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 #ifndef PERL_REGCOMP_H_
12 #define PERL_REGCOMP_H_
14 #include "regcharclass.h"
16 /* Convert branch sequences to more efficient trie ops? */
17 #define PERL_ENABLE_TRIE_OPTIMISATION 1
19 /* Be really aggressive about optimising patterns with trie sequences? */
20 #define PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION 1
22 /* Should the optimiser take positive assertions into account? */
23 #define PERL_ENABLE_POSITIVE_ASSERTION_STUDY 0
25 /* Not for production use: */
26 #define PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS 0
28 /* Activate offsets code - set to if 1 to enable */
30 #define RE_TRACK_PATTERN_OFFSETS
34 * The "internal use only" fields in regexp.h are present to pass info from
35 * compile to execute that permits the execute phase to run lots faster on
36 * simple cases. They are:
38 * regstart sv that must begin a match; NULL if none obvious
39 * reganch is the match anchored (at beginning-of-line only)?
40 * regmust string (pointer into program) that match must include, or NULL
41 * [regmust changed to SV* for bminstr()--law]
42 * regmlen length of regmust string
43 * [regmlen not used currently]
45 * Regstart and reganch permit very fast decisions on suitable starting points
46 * for a match, cutting down the work a lot. Regmust permits fast rejection
47 * of lines that cannot possibly match. The regmust tests are costly enough
48 * that pregcomp() supplies a regmust only if the r.e. contains something
49 * potentially expensive (at present, the only such thing detected is * or +
50 * at the start of the r.e., which can involve a lot of backup). Regmlen is
51 * supplied because the test in pregexec() needs it and pregcomp() is computing
53 * [regmust is now supplied always. The tests that use regmust have a
54 * heuristic that disables the test if it usually matches.]
56 * [In fact, we now use regmust in many cases to locate where the search
57 * starts in the string, so if regback is >= 0, the regmust search is never
58 * wasted effort. The regback variable says how many characters back from
59 * where regmust matched is the earliest possible start of the match.
60 * For instance, /[a-z].foo/ has a regmust of 'foo' and a regback of 2.]
64 * Structure for regexp "program". This is essentially a linear encoding
65 * of a nondeterministic finite-state machine (aka syntax charts or
66 * "railroad normal form" in parsing technology). Each node is an opcode
67 * plus a "next" pointer, possibly plus an operand. "Next" pointers of
68 * all nodes except BRANCH implement concatenation; a "next" pointer with
69 * a BRANCH on both ends of it is connecting two alternatives. (Here we
70 * have one of the subtle syntax dependencies: an individual BRANCH (as
71 * opposed to a collection of them) is never concatenated with anything
72 * because of operator precedence.) The operand of some types of node is
73 * a literal string; for others, it is a node leading into a sub-FSM. In
74 * particular, the operand of a BRANCH node is the first node of the branch.
75 * (NB this is *not* a tree structure: the tail of the branch connects
76 * to the thing following the set of BRANCHes.) The opcodes are defined
77 * in regnodes.h which is generated from regcomp.sym by regcomp.pl.
81 * A node is one char of opcode followed by two chars of "next" pointer.
82 * "Next" pointers are stored as two 8-bit pieces, high order first. The
83 * value is a positive offset from the opcode of the node containing it.
84 * An operand, if any, simply follows the node. (Note that much of the
85 * code generation knows about this implicit relationship.)
87 * Using two bytes for the "next" pointer is vast overkill for most things,
88 * but allows patterns to get big without disasters.
90 * [The "next" pointer is always aligned on an even
91 * boundary, and reads the offset directly as a short.]
94 /* This is the stuff that used to live in regexp.h that was truly
95 private to the engine itself. It now lives here. */
97 typedef struct regexp_internal {
98 int name_list_idx; /* Optional data index of an array of paren names */
100 U32 *offsets; /* offset annotations 20001228 MJD
101 data about mapping the program to the
103 offsets[0] is proglen when this is used
108 regnode *regstclass; /* Optional startclass as identified or constructed
110 struct reg_data *data; /* Additional miscellaneous data used by the program.
111 Used to make it easier to clone and free arbitrary
112 data that the regops need. Often the ARG field of
113 a regop is an index into this structure */
114 struct reg_code_blocks *code_blocks;/* positions of literal (?{}) */
115 regnode program[1]; /* Unwarranted chumminess with compiler. */
118 #define RXi_SET(x,y) (x)->pprivate = (void*)(y)
119 #define RXi_GET(x) ((regexp_internal *)((x)->pprivate))
120 #define RXi_GET_DECL(r,ri) regexp_internal *ri = RXi_GET(r)
122 * Flags stored in regexp->intflags
123 * These are used only internally to the regexp engine
125 * See regexp.h for flags used externally to the regexp engine
127 #define RXp_INTFLAGS(rx) ((rx)->intflags)
128 #define RX_INTFLAGS(prog) RXp_INTFLAGS(ReANY(prog))
130 #define PREGf_SKIP 0x00000001
131 #define PREGf_IMPLICIT 0x00000002 /* Converted .* to ^.* */
132 #define PREGf_NAUGHTY 0x00000004 /* how exponential is this pattern? */
133 #define PREGf_VERBARG_SEEN 0x00000008
134 #define PREGf_CUTGROUP_SEEN 0x00000010
135 #define PREGf_USE_RE_EVAL 0x00000020 /* compiled with "use re 'eval'" */
