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.
11 #if ! defined(PERL_REGCOMP_H_) && ( defined(PERL_CORE) \
12 || defined(PERL_EXT_RE_BUILD))
13 #define PERL_REGCOMP_H_
15 #include "regcharclass.h"
17 /* Convert branch sequences to more efficient trie ops? */
18 #define PERL_ENABLE_TRIE_OPTIMISATION 1
20 /* Be really aggressive about optimising patterns with trie sequences? */
21 #define PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION 1
23 /* Should the optimiser take positive assertions into account? */
24 #define PERL_ENABLE_POSITIVE_ASSERTION_STUDY 0
26 /* Not for production use: */
27 #define PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS 0
30 * Structure for regexp "program". This is essentially a linear encoding
31 * of a nondeterministic finite-state machine (aka syntax charts or
32 * "railroad normal form" in parsing technology). Each node is an opcode
33 * plus a "next" pointer, possibly plus an operand. "Next" pointers of
34 * all nodes except BRANCH implement concatenation; a "next" pointer with
35 * a BRANCH on both ends of it is connecting two alternatives. (Here we
36 * have one of the subtle syntax dependencies: an individual BRANCH (as
37 * opposed to a collection of them) is never concatenated with anything
38 * because of operator precedence.) The operand of some types of node is
39 * a literal string; for others, it is a node leading into a sub-FSM. In
40 * particular, the operand of a BRANCH node is the first node of the branch.
41 * (NB this is *not* a tree structure: the tail of the branch connects
42 * to the thing following the set of BRANCHes.) The opcodes are defined
43 * in regnodes.h which is generated from regcomp.sym by regcomp.pl.
47 * A node is one char of opcode followed by two chars of "next" pointer.
48 * "Next" pointers are stored as two 8-bit pieces, high order first. The
49 * value is a positive offset from the opcode of the node containing it.
50 * An operand, if any, simply follows the node. (Note that much of the
51 * code generation knows about this implicit relationship.)
53 * Using two bytes for the "next" pointer is vast overkill for most things,
54 * but allows patterns to get big without disasters.
56 * [The "next" pointer is always aligned on an even
57 * boundary, and reads the offset directly as a short.]
60 /* This is the stuff that used to live in regexp.h that was truly
61 private to the engine itself. It now lives here. */
63 typedef struct regexp_internal {
64 regnode *regstclass; /* Optional startclass as identified or constructed
66 struct reg_data *data; /* Additional miscellaneous data used by the program.
67 Used to make it easier to clone and free arbitrary
68 data that the regops need. Often the ARG field of
69 a regop is an index into this structure. NOTE the
70 0th element of this structure is NEVER used and is
71 strictly reserved for internal purposes. */
72 struct reg_code_blocks *code_blocks;/* positions of literal (?{}) */
73 U32 proglen; /* size of the compiled program in regnodes */
74 U32 name_list_idx; /* Optional data index of an array of paren names,
75 only valid when RXp_PAREN_NAMES(prog) is true,
76 0 means "no value" like any other index into the
78 regnode program[1]; /* Unwarranted chumminess with compiler. */
81 #define RXi_SET(x,y) (x)->pprivate = (void*)(y)
82 #define RXi_GET(x) ((regexp_internal *)((x)->pprivate))
83 #define RXi_GET_DECL(r,ri) regexp_internal *ri = RXi_GET(r)
85 * Flags stored in regexp->intflags
86 * These are used only internally to the regexp engine
88 * See regexp.h for flags used externally to the regexp engine
90 #define RXp_INTFLAGS(rx) ((rx)->intflags)
91 #define RX_INTFLAGS(prog) RXp_INTFLAGS(ReANY(prog))
93 #define PREGf_SKIP 0x00000001
94 #define PREGf_IMPLICIT 0x00000002 /* Converted .* to ^.* */
95 #define PREGf_NAUGHTY 0x00000004 /* how exponential is this pattern? */
96 #define PREGf_VERBARG_SEEN 0x00000008
97 #define PREGf_CUTGROUP_SEEN 0x00000010
98 #define PREGf_USE_RE_EVAL 0x00000020 /* compiled with "use re 'eval'" */
99 /* these used to be extflags, but are now intflags */
100 #define PREGf_NOSCAN 0x00000040
102 #define PREGf_GPOS_SEEN 0x00000100
103 #define PREGf_GPOS_FLOAT 0x00000200
105 #define PREGf_ANCH_MBOL 0x00000400
106 #define PREGf_ANCH_SBOL 0x00000800
107 #define PREGf_ANCH_GPOS 0x00001000
108 #define PREGf_RECURSE_SEEN 0x00002000
111 ( PREGf_ANCH_SBOL | PREGf_ANCH_GPOS | PREGf_ANCH_MBOL )
113 /* this is where the old regcomp.h started */
115 struct regnode_string {
122 struct regnode_lstring { /* Constructed this way to keep the string aligned. */
126 U32 str_len; /* Only 18 bits allowed before would overflow 'next_off' */
130 struct regnode_anyofhs { /* Constructed this way to keep the string aligned. */
134 U32 arg1; /* set by set_ANYOF_arg() */
138 /* Argument bearing node - workhorse,
139 arg1 is often for the data field */
147 /* Node whose argument is 'SV *'. This needs to be used very carefully in
148 * situations where pointers won't become invalid because of, say re-mallocs */
156 /* Similar to a regnode_1 but with an extra signed argument */
165 /* 'Two field' -- Two 16 bit unsigned args */
174 #define ANYOF_BITMAP_SIZE (NUM_ANYOF_CODE_POINTS / 8) /* 8 bits/Byte */
176 /* Note that these form structs which are supersets of the next smaller one, by
177 * appending fields. Alignment problems can occur if one of those optional
178 * fields requires stricter alignment than the base struct. And formal
179 * parameters that can really be two or more of the structs should be
180 * declared as the smallest one it could be. See commit message for
181 * 7dcac5f6a5195002b55c935ee1d67f67e1df280b. Regnode allocation is done
182 * without regard to alignment, and changing it to would also require changing
183 * the code that inserts and deletes regnodes. The basic single-argument
184 * regnode has a U32, which is what reganode() allocates as a unit. Therefore
185 * no field can require stricter alignment than U32. */
187 /* also used by trie */
188 struct regnode_charclass {
192 U32 arg1; /* set by set_ANYOF_arg() */
193 char bitmap[ANYOF_BITMAP_SIZE]; /* only compile-time */
196 /* has runtime (locale) \d, \w, ..., [:posix:] classes */
197 struct regnode_charclass_posixl {
198 U8 flags; /* ANYOF_MATCHES_POSIXL bit must go here */
202 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time ... */
203 U32 classflags; /* and run-time */
206 /* A synthetic start class (SSC); is a regnode_charclass_posixl_fold, plus an
207 * extra SV*, used only during its construction and which is not used by
208 * regexec.c. Note that the 'next_off' field is unused, as the SSC stands
209 * alone, so there is never a next node. Also, there is no alignment issue,
210 * because these are declared or allocated as a complete unit so the compiler
211 * takes care of alignment. This is unlike the other regnodes which are
212 * allocated in terms of multiples of a single-argument regnode. SSC nodes can
213 * have a pointer field because there is no alignment issue, and because it is
214 * set to NULL after construction, before any cloning of the pattern */
216 U8 flags; /* ANYOF_MATCHES_POSIXL bit must go here */
220 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time ... */
221 U32 classflags; /* ... and run-time */
223 /* Auxiliary, only used during construction; NULL afterwards: list of code
228 /* We take advantage of 'next_off' not otherwise being used in the SSC by
229 * actually using it: by setting it to 1. This allows us to test and
230 * distinguish between an SSC and other ANYOF node types, as 'next_off' cannot
231 * otherwise be 1, because it is the offset to the next regnode expressed in
232 * units of regnodes. Since an ANYOF node contains extra fields, it adds up
233 * to 12 regnode units on 32-bit systems, (hence the minimum this can be (if
234 * not 0) is 11 there. Even if things get tightly packed on a 64-bit system,
235 * it still would be more than 1. */
236 #define set_ANYOF_SYNTHETIC(n) STMT_START{ OP(n) = ANYOF; \
239 #define is_ANYOF_SYNTHETIC(n) (PL_regkind[OP(n)] == ANYOF && NEXT_OFF(n) == 1)
241 /* XXX fix this description.
242 Impose a limit of REG_INFTY on various pattern matching operations
243 to limit stack growth and to avoid "infinite" recursions.
245 /* The default size for REG_INFTY is U16_MAX, which is the same as
246 USHORT_MAX (see perl.h). Unfortunately U16 isn't necessarily 16 bits
247 (see handy.h). On the Cray C90, sizeof(short)==4 and hence U16_MAX is
248 ((1<<32)-1), while on the Cray T90, sizeof(short)==8 and U16_MAX is
249 ((1<<64)-1). To limit stack growth to reasonable sizes, supply a
251 --Andy Dougherty 11 June 1998
255 # define REG_INFTY nBIT_UMAX(16)
260 # define REG_INFTY U16_MAX
263 #define ARG_VALUE(arg) (arg)
264 #define ARG__SET(arg,val) ((arg) = (val))
270 #define ARG(p) ARG_VALUE(ARG_LOC(p))
271 #define ARGp(p) ARG_VALUE(ARGp_LOC(p))
272 #define ARG1(p) ARG_VALUE(ARG1_LOC(p))
273 #define ARG2(p) ARG_VALUE(ARG2_LOC(p))
274 #define ARG2L(p) ARG_VALUE(ARG2L_LOC(p))
276 #define ARG_SET(p, val) ARG__SET(ARG_LOC(p), (val))
277 #define ARGp_SET(p, val) ARG__SET(ARGp_LOC(p), (val))
278 #define ARG1_SET(p, val) ARG__SET(ARG1_LOC(p), (val))
279 #define ARG2_SET(p, val) ARG__SET(ARG2_LOC(p), (val))
280 #define ARG2L_SET(p, val) ARG__SET(ARG2L_LOC(p), (val))
285 #define NEXT_OFF(p) ((p)->next_off)
286 #define NODE_ALIGN(node)
287 /* the following define was set to 0xde in 075abff3
288 * as part of some linting logic. I have set it to 0
289 * as otherwise in every place where we /might/ set flags
290 * we have to set it 0 explicitly, which duplicates
291 * assignments and IMO adds an unacceptable level of
292 * surprise to working in the regex engine. If this
293 * is changed from 0 then at the very least make sure
294 * that SBOL for /^/ sets the flags to 0 explicitly.
296 #define NODE_ALIGN_FILL(node) ((node)->flags = 0)
298 #define SIZE_ALIGN NODE_ALIGN
304 #define OP(p) ((p)->type)
305 #define FLAGS(p) ((p)->flags) /* Caution: Doesn't apply to all \
306 regnode types. For some, it's the \
307 character set of the regnode */
308 #define STR_LENs(p) (__ASSERT_(OP(p) != LEXACT && OP(p) != LEXACT_REQ8) \
309 ((struct regnode_string *)p)->str_len)
310 #define STRINGs(p) (__ASSERT_(OP(p) != LEXACT && OP(p) != LEXACT_REQ8) \
311 ((struct regnode_string *)p)->string)
312 #define OPERANDs(p) STRINGs(p)
314 /* Long strings. Currently limited to length 18 bits, which handles a 262000
315 * byte string. The limiting factor is the 16 bit 'next_off' field, which
316 * points to the next regnode, so the furthest away it can be is 2**16. On
317 * most architectures, regnodes are 2**2 bytes long, so that yields 2**18
318 * bytes. Should a longer string be desired, we could increase it to 26 bits
319 * fairly easily, by changing this node to have longj type which causes the ARG
320 * field to be used for the link to the next regnode (although code would have
321 * to be changed to account for this), and then use a combination of the flags
322 * and next_off fields for the length. To get 34 bit length, also change the
323 * node to be an ARG2L, using the second 32 bit field for the length, and not
324 * using the flags nor next_off fields at all. One could have an llstring node
325 * and even an lllstring type. */
326 #define STR_LENl(p) (__ASSERT_(OP(p) == LEXACT || OP(p) == LEXACT_REQ8) \
327 (((struct regnode_lstring *)p)->str_len))
328 #define STRINGl(p) (__ASSERT_(OP(p) == LEXACT || OP(p) == LEXACT_REQ8) \
329 (((struct regnode_lstring *)p)->string))
330 #define OPERANDl(p) STRINGl(p)
332 #define STR_LEN(p) ((OP(p) == LEXACT || OP(p) == LEXACT_REQ8) \
333 ? STR_LENl(p) : STR_LENs(p))
334 #define STRING(p) ((OP(p) == LEXACT || OP(p) == LEXACT_REQ8) \
335 ? STRINGl(p) : STRINGs(p))
336 #define OPERAND(p) STRING(p)
338 /* The number of (smallest) regnode equivalents that a string of length l bytes
340 #define STR_SZ(l) (((l) + sizeof(regnode) - 1) / sizeof(regnode))
342 /* The number of (smallest) regnode equivalents that the EXACTISH node 'p'
344 #define NODE_SZ_STR(p) (STR_SZ(STR_LEN(p)) + 1 + regarglen[(p)->type])
346 #define setSTR_LEN(p,v) \
348 if (OP(p) == LEXACT || OP(p) == LEXACT_REQ8) \
349 ((struct regnode_lstring *)(p))->str_len = (v); \
351 ((struct regnode_string *)(p))->str_len = (v); \
354 #define ANYOFR_BASE_BITS 20
355 #define ANYOFRbase(p) (ARG(p) & nBIT_MASK(ANYOFR_BASE_BITS))
356 #define ANYOFRdelta(p) (ARG(p) >> ANYOFR_BASE_BITS)
363 #define NODE_ALIGN(node)
364 #define ARG_LOC(p) (((struct regnode_1 *)p)->arg1)
365 #define ARGp_LOC(p) (((struct regnode_p *)p)->arg1)
366 #define ARG1_LOC(p) (((struct regnode_2 *)p)->arg1)
367 #define ARG2_LOC(p) (((struct regnode_2 *)p)->arg2)
368 #define ARG2L_LOC(p) (((struct regnode_2L *)p)->arg2)
370 #define NODE_STEP_REGNODE 1 /* sizeof(regnode)/sizeof(regnode) */
371 #define EXTRA_STEP_2ARGS EXTRA_SIZE(struct regnode_2)
373 #define NEXTOPER(p) ((p) + NODE_STEP_REGNODE)
374 #define PREVOPER(p) ((p) - NODE_STEP_REGNODE)
376 #define FILL_NODE(offset, op) \
378 OP(REGNODE_p(offset)) = op; \
379 NEXT_OFF(REGNODE_p(offset)) = 0; \
381 #define FILL_ADVANCE_NODE(offset, op) \
383 FILL_NODE(offset, op); \
386 #define FILL_ADVANCE_NODE_ARG(offset, op, arg) \
388 ARG_SET(REGNODE_p(offset), arg); \
389 FILL_ADVANCE_NODE(offset, op); \
390 /* This is used generically for other operations \
391 * that have a longer argument */ \
392 (offset) += regarglen[op]; \
394 #define FILL_ADVANCE_NODE_ARGp(offset, op, arg) \
396 ARGp_SET(REGNODE_p(offset), arg); \
397 FILL_ADVANCE_NODE(offset, op); \
398 (offset) += regarglen[op]; \
400 #define FILL_ADVANCE_NODE_2L_ARG(offset, op, arg1, arg2) \
402 ARG_SET(REGNODE_p(offset), arg1); \
403 ARG2L_SET(REGNODE_p(offset), arg2); \
404 FILL_ADVANCE_NODE(offset, op); \
408 #define REG_MAGIC 0234
410 /* An ANYOF node is basically a bitmap with the index being a code point. If
411 * the bit for that code point is 1, the code point matches; if 0, it doesn't
412 * match (complemented if inverted). There is an additional mechanism to deal
413 * with cases where the bitmap is insufficient in and of itself. This #define
414 * indicates if the bitmap does fully represent what this ANYOF node can match.
415 * The ARG is set to this special value (since 0, 1, ... are legal, but will
416 * never reach this high). */
417 #define ANYOF_ONLY_HAS_BITMAP ((U32) -1)
419 /* When the bitmap isn't completely sufficient for handling the ANYOF node,
420 * flags (in node->flags of the ANYOF node) get set to indicate this. These
421 * are perennially in short supply. Beyond several cases where warnings need
422 * to be raised under certain circumstances, currently, there are six cases
423 * where the bitmap alone isn't sufficient. We could use six flags to
424 * represent the 6 cases, but to save flags bits, we play some games. The
427 * 1) The bitmap has a compiled-in very finite size. So something else needs
428 * to be used to specify if a code point that is too large for the bitmap
429 * actually matches. The mechanism currently is an inversion
430 * list. ANYOF_ONLY_HAS_BITMAP, described above, being TRUE indicates
431 * there are no matches of too-large code points. But if it is FALSE,
432 * then almost certainly there are matches too large for the bitmap. (The
433 * other cases, described below, either imply this one or are extremely
434 * rare in practice.) So we can just assume that a too-large code point
435 * will need something beyond the bitmap if ANYOF_ONLY_HAS_BITMAP is
436 * FALSE, instead of having a separate flag for this.
437 * 2) A subset of item 1) is if all possible code points outside the bitmap
438 * match. This is a common occurrence when the class is complemented,
439 * like /[^ij]/. Therefore a bit is reserved to indicate this,
440 * rather than having an inversion list created,
441 * ANYOF_MATCHES_ALL_ABOVE_BITMAP.
442 * 3) Under /d rules, it can happen that code points that are in the upper
443 * latin1 range (\x80-\xFF or their equivalents on EBCDIC platforms) match
444 * only if the runtime target string being matched against is UTF-8. For
445 * example /[\w[:punct:]]/d. This happens only for posix classes (with a
446 * couple of exceptions, like \d where it doesn't happen), and all such
447 * ones also have above-bitmap matches. Thus, 3) implies 1) as well.
448 * Note that /d rules are no longer encouraged; 'use 5.14' or higher
449 * deselects them. But a flag is required so that they can be properly
450 * handled. But it can be a shared flag: see 5) below.
451 * 4) Also under /d rules, something like /[\Wfoo]/ will match everything in
452 * the \x80-\xFF range, unless the string being matched against is UTF-8.
453 * An inversion list could be created for this case, but this is
454 * relatively common, and it turns out that it's all or nothing: if any
455 * one of these code points matches, they all do. Hence a single bit
456 * suffices. We use a shared flag that doesn't take up space by itself:
457 * ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER. This
458 * also implies 1), with one exception: [:^cntrl:].
459 * 5) A user-defined \p{} property may not have been defined by the time the
460 * regex is compiled. In this case, we don't know until runtime what it
461 * will match, so we have to assume it could match anything, including
462 * code points that ordinarily would be in the bitmap. A flag bit is
463 * necessary to indicate this, though it can be shared with the item 3)
464 * flag, as that only occurs under /d, and this only occurs under non-d.
465 * This case is quite uncommon in the field, and the /(?[ ...])/ construct
466 * is a better way to accomplish what this feature does. This case also
468 * ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP
469 * is the shared flag.
470 * 6) /[foo]/il may have folds that are only valid if the runtime locale is a
471 * UTF-8 one. These are quite rare, so it would be good to avoid the
472 * expense of looking for them. But /l matching is slow anyway, and we've
473 * traditionally not worried too much about its performance. And this
474 * condition requires the ANYOFL_FOLD flag to be set, so testing for
475 * that flag would be sufficient to rule out most cases of this. So it is
476 * unclear if this should have a flag or not. But, this flag can be
477 * shared with another, so it doesn't occupy extra space.
479 * At the moment, there is one spare bit, but this could be increased by
482 * If just one more bit is needed, as of this writing it seems to khw that the
483 * best choice would be to make ANYOF_MATCHES_ALL_ABOVE_BITMAP not a flag, but
486 * #define ANYOF_MATCHES_ALL_ABOVE_BITMAP ((U32) -2)
488 * and access it through the ARG like ANYOF_ONLY_HAS_BITMAP is. This flag is
489 * used by all ANYOF node types, and it could be used to avoid calling the
490 * handler function, as the macro REGINCLASS in regexec.c does now for other
493 * Another possibility is based on the fact that ANYOF_MATCHES_POSIXL is
494 * redundant with the node type ANYOFPOSIXL. That flag could be removed, but
495 * at the expense of extra code in regexec.c. The flag has been retained
496 * because it allows us to see if we need to call reginclass, or just use the
497 * bitmap in one test.
499 * If this is done, an extension would be to make all ANYOFL nodes contain the
500 * extra 32 bits that ANYOFPOSIXL ones do. The posix flags only occupy 30
501 * bits, so the ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD flags
502 * and ANYOFL_FOLD could be moved to that extra space, but it would mean extra
503 * instructions, as there are currently places in the code that assume those
506 * All told, 5 bits could be available for other uses if all of the above were
509 * Some flags are not used in synthetic start class (SSC) nodes, so could be
510 * shared should new flags be needed for SSCs, like SSC_MATCHES_EMPTY_STRING
513 /* If this is set, the result of the match should be complemented. regexec.c
514 * is expecting this to be in the low bit. Never in an SSC */
515 #define ANYOF_INVERT 0x01
517 /* For the SSC node only, which cannot be inverted, so is shared with that bit.
518 * This is used only during regex compilation. */
519 #define SSC_MATCHES_EMPTY_STRING ANYOF_INVERT
521 /* Set if this is a regnode_charclass_posixl vs a regnode_charclass. This
522 * is used for runtime \d, \w, [:posix:], ..., which are used only in locale
523 * and the optimizer's synthetic start class. Non-locale \d, etc are resolved
524 * at compile-time. Only set under /l; can be in SSC */
525 #define ANYOF_MATCHES_POSIXL 0x02
527 /* The fold is calculated and stored in the bitmap where possible at compile
528 * time. However under locale, the actual folding varies depending on
529 * what the locale is at the time of execution, so it has to be deferred until
530 * then. Only set under /l; never in an SSC */
531 #define ANYOFL_FOLD 0x04
533 /* Shared bit set only with ANYOFL and SSC nodes:
534 * If ANYOFL_FOLD is set, this flag indicates there are potential matches
535 * valid only if the locale is a UTF-8 one.
536 * If ANYOFL_FOLD is NOT set, this flag means to warn if the runtime locale
537 * isn't a UTF-8 one (and the generated node assumes a UTF-8 locale).
538 * None of INVERT, POSIXL,
539 * ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP
541 #define ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD 0x08
543 /* Convenience macros for teasing apart the meanings when reading the above bit
545 #define ANYOFL_SOME_FOLDS_ONLY_IN_UTF8_LOCALE(flags) \
546 ((flags & ( ANYOFL_FOLD /* Both bits are set */ \
547 |ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD)) \
549 |ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD))
551 #define ANYOFL_UTF8_LOCALE_REQD(flags) \
552 ((flags & ( ANYOFL_FOLD /* Only REQD bit is set */ \
553 |ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD)) \
554 == ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD)
556 /* Spare: Be sure to change ANYOF_FLAGS_ALL if this gets used 0x10 */
558 /* If set, the node matches every code point NUM_ANYOF_CODE_POINTS and above.
559 * Can be in an SSC */
560 #define ANYOF_MATCHES_ALL_ABOVE_BITMAP 0x20
563 * Under /d it means the ANYOFD node matches more things if the target
564 * string is encoded in UTF-8; any such things will be non-ASCII,
565 * characters that are < 256, and can be accessed via the inversion
567 * When not under /d, it means the ANYOF node contains a user-defined
568 * property that wasn't yet defined at the time the regex was compiled,
569 * and so must be looked up at runtime, by creating an inversion list.
570 * (These uses are mutually exclusive because a user-defined property is
571 * specified by \p{}, and \p{} implies /u which deselects /d). The long macro
572 * name is to make sure that you are cautioned about its shared nature. Only
573 * the non-/d meaning can be in an SSC */
574 #define ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP 0x40
577 * Under /d it means the ANYOFD node matches all non-ASCII Latin1
578 * characters when the target string is not in utf8.
579 * When not under /d, it means the ANYOF node should raise a warning if
580 * matching against an above-Unicode code point.
581 * (These uses are mutually exclusive because the warning requires a \p{}, and
582 * \p{} implies /u which deselects /d). An SSC node only has this bit set if
583 * what is meant is the warning. The long macro name is to make sure that you
584 * are cautioned about its shared nature */
585 #define ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER 0x80
587 #define ANYOF_FLAGS_ALL ((U8) ~0x10)
589 #define ANYOF_LOCALE_FLAGS (ANYOFL_FOLD | ANYOF_MATCHES_POSIXL)
591 /* These are the flags that apply to both regular ANYOF nodes and synthetic
592 * start class nodes during construction of the SSC. During finalization of
593 * the SSC, other of the flags may get added to it */
594 #define ANYOF_COMMON_FLAGS 0
596 /* Character classes for node->classflags of ANYOF */
597 /* Should be synchronized with a table in regprop() */
598 /* 2n should be the normal one, paired with its complement at 2n+1 */
600 #define ANYOF_ALPHA ((_CC_ALPHA) * 2)
601 #define ANYOF_NALPHA ((ANYOF_ALPHA) + 1)
602 #define ANYOF_ALPHANUMERIC ((_CC_ALPHANUMERIC) * 2) /* [[:alnum:]] isalnum(3), utf8::IsAlnum */
603 #define ANYOF_NALPHANUMERIC ((ANYOF_ALPHANUMERIC) + 1)
604 #define ANYOF_ASCII ((_CC_ASCII) * 2)
605 #define ANYOF_NASCII ((ANYOF_ASCII) + 1)
606 #define ANYOF_BLANK ((_CC_BLANK) * 2) /* GNU extension: space and tab: non-vertical space */
607 #define ANYOF_NBLANK ((ANYOF_BLANK) + 1)
608 #define ANYOF_CASED ((_CC_CASED) * 2) /* Pseudo class for [:lower:] or
609 [:upper:] under /i */
610 #define ANYOF_NCASED ((ANYOF_CASED) + 1)
611 #define ANYOF_CNTRL ((_CC_CNTRL) * 2)
612 #define ANYOF_NCNTRL ((ANYOF_CNTRL) + 1)
613 #define ANYOF_DIGIT ((_CC_DIGIT) * 2) /* \d */
614 #define ANYOF_NDIGIT ((ANYOF_DIGIT) + 1)
615 #define ANYOF_GRAPH ((_CC_GRAPH) * 2)
616 #define ANYOF_NGRAPH ((ANYOF_GRAPH) + 1)
617 #define ANYOF_LOWER ((_CC_LOWER) * 2)
618 #define ANYOF_NLOWER ((ANYOF_LOWER) + 1)
619 #define ANYOF_PRINT ((_CC_PRINT) * 2)
620 #define ANYOF_NPRINT ((ANYOF_PRINT) + 1)
621 #define ANYOF_PUNCT ((_CC_PUNCT) * 2)
622 #define ANYOF_NPUNCT ((ANYOF_PUNCT) + 1)
623 #define ANYOF_SPACE ((_CC_SPACE) * 2) /* \s */
624 #define ANYOF_NSPACE ((ANYOF_SPACE) + 1)
625 #define ANYOF_UPPER ((_CC_UPPER) * 2)
626 #define ANYOF_NUPPER ((ANYOF_UPPER) + 1)
627 #define ANYOF_WORDCHAR ((_CC_WORDCHAR) * 2) /* \w, PL_utf8_alnum, utf8::IsWord, ALNUM */
628 #define ANYOF_NWORDCHAR ((ANYOF_WORDCHAR) + 1)
629 #define ANYOF_XDIGIT ((_CC_XDIGIT) * 2)
630 #define ANYOF_NXDIGIT ((ANYOF_XDIGIT) + 1)
632 /* pseudo classes below this, not stored in the class bitmap, but used as flags
633 during compilation of char classes */
635 #define ANYOF_VERTWS ((_CC_VERTSPACE) * 2)
636 #define ANYOF_NVERTWS ((ANYOF_VERTWS)+1)
638 /* It is best if this is the last one, as all above it are stored as bits in a
639 * bitmap, and it isn't part of that bitmap */
640 #if _CC_VERTSPACE != _HIGHEST_REGCOMP_DOT_H_SYNC
641 # error Problem with handy.h _HIGHEST_REGCOMP_DOT_H_SYNC #define
644 #define ANYOF_POSIXL_MAX (ANYOF_VERTWS) /* So upper loop limit is written:
646 * Hence doesn't include VERTWS, as that
647 * is a pseudo class */
648 #define ANYOF_MAX ANYOF_POSIXL_MAX
650 #if (ANYOF_POSIXL_MAX > 32) /* Must fit in 32-bit word */
651 # error Problem with handy.h _CC_foo #defines
654 #define ANYOF_HORIZWS ((ANYOF_POSIXL_MAX)+2) /* = (ANYOF_NVERTWS + 1) */
655 #define ANYOF_NHORIZWS ((ANYOF_POSIXL_MAX)+3)
657 #define ANYOF_UNIPROP ((ANYOF_POSIXL_MAX)+4) /* Used to indicate a Unicode
658 property: \p{} or \P{} */
660 /* Backward source code compatibility. */
662 #define ANYOF_ALNUML ANYOF_ALNUM
663 #define ANYOF_NALNUML ANYOF_NALNUM
664 #define ANYOF_SPACEL ANYOF_SPACE
665 #define ANYOF_NSPACEL ANYOF_NSPACE
666 #define ANYOF_ALNUM ANYOF_WORDCHAR
667 #define ANYOF_NALNUM ANYOF_NWORDCHAR
669 /* Utility macros for the bitmap and classes of ANYOF */
671 #define ANYOF_FLAGS(p) ((p)->flags)
673 #define ANYOF_BIT(c) (1U << ((c) & 7))
675 #define ANYOF_POSIXL_BITMAP(p) (((regnode_charclass_posixl*) (p))->classflags)
677 #define POSIXL_SET(field, c) ((field) |= (1U << (c)))
678 #define ANYOF_POSIXL_SET(p, c) POSIXL_SET(ANYOF_POSIXL_BITMAP(p), (c))
680 #define POSIXL_CLEAR(field, c) ((field) &= ~ (1U <<(c)))
681 #define ANYOF_POSIXL_CLEAR(p, c) POSIXL_CLEAR(ANYOF_POSIXL_BITMAP(p), (c))
683 #define POSIXL_TEST(field, c) ((field) & (1U << (c)))
684 #define ANYOF_POSIXL_TEST(p, c) POSIXL_TEST(ANYOF_POSIXL_BITMAP(p), (c))
686 #define POSIXL_ZERO(field) STMT_START { (field) = 0; } STMT_END
687 #define ANYOF_POSIXL_ZERO(ret) POSIXL_ZERO(ANYOF_POSIXL_BITMAP(ret))
689 #define ANYOF_POSIXL_SET_TO_BITMAP(p, bits) \
690 STMT_START { ANYOF_POSIXL_BITMAP(p) = (bits); } STMT_END
692 /* Shifts a bit to get, eg. 0x4000_0000, then subtracts 1 to get 0x3FFF_FFFF */
693 #define ANYOF_POSIXL_SETALL(ret) \
695 ANYOF_POSIXL_BITMAP(ret) = nBIT_MASK(ANYOF_POSIXL_MAX); \
697 #define ANYOF_CLASS_SETALL(ret) ANYOF_POSIXL_SETALL(ret)
699 #define ANYOF_POSIXL_TEST_ANY_SET(p) \
700 ((ANYOF_FLAGS(p) & ANYOF_MATCHES_POSIXL) && ANYOF_POSIXL_BITMAP(p))
701 #define ANYOF_CLASS_TEST_ANY_SET(p) ANYOF_POSIXL_TEST_ANY_SET(p)
703 /* Since an SSC always has this field, we don't have to test for that; nor do
704 * we want to because the bit isn't set for SSC during its construction */
705 #define ANYOF_POSIXL_SSC_TEST_ANY_SET(p) \
706 cBOOL(((regnode_ssc*)(p))->classflags)
707 #define ANYOF_POSIXL_SSC_TEST_ALL_SET(p) /* Are all bits set? */ \
708 (((regnode_ssc*) (p))->classflags \
709 == nBIT_MASK(ANYOF_POSIXL_MAX))
711 #define ANYOF_POSIXL_TEST_ALL_SET(p) \
712 ((ANYOF_FLAGS(p) & ANYOF_MATCHES_POSIXL) \
713 && ANYOF_POSIXL_BITMAP(p) == nBIT_MASK(ANYOF_POSIXL_MAX))
715 #define ANYOF_POSIXL_OR(source, dest) STMT_START { (dest)->classflags |= (source)->classflags ; } STMT_END
716 #define ANYOF_CLASS_OR(source, dest) ANYOF_POSIXL_OR((source), (dest))
718 #define ANYOF_POSIXL_AND(source, dest) STMT_START { (dest)->classflags &= (source)->classflags ; } STMT_END
720 #define ANYOF_BITMAP_ZERO(ret) Zero(((regnode_charclass*)(ret))->bitmap, ANYOF_BITMAP_SIZE, char)
721 #define ANYOF_BITMAP(p) ((regnode_charclass*)(p))->bitmap
722 #define ANYOF_BITMAP_BYTE(p, c) BITMAP_BYTE(ANYOF_BITMAP(p), c)
723 #define ANYOF_BITMAP_SET(p, c) (ANYOF_BITMAP_BYTE(p, c) |= ANYOF_BIT(c))
724 #define ANYOF_BITMAP_CLEAR(p,c) (ANYOF_BITMAP_BYTE(p, c) &= ~ANYOF_BIT(c))
725 #define ANYOF_BITMAP_TEST(p, c) cBOOL(ANYOF_BITMAP_BYTE(p, c) & ANYOF_BIT(c))
727 #define ANYOF_BITMAP_SETALL(p) \
728 memset (ANYOF_BITMAP(p), 255, ANYOF_BITMAP_SIZE)
729 #define ANYOF_BITMAP_CLEARALL(p) \
730 Zero (ANYOF_BITMAP(p), ANYOF_BITMAP_SIZE)
733 * Utility definitions.
736 # define UCHARAT(p) ((int)*(const U8*)(p))
738 # define UCHARAT(p) ((int)*(p)&CHARMASK)
741 /* Number of regnode equivalents that 'guy' occupies beyond the size of the
742 * smallest regnode. */
743 #define EXTRA_SIZE(guy) ((sizeof(guy)-1)/sizeof(struct regnode))
745 #define REG_ZERO_LEN_SEEN 0x00000001
746 #define REG_LOOKBEHIND_SEEN 0x00000002
747 #define REG_GPOS_SEEN 0x00000004
749 #define REG_RECURSE_SEEN 0x00000020
750 #define REG_TOP_LEVEL_BRANCHES_SEEN 0x00000040
751 #define REG_VERBARG_SEEN 0x00000080
752 #define REG_CUTGROUP_SEEN 0x00000100
753 #define REG_RUN_ON_COMMENT_SEEN 0x00000200
754 #define REG_UNFOLDED_MULTI_SEEN 0x00000400
756 #define REG_UNBOUNDED_QUANTIFIER_SEEN 0x00001000
761 #ifdef PLUGGABLE_RE_EXTENSION
762 #include "re_nodes.h"
764 #include "regnodes.h"
767 #ifndef PLUGGABLE_RE_EXTENSION
769 EXTCONST regexp_engine PL_core_reg_engine;
771 EXTCONST regexp_engine PL_core_reg_engine = {
774 Perl_re_intuit_start,
775 Perl_re_intuit_string,
776 Perl_regfree_internal,
777 Perl_reg_numbered_buff_fetch,
778 Perl_reg_numbered_buff_store,
779 Perl_reg_numbered_buff_length,
781 Perl_reg_named_buff_iter,
783 #if defined(USE_ITHREADS)
784 Perl_regdupe_internal,
789 #endif /* PLUGGABLE_RE_EXTENSION */
795 /* .what is a character array with one character for each member of .data
796 * The character describes the function of the corresponding .data item:
797 * a - AV for paren_name_list under DEBUGGING
798 * f - start-class data for regstclass optimization
799 * l - start op for literal (?{EVAL}) item
800 * L - start op for literal (?{EVAL}) item, with separate CV (qr//)
801 * r - pointer to an embedded code-containing qr, e.g. /ab$qr/
802 * s - inversion list for Unicode-style character class, and the
803 * multicharacter strings resulting from casefolding the single-character
804 * entries in the character class
806 * u - trie struct's widecharmap (a HV, so can't share, must dup)
807 * also used for revcharmap and words under DEBUGGING
808 * T - aho-trie struct
809 * S - sv for named capture lookup
810 * 20010712 mjd@plover.com
811 * (Remember to update re_dup() and pregfree() if you add any items.)
819 /* Code in S_to_utf8_substr() and S_to_byte_substr() in regexec.c accesses
820 anchored* and float* via array indexes 0 and 1. */
821 #define anchored_substr substrs->data[0].substr
822 #define anchored_utf8 substrs->data[0].utf8_substr
823 #define anchored_offset substrs->data[0].min_offset
824 #define anchored_end_shift substrs->data[0].end_shift
826 #define float_substr substrs->data[1].substr
827 #define float_utf8 substrs->data[1].utf8_substr
828 #define float_min_offset substrs->data[1].min_offset
829 #define float_max_offset substrs->data[1].max_offset
830 #define float_end_shift substrs->data[1].end_shift
832 #define check_substr substrs->data[2].substr
833 #define check_utf8 substrs->data[2].utf8_substr
834 #define check_offset_min substrs->data[2].min_offset
835 #define check_offset_max substrs->data[2].max_offset
836 #define check_end_shift substrs->data[2].end_shift
838 #define RX_ANCHORED_SUBSTR(rx) (ReANY(rx)->anchored_substr)
839 #define RX_ANCHORED_UTF8(rx) (ReANY(rx)->anchored_utf8)
840 #define RX_FLOAT_SUBSTR(rx) (ReANY(rx)->float_substr)
841 #define RX_FLOAT_UTF8(rx) (ReANY(rx)->float_utf8)
843 /* trie related stuff */
845 /* a transition record for the state machine. the
846 check field determines which state "owns" the
847 transition. the char the transition is for is
848 determined by offset from the owning states base
849 field. the next field determines which state
850 is to be transitioned to if any.
852 struct _reg_trie_trans {
857 /* a transition list element for the list based representation */
858 struct _reg_trie_trans_list_elem {
862 typedef struct _reg_trie_trans_list_elem reg_trie_trans_le;
864 /* a state for compressed nodes. base is an offset
865 into an array of reg_trie_trans array. If wordnum is
866 nonzero the state is accepting. if base is zero then
867 the state has no children (and will be accepting)
869 struct _reg_trie_state {
873 reg_trie_trans_le* list;
877 /* info per word; indexed by wordnum */
879 U16 prev; /* previous word in acceptance chain; eg in
880 * zzz|abc|ab/ after matching the chars abc, the
881 * accepted word is #2, and the previous accepted
883 U32 len; /* how many chars long is this word? */
884 U32 accept; /* accept state for this word */
888 typedef struct _reg_trie_state reg_trie_state;
889 typedef struct _reg_trie_trans reg_trie_trans;
892 /* anything in here that needs to be freed later
893 should be dealt with in pregfree.
894 refcount is first in both this and _reg_ac_data to allow a space
895 optimisation in Perl_regdupe. */
896 struct _reg_trie_data {
897 U32 refcount; /* number of times this trie is referenced */
898 U32 lasttrans; /* last valid transition element */
899 U16 *charmap; /* byte to charid lookup array */
900 reg_trie_state *states; /* state data */
901 reg_trie_trans *trans; /* array of transition elements */
902 char *bitmap; /* stclass bitmap */
903 U16 *jump; /* optional 1 indexed array of offsets before tail
904 for the node following a given word. */
905 reg_trie_wordinfo *wordinfo; /* array of info per word */
906 U16 uniquecharcount; /* unique chars in trie (width of trans table) */
907 U32 startstate; /* initial state - used for common prefix optimisation */
908 STRLEN minlen; /* minimum length of words in trie - build/opt only? */
909 STRLEN maxlen; /* maximum length of words in trie - build/opt only? */
910 U32 prefixlen; /* #chars in common prefix */
911 U32 statecount; /* Build only - number of states in the states array
912 (including the unused zero state) */
913 U32 wordcount; /* Build only */
915 STRLEN charcount; /* Build only */
918 /* There is one (3 under DEBUGGING) pointers that logically belong in this
919 structure, but are held outside as they need duplication on thread cloning,
920 whereas the rest of the structure can be read only:
921 HV *widecharmap; code points > 255 to charid
923 AV *words; Array of words contained in trie, for dumping
924 AV *revcharmap; Map of each charid back to its character representation
928 #define TRIE_WORDS_OFFSET 2
930 typedef struct _reg_trie_data reg_trie_data;
932 /* refcount is first in both this and _reg_trie_data to allow a space
933 optimisation in Perl_regdupe. */
934 struct _reg_ac_data {
938 reg_trie_state *states;
940 typedef struct _reg_ac_data reg_ac_data;
942 /* ANY_BIT doesn't use the structure, so we can borrow it here.
943 This is simpler than refactoring all of it as wed end up with
944 three different sets... */
946 #define TRIE_BITMAP(p) (((reg_trie_data *)(p))->bitmap)
947 #define TRIE_BITMAP_BYTE(p, c) BITMAP_BYTE(TRIE_BITMAP(p), c)
948 #define TRIE_BITMAP_SET(p, c) (TRIE_BITMAP_BYTE(p, c) |= ANYOF_BIT((U8)c))
949 #define TRIE_BITMAP_CLEAR(p,c) (TRIE_BITMAP_BYTE(p, c) &= ~ANYOF_BIT((U8)c))
950 #define TRIE_BITMAP_TEST(p, c) (TRIE_BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
952 #define IS_ANYOF_TRIE(op) ((op)==TRIEC || (op)==AHOCORASICKC)
953 #define IS_TRIE_AC(op) ((op)>=AHOCORASICK)
956 #define BITMAP_BYTE(p, c) (( (U8*) p) [ ( ( (UV) (c)) >> 3) ] )
957 #define BITMAP_TEST(p, c) (BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
959 /* these defines assume uniquecharcount is the correct variable, and state may be evaluated twice */
960 #define TRIE_NODENUM(state) (((state)-1)/(trie->uniquecharcount)+1)
961 #define SAFE_TRIE_NODENUM(state) ((state) ? (((state)-1)/(trie->uniquecharcount)+1) : (state))
962 #define TRIE_NODEIDX(state) ((state) ? (((state)-1)*(trie->uniquecharcount)+1) : (state))
965 #define TRIE_CHARCOUNT(trie) ((trie)->charcount)
967 #define TRIE_CHARCOUNT(trie) (trie_charcount)
970 #define RE_TRIE_MAXBUF_INIT 65536
971 #define RE_TRIE_MAXBUF_NAME "\022E_TRIE_MAXBUF"
972 #define RE_DEBUG_FLAGS "\022E_DEBUG_FLAGS"
974 #define RE_COMPILE_RECURSION_INIT 1000
975 #define RE_COMPILE_RECURSION_LIMIT "\022E_COMPILE_RECURSION_LIMIT"
979 RE_DEBUG_FLAGS is used to control what debug output is emitted
980 its divided into three groups of options, some of which interact.
981 The three groups are: Compile, Execute, Extra. There is room for a
982 further group, as currently only the low three bytes are used.
1001 If you modify any of these make sure you make corresponding changes to
1002 re.pm, especially to the documentation.
1008 #define RE_DEBUG_COMPILE_MASK 0x0000FF
1009 #define RE_DEBUG_COMPILE_PARSE 0x000001
1010 #define RE_DEBUG_COMPILE_OPTIMISE 0x000002
1011 #define RE_DEBUG_COMPILE_TRIE 0x000004
1012 #define RE_DEBUG_COMPILE_DUMP 0x000008
1013 #define RE_DEBUG_COMPILE_FLAGS 0x000010
1014 #define RE_DEBUG_COMPILE_TEST 0x000020
1017 #define RE_DEBUG_EXECUTE_MASK 0x00FF00
1018 #define RE_DEBUG_EXECUTE_INTUIT 0x000100
1019 #define RE_DEBUG_EXECUTE_MATCH 0x000200
1020 #define RE_DEBUG_EXECUTE_TRIE 0x000400
1023 #define RE_DEBUG_EXTRA_MASK 0x3FF0000
1024 #define RE_DEBUG_EXTRA_TRIE 0x0010000
1025 #define RE_DEBUG_EXTRA_STATE 0x0080000
1026 #define RE_DEBUG_EXTRA_OPTIMISE 0x0100000
1027 #define RE_DEBUG_EXTRA_BUFFERS 0x0400000
1028 #define RE_DEBUG_EXTRA_GPOS 0x0800000
1029 #define RE_DEBUG_EXTRA_DUMP_PRE_OPTIMIZE 0x1000000
1030 #define RE_DEBUG_EXTRA_WILDCARD 0x2000000
1032 #define RE_DEBUG_EXTRA_STACK 0x0280000
1034 #define RE_DEBUG_FLAG(x) (re_debug_flags & (x))
1036 #define DEBUG_COMPILE_r(x) DEBUG_r( \
1037 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_COMPILE_MASK)) x )
1038 #define DEBUG_PARSE_r(x) DEBUG_r( \
1039 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_COMPILE_PARSE)) x )
1040 #define DEBUG_OPTIMISE_r(x) DEBUG_r( \
1041 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_COMPILE_OPTIMISE)) x )
1042 #define DEBUG_DUMP_r(x) DEBUG_r( \
1043 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_COMPILE_DUMP)) x )
1044 #define DEBUG_TRIE_COMPILE_r(x) DEBUG_r( \
1045 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_COMPILE_TRIE)) x )
1046 #define DEBUG_FLAGS_r(x) DEBUG_r( \
1047 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_COMPILE_FLAGS)) x )
1048 #define DEBUG_TEST_r(x) DEBUG_r( \
1049 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_COMPILE_TEST)) x )
1051 #define DEBUG_EXECUTE_r(x) DEBUG_r( \
1052 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXECUTE_MASK)) x )
1053 #define DEBUG_INTUIT_r(x) DEBUG_r( \
1054 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXECUTE_INTUIT)) x )
1055 #define DEBUG_MATCH_r(x) DEBUG_r( \
1056 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXECUTE_MATCH)) x )
1057 #define DEBUG_TRIE_EXECUTE_r(x) DEBUG_r( \
1058 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXECUTE_TRIE)) x )
1061 #define DEBUG_EXTRA_r(x) DEBUG_r( \
1062 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_MASK)) x )
1063 #define DEBUG_STATE_r(x) DEBUG_r( \
1064 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_STATE)) x )
1065 #define DEBUG_STACK_r(x) DEBUG_r( \
1066 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_STACK)) x )
1067 #define DEBUG_BUFFERS_r(x) DEBUG_r( \
1068 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_BUFFERS)) x )
1070 #define DEBUG_OPTIMISE_MORE_r(x) DEBUG_r( \
1071 if (DEBUG_v_TEST || ((RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE) == \
1072 RE_DEBUG_FLAG(RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE))) x )
1073 #define DEBUG_TRIE_COMPILE_MORE_r(x) DEBUG_TRIE_COMPILE_r( \
1074 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_TRIE)) x )
1075 #define DEBUG_TRIE_EXECUTE_MORE_r(x) DEBUG_TRIE_EXECUTE_r( \
1076 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_TRIE)) x )
1078 #define DEBUG_TRIE_r(x) DEBUG_r( \
1079 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_COMPILE_TRIE \
1080 | RE_DEBUG_EXECUTE_TRIE )) x )
1081 #define DEBUG_GPOS_r(x) DEBUG_r( \
1082 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_GPOS)) x )
1084 #define DEBUG_DUMP_PRE_OPTIMIZE_r(x) DEBUG_r( \
1085 if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_DUMP_PRE_OPTIMIZE)) x )
1087 /* initialization */
1088 /* Get the debug flags for code not in regcomp.c nor regexec.c. This doesn't
1089 * initialize the variable if it isn't already there, instead it just assumes
1090 * the flags are 0 */
1091 #define DECLARE_AND_GET_RE_DEBUG_FLAGS_NON_REGEX \
1092 volatile IV re_debug_flags = 0; PERL_UNUSED_VAR(re_debug_flags); \
1094 SV * re_debug_flags_sv = NULL; \
1095 /* get_sv() can return NULL during global destruction. */ \
1096 re_debug_flags_sv = PL_curcop ? get_sv(RE_DEBUG_FLAGS, GV_ADD) : NULL; \
1097 if (re_debug_flags_sv && SvIOK(re_debug_flags_sv)) \
1098 re_debug_flags=SvIV(re_debug_flags_sv); \
1104 /* For use in regcomp.c and regexec.c, Get the debug flags, and initialize to
1105 * the defaults if not done already */
1106 #define DECLARE_AND_GET_RE_DEBUG_FLAGS \
1107 volatile IV re_debug_flags = 0; PERL_UNUSED_VAR(re_debug_flags); \
1109 SV * re_debug_flags_sv = NULL; \
1110 /* get_sv() can return NULL during global destruction. */ \
1111 re_debug_flags_sv = PL_curcop ? get_sv(RE_DEBUG_FLAGS, GV_ADD) : NULL; \
1112 if (re_debug_flags_sv) { \
1113 if (!SvIOK(re_debug_flags_sv)) /* If doesnt exist set to default */\
1114 sv_setuv(re_debug_flags_sv, \
1115 /* These defaults should be kept in sync with re.pm */ \
1116 RE_DEBUG_COMPILE_DUMP | RE_DEBUG_EXECUTE_MASK ); \
1117 re_debug_flags=SvIV(re_debug_flags_sv); \
1121 #define isDEBUG_WILDCARD (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_WILDCARD))
1123 #define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) \
1124 const char * const rpv = \
1125 pv_pretty((dsv), (pv), (l), (m), \
1126 PL_colors[(c1)],PL_colors[(c2)], \
1127 PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ); \
1128 const int rlen = SvCUR(dsv)
1130 /* This is currently unsed in the core */
1131 #define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m) \
1132 const char * const rpv = \
1133 pv_pretty((dsv), (SvPV_nolen_const(sv)), (SvCUR(sv)), (m), \
1134 PL_colors[(c1)],PL_colors[(c2)], \
1135 PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) )
1137 #define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) \
1138 const char * const rpv = \
1139 pv_pretty((dsv), (pv), (l), (m), \
1140 PL_colors[0], PL_colors[1], \
1141 ( PERL_PV_PRETTY_QUOTE | PERL_PV_ESCAPE_RE | PERL_PV_ESCAPE_NONASCII | PERL_PV_PRETTY_ELLIPSES | \
1142 ((isuni) ? PERL_PV_ESCAPE_UNI : 0)) \
1145 #define RE_SV_DUMPLEN(ItEm) (SvCUR(ItEm) - (SvTAIL(ItEm)!=0))
1146 #define RE_SV_TAIL(ItEm) (SvTAIL(ItEm) ? "$" : "")
1148 #else /* if not DEBUGGING */
1150 #define DECLARE_AND_GET_RE_DEBUG_FLAGS dNOOP
1151 #define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) dNOOP
1152 #define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m)
1153 #define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) dNOOP
1154 #define RE_SV_DUMPLEN(ItEm)
1155 #define RE_SV_TAIL(ItEm)
1156 #define isDEBUG_WILDCARD 0
1158 #endif /* DEBUG RELATED DEFINES */
1160 #define FIRST_NON_ASCII_DECIMAL_DIGIT 0x660 /* ARABIC_INDIC_DIGIT_ZERO */
1163 TRADITIONAL_BOUND = _CC_WORDCHAR,
1170 /* This unpacks the FLAGS field of ANYOF[HR]x nodes. The value it contains
1171 * gives the strict lower bound for the UTF-8 start byte of any code point
1172 * matchable by the node, and a loose upper bound as well.
1174 * The low bound is stored as 0xC0 + ((the upper 6 bits) >> 2)
1175 * The loose upper bound is determined from the lowest 2 bits and the low bound
1176 * (called x) as follows:
1178 * 11 The upper limit of the range can be as much as (EF - x) / 8
1179 * 10 The upper limit of the range can be as much as (EF - x) / 4
1180 * 01 The upper limit of the range can be as much as (EF - x) / 2
1181 * 00 The upper limit of the range can be as much as EF
1183 * For motivation of this design, see commit message in
1184 * 3146c00a633e9cbed741e10146662fbcedfdb8d3 */
1186 # define MAX_ANYOF_HRx_BYTE 0xF4
1188 # define MAX_ANYOF_HRx_BYTE 0xEF
1190 #define LOWEST_ANYOF_HRx_BYTE(b) (((b) >> 2) + 0xC0)
1191 #define HIGHEST_ANYOF_HRx_BYTE(b) \
1192 (LOWEST_ANYOF_HRx_BYTE(b) \
1193 + ((MAX_ANYOF_HRx_BYTE - LOWEST_ANYOF_HRx_BYTE(b)) >> ((b) & 3)))
1195 #endif /* PERL_REGCOMP_H_ */
1198 * ex: set ts=8 sts=4 sw=4 et: