_charclass_posixl*) (p))->classflags |= (1U << (c)))
643 #define ANYOF_CLASS_SET(p, c) ANYOF_POSIXL_SET((p), (c))
645 #define ANYOF_POSIXL_CLEAR(p, c) (((regnode_charclass_posixl*) (p))->classflags &= ~ (1U <<(c)))
646 #define ANYOF_CLASS_CLEAR(p, c) ANYOF_POSIXL_CLEAR((p), (c))
648 #define ANYOF_POSIXL_TEST(p, c) (((regnode_charclass_posixl*) (p))->classflags & (1U << (c)))
649 #define ANYOF_CLASS_TEST(p, c) ANYOF_POSIXL_TEST((p), (c))
651 #define ANYOF_POSIXL_ZERO(ret) STMT_START { ((regnode_charclass_posixl*) (ret))->classflags = 0; } STMT_END
652 #define ANYOF_CLASS_ZERO(ret) ANYOF_POSIXL_ZERO(ret)
654 /* Shifts a bit to get, eg. 0x4000_0000, then subtracts 1 to get 0x3FFF_FFFF */
655 #define ANYOF_POSIXL_SETALL(ret) STMT_START { ((regnode_charclass_posixl*) (ret))->classflags = ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1; } STMT_END
656 #define ANYOF_CLASS_SETALL(ret) ANYOF_POSIXL_SETALL(ret)
658 #define ANYOF_POSIXL_TEST_ANY_SET(p) \
659 ((ANYOF_FLAGS(p) & ANYOF_MATCHES_POSIXL) \
660 && (((regnode_charclass_posixl*)(p))->classflags))
661 #define ANYOF_CLASS_TEST_ANY_SET(p) ANYOF_POSIXL_TEST_ANY_SET(p)
663 /* Since an SSC always has this field, we don't have to test for that; nor do
664 * we want to because the bit isn't set for SSC during its construction */
665 #define ANYOF_POSIXL_SSC_TEST_ANY_SET(p) \
666 cBOOL(((regnode_ssc*)(p))->classflags)
667 #define ANYOF_POSIXL_SSC_TEST_ALL_SET(p) /* Are all bits set? */ \
668 (((regnode_ssc*) (p))->classflags \
669 == ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1)
671 #define ANYOF_POSIXL_TEST_ALL_SET(p) \
672 ((ANYOF_FLAGS(p) & ANYOF_MATCHES_POSIXL) \
673 && ((regnode_charclass_posixl*) (p))->classflags \
674 == ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1)
676 #define ANYOF_POSIXL_OR(source, dest) STMT_START { (dest)->classflags |= (source)->classflags ; } STMT_END
677 #define ANYOF_CLASS_OR(source, dest) ANYOF_POSIXL_OR((source), (dest))
679 #define ANYOF_POSIXL_AND(source, dest) STMT_START { (dest)->classflags &= (source)->classflags ; } STMT_END
681 #define ANYOF_BITMAP_ZERO(ret) Zero(((regnode_charclass*)(ret))->bitmap, ANYOF_BITMAP_SIZE, char)
682 #define ANYOF_BITMAP(p) ((regnode_charclass*)(p))->bitmap
683 #define ANYOF_BITMAP_BYTE(p, c) BITMAP_BYTE(ANYOF_BITMAP(p), c)
684 #define ANYOF_BITMAP_SET(p, c) (ANYOF_BITMAP_BYTE(p, c) |= ANYOF_BIT(c))
685 #define ANYOF_BITMAP_CLEAR(p,c) (ANYOF_BITMAP_BYTE(p, c) &= ~ANYOF_BIT(c))
686 #define ANYOF_BITMAP_TEST(p, c) cBOOL(ANYOF_BITMAP_BYTE(p, c) & ANYOF_BIT(c))
688 #define ANYOF_BITMAP_SETALL(p) \
689 memset (ANYOF_BITMAP(p), 255, ANYOF_BITMAP_SIZE)
690 #define ANYOF_BITMAP_CLEARALL(p) \
691 Zero (ANYOF_BITMAP(p), ANYOF_BITMAP_SIZE)
693 #define ANYOF_SKIP EXTRA_SIZE(regnode_charclass)
694 #define ANYOF_POSIXL_SKIP EXTRA_SIZE(regnode_charclass_posixl)
697 * Utility definitions.
700 # define UCHARAT(p) ((int)*(const U8*)(p))
702 # define UCHARAT(p) ((int)*(p)&CHARMASK)
705 #define EXTRA_SIZE(guy) ((sizeof(guy)-1)/sizeof(struct regnode))
707 #define REG_ZERO_LEN_SEEN 0x00000001
708 #define REG_LOOKBEHIND_SEEN 0x00000002
709 #define REG_GPOS_SEEN 0x00000004
711 #define REG_RECURSE_SEEN 0x00000020
712 #define REG_TOP_LEVEL_BRANCHES_SEEN 0x00000040
713 #define REG_VERBARG_SEEN 0x00000080
714 #define REG_CUTGROUP_SEEN 0x00000100
715 #define REG_RUN_ON_COMMENT_SEEN 0x00000200
716 #define REG_UNFOLDED_MULTI_SEEN 0x00000400
718 #define REG_UNBOUNDED_QUANTIFIER_SEEN 0x00001000
723 #ifdef PLUGGABLE_RE_EXTENSION
724 #include "re_nodes.h"
726 #include "regnodes.h"
729 #ifndef PLUGGABLE_RE_EXTENSION
731 EXTCONST regexp_engine PL_core_reg_engine;
733 EXTCONST regexp_engine PL_core_reg_engine = {
736 Perl_re_intuit_start,
737 Perl_re_intuit_string,
738 Perl_regfree_internal,
739 Perl_reg_numbered_buff_fetch,
740 Perl_reg_numbered_buff_store,
741 Perl_reg_numbered_buff_length,
743 Perl_reg_named_buff_iter,
745 #if defined(USE_ITHREADS)
746 Perl_regdupe_internal,
751 #endif /* PLUGGABLE_RE_EXTENSION */
757 /* .what is a character array with one character for each member of .data
758 * The character describes the function of the corresponding .data item:
759 * a - AV for paren_name_list under DEBUGGING
760 * f - start-class data for regstclass optimization
761 * l - start op for literal (?{EVAL}) item
762 * L - start op for literal (?{EVAL}) item, with separate CV (qr//)
763 * r - pointer to an embedded code-containing qr, e.g. /ab$qr/
764 * s - swash for Unicode-style character class, and the multicharacter
765 * strings resulting from casefolding the single-character entries
766 * in the character class
768 * u - trie struct's widecharmap (a HV, so can't share, must dup)
769 * also used for revcharmap and words under DEBUGGING
770 * T - aho-trie struct
771 * S - sv for named capture lookup
772 * 20010712 mjd@plover.com
773 * (Remember to update re_dup() and pregfree() if you add any items.)
781 /* Code in S_to_utf8_substr() and S_to_byte_substr() in regexec.c accesses
782 anchored* and float* via array indexes 0 and 1. */
783 #define anchored_substr substrs->data[0].substr
784 #define anchored_utf8 substrs->data[0].utf8_substr
785 #define anchored_offset substrs->data[0].min_offset
786 #define anchored_end_shift substrs->data[0].end_shift
788 #define float_substr substrs->data[1].substr
789 #define float_utf8 substrs->data[1].utf8_substr
790 #define float_min_offset substrs->data[1].min_offset
791 #define float_max_offset substrs->data[1].max_offset
792 #define float_end_shift substrs->data[1].end_shift
794 #define check_substr substrs->data[2].substr
795 #define check_utf8 substrs->data[2].utf8_substr
796 #define check_offset_min substrs->data[2].min_offset
797 #define check_offset_max substrs->data[2].max_offset
798 #define check_end_shift substrs->data[2].end_shift
800 #define RX_ANCHORED_SUBSTR(rx) (ReANY(rx)->anchored_substr)
801 #define RX_ANCHORED_UTF8(rx) (ReANY(rx)->anchored_utf8)
802 #define RX_FLOAT_SUBSTR(rx) (ReANY(rx)->float_substr)
803 #define RX_FLOAT_UTF8(rx) (ReANY(rx)->float_utf8)
805 /* trie related stuff */
807 /* a transition record for the state machine. the
808 check field determines which state "owns" the
809 transition. the char the transition is for is
810 determined by offset from the owning states base
811 field. the next field determines which state
812 is to be transitioned to if any.