136 /* these used to be extflags, but are now intflags */
137 #define PREGf_NOSCAN 0x00000040
139 #define PREGf_GPOS_SEEN 0x00000100
140 #define PREGf_GPOS_FLOAT 0x00000200
142 #define PREGf_ANCH_MBOL 0x00000400
143 #define PREGf_ANCH_SBOL 0x00000800
144 #define PREGf_ANCH_GPOS 0x00001000
145 #define PREGf_RECURSE_SEEN 0x00002000
148 ( PREGf_ANCH_SBOL | PREGf_ANCH_GPOS | PREGf_ANCH_MBOL )
150 /* this is where the old regcomp.h started */
152 struct regnode_string {
159 struct regnode_lstring { /* Constructed this way to keep the string aligned. */
163 U32 str_len; /* Only 18 bits allowed before would overflow 'next_off' */
167 struct regnode_anyofhs { /* Constructed this way to keep the string aligned. */
171 U32 arg1; /* set by set_ANYOF_arg() */
175 /* Argument bearing node - workhorse,
176 arg1 is often for the data field */
184 /* Similar to a regnode_1 but with an extra signed argument */
193 /* 'Two field' -- Two 16 bit unsigned args */
202 #define ANYOF_BITMAP_SIZE (NUM_ANYOF_CODE_POINTS / 8) /* 8 bits/Byte */
204 /* Note that these form structs which are supersets of the next smaller one, by
205 * appending fields. Alignment problems can occur if one of those optional
206 * fields requires stricter alignment than the base struct. And formal
207 * parameters that can really be two or more of the structs should be
208 * declared as the smallest one it could be. See commit message for
209 * 7dcac5f6a5195002b55c935ee1d67f67e1df280b. Regnode allocation is done
210 * without regard to alignment, and changing it to would also require changing
211 * the code that inserts and deletes regnodes. The basic single-argument
212 * regnode has a U32, which is what reganode() allocates as a unit. Therefore
213 * no field can require stricter alignment than U32. */
215 /* also used by trie */
216 struct regnode_charclass {
220 U32 arg1; /* set by set_ANYOF_arg() */
221 char bitmap[ANYOF_BITMAP_SIZE]; /* only compile-time */
224 /* has runtime (locale) \d, \w, ..., [:posix:] classes */
225 struct regnode_charclass_posixl {
226 U8 flags; /* ANYOF_MATCHES_POSIXL bit must go here */
230 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time ... */
231 U32 classflags; /* and run-time */
234 /* A synthetic start class (SSC); is a regnode_charclass_posixl_fold, plus an
235 * extra SV*, used only during its construction and which is not used by
236 * regexec.c. Note that the 'next_off' field is unused, as the SSC stands
237 * alone, so there is never a next node. Also, there is no alignment issue,
238 * because these are declared or allocated as a complete unit so the compiler
239 * takes care of alignment. This is unlike the other regnodes which are
240 * allocated in terms of multiples of a single-argument regnode. SSC nodes can
241 * have a pointer field because there is no alignment issue, and because it is
242 * set to NULL after construction, before any cloning of the pattern */
244 U8 flags; /* ANYOF_MATCHES_POSIXL bit must go here */
248 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time ... */
249 U32 classflags; /* ... and run-time */
251 /* Auxiliary, only used during construction; NULL afterwards: list of code
256 /* We take advantage of 'next_off' not otherwise being used in the SSC by
257 * actually using it: by setting it to 1. This allows us to test and
258 * distinguish between an SSC and other ANYOF node types, as 'next_off' cannot
259 * otherwise be 1, because it is the offset to the next regnode expressed in
260 * units of regnodes. Since an ANYOF node contains extra fields, it adds up
261 * to 12 regnode units on 32-bit systems, (hence the minimum this can be (if
262 * not 0) is 11 there. Even if things get tightly packed on a 64-bit system,
263 * it still would be more than 1. */
264 #define set_ANYOF_SYNTHETIC(n) STMT_START{ OP(n) = ANYOF; \
267 #define is_ANYOF_SYNTHETIC(n) (PL_regkind[OP(n)] == ANYOF && NEXT_OFF(n) == 1)
269 /* XXX fix this description.
270 Impose a limit of REG_INFTY on various pattern matching operations
271 to limit stack growth and to avoid "infinite" recursions.
273 /* The default size for REG_INFTY is U16_MAX, which is the same as
274 USHORT_MAX (see perl.h). Unfortunately U16 isn't necessarily 16 bits
275 (see handy.h). On the Cray C90, sizeof(short)==4 and hence U16_MAX is
276 ((1<<32)-1), while on the Cray T90, sizeof(short)==8 and U16_MAX is
277 ((1<<64)-1). To limit stack growth to reasonable sizes, supply a
279 --Andy Dougherty 11 June 1998
283 # define REG_INFTY ((1<<16)-1)
288 # define REG_INFTY U16_MAX
291 #define ARG_VALUE(arg) (arg)
292 #define ARG__SET(arg,val) ((arg) = (val))
298 #define ARG(p) ARG_VALUE(ARG_LOC(p))
299 #define ARG1(p) ARG_VALUE(ARG1_LOC(p))
300 #define ARG2(p) ARG_VALUE(ARG2_LOC(p))
301 #define ARG2L(p) ARG_VALUE(ARG2L_LOC(p))
303 #define ARG_SET(p, val) ARG__SET(ARG_LOC(p), (val))
304 #define ARG1_SET(p, val) ARG__SET(ARG1_LOC(p), (val))
305 #define ARG2_SET(p, val) ARG__SET(ARG2_LOC(p), (val))
306 #define ARG2L_SET(p, val) ARG__SET(ARG2L_LOC(p), (val))
311 #define NEXT_OFF(p) ((p)->next_off)
312 #define NODE_ALIGN(node)
313 /* the following define was set to 0xde in 075abff3
314 * as part of some linting logic. I have set it to 0
315 * as otherwise in every place where we /might/ set flags
316 * we have to set it 0 explicitly, which duplicates
317 * assignments and IMO adds an unacceptable level of
318 * surprise to working in the regex engine. If this
319 * is changed from 0 then at the very least make sure
320 * that SBOL for /^/ sets the flags to 0 explicitly.
322 #define NODE_ALIGN_FILL(node) ((node)->flags = 0)
324 #define SIZE_ALIGN NODE_ALIGN
330 #define OP(p) ((p)->type)
331 #define FLAGS(p) ((p)->flags) /* Caution: Doesn't apply to all \
332 regnode types. For some, it's the \
333 character set of the regnode */
334 #define STR_LENs(p) (__ASSERT_(OP(p) != LEXACT && OP(p) != LEXACT_REQ8) \
335 ((struct regnode_string *)p)->str_len)
336 #define STRINGs(p) (__ASSERT_(OP(p) != LEXACT && OP(p) != LEXACT_REQ8) \
337 ((struct regnode_string *)p)->string)
338 #define OPERANDs(p) STRINGs(p)
340 /* Long strings. Currently limited to length 18 bits, which handles a 262000
341 * byte string. The limiting factor is the 16 bit 'next_off' field, which
342 * points to the next regnode, so the furthest away it can be is 2**16. On
343 * most architectures, regnodes are 2**2 bytes long, so that yields 2**18
344 * bytes. Should a longer string be desired, we could increase it to 26 bits
345 * fairly easily, by changing this node to have longj type which causes the ARG
346 * field to be used for the link to the next regnode (although code would have
347 * to be changed to account for this), and then use a combination of the flags
348 * and next_off fields for the length. To get 34 bit length, also change the
349 * node to be an ARG2L, using the second 32 bit field for the length, and not
350 * using the flags nor next_off fields at all. One could have an llstring node
351 * and even an lllstring type. */
352 #define STR_LENl(p) (__ASSERT_(OP(p) == LEXACT || OP(p) == LEXACT_REQ8) \
353 (((struct regnode_lstring *)p)->str_len))
354 #define STRINGl(p) (__ASSERT_(OP(p) == LEXACT || OP(p) == LEXACT_REQ8) \
355 (((struct regnode_lstring *)p)->string))
356 #define OPERANDl(p) STRINGl(p)
358 #define STR_LEN(p) ((OP(p) == LEXACT || OP(p) == LEXACT_REQ8) \
359 ? STR_LENl(p) : STR_LENs(p))
360 #define STRING(p) ((OP(p) == LEXACT || OP(p) == LEXACT_REQ8) \
361 ? STRINGl(p) : STRINGs(p))
362 #define OPERAND(p) STRING(p)
364 /* The number of (smallest) regnode equivalents that a string of length l bytes
366 #define STR_SZ(l) (((l) + sizeof(regnode) - 1) / sizeof(regnode))
368 /* The number of (smallest) regnode equivalents that the EXACTISH node 'p'
370 #define NODE_SZ_STR(p) (STR_SZ(STR_LEN(p)) + 1 + regarglen[(p)->type])
372 #define setSTR_LEN(p,v) \
374 if (OP(p) == LEXACT || OP(p) == LEXACT_REQ8) \
375 ((struct regnode_lstring *)(p))->str_len = (v); \
377 ((struct regnode_string *)(p))->str_len = (v); \
380 #define ANYOFR_BASE_BITS 20
381 #define ANYOFRbase(p) (ARG(p) & ((1 << ANYOFR_BASE_BITS) - 1))
382 #define ANYOFRdelta(p) (ARG(p) >> ANYOFR_BASE_BITS)
389 #define NODE_ALIGN(node)
390 #define ARG_LOC(p) (((struct regnode_1 *)p)->arg1)
391 #define ARG1_LOC(p) (((struct regnode_2 *)p)->arg1)
392 #define ARG2_LOC(p) (((struct regnode_2 *)p)->arg2)
393 #define ARG2L_LOC(p) (((struct regnode_2L *)p)->arg2)
395 #define NODE_STEP_REGNODE 1 /* sizeof(regnode)/sizeof(regnode) */
396 #define EXTRA_STEP_2ARGS EXTRA_SIZE(struct regnode_2)
398 #define NEXTOPER(p) ((p) + NODE_STEP_REGNODE)
399 #define PREVOPER(p) ((p) - NODE_STEP_REGNODE)
401 #define FILL_NODE(offset, op) \
403 OP(REGNODE_p(offset)) = op; \
404 NEXT_OFF(REGNODE_p(offset)) = 0; \
406 #define FILL_ADVANCE_NODE(offset, op) \
408 FILL_NODE(offset, op); \
411 #define FILL_ADVANCE_NODE_ARG(offset, op, arg) \
413 ARG_SET(REGNODE_p(offset), arg); \
414 FILL_ADVANCE_NODE(offset, op); \
415 /* This is used generically for other operations \
416 * that have a longer argument */ \
417 (offset) += regarglen[op]; \
419 #define FILL_ADVANCE_NODE_2L_ARG(offset, op, arg1, arg2) \
421 ARG_SET(REGNODE_p(offset), arg1); \
422 ARG2L_SET(REGNODE_p(offset), arg2); \
423 FILL_ADVANCE_NODE(offset, op); \
427 #define REG_MAGIC 0234
429 /* An ANYOF node is basically a bitmap with the index being a code point. If
430 * the bit for that code point is 1, the code point matches; if 0, it doesn't
431 * match (complemented if inverted). There is an additional mechanism to deal
432 * with cases where the bitmap is insufficient in and of itself. This #define
433 * indicates if the bitmap does fully represent what this ANYOF node can match.
434 * The ARG is set to this special value (since 0, 1, ... are legal, but will
435 * never reach this high). */
436 #define ANYOF_ONLY_HAS_BITMAP ((U32) -1)
438 /* When the bitmap isn't completely sufficient for handling the ANYOF node,
439 * flags (in node->flags of the ANYOF node) get set to indicate this. These
440 * are perennially in short supply. Beyond several cases where warnings need
441 * to be raised under certain circumstances, currently, there are six cases
442 * where the bitmap alone isn't sufficient. We could use six flags to
443 * represent the 6 cases, but to save flags bits, we play some games. The
446 * 1) The bitmap has a compiled-in very finite size. So something else needs
447 * to be used to specify if a code point that is too large for the bitmap
448 * actually matches. The mechanism currently is an inversion
449 * list. ANYOF_ONLY_HAS_BITMAP, described above, being TRUE indicates
450 * there are no matches of too-large code points. But if it is FALSE,
451 * then almost certainly there are matches too large for the bitmap. (The
452 * other cases, described below, either imply this one or are extremely
453 * rare in practice.) So we can just assume that a too-large code point
454 * will need something beyond the bitmap if ANYOF_ONLY_HAS_BITMAP is
455 * FALSE, instead of having a separate flag for this.
456 * 2) A subset of item 1) is if all possible code points outside the bitmap
457 * match. This is a common occurrence when the class is complemented,
458 * like /[^ij]/. Therefore a bit is reserved to indicate this,
459 * rather than having an inversion list created,
460 * ANYOF_MATCHES_ALL_ABOVE_BITMAP.
461 * 3) Under /d rules, it can happen that code points that are in the upper
462 * latin1 range (\x80-\xFF or their equivalents on EBCDIC platforms) match
463 * only if the runtime target string being matched against is UTF-8. For
464 * example /[\w[:punct:]]/d. This happens only for posix classes (with a
465 * couple of exceptions, like \d where it doesn't happen), and all such
466 * ones also have above-bitmap matches. Thus, 3) implies 1) as well.
467 * Note that /d rules are no longer encouraged; 'use 5.14' or higher
468 * deselects them. But a flag is required so that they can be properly
469 * handled. But it can be a shared flag: see 5) below.
470 * 4) Also under /d rules, something like /[\Wfoo]/ will match everything in
471 * the \x80-\xFF range, unless the string being matched against is UTF-8.
472 * An inversion list could be created for this case, but this is
473 * relatively common, and it turns out that it's all or nothing: if any
474 * one of these code points matches, they all do. Hence a single bit
475 * suffices. We use a shared flag that doesn't take up space by itself:
476 * ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER. This
477 * also implies 1), with one exception: [:^cntrl:].
478 * 5) A user-defined \p{} property may not have been defined by the time the
479 * regex is compiled. In this case, we don't know until runtime what it
480 * will match, so we have to assume it could match anything, including
481 * code points that ordinarily would be in the bitmap. A flag bit is
482 * necessary to indicate this , though it can be shared with the item 3)
483 * flag, as that only occurs under /d, and this only occurs under non-d.
484 * This case is quite uncommon in the field, and the /(?[ ...])/ construct
485 * is a better way to accomplish what this feature does. This case also
487 * ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP
488 * is the shared flag.
489 * 6) /[foo]/il may have folds that are only valid if the runtime locale is a
490 * UTF-8 one. These are quite rare, so it would be good to avoid the
491 * expense of looking for them. But /l matching is slow anyway, and we've
492 * traditionally not worried too much about its performance. And this
493 * condition requires the ANYOFL_FOLD flag to be set, so testing for
494 * that flag would be sufficient to rule out most cases of this. So it is
495 * unclear if this should have a flag or not. But, this flag can be
496 * shared with another, so it doesn't occupy extra space.
498 * At the moment, there is one spare bit, but this could be increased by
501 * If just one more bit is needed, as of this writing it seems to khw that the
502 * best choice would be to make ANYOF_MATCHES_ALL_ABOVE_BITMAP not a flag, but
505 * #define ANYOF_MATCHES_ALL_ABOVE_BITMAP ((U32) -2)
507 * and access it through the ARG like ANYOF_ONLY_HAS_BITMAP is. This flag is
508 * used by all ANYOF node types, and it could be used to avoid calling the
509 * handler function, as the macro REGINCLASS in regexec.c does now for other
512 * Another possibility is based on the fact that ANYOF_MATCHES_POSIXL is
513 * redundant with the node type ANYOFPOSIXL. That flag could be removed, but
514 * at the expense of extra code in regexec.c. The flag has been retained
515 * because it allows us to see if we need to call reginsert, or just use the
516 * bitmap in one test.
518 * If this is done, an extension would be to make all ANYOFL nodes contain the
519 * extra 32 bits that ANYOFPOSIXL ones do. The posix flags only occupy 30
520 * bits, so the ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD flags
521 * and ANYOFL_FOLD could be moved to that extra space, but it would mean extra
522 * instructions, as there are currently places in the code that assume those
525 * All told, 5 bits could be available for other uses if all of the above were
528 * Some flags are not used in synthetic start class (SSC) nodes, so could be
529 * shared should new flags be needed for SSCs, like SSC_MATCHES_EMPTY_STRING
532 /* If this is set, the result of the match should be complemented. regexec.c
533 * is expecting this to be in the low bit. Never in an SSC */
534 #define ANYOF_INVERT 0x01
536 /* For the SSC node only, which cannot be inverted, so is shared with that bit.
537 * This is used only during regex compilation. */
538 #define SSC_MATCHES_EMPTY_STRING ANYOF_INVERT
540 /* Set if this is a regnode_charclass_posixl vs a regnode_charclass. This
541 * is used for runtime \d, \w, [:posix:], ..., which are used only in locale
542 * and the optimizer's synthetic start class. Non-locale \d, etc are resolved
543 * at compile-time. Only set under /l; can be in SSC */
544 #define ANYOF_MATCHES_POSIXL 0x02
546 /* The fold is calculated and stored in the bitmap where possible at compile
547 * time. However under locale, the actual folding varies depending on
548 * what the locale is at the time of execution, so it has to be deferred until
549 * then. Only set under /l; never in an SSC */
550 #define ANYOFL_FOLD 0x04
552 /* Shared bit set only with ANYOFL and SSC nodes:
553 * If ANYOFL_FOLD is set, this flag indicates there are potential matches
554 * valid only if the locale is a UTF-8 one.
555 * If ANYOFL_FOLD is NOT set, this flag means to warn if the runtime locale
556 * isn't a UTF-8 one (and the generated node assumes a UTF-8 locale).
557 * None of INVERT, POSIXL,
558 * ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP
560 #define ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD 0x08
562 /* Convenience macros for teasing apart the meanings when reading the above bit
564 #define ANYOFL_SOME_FOLDS_ONLY_IN_UTF8_LOCALE(flags) \
565 ((flags & ( ANYOFL_FOLD /* Both bits are set */ \
566 |ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD)) \
568 |ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD))
570 #define ANYOFL_UTF8_LOCALE_REQD(flags) \
571 ((flags & ( ANYOFL_FOLD /* Only REQD bit is set */ \
572 |ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD)) \
573 == ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD)
575 /* Spare: Be sure to change ANYOF_FLAGS_ALL if this gets used 0x10 */
577 /* If set, the node matches every code point NUM_ANYOF_CODE_POINTS and above.
578 * Can be in an SSC */
579 #define ANYOF_MATCHES_ALL_ABOVE_BITMAP 0x20
582 * Under /d it means the ANYOFD node matches more things if the target
583 * string is encoded in UTF-8; any such things will be non-ASCII,
584 * characters that are < 256, and can be accessed via the inversion
586 * When not under /d, it means the ANYOF node contains a user-defined
587 * property that wasn't yet defined at the time the regex was compiled,
588 * and so must be looked up at runtime, by creating an inversion list.
589 * (These uses are mutually exclusive because a user-defined property is
590 * specified by \p{}, and \p{} implies /u which deselects /d). The long macro
591 * name is to make sure that you are cautioned about its shared nature. Only
592 * the non-/d meaning can be in an SSC */
593 #define ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP 0x40
596 * Under /d it means the ANYOFD node matches all non-ASCII Latin1
597 * characters when the target string is not in utf8.
598 * When not under /d, it means the ANYOF node should raise a warning if
599 * matching against an above-Unicode code point.
600 * (These uses are mutually exclusive because the warning requires a \p{}, and
601 * \p{} implies /u which deselects /d). An SSC node only has this bit set if
602 * what is meant is the warning. The long macro name is to make sure that you
603 * are cautioned about its shared nature */
604 #define ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER 0x80
606 #define ANYOF_FLAGS_ALL (0xff & ~0x10)
608 #define ANYOF_LOCALE_FLAGS (ANYOFL_FOLD | ANYOF_MATCHES_POSIXL)
610 /* These are the flags that apply to both regular ANYOF nodes and synthetic
611 * start class nodes during construction of the SSC. During finalization of
612 * the SSC, other of the flags may get added to it */
613 #define ANYOF_COMMON_FLAGS 0
615 /* Character classes for node->classflags of ANYOF */
616 /* Should be synchronized with a table in regprop() */
617 /* 2n should be the normal one, paired with its complement at 2n+1 */
619 #define ANYOF_ALPHA ((_CC_ALPHA) * 2)
620 #define ANYOF_NALPHA ((ANYOF_ALPHA) + 1)
621 #define ANYOF_ALPHANUMERIC ((_CC_ALPHANUMERIC) * 2) /* [[:alnum:]] isalnum(3), utf8::IsAlnum */
622 #define ANYOF_NALPHANUMERIC ((ANYOF_ALPHANUMERIC) + 1)
623 #define ANYOF_ASCII ((_CC_ASCII) * 2)
624 #define ANYOF_NASCII ((ANYOF_ASCII) + 1)
625 #define ANYOF_BLANK ((_CC_BLANK) * 2) /* GNU extension: space and tab: non-vertical space */
626 #define ANYOF_NBLANK ((ANYOF_BLANK) + 1)
627 #define ANYOF_CASED ((_CC_CASED) * 2) /* Pseudo class for [:lower:] or
628 [:upper:] under /i */
629 #define ANYOF_NCASED ((ANYOF_CASED) + 1)
630 #define ANYOF_CNTRL ((_CC_CNTRL) * 2)
631 #define ANYOF_NCNTRL ((ANYOF_CNTRL) + 1)
632 #define ANYOF_DIGIT ((_CC_DIGIT) * 2) /* \d */
633 #define ANYOF_NDIGIT ((ANYOF_DIGIT) + 1)
634 #define ANYOF_GRAPH ((_CC_GRAPH) * 2)
635 #define ANYOF_NGRAPH ((ANYOF_GRAPH) + 1)
636 #define ANYOF_LOWER ((_CC_LOWER) * 2)
637 #define ANYOF_NLOWER ((ANYOF_LOWER) + 1)
638 #define ANYOF_PRINT ((_CC_PRINT) * 2)
639 #define ANYOF_NPRINT ((ANYOF_PRINT) + 1)
640 #define ANYOF_PUNCT ((_CC_PUNCT) * 2)
641 #define ANYOF_NPUNCT ((ANYOF_PUNCT) + 1)
642 #define ANYOF_SPACE ((_CC_SPACE) * 2) /* \s */
643 #define ANYOF_NSPACE ((ANYOF_SPACE) + 1)
644 #define ANYOF_UPPER ((_CC_UPPER) * 2)
645 #define ANYOF_NUPPER ((ANYOF_UPPER) + 1)
646 #define ANYOF_WORDCHAR ((_CC_WORDCHAR) * 2) /* \w, PL_utf8_alnum, utf8::IsWord, ALNUM */
647 #define ANYOF_NWORDCHAR ((ANYOF_WORDCHAR) + 1)
648 #define ANYOF_XDIGIT ((_CC_XDIGIT) * 2)
649 #define ANYOF_NXDIGIT ((ANYOF_XDIGIT) + 1)
651 /* pseudo classes below this, not stored in the class bitmap, but used as flags
652 during compilation of char classes */
654 #define ANYOF_VERTWS ((_CC_VERTSPACE) * 2)
655 #define ANYOF_NVERTWS ((ANYOF_VERTWS)+1)
657 /* It is best if this is the last one, as all above it are stored as bits in a
658 * bitmap, and it isn't part of that bitmap */
659 #if _CC_VERTSPACE != _HIGHEST_REGCOMP_DOT_H_SYNC
660 # error Problem with handy.h _HIGHEST_REGCOMP_DOT_H_SYNC #define
663 #define ANYOF_POSIXL_MAX (ANYOF_VERTWS) /* So upper loop limit is written:
665 * Hence doesn't include VERTWS, as that
666 * is a pseudo class */
667 #define ANYOF_MAX ANYOF_POSIXL_MAX
669 #if (ANYOF_POSIXL_MAX > 32) /* Must fit in 32-bit word */
670 # error Problem with handy.h _CC_foo #defines
673 #define ANYOF_HORIZWS ((ANYOF_POSIXL_MAX)+2) /* = (ANYOF_NVERTWS + 1) */
674 #define ANYOF_NHORIZWS ((ANYOF_POSIXL_MAX)+3)
676 #define ANYOF_UNIPROP ((ANYOF_POSIXL_MAX)+4) /* Used to indicate a Unicode
677 property: \p{} or \P{} */
679 /* Backward source code compatibility. */
681 #define ANYOF_ALNUML ANYOF_ALNUM
682 #define ANYOF_NALNUML ANYOF_NALNUM
683 #define ANYOF_SPACEL ANYOF_SPACE
684 #define ANYOF_NSPACEL ANYOF_NSPACE
685 #define ANYOF_ALNUM ANYOF_WORDCHAR
686 #define ANYOF_NALNUM ANYOF_NWORDCHAR
688 /* Utility macros for the bitmap and classes of ANYOF */
690 #define ANYOF_FLAGS(p) ((p)->flags)
692 #define ANYOF_BIT(c) (1U << ((c) & 7))
694 #define POSIXL_SET(field, c) ((field) |= (1U << (c)))
695 #define ANYOF_POSIXL_SET(p, c) POSIXL_SET(((regnode_charclass_posixl*) (p))->classflags, (c))
697 #define POSIXL_CLEAR(field, c) ((field) &= ~ (1U <<(c)))
698 #define ANYOF_POSIXL_CLEAR(p, c) POSIXL_CLEAR(((regnode_charclass_posixl*) (p))->classflags, (c))
700 #define POSIXL_TEST(field, c) ((field) & (1U << (c)))
701 #define ANYOF_POSIXL_TEST(p, c) POSIXL_TEST(((regnode_charclass_posixl*) (p))->classflags, (c))
703 #define POSIXL_ZERO(field) STMT_START { (field) = 0; } STMT_END
704 #define ANYOF_POSIXL_ZERO(ret) POSIXL_ZERO(((regnode_charclass_posixl*) (ret))->classflags)
706 #define ANYOF_POSIXL_SET_TO_BITMAP(p, bits) \
708 ((regnode_charclass_posixl*) (p))->classflags = (bits); \
711 /* Shifts a bit to get, eg. 0x4000_0000, then subtracts 1 to get 0x3FFF_FFFF */
712 #define ANYOF_POSIXL_SETALL(ret) STMT_START { ((regnode_charclass_posixl*) (ret))->classflags = ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1; } STMT_END
713 #define ANYOF_CLASS_SETALL(ret) ANYOF_POSIXL_SETALL(ret)
715 #define ANYOF_POSIXL_TEST_ANY_SET(p) \
716 ((ANYOF_FLAGS(p) & ANYOF_MATCHES_POSIXL) \
717 && (((regnode_charclass_posixl*)(p))->classflags))
718 #define ANYOF_CLASS_TEST_ANY_SET(p) ANYOF_POSIXL_TEST_ANY_SET(p)
720 /* Since an SSC always has this field, we don't have to test for that; nor do
721 * we want to because the bit isn't set for SSC during its construction */
722 #define ANYOF_POSIXL_SSC_TEST_ANY_SET(p) \
723 cBOOL(((regnode_ssc*)(p))->classflags)
724 #define ANYOF_POSIXL_SSC_TEST_ALL_SET(p) /* Are all bits set? */ \
725 (((regnode_ssc*) (p))->classflags \
726 == ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1)
728 #define ANYOF_POSIXL_TEST_ALL_SET(p) \
729 ((ANYOF_FLAGS(p) & ANYOF_MATCHES_POSIXL) \
730 && ((regnode_charclass_posixl*) (p))->classflags \
731 == ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1)
733 #define ANYOF_POSIXL_OR(source, dest) STMT_START { (dest)->classflags |= (source)->classflags ; } STMT_END
734 #define ANYOF_CLASS_OR(source, dest) ANYOF_POSIXL_OR((source), (dest))
736 #define ANYOF_POSIXL_AND(source, dest) STMT_START { (dest)->classflags &= (source)->classflags ; } STMT_END
738 #define ANYOF_BITMAP_ZERO(ret) Zero(((regnode_charclass*)(ret))->bitmap, ANYOF_BITMAP_SIZE, char)
739 #define ANYOF_BITMAP(p) ((regnode_charclass*)(p))->bitmap
740 #define ANYOF_BITMAP_BYTE(p, c) BITMAP_BYTE(ANYOF_BITMAP(p), c)
741 #define ANYOF_BITMAP_SET(p, c) (ANYOF_BITMAP_BYTE(p, c) |= ANYOF_BIT(c))
742 #define ANYOF_BITMAP_CLEAR(p,c) (ANYOF_BITMAP_BYTE(p, c) &= ~ANYOF_BIT(c))
743 #define ANYOF_BITMAP_TEST(p, c) cBOOL(ANYOF_BITMAP_BYTE(p, c) & ANYOF_BIT(c))
745 #define ANYOF_BITMAP_SETALL(p) \
746 memset (ANYOF_BITMAP(p), 255, ANYOF_BITMAP_SIZE)
747 #define ANYOF_BITMAP_CLEARALL(p) \
748 Zero (ANYOF_BITMAP(p), ANYOF_BITMAP_SIZE)
751 * Utility definitions.
754 # define UCHARAT(p) ((int)*(const U8*)(p))
756 # define UCHARAT(p) ((int)*(p)&CHARMASK)
759 /* Number of regnode equivalents that 'guy' occupies beyond the size of the
760 * smallest regnode. */
761 #define EXTRA_SIZE(guy) ((sizeof(guy)-1)/sizeof(struct regnode))
763 #define REG_ZERO_LEN_SEEN 0x00000001
764 #define REG_LOOKBEHIND_SEEN 0x00000002
765 #define REG_GPOS_SEEN 0x00000004
767 #define REG_RECURSE_SEEN 0x00000020
768 #define REG_TOP_LEVEL_BRANCHES_SEEN 0x00000040
769 #define REG_VERBARG_SEEN 0x00000080
770 #define REG_CUTGROUP_SEEN 0x00000100
771 #define REG_RUN_ON_COMMENT_SEEN 0x00000200
772 #define REG_UNFOLDED_MULTI_SEEN 0x00000400
774 #define REG_UNBOUNDED_QUANTIFIER_SEEN 0x00001000
779 #ifdef PLUGGABLE_RE_EXTENSION
780 #include "re_nodes.h"
782 #include "regnodes.h"
785 #ifndef PLUGGABLE_RE_EXTENSION
787 EXTCONST regexp_engine PL_core_reg_engine;
789 EXTCONST regexp_engine PL_core_reg_engine = {
792 Perl_re_intuit_start,
793 Perl_re_intuit_string,
794 Perl_regfree_internal,
795 Perl_reg_numbered_buff_fetch,
796 Perl_reg_numbered_buff_store,
797 Perl_reg_numbered_buff_length,
799 Perl_reg_named_buff_iter,
801 #if defined(USE_ITHREADS)
802 Perl_regdupe_internal,
807 #endif /* PLUGGABLE_RE_EXTENSION */
813 /* .what is a character array with one character for each member of .data
814 * The character describes the function of the corresponding .data item:
815 * a - AV for paren_name_list under DEBUGGING
816 * f - start-class data for regstclass optimization
817 * l - start op for literal (?{EVAL}) item
818 * L - start op for literal (?{EVAL}) item, with separate CV (qr//)
819 * r - pointer to an embedded code-containing qr, e.g. /ab$qr/
820 * s - inversion list for Unicode-style character class, and the
821 * multicharacter strings resulting from casefolding the single-character
822 * entries in the character class
824 * u - trie struct's widecharmap (a HV, so can't share, must dup)
825 * also used for revcharmap and words under DEBUGGING
826 * T - aho-trie struct
827 * S - sv for named capture lookup
828 * 20010712 mjd@plover.com
829 * (Remember to update re_dup() and pregfree() if you add any items.)
837 /* Code in S_to_utf8_substr() and S_to_byte_substr() in regexec.c accesses
838 anchored* and float* via array indexes 0 and 1. */
839 #define anchored_substr substrs->data[0].substr
840 #define anchored_utf8 substrs->data[0].utf8_substr
841 #define anchored_offset substrs->data[0].min_offset
842 #define anchored_end_shift substrs->data[0].end_shift
844 #define float_substr substrs->data[1].substr
845 #define float_utf8 substrs->data[1].utf8_substr
846 #define float_min_offset substrs->data[1].min_offset
847 #define float_max_offset substrs->data[1].max_offset
848 #define float_end_shift substrs->data[1].end_shift
850 #define check_substr substrs->data[2].substr
851 #define check_utf8 substrs->data[2].utf8_substr
852 #define check_offset_min substrs->data[2].min_offset
853 #define check_offset_max substrs->data[2].max_offset
854 #define check_end_shift substrs->data[2].end_shift
856 #define RX_ANCHORED_SUBSTR(rx) (ReANY(rx)->anchored_substr)
857 #define RX_ANCHORED_UTF8(rx) (ReANY(rx)->anchored_utf8)
858 #define RX_FLOAT_SUBSTR(rx) (ReANY(rx)->float_substr)
859 #define RX_FLOAT_UTF8(rx) (ReANY(rx)->float_utf8)
861 /* trie related stuff */
863 /* a transition record for the state machine. the
864 check field determines which state "owns" the
865 transition. the char the transition is for is
866 determined by offset from the owning states base
867 field. the next field determines which state
868 is to be transitioned to if any.
870 struct _reg_trie_trans {
875 /* a transition list element for the list based representation */
876 struct _reg_trie_trans_list_elem {
880 typedef struct _reg_trie_trans_list_elem reg_trie_trans_le;
882 /* a state for compressed nodes. base is an offset
883 into an array of reg_trie_trans array. If wordnum is
884 nonzero the state is accepting. if base is zero then
885 the state has no children (and will be accepting)
887 struct _reg_trie_state {
891 reg_trie_trans_le* list;
895 /* info per word; indexed by wordnum */
897 U16 prev; /* previous word in acceptance chain; eg in
898 * zzz|abc|ab/ after matching the chars abc, the
899 * accepted word is #2, and the previous accepted
901 U32 len; /* how many chars long is this word? */
902 U32 accept; /* accept state for this word */
906 typedef struct _reg_trie_state reg_trie_state;
907 typedef struct _reg_trie_trans reg_trie_trans;
910 /* anything in here that needs to be freed later
911 should be dealt with in pregfree.
912 refcount is first in both this and _reg_ac_data to allow a space
913 optimisation in Perl_regdupe. */
914 struct _reg_trie_data {
915 U32 refcount; /* number of times this trie is referenced */
916 U32 lasttrans; /* last valid transition element */
917 U16 *charmap; /* byte to charid lookup array */
918 reg_trie_state *states; /* state data */
919 reg_trie_trans *trans; /* array of transition elements */
920 char *bitmap; /* stclass bitmap */
921 U16 *jump; /* optional 1 indexed array of offsets before tail
922 for the node following a given word. */
923 reg_trie_wordinfo *wordinfo; /* array of info per word */
924 U16 uniquecharcount; /* unique chars in trie (width of trans table) */
925 U32 startstate; /* initial state - used for common prefix optimisation */
926 STRLEN minlen; /* minimum length of words in trie - build/opt only? */
927 STRLEN maxlen; /* maximum length of words in trie - build/opt only? */
928 U32 prefixlen; /* #chars in common prefix */
929 U32 statecount; /* Build only - number of states in the states array
930 (including the unused zero state) */
931 U32 wordcount; /* Build only */
933 STRLEN charcount; /* Build only */
936 /* There is one (3 under DEBUGGING) pointers that logically belong in this
937 structure, but are held outside as they need duplication on thread cloning,
938 whereas the rest of the structure can be read only:
939 HV *widecharmap; code points > 255 to charid
941 AV *words; Array of words contained in trie, for dumping
942 AV *revcharmap; Map of each charid back to its character representation
946 #define TRIE_WORDS_OFFSET 2
948 typedef struct _reg_trie_data reg_trie_data;
950 /* refcount is first in both this and _reg_trie_data to allow a space
951 optimisation in Perl_regdupe. */
952 struct _reg_ac_data {
956 reg_trie_state *states;
958 typedef struct _reg_ac_data reg_ac_data;
960 /* ANY_BIT doesn't use the structure, so we can borrow it here.
961 This is simpler than refactoring all of it as wed end up with
962 three different sets... */
964 #define TRIE_BITMAP(p) (((reg_trie_data *)(p))->bitmap)
965 #define TRIE_BITMAP_BYTE(p, c) BITMAP_BYTE(TRIE_BITMAP(p), c)
966 #define TRIE_BITMAP_SET(p, c) (TRIE_BITMAP_BYTE(p, c) |= ANYOF_BIT((U8)c))
967 #define TRIE_BITMAP_CLEAR(p,c) (TRIE_BITMAP_BYTE(p, c) &= ~ANYOF_BIT((U8)c))
968 #define TRIE_BITMAP_TEST(p, c) (TRIE_BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
970 #define IS_ANYOF_TRIE(op) ((op)==TRIEC || (op)==AHOCORASICKC)
971 #define IS_TRIE_AC(op) ((op)>=AHOCORASICK)
974 #define BITMAP_BYTE(p, c) (( (U8*) p) [ ( ( (UV) (c)) >> 3) ] )
975 #define BITMAP_TEST(p, c) (BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
977 /* these defines assume uniquecharcount is the correct variable, and state may be evaluated twice */
978 #define TRIE_NODENUM(state) (((state)-1)/(trie->uniquecharcount)+1)
979 #define SAFE_TRIE_NODENUM(state) ((state) ? (((state)-1)/(trie->uniquecharcount)+1) : (state))
980 #define TRIE_NODEIDX(state) ((state) ? (((state)-1)*(trie->uniquecharcount)+1) : (state))
983 #define TRIE_CHARCOUNT(trie) ((trie)->charcount)
985 #define TRIE_CHARCOUNT(trie) (trie_charcount)
988 #define RE_TRIE_MAXBUF_INIT 65536
989 #define RE_TRIE_MAXBUF_NAME "\022E_TRIE_MAXBUF"
990 #define RE_DEBUG_FLAGS "\022E_DEBUG_FLAGS"
992 #define RE_COMPILE_RECURSION_INIT 1000
993 #define RE_COMPILE_RECURSION_LIMIT "\022E_COMPILE_RECURSION_LIMIT"
997 RE_DEBUG_FLAGS is used to control what debug output is emitted
998 its divided into three groups of options, some of which interact.
999 The three groups are: Compile, Execute, Extra. There is room for a
1000 further group, as currently only the low three bytes are used.
1021 If you modify any of these make sure you make corresponding changes to
1022 re.pm, especially to the documentation.
1028 #define RE_DEBUG_COMPILE_MASK 0x0000FF
1029 #define RE_DEBUG_COMPILE_PARSE 0x000001
1030 #define RE_DEBUG_COMPILE_OPTIMISE 0x000002
1031 #define RE_DEBUG_COMPILE_TRIE 0x000004
1032 #define RE_DEBUG_COMPILE_DUMP 0x000008
1033 #define RE_DEBUG_COMPILE_FLAGS 0x000010
1034 #define RE_DEBUG_COMPILE_TEST 0x000020
1037 #define RE_DEBUG_EXECUTE_MASK 0x00FF00
1038 #define RE_DEBUG_EXECUTE_INTUIT 0x000100
1039 #define RE_DEBUG_EXECUTE_MATCH 0x000200
1040 #define RE_DEBUG_EXECUTE_TRIE 0x000400
1043 #define RE_DEBUG_EXTRA_MASK 0x1FF0000
1044 #define RE_DEBUG_EXTRA_TRIE 0x0010000
1045 #define RE_DEBUG_EXTRA_OFFSETS 0x0020000
1046 #define RE_DEBUG_EXTRA_OFFDEBUG 0x0040000
1047 #define RE_DEBUG_EXTRA_STATE 0x0080000
1048 #define RE_DEBUG_EXTRA_OPTIMISE 0x0100000
1049 #define RE_DEBUG_EXTRA_BUFFERS 0x0400000
1050 #define RE_DEBUG_EXTRA_GPOS 0x0800000
1051 #define RE_DEBUG_EXTRA_DUMP_PRE_OPTIMIZE 0x1000000
1053 #define RE_DEBUG_EXTRA_STACK 0x0280000
1055 #define RE_DEBUG_FLAG(x) (re_debug_flags & x)
1057 #define DEBUG_COMPILE_r(x) DEBUG_r( \
1058 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_COMPILE_MASK)) x )
1059 #define DEBUG_PARSE_r(x) DEBUG_r( \
1060 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_COMPILE_PARSE)) x )
1061 #define DEBUG_OPTIMISE_r(x) DEBUG_r( \
1062 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_COMPILE_OPTIMISE)) x )
1063 #define DEBUG_DUMP_r(x) DEBUG_r( \
1064 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_COMPILE_DUMP)) x )
1065 #define DEBUG_TRIE_COMPILE_r(x) DEBUG_r( \
1066 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_COMPILE_TRIE)) x )
1067 #define DEBUG_FLAGS_r(x) DEBUG_r( \
1068 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_COMPILE_FLAGS)) x )
1069 #define DEBUG_TEST_r(x) DEBUG_r( \
1070 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_COMPILE_TEST)) x )
1072 #define DEBUG_EXECUTE_r(x) DEBUG_r( \
1073 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXECUTE_MASK)) x )
1074 #define DEBUG_INTUIT_r(x) DEBUG_r( \
1075 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXECUTE_INTUIT)) x )
1076 #define DEBUG_MATCH_r(x) DEBUG_r( \
1077 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXECUTE_MATCH)) x )
1078 #define DEBUG_TRIE_EXECUTE_r(x) DEBUG_r( \
1079 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXECUTE_TRIE)) x )
1082 #define DEBUG_EXTRA_r(x) DEBUG_r( \
1083 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_MASK)) x )
1084 #define DEBUG_OFFSETS_r(x) DEBUG_r( \
1085 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_OFFSETS)) x )
1086 #define DEBUG_STATE_r(x) DEBUG_r( \
1087 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_STATE)) x )
1088 #define DEBUG_STACK_r(x) DEBUG_r( \
1089 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_STACK)) x )
1090 #define DEBUG_BUFFERS_r(x) DEBUG_r( \
1091 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_BUFFERS)) x )
1093 #define DEBUG_OPTIMISE_MORE_r(x) DEBUG_r( \
1094 if (DEBUG_v_TEST || ((RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE) == \
1095 (re_debug_flags & (RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE)))) x )
1096 #define MJD_OFFSET_DEBUG(x) DEBUG_r( \
1097 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_OFFDEBUG)) \
1098 Perl_warn_nocontext x )
1099 #define DEBUG_TRIE_COMPILE_MORE_r(x) DEBUG_TRIE_COMPILE_r( \
1100 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_TRIE)) x )
1101 #define DEBUG_TRIE_EXECUTE_MORE_r(x) DEBUG_TRIE_EXECUTE_r( \
1102 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_TRIE)) x )
1104 #define DEBUG_TRIE_r(x) DEBUG_r( \
1105 if (DEBUG_v_TEST || (re_debug_flags & (RE_DEBUG_COMPILE_TRIE \
1106 | RE_DEBUG_EXECUTE_TRIE ))) x )
1107 #define DEBUG_GPOS_r(x) DEBUG_r( \
1108 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_GPOS)) x )
1110 #define DEBUG_DUMP_PRE_OPTIMIZE_r(x) DEBUG_r( \
1111 if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_DUMP_PRE_OPTIMIZE)) x )
1113 /* initialization */
1114 /* get_sv() can return NULL during global destruction. */
1115 #define GET_RE_DEBUG_FLAGS DEBUG_r({ \
1116 SV * re_debug_flags_sv = NULL; \
1117 re_debug_flags_sv = PL_curcop ? get_sv(RE_DEBUG_FLAGS, GV_ADD) : NULL; \
1118 if (re_debug_flags_sv) { \
1119 if (!SvIOK(re_debug_flags_sv)) \
1120 sv_setuv(re_debug_flags_sv, RE_DEBUG_COMPILE_DUMP | RE_DEBUG_EXECUTE_MASK ); \
1121 re_debug_flags=SvIV(re_debug_flags_sv); \
1127 #define GET_RE_DEBUG_FLAGS_DECL volatile IV re_debug_flags = 0; \
1128 PERL_UNUSED_VAR(re_debug_flags); GET_RE_DEBUG_FLAGS;
1130 #define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) \
1131 const char * const rpv = \
1132 pv_pretty((dsv), (pv), (l), (m), \
1133 PL_colors[(c1)],PL_colors[(c2)], \
1134 PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ); \
1135 const int rlen = SvCUR(dsv)
1137 /* This is currently unsed in the core */
1138 #define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m) \
1139 const char * const rpv = \
1140 pv_pretty((dsv), (SvPV_nolen_const(sv)), (SvCUR(sv)), (m), \
1141 PL_colors[(c1)],PL_colors[(c2)], \
1142 PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) )
1144 #define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) \
1145 const char * const rpv = \
1146 pv_pretty((dsv), (pv), (l), (m), \
1147 PL_colors[0], PL_colors[1], \
1148 ( PERL_PV_PRETTY_QUOTE | PERL_PV_ESCAPE_RE | PERL_PV_ESCAPE_NONASCII | PERL_PV_PRETTY_ELLIPSES | \
1149 ((isuni) ? PERL_PV_ESCAPE_UNI : 0)) \
1152 #define RE_SV_DUMPLEN(ItEm) (SvCUR(ItEm) - (SvTAIL(ItEm)!=0))
1153 #define RE_SV_TAIL(ItEm) (SvTAIL(ItEm) ? "$" : "")
1155 #else /* if not DEBUGGING */
1157 #define GET_RE_DEBUG_FLAGS_DECL
1158 #define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2)
1159 #define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m)
1160 #define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m)
1161 #define RE_SV_DUMPLEN(ItEm)
1162 #define RE_SV_TAIL(ItEm)
1164 #endif /* DEBUG RELATED DEFINES */
1166 #define FIRST_NON_ASCII_DECIMAL_DIGIT 0x660 /* ARABIC_INDIC_DIGIT_ZERO */
1169 TRADITIONAL_BOUND = _CC_WORDCHAR,
1176 /* This unpacks the FLAGS field of ANYOF[HR]x nodes. The value it contains
1177 * gives the strict lower bound for the UTF-8 start byte of any code point
1178 * matchable by the node, and a loose upper bound as well.
1180 * The low bound is stored in the upper 6 bits, plus 0xC0.
1181 * The loose upper bound is determined from the lowest 2 bits and the low bound
1182 * (called x) as follows:
1184 * 11 The upper limit of the range can be as much as (EF - x) / 8
1185 * 10 The upper limit of the range can be as much as (EF - x) / 4
1186 * 01 The upper limit of the range can be as much as (EF - x) / 2
1187 * 00 The upper limit of the range can be as much as EF
1189 * For motivation of this design, see commit message in
1190 * 3146c00a633e9cbed741e10146662fbcedfdb8d3 */
1192 # define MAX_ANYOF_HRx_BYTE 0xF4
1194 # define MAX_ANYOF_HRx_BYTE 0xEF
1196 #define LOWEST_ANYOF_HRx_BYTE(b) (((b) >> 2) + 0xC0)
1197 #define HIGHEST_ANYOF_HRx_BYTE(b) \
1198 (LOWEST_ANYOF_HRx_BYTE(b) \
1199 + ((MAX_ANYOF_HRx_BYTE - LOWEST_ANYOF_HRx_BYTE(b)) >> ((b) & 3)))
1201 #endif /* PERL_REGCOMP_H_ */
1204 * ex: set ts=8 sts=4 sw=4 et: