3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
4 * 2000, 2001, 2002, 2003, 2005, 2006 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 typedef OP OP_4tree; /* Will be redefined later. */
14 /* Convert branch sequences to more efficient trie ops? */
15 #define PERL_ENABLE_TRIE_OPTIMISATION 1
17 /* Be really agressive about optimising patterns with trie sequences? */
18 #define PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION 1
20 /* Should the optimiser take positive assertions into account? */
21 #define PERL_ENABLE_POSITIVE_ASSERTION_STUDY 0
23 /* Not for production use: */
24 #define PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS 0
26 /* Activate offsets code - set to if 1 to enable */
28 #define RE_TRACK_PATTERN_OFFSETS
31 /* Unless the next line is uncommented it is illegal to combine lazy
32 matching with possessive matching. Frankly it doesn't make much sense
33 to allow it as X*?+ matches nothing, X+?+ matches a single char only,
34 and X{min,max}?+ matches min times only.
36 /* #define REG_ALLOW_MINMOD_SUSPEND */
39 * The "internal use only" fields in regexp.h are present to pass info from
40 * compile to execute that permits the execute phase to run lots faster on
41 * simple cases. They are:
43 * regstart sv that must begin a match; NULL if none obvious
44 * reganch is the match anchored (at beginning-of-line only)?
45 * regmust string (pointer into program) that match must include, or NULL
46 * [regmust changed to SV* for bminstr()--law]
47 * regmlen length of regmust string
48 * [regmlen not used currently]
50 * Regstart and reganch permit very fast decisions on suitable starting points
51 * for a match, cutting down the work a lot. Regmust permits fast rejection
52 * of lines that cannot possibly match. The regmust tests are costly enough
53 * that pregcomp() supplies a regmust only if the r.e. contains something
54 * potentially expensive (at present, the only such thing detected is * or +
55 * at the start of the r.e., which can involve a lot of backup). Regmlen is
56 * supplied because the test in pregexec() needs it and pregcomp() is computing
58 * [regmust is now supplied always. The tests that use regmust have a
59 * heuristic that disables the test if it usually matches.]
61 * [In fact, we now use regmust in many cases to locate where the search
62 * starts in the string, so if regback is >= 0, the regmust search is never
63 * wasted effort. The regback variable says how many characters back from
64 * where regmust matched is the earliest possible start of the match.
65 * For instance, /[a-z].foo/ has a regmust of 'foo' and a regback of 2.]
69 * Structure for regexp "program". This is essentially a linear encoding
70 * of a nondeterministic finite-state machine (aka syntax charts or
71 * "railroad normal form" in parsing technology). Each node is an opcode
72 * plus a "next" pointer, possibly plus an operand. "Next" pointers of
73 * all nodes except BRANCH implement concatenation; a "next" pointer with
74 * a BRANCH on both ends of it is connecting two alternatives. (Here we
75 * have one of the subtle syntax dependencies: an individual BRANCH (as
76 * opposed to a collection of them) is never concatenated with anything
77 * because of operator precedence.) The operand of some types of node is
78 * a literal string; for others, it is a node leading into a sub-FSM. In
79 * particular, the operand of a BRANCH node is the first node of the branch.
80 * (NB this is *not* a tree structure: the tail of the branch connects
81 * to the thing following the set of BRANCHes.) The opcodes are defined
82 * in regnodes.h which is generated from regcomp.sym by regcomp.pl.
86 * A node is one char of opcode followed by two chars of "next" pointer.
87 * "Next" pointers are stored as two 8-bit pieces, high order first. The
88 * value is a positive offset from the opcode of the node containing it.
89 * An operand, if any, simply follows the node. (Note that much of the
90 * code generation knows about this implicit relationship.)
92 * Using two bytes for the "next" pointer is vast overkill for most things,
93 * but allows patterns to get big without disasters.
95 * [The "next" pointer is always aligned on an even
96 * boundary, and reads the offset directly as a short. Also, there is no
97 * special test to reverse the sign of BACK pointers since the offset is
101 /* This is the stuff that used to live in regexp.h that was truly
102 private to the engine itself. It now lives here. */
106 typedef struct regexp_internal {
107 int name_list_idx; /* Optional data index of an array of paren names */
109 U32 *offsets; /* offset annotations 20001228 MJD
110 data about mapping the program to the
112 offsets[0] is proglen when this is used
117 regnode *regstclass; /* Optional startclass as identified or constructed
119 struct reg_data *data; /* Additional miscellaneous data used by the program.
120 Used to make it easier to clone and free arbitrary
121 data that the regops need. Often the ARG field of
122 a regop is an index into this structure */
123 regnode program[1]; /* Unwarranted chumminess with compiler. */
126 #define RXi_SET(x,y) (x)->pprivate = (void*)(y)
127 #define RXi_GET(x) ((regexp_internal *)((x)->pprivate))
128 #define RXi_GET_DECL(r,ri) regexp_internal *ri = RXi_GET(r)
130 * Flags stored in regexp->intflags
131 * These are used only internally to the regexp engine
133 * See regexp.h for flags used externally to the regexp engine
135 #define PREGf_SKIP 0x00000001
136 #define PREGf_IMPLICIT 0x00000002 /* Converted .* to ^.* */
137 #define PREGf_NAUGHTY 0x00000004 /* how exponential is this pattern? */
138 #define PREGf_VERBARG_SEEN 0x00000008
139 #define PREGf_CUTGROUP_SEEN 0x00000010
142 /* this is where the old regcomp.h started */
144 struct regnode_string {
151 /* Argument bearing node - workhorse,
152 arg1 is often for the data field */
160 /* Similar to a regnode_1 but with an extra signed argument */
169 /* 'Two field' -- Two 16 bit unsigned args */
179 #define ANYOF_BITMAP_SIZE 32 /* 256 b/(8 b/B) */
180 #define ANYOF_CLASSBITMAP_SIZE 4 /* up to 32 (8*4) named classes */
182 /* also used by trie */
183 struct regnode_charclass {
188 char bitmap[ANYOF_BITMAP_SIZE]; /* only compile-time */
191 struct regnode_charclass_class { /* has [[:blah:]] classes */
192 U8 flags; /* should have ANYOF_CLASS here */
196 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time */
197 char classflags[ANYOF_CLASSBITMAP_SIZE]; /* and run-time */
200 /* XXX fix this description.
201 Impose a limit of REG_INFTY on various pattern matching operations
202 to limit stack growth and to avoid "infinite" recursions.
204 /* The default size for REG_INFTY is I16_MAX, which is the same as
205 SHORT_MAX (see perl.h). Unfortunately I16 isn't necessarily 16 bits
206 (see handy.h). On the Cray C90, sizeof(short)==4 and hence I16_MAX is
207 ((1<<31)-1), while on the Cray T90, sizeof(short)==8 and I16_MAX is
208 ((1<<63)-1). To limit stack growth to reasonable sizes, supply a
210 --Andy Dougherty 11 June 1998
214 # define REG_INFTY ((1<<15)-1)
219 # define REG_INFTY I16_MAX
222 #define ARG_VALUE(arg) (arg)
223 #define ARG__SET(arg,val) ((arg) = (val))
229 #define ARG(p) ARG_VALUE(ARG_LOC(p))
230 #define ARG1(p) ARG_VALUE(ARG1_LOC(p))
231 #define ARG2(p) ARG_VALUE(ARG2_LOC(p))
232 #define ARG2L(p) ARG_VALUE(ARG2L_LOC(p))
234 #define ARG_SET(p, val) ARG__SET(ARG_LOC(p), (val))
235 #define ARG1_SET(p, val) ARG__SET(ARG1_LOC(p), (val))
236 #define ARG2_SET(p, val) ARG__SET(ARG2_LOC(p), (val))
237 #define ARG2L_SET(p, val) ARG__SET(ARG2L_LOC(p), (val))
242 #define NEXT_OFF(p) ((p)->next_off)
243 #define NODE_ALIGN(node)
244 #define NODE_ALIGN_FILL(node) ((node)->flags = 0xde) /* deadbeef */
246 #define SIZE_ALIGN NODE_ALIGN
253 #define OP(p) ((p)->type)
254 #define OPERAND(p) (((struct regnode_string *)p)->string)
255 #define MASK(p) ((char*)OPERAND(p))
256 #define STR_LEN(p) (((struct regnode_string *)p)->str_len)
257 #define STRING(p) (((struct regnode_string *)p)->string)
258 #define STR_SZ(l) ((l + sizeof(regnode) - 1) / sizeof(regnode))
259 #define NODE_SZ_STR(p) (STR_SZ(STR_LEN(p))+1)
266 #define NODE_ALIGN(node)
267 #define ARG_LOC(p) (((struct regnode_1 *)p)->arg1)
268 #define ARG1_LOC(p) (((struct regnode_2 *)p)->arg1)
269 #define ARG2_LOC(p) (((struct regnode_2 *)p)->arg2)
270 #define ARG2L_LOC(p) (((struct regnode_2L *)p)->arg2)
272 #define NODE_STEP_REGNODE 1 /* sizeof(regnode)/sizeof(regnode) */
273 #define EXTRA_STEP_2ARGS EXTRA_SIZE(struct regnode_2)
275 #define NODE_STEP_B 4
277 #define NEXTOPER(p) ((p) + NODE_STEP_REGNODE)
278 #define PREVOPER(p) ((p) - NODE_STEP_REGNODE)
280 #define FILL_ADVANCE_NODE(ptr, op) STMT_START { \
281 (ptr)->type = op; (ptr)->next_off = 0; (ptr)++; } STMT_END
282 #define FILL_ADVANCE_NODE_ARG(ptr, op, arg) STMT_START { \
283 ARG_SET(ptr, arg); FILL_ADVANCE_NODE(ptr, op); (ptr) += 1; } STMT_END
285 #define REG_MAGIC 0234
287 #define SIZE_ONLY (RExC_emit == &PL_regdummy)
289 /* Flags for node->flags of ANYOF */
291 #define ANYOF_CLASS 0x08 /* has [[:blah:]] classes */
292 #define ANYOF_INVERT 0x04
293 #define ANYOF_FOLD 0x02
294 #define ANYOF_LOCALE 0x01
296 /* Used for regstclass only */
297 #define ANYOF_EOS 0x10 /* Can match an empty string too */
299 /* There is a character or a range past 0xff */
300 #define ANYOF_UNICODE 0x20
301 #define ANYOF_UNICODE_ALL 0x40 /* Can match any char past 0xff */
303 /* size of node is large (includes class pointer) */
304 #define ANYOF_LARGE 0x80
306 /* Are there any runtime flags on in this node? */
307 #define ANYOF_RUNTIME(s) (ANYOF_FLAGS(s) & 0x0f)
309 #define ANYOF_FLAGS_ALL 0xff
311 /* Character classes for node->classflags of ANYOF */
312 /* Should be synchronized with a table in regprop() */
313 /* 2n should pair with 2n+1 */
315 #define ANYOF_ALNUM 0 /* \w, PL_utf8_alnum, utf8::IsWord, ALNUM */
316 #define ANYOF_NALNUM 1
317 #define ANYOF_SPACE 2 /* \s */
318 #define ANYOF_NSPACE 3
319 #define ANYOF_DIGIT 4
320 #define ANYOF_NDIGIT 5
321 #define ANYOF_ALNUMC 6 /* isalnum(3), utf8::IsAlnum, ALNUMC */
322 #define ANYOF_NALNUMC 7
323 #define ANYOF_ALPHA 8
324 #define ANYOF_NALPHA 9
325 #define ANYOF_ASCII 10
326 #define ANYOF_NASCII 11
327 #define ANYOF_CNTRL 12
328 #define ANYOF_NCNTRL 13
329 #define ANYOF_GRAPH 14
330 #define ANYOF_NGRAPH 15
331 #define ANYOF_LOWER 16
332 #define ANYOF_NLOWER 17
333 #define ANYOF_PRINT 18
334 #define ANYOF_NPRINT 19
335 #define ANYOF_PUNCT 20
336 #define ANYOF_NPUNCT 21
337 #define ANYOF_UPPER 22
338 #define ANYOF_NUPPER 23
339 #define ANYOF_XDIGIT 24
340 #define ANYOF_NXDIGIT 25
341 #define ANYOF_PSXSPC 26 /* POSIX space: \s plus the vertical tab */
342 #define ANYOF_NPSXSPC 27
343 #define ANYOF_BLANK 28 /* GNU extension: space and tab: non-vertical space */
344 #define ANYOF_NBLANK 29
348 /* Backward source code compatibility. */
350 #define ANYOF_ALNUML ANYOF_ALNUM
351 #define ANYOF_NALNUML ANYOF_NALNUM
352 #define ANYOF_SPACEL ANYOF_SPACE
353 #define ANYOF_NSPACEL ANYOF_NSPACE
355 /* Utility macros for the bitmap and classes of ANYOF */
357 #define ANYOF_SIZE (sizeof(struct regnode_charclass))
358 #define ANYOF_CLASS_SIZE (sizeof(struct regnode_charclass_class))
360 #define ANYOF_FLAGS(p) ((p)->flags)
362 #define ANYOF_BIT(c) (1 << ((c) & 7))
364 #define ANYOF_CLASS_BYTE(p, c) (((struct regnode_charclass_class*)(p))->classflags[((c) >> 3) & 3])
365 #define ANYOF_CLASS_SET(p, c) (ANYOF_CLASS_BYTE(p, c) |= ANYOF_BIT(c))
366 #define ANYOF_CLASS_CLEAR(p, c) (ANYOF_CLASS_BYTE(p, c) &= ~ANYOF_BIT(c))
367 #define ANYOF_CLASS_TEST(p, c) (ANYOF_CLASS_BYTE(p, c) & ANYOF_BIT(c))
369 #define ANYOF_CLASS_ZERO(ret) Zero(((struct regnode_charclass_class*)(ret))->classflags, ANYOF_CLASSBITMAP_SIZE, char)
370 #define ANYOF_BITMAP_ZERO(ret) Zero(((struct regnode_charclass*)(ret))->bitmap, ANYOF_BITMAP_SIZE, char)
372 #define ANYOF_BITMAP(p) (((struct regnode_charclass*)(p))->bitmap)
373 #define ANYOF_BITMAP_BYTE(p, c) (ANYOF_BITMAP(p)[(((U8)(c)) >> 3) & 31])
374 #define ANYOF_BITMAP_SET(p, c) (ANYOF_BITMAP_BYTE(p, c) |= ANYOF_BIT(c))
375 #define ANYOF_BITMAP_CLEAR(p,c) (ANYOF_BITMAP_BYTE(p, c) &= ~ANYOF_BIT(c))
376 #define ANYOF_BITMAP_TEST(p, c) (ANYOF_BITMAP_BYTE(p, c) & ANYOF_BIT(c))
378 #define ANYOF_BITMAP_SETALL(p) \
379 memset (ANYOF_BITMAP(p), 255, ANYOF_BITMAP_SIZE)
380 #define ANYOF_BITMAP_CLEARALL(p) \
381 Zero (ANYOF_BITMAP(p), ANYOF_BITMAP_SIZE)
382 /* Check that all 256 bits are all set. Used in S_cl_is_anything() */
383 #define ANYOF_BITMAP_TESTALLSET(p) \
384 memEQ (ANYOF_BITMAP(p), "\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377", ANYOF_BITMAP_SIZE)
386 #define ANYOF_SKIP ((ANYOF_SIZE - 1)/sizeof(regnode))
387 #define ANYOF_CLASS_SKIP ((ANYOF_CLASS_SIZE - 1)/sizeof(regnode))
388 #define ANYOF_CLASS_ADD_SKIP (ANYOF_CLASS_SKIP - ANYOF_SKIP)
392 * Utility definitions.
395 # define UCHARAT(p) ((int)*(const U8*)(p))
397 # define UCHARAT(p) ((int)*(p)&CHARMASK)
400 #define EXTRA_SIZE(guy) ((sizeof(guy)-1)/sizeof(struct regnode))
402 #define REG_SEEN_ZERO_LEN 0x00000001
403 #define REG_SEEN_LOOKBEHIND 0x00000002
404 #define REG_SEEN_GPOS 0x00000004
405 #define REG_SEEN_EVAL 0x00000008
406 #define REG_SEEN_CANY 0x00000010
407 #define REG_SEEN_SANY REG_SEEN_CANY /* src bckwrd cmpt */
408 #define REG_SEEN_RECURSE 0x00000020
409 #define REG_TOP_LEVEL_BRANCHES 0x00000040
410 #define REG_SEEN_VERBARG 0x00000080
411 #define REG_SEEN_CUTGROUP 0x00000100
412 #define REG_SEEN_RUN_ON_COMMENT 0x00000200
416 #ifdef PLUGGABLE_RE_EXTENSION
417 #include "re_nodes.h"
419 #include "regnodes.h"
422 /* The following have no fixed length. U8 so we can do strchr() on it. */
424 EXTCONST U8 PL_varies[];
426 EXTCONST U8 PL_varies[] = {
427 BRANCH, BACK, STAR, PLUS, CURLY, CURLYX, REF, REFF, REFFL,
428 WHILEM, CURLYM, CURLYN, BRANCHJ, IFTHEN, SUSPEND, CLUMP,
434 /* The following always have a length of 1. U8 we can do strchr() on it. */
435 /* (Note that length 1 means "one character" under UTF8, not "one octet".) */
437 EXTCONST U8 PL_simple[];
439 EXTCONST U8 PL_simple[] = {
451 #ifndef PLUGGABLE_RE_EXTENSION
453 EXTCONST regexp_engine PL_core_reg_engine;
455 EXTCONST regexp_engine PL_core_reg_engine = {
458 Perl_re_intuit_start,
459 Perl_re_intuit_string,
460 Perl_regfree_internal,
461 Perl_reg_numbered_buff_get,
462 Perl_reg_named_buff_get,
463 #if defined(USE_ITHREADS)
464 Perl_regdupe_internal
468 #endif /* PLUGGABLE_RE_EXTENSION */
474 /* .what is a character array with one character for each member of .data
475 * The character describes the function of the corresponding .data item:
476 * f - start-class data for regstclass optimization
477 * n - Root of op tree for (?{EVAL}) item
478 * o - Start op for (?{EVAL}) item
479 * p - Pad for (?{EVAL}) item
480 * s - swash for unicode-style character class, and the multicharacter
481 * strings resulting from casefolding the single-character entries
482 * in the character class
484 * u - trie struct's widecharmap (a HV, so can't share, must dup)
485 * also used for revcharmap and words under DEBUGGING
486 * T - aho-trie struct
487 * S - sv for named capture lookup
488 * 20010712 mjd@plover.com
489 * (Remember to update re_dup() and pregfree() if you add any items.)
497 /* Code in S_to_utf8_substr() and S_to_byte_substr() in regexec.c accesses
498 anchored* and float* via array indexes 0 and 1. */
499 #define anchored_substr substrs->data[0].substr
500 #define anchored_utf8 substrs->data[0].utf8_substr
501 #define anchored_offset substrs->data[0].min_offset
502 #define anchored_end_shift substrs->data[0].end_shift
504 #define float_substr substrs->data[1].substr
505 #define float_utf8 substrs->data[1].utf8_substr
506 #define float_min_offset substrs->data[1].min_offset
507 #define float_max_offset substrs->data[1].max_offset
508 #define float_end_shift substrs->data[1].end_shift
510 #define check_substr substrs->data[2].substr
511 #define check_utf8 substrs->data[2].utf8_substr
512 #define check_offset_min substrs->data[2].min_offset
513 #define check_offset_max substrs->data[2].max_offset
514 #define check_end_shift substrs->data[2].end_shift
518 /* trie related stuff */
520 /* a transition record for the state machine. the
521 check field determines which state "owns" the
522 transition. the char the transition is for is
523 determined by offset from the owning states base
524 field. the next field determines which state
525 is to be transitioned to if any.
527 struct _reg_trie_trans {
532 /* a transition list element for the list based representation */
533 struct _reg_trie_trans_list_elem {
537 typedef struct _reg_trie_trans_list_elem reg_trie_trans_le;
539 /* a state for compressed nodes. base is an offset
540 into an array of reg_trie_trans array. If wordnum is
541 nonzero the state is accepting. if base is zero then
542 the state has no children (and will be accepting)
544 struct _reg_trie_state {
548 reg_trie_trans_le* list;
554 typedef struct _reg_trie_state reg_trie_state;
555 typedef struct _reg_trie_trans reg_trie_trans;
558 /* anything in here that needs to be freed later
559 should be dealt with in pregfree.
560 refcount is first in both this and _reg_ac_data to allow a space
561 optimisation in Perl_regdupe. */
562 struct _reg_trie_data {
563 U32 refcount; /* number of times this trie is referenced */
564 U32 lasttrans; /* last valid transition element */
565 U16 *charmap; /* byte to charid lookup array */
566 reg_trie_state *states; /* state data */
567 reg_trie_trans *trans; /* array of transition elements */
568 char *bitmap; /* stclass bitmap */
569 U32 *wordlen; /* array of lengths of words */
570 U16 *jump; /* optional 1 indexed array of offsets before tail
571 for the node following a given word. */
572 U16 *nextword; /* optional 1 indexed array to support linked list
573 of duplicate wordnums */
574 U16 uniquecharcount; /* unique chars in trie (width of trans table) */
575 U32 startstate; /* initial state - used for common prefix optimisation */
576 STRLEN minlen; /* minimum length of words in trie - build/opt only? */
577 STRLEN maxlen; /* maximum length of words in trie - build/opt only? */
578 U32 statecount; /* Build only - number of states in the states array
579 (including the unused zero state) */
580 U32 wordcount; /* Build only */
582 STRLEN charcount; /* Build only */
585 /* There is one (3 under DEBUGGING) pointers that logically belong in this
586 structure, but are held outside as they need duplication on thread cloning,
587 whereas the rest of the structure can be read only:
588 HV *widecharmap; code points > 255 to charid
590 AV *words; Array of words contained in trie, for dumping
591 AV *revcharmap; Map of each charid back to its character representation
595 #define TRIE_WORDS_OFFSET 2
597 typedef struct _reg_trie_data reg_trie_data;
599 /* refcount is first in both this and _reg_trie_data to allow a space
600 optimisation in Perl_regdupe. */
601 struct _reg_ac_data {
605 reg_trie_state *states;
607 typedef struct _reg_ac_data reg_ac_data;
609 /* ANY_BIT doesnt use the structure, so we can borrow it here.
610 This is simpler than refactoring all of it as wed end up with
611 three different sets... */
613 #define TRIE_BITMAP(p) (((reg_trie_data *)(p))->bitmap)
614 #define TRIE_BITMAP_BYTE(p, c) (TRIE_BITMAP(p)[(((U8)(c)) >> 3) & 31])
615 #define TRIE_BITMAP_SET(p, c) (TRIE_BITMAP_BYTE(p, c) |= ANYOF_BIT((U8)c))
616 #define TRIE_BITMAP_CLEAR(p,c) (TRIE_BITMAP_BYTE(p, c) &= ~ANYOF_BIT((U8)c))
617 #define TRIE_BITMAP_TEST(p, c) (TRIE_BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
619 #define IS_ANYOF_TRIE(op) ((op)==TRIEC || (op)==AHOCORASICKC)
620 #define IS_TRIE_AC(op) ((op)>=AHOCORASICK)
623 #define BITMAP_BYTE(p, c) (((U8*)p)[(((U8)(c)) >> 3) & 31])
624 #define BITMAP_TEST(p, c) (BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
626 /* these defines assume uniquecharcount is the correct variable, and state may be evaluated twice */
627 #define TRIE_NODENUM(state) (((state)-1)/(trie->uniquecharcount)+1)
628 #define SAFE_TRIE_NODENUM(state) ((state) ? (((state)-1)/(trie->uniquecharcount)+1) : (state))
629 #define TRIE_NODEIDX(state) ((state) ? (((state)-1)*(trie->uniquecharcount)+1) : (state))
632 #define TRIE_CHARCOUNT(trie) ((trie)->charcount)
634 #define TRIE_CHARCOUNT(trie) (trie_charcount)
637 #define RE_TRIE_MAXBUF_INIT 65536
638 #define RE_TRIE_MAXBUF_NAME "\022E_TRIE_MAXBUF"
639 #define RE_DEBUG_FLAGS "\022E_DEBUG_FLAGS"
643 RE_DEBUG_FLAGS is used to control what debug output is emitted
644 its divided into three groups of options, some of which interact.
645 The three groups are: Compile, Execute, Extra. There is room for a
646 further group, as currently only the low three bytes are used.
667 If you modify any of these make sure you make corresponding changes to
668 re.pm, especially to the documentation.
674 #define RE_DEBUG_COMPILE_MASK 0x0000FF
675 #define RE_DEBUG_COMPILE_PARSE 0x000001
676 #define RE_DEBUG_COMPILE_OPTIMISE 0x000002
677 #define RE_DEBUG_COMPILE_TRIE 0x000004
678 #define RE_DEBUG_COMPILE_DUMP 0x000008
681 #define RE_DEBUG_EXECUTE_MASK 0x00FF00
682 #define RE_DEBUG_EXECUTE_INTUIT 0x000100
683 #define RE_DEBUG_EXECUTE_MATCH 0x000200
684 #define RE_DEBUG_EXECUTE_TRIE 0x000400
687 #define RE_DEBUG_EXTRA_MASK 0xFF0000
688 #define RE_DEBUG_EXTRA_TRIE 0x010000
689 #define RE_DEBUG_EXTRA_OFFSETS 0x020000
690 #define RE_DEBUG_EXTRA_OFFDEBUG 0x040000
691 #define RE_DEBUG_EXTRA_STATE 0x080000
692 #define RE_DEBUG_EXTRA_OPTIMISE 0x100000
693 #define RE_DEBUG_EXTRA_BUFFERS 0x400000
695 #define RE_DEBUG_EXTRA_STACK 0x280000
697 #define RE_DEBUG_FLAG(x) (re_debug_flags & x)
699 #define DEBUG_COMPILE_r(x) DEBUG_r( \
700 if (re_debug_flags & RE_DEBUG_COMPILE_MASK) x )
701 #define DEBUG_PARSE_r(x) DEBUG_r( \
702 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x )
703 #define DEBUG_OPTIMISE_r(x) DEBUG_r( \
704 if (re_debug_flags & RE_DEBUG_COMPILE_OPTIMISE) x )
705 #define DEBUG_PARSE_r(x) DEBUG_r( \
706 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x )
707 #define DEBUG_DUMP_r(x) DEBUG_r( \
708 if (re_debug_flags & RE_DEBUG_COMPILE_DUMP) x )
709 #define DEBUG_TRIE_COMPILE_r(x) DEBUG_r( \
710 if (re_debug_flags & RE_DEBUG_COMPILE_TRIE) x )
713 #define DEBUG_EXECUTE_r(x) DEBUG_r( \
714 if (re_debug_flags & RE_DEBUG_EXECUTE_MASK) x )
715 #define DEBUG_INTUIT_r(x) DEBUG_r( \
716 if (re_debug_flags & RE_DEBUG_EXECUTE_INTUIT) x )
717 #define DEBUG_MATCH_r(x) DEBUG_r( \
718 if (re_debug_flags & RE_DEBUG_EXECUTE_MATCH) x )
719 #define DEBUG_TRIE_EXECUTE_r(x) DEBUG_r( \
720 if (re_debug_flags & RE_DEBUG_EXECUTE_TRIE) x )
723 #define DEBUG_EXTRA_r(x) DEBUG_r( \
724 if (re_debug_flags & RE_DEBUG_EXTRA_MASK) x )
725 #define DEBUG_OFFSETS_r(x) DEBUG_r( \
726 if (re_debug_flags & RE_DEBUG_EXTRA_OFFSETS) x )
727 #define DEBUG_STATE_r(x) DEBUG_r( \
728 if (re_debug_flags & RE_DEBUG_EXTRA_STATE) x )
729 #define DEBUG_STACK_r(x) DEBUG_r( \
730 if (re_debug_flags & RE_DEBUG_EXTRA_STACK) x )
731 #define DEBUG_BUFFERS_r(x) DEBUG_r( \
732 if (re_debug_flags & RE_DEBUG_EXTRA_BUFFERS) x )
734 #define DEBUG_OPTIMISE_MORE_r(x) DEBUG_r( \
735 if ((RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE) == \
736 (re_debug_flags & (RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE)) ) x )
737 #define MJD_OFFSET_DEBUG(x) DEBUG_r( \
738 if (re_debug_flags & RE_DEBUG_EXTRA_OFFDEBUG) \
739 Perl_warn_nocontext x )
740 #define DEBUG_TRIE_COMPILE_MORE_r(x) DEBUG_TRIE_COMPILE_r( \
741 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x )
742 #define DEBUG_TRIE_EXECUTE_MORE_r(x) DEBUG_TRIE_EXECUTE_r( \
743 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x )
745 #define DEBUG_TRIE_r(x) DEBUG_r( \
746 if (re_debug_flags & (RE_DEBUG_COMPILE_TRIE \
747 | RE_DEBUG_EXECUTE_TRIE )) x )
750 /* get_sv() can return NULL during global destruction. */
751 #define GET_RE_DEBUG_FLAGS DEBUG_r({ \
752 SV * re_debug_flags_sv = NULL; \
753 re_debug_flags_sv = get_sv(RE_DEBUG_FLAGS, 1); \
754 if (re_debug_flags_sv) { \
755 if (!SvIOK(re_debug_flags_sv)) \
756 sv_setuv(re_debug_flags_sv, RE_DEBUG_COMPILE_DUMP | RE_DEBUG_EXECUTE_MASK ); \
757 re_debug_flags=SvIV(re_debug_flags_sv); \
763 #define GET_RE_DEBUG_FLAGS_DECL IV re_debug_flags = 0; GET_RE_DEBUG_FLAGS;
765 #define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) \
766 const char * const rpv = \
767 pv_pretty((dsv), (pv), (l), (m), \
768 PL_colors[(c1)],PL_colors[(c2)], \
769 PERL_PV_ESCAPE_RE |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ); \
770 const int rlen = SvCUR(dsv)
772 #define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m) \
773 const char * const rpv = \
774 pv_pretty((dsv), (SvPV_nolen_const(sv)), (SvCUR(sv)), (m), \
775 PL_colors[(c1)],PL_colors[(c2)], \
776 PERL_PV_ESCAPE_RE |((isuni) ? PERL_PV_ESCAPE_UNI : 0) )
778 #define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) \
779 const char * const rpv = \
780 pv_pretty((dsv), (pv), (l), (m), \
781 PL_colors[0], PL_colors[1], \
782 ( PERL_PV_PRETTY_QUOTE | PERL_PV_ESCAPE_RE | PERL_PV_PRETTY_ELIPSES | \
783 ((isuni) ? PERL_PV_ESCAPE_UNI : 0)) \
786 #define RE_SV_DUMPLEN(ItEm) (SvCUR(ItEm) - (SvTAIL(ItEm)!=0))
787 #define RE_SV_TAIL(ItEm) (SvTAIL(ItEm) ? "$" : "")
789 #else /* if not DEBUGGING */
791 #define GET_RE_DEBUG_FLAGS_DECL
792 #define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2)
793 #define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m)
794 #define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m)
795 #define RE_SV_DUMPLEN(ItEm)
796 #define RE_SV_TAIL(ItEm)
798 #endif /* DEBUG RELATED DEFINES */