814 struct _reg_trie_trans {
819 /* a transition list element for the list based representation */
820 struct _reg_trie_trans_list_elem {
824 typedef struct _reg_trie_trans_list_elem reg_trie_trans_le;
826 /* a state for compressed nodes. base is an offset
827 into an array of reg_trie_trans array. If wordnum is
828 nonzero the state is accepting. if base is zero then
829 the state has no children (and will be accepting)
831 struct _reg_trie_state {
835 reg_trie_trans_le* list;
839 /* info per word; indexed by wordnum */
841 U16 prev; /* previous word in acceptance chain; eg in
842 * zzz|abc|ab/ after matching the chars abc, the
843 * accepted word is #2, and the previous accepted
845 U32 len; /* how many chars long is this word? */
846 U32 accept; /* accept state for this word */
850 typedef struct _reg_trie_state reg_trie_state;
851 typedef struct _reg_trie_trans reg_trie_trans;
854 /* anything in here that needs to be freed later
855 should be dealt with in pregfree.
856 refcount is first in both this and _reg_ac_data to allow a space
857 optimisation in Perl_regdupe. */
858 struct _reg_trie_data {
859 U32 refcount; /* number of times this trie is referenced */
860 U32 lasttrans; /* last valid transition element */
861 U16 *charmap; /* byte to charid lookup array */
862 reg_trie_state *states; /* state data */
863 reg_trie_trans *trans; /* array of transition elements */
864 char *bitmap; /* stclass bitmap */
865 U16 *jump; /* optional 1 indexed array of offsets before tail
866 for the node following a given word. */
867 reg_trie_wordinfo *wordinfo; /* array of info per word */
868 U16 uniquecharcount; /* unique chars in trie (width of trans table) */
869 U32 startstate; /* initial state - used for common prefix optimisation */
870 STRLEN minlen; /* minimum length of words in trie - build/opt only? */
871 STRLEN maxlen; /* maximum length of words in trie - build/opt only? */
872 U32 prefixlen; /* #chars in common prefix */
873 U32 statecount; /* Build only - number of states in the states array
874 (including the unused zero state) */
875 U32 wordcount; /* Build only */
877 STRLEN charcount; /* Build only */
880 /* There is one (3 under DEBUGGING) pointers that logically belong in this
881 structure, but are held outside as they need duplication on thread cloning,
882 whereas the rest of the structure can be read only:
883 HV *widecharmap; code points > 255 to charid
885 AV *words; Array of words contained in trie, for dumping
886 AV *revcharmap; Map of each charid back to its character representation
890 #define TRIE_WORDS_OFFSET 2
892 typedef struct _reg_trie_data reg_trie_data;
894 /* refcount is first in both this and _reg_trie_data to allow a space
895 optimisation in Perl_regdupe. */
896 struct _reg_ac_data {
900 reg_trie_state *states;
902 typedef struct _reg_ac_data reg_ac_data;
904 /* ANY_BIT doesn't use the structure, so we can borrow it here.
905 This is simpler than refactoring all of it as wed end up with
906 three different sets... */
908 #define TRIE_BITMAP(p) (((reg_trie_data *)(p))->bitmap)
909 #define TRIE_BITMAP_BYTE(p, c) BITMAP_BYTE(TRIE_BITMAP(p), c)
910 #define TRIE_BITMAP_SET(p, c) (TRIE_BITMAP_BYTE(p, c) |= ANYOF_BIT((U8)c))
911 #define TRIE_BITMAP_CLEAR(p,c) (TRIE_BITMAP_BYTE(p, c) &= ~ANYOF_BIT((U8)c))
912 #define TRIE_BITMAP_TEST(p, c) (TRIE_BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
914 #define IS_ANYOF_TRIE(op) ((op)==TRIEC || (op)==AHOCORASICKC)
915 #define IS_TRIE_AC(op) ((op)>=AHOCORASICK)
918 #define BITMAP_BYTE(p, c) (( (U8*) p) [ ( ( (UV) (c)) >> 3) ] )
919 #define BITMAP_TEST(p, c) (BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
921 /* these defines assume uniquecharcount is the correct variable, and state may be evaluated twice */
922 #define TRIE_NODENUM(state) (((state)-1)/(trie->uniquecharcount)+1)
923 #define SAFE_TRIE_NODENUM(state) ((state) ? (((state)-1)/(trie->uniquecharcount)+1) : (state))
924 #define TRIE_NODEIDX(state) ((state) ? (((state)-1)*(trie->uniquecharcount)+1) : (state))
927 #define TRIE_CHARCOUNT(trie) ((trie)->charcount)
929 #define TRIE_CHARCOUNT(trie) (trie_charcount)
932 #define RE_TRIE_MAXBUF_INIT 65536
933 #define RE_TRIE_MAXBUF_NAME "\022E_TRIE_MAXBUF"
934 #define RE_DEBUG_FLAGS "\022E_DEBUG_FLAGS"
938 RE_DEBUG_FLAGS is used to control what debug output is emitted
939 its divided into three groups of options, some of which interact.
940 The three groups are: Compile, Execute, Extra. There is room for a
941 further group, as currently only the low three bytes are used.
962 If you modify any of these make sure you make corresponding changes to
963 re.pm, especially to the documentation.
969 #define RE_DEBUG_COMPILE_MASK 0x0000FF
970 #define RE_DEBUG_COMPILE_PARSE 0x000001
971 #define RE_DEBUG_COMPILE_OPTIMISE 0x000002
972 #define RE_DEBUG_COMPILE_TRIE 0x000004
973 #define RE_DEBUG_COMPILE_DUMP 0x000008
974 #define RE_DEBUG_COMPILE_FLAGS 0x000010
975 #define RE_DEBUG_COMPILE_TEST 0x000020
978 #define RE_DEBUG_EXECUTE_MASK 0x00FF00
979 #define RE_DEBUG_EXECUTE_INTUIT 0x000100
980 #define RE_DEBUG_EXECUTE_MATCH 0x000200
981 #define RE_DEBUG_EXECUTE_TRIE 0x000400
984 #define RE_DEBUG_EXTRA_MASK 0xFF0000
985 #define RE_DEBUG_EXTRA_TRIE 0x010000
986 #define RE_DEBUG_EXTRA_OFFSETS 0x020000
987 #define RE_DEBUG_EXTRA_OFFDEBUG 0x040000
988 #define RE_DEBUG_EXTRA_STATE 0x080000
989 #define RE_DEBUG_EXTRA_OPTIMISE 0x100000
990 #define RE_DEBUG_EXTRA_BUFFERS 0x400000
991 #define RE_DEBUG_EXTRA_GPOS 0x800000
993 #define RE_DEBUG_EXTRA_STACK 0x280000
995 #define RE_DEBUG_FLAG(x) (re_debug_flags & x)
997 #define DEBUG_COMPILE_r(x) DEBUG_r( \
998 if (re_debug_flags & RE_DEBUG_COMPILE_MASK) x )
999 #define DEBUG_PARSE_r(x) DEBUG_r( \
1000 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x )
1001 #define DEBUG_OPTIMISE_r(x) DEBUG_r( \
1002 if (re_debug_flags & RE_DEBUG_COMPILE_OPTIMISE) x )
1003 #define DEBUG_PARSE_r(x) DEBUG_r( \
1004 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x )
1005 #define DEBUG_DUMP_r(x) DEBUG_r( \
1006 if (re_debug_flags & RE_DEBUG_COMPILE_DUMP) x )
1007 #define DEBUG_TRIE_COMPILE_r(x) DEBUG_r( \
1008 if (re_debug_flags & RE_DEBUG_COMPILE_TRIE) x )
1009 #define DEBUG_FLAGS_r(x) DEBUG_r( \
1010 if (re_debug_flags & RE_DEBUG_COMPILE_FLAGS) x )
1011 #define DEBUG_TEST_r(x) DEBUG_r( \
1012 if (re_debug_flags & RE_DEBUG_COMPILE_TEST) x )
1014 #define DEBUG_EXECUTE_r(x) DEBUG_r( \
1015 if (re_debug_flags & RE_DEBUG_EXECUTE_MASK) x )
1016 #define DEBUG_INTUIT_r(x) DEBUG_r( \
1017 if (re_debug_flags & RE_DEBUG_EXECUTE_INTUIT) x )
1018 #define DEBUG_MATCH_r(x) DEBUG_r( \
1019 if (re_debug_flags & RE_DEBUG_EXECUTE_MATCH) x )
1020 #define DEBUG_TRIE_EXECUTE_r(x) DEBUG_r( \
1021 if (re_debug_flags & RE_DEBUG_EXECUTE_TRIE) x )
1024 #define DEBUG_EXTRA_r(x) DEBUG_r( \
1025 if (re_debug_flags & RE_DEBUG_EXTRA_MASK) x )
1026 #define DEBUG_OFFSETS_r(x) DEBUG_r( \
1027 if (re_debug_flags & RE_DEBUG_EXTRA_OFFSETS) x )
1028 #define DEBUG_STATE_r(x) DEBUG_r( \
1029 if (re_debug_flags & RE_DEBUG_EXTRA_STATE) x )
1030 #define DEBUG_STACK_r(x) DEBUG_r( \
1031 if (re_debug_flags & RE_DEBUG_EXTRA_STACK) x )
1032 #define DEBUG_BUFFERS_r(x) DEBUG_r( \
1033 if (re_debug_flags & RE_DEBUG_EXTRA_BUFFERS) x )
1035 #define DEBUG_OPTIMISE_MORE_r(x) DEBUG_r( \
1036 if ((RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE) == \
1037 (re_debug_flags & (RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE)) ) x )
1038 #define MJD_OFFSET_DEBUG(x) DEBUG_r( \
1039 if (re_debug_flags & RE_DEBUG_EXTRA_OFFDEBUG) \
1040 Perl_warn_nocontext x )
1041 #define DEBUG_TRIE_COMPILE_MORE_r(x) DEBUG_TRIE_COMPILE_r( \
1042 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x )
1043 #define DEBUG_TRIE_EXECUTE_MORE_r(x) DEBUG_TRIE_EXECUTE_r( \
1044 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x )
1046 #define DEBUG_TRIE_r(x) DEBUG_r( \
1047 if (re_debug_flags & (RE_DEBUG_COMPILE_TRIE \
1048 | RE_DEBUG_EXECUTE_TRIE )) x )
1049 #define DEBUG_GPOS_r(x) DEBUG_r( \
1050 if (re_debug_flags & RE_DEBUG_EXTRA_GPOS) x )
1052 /* initialization */
1053 /* get_sv() can return NULL during global destruction. */
1054 #define GET_RE_DEBUG_FLAGS DEBUG_r({ \
1055 SV * re_debug_flags_sv = NULL; \
1056 re_debug_flags_sv = PL_curcop ? get_sv(RE_DEBUG_FLAGS, GV_ADD) : NULL; \
1057 if (re_debug_flags_sv) { \
1058 if (!SvIOK(re_debug_flags_sv)) \
1059 sv_setuv(re_debug_flags_sv, RE_DEBUG_COMPILE_DUMP | RE_DEBUG_EXECUTE_MASK ); \
1060 re_debug_flags=SvIV(re_debug_flags_sv); \
1066 #define GET_RE_DEBUG_FLAGS_DECL volatile IV re_debug_flags = 0; \
1067 PERL_UNUSED_VAR(re_debug_flags); GET_RE_DEBUG_FLAGS;
1069 #define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) \
1070 const char * const rpv = \
1071 pv_pretty((dsv), (pv), (l), (m), \
1072 PL_colors[(c1)],PL_colors[(c2)], \
1073 PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ); \
1074 const int rlen = SvCUR(dsv)
1076 /* This is currently unsed in the core */
1077 #define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m) \
1078 const char * const rpv = \
1079 pv_pretty((dsv), (SvPV_nolen_const(sv)), (SvCUR(sv)), (m), \
1080 PL_colors[(c1)],PL_colors[(c2)], \
1081 PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) )
1083 #define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) \
1084 const char * const rpv = \
1085 pv_pretty((dsv), (pv), (l), (m), \
1086 PL_colors[0], PL_colors[1], \
1087 ( PERL_PV_PRETTY_QUOTE | PERL_PV_ESCAPE_RE | PERL_PV_ESCAPE_NONASCII | PERL_PV_PRETTY_ELLIPSES | \
1088 ((isuni) ? PERL_PV_ESCAPE_UNI : 0)) \
1091 #define RE_SV_DUMPLEN(ItEm) (SvCUR(ItEm) - (SvTAIL(ItEm)!=0))
1092 #define RE_SV_TAIL(ItEm) (SvTAIL(ItEm) ? "$" : "")
1094 #else /* if not DEBUGGING */
1096 #define GET_RE_DEBUG_FLAGS_DECL
1097 #define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2)
1098 #define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m)
1099 #define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m)
1100 #define RE_SV_DUMPLEN(ItEm)
1101 #define RE_SV_TAIL(ItEm)
1103 #endif /* DEBUG RELATED DEFINES */
1105 #define FIRST_NON_ASCII_DECIMAL_DIGIT 0x660 /* ARABIC_INDIC_DIGIT_ZERO */
1108 TRADITIONAL_BOUND = _CC_WORDCHAR,
1116 * ex: set ts=8 sts=4 sw=4 et: