This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Introduction of d_pseudofork
[perl5.git] / regcomp.h
CommitLineData
a0d0e21e 1/* regcomp.h
d6376244 2 *
4bb101f2 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
1d325971 4 * 2000, 2001, 2002, 2003, 2005 by Larry Wall and others
d6376244
JH
5 *
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.
8 *
a687059c
LW
9 */
10
c277df42
IZ
11typedef OP OP_4tree; /* Will be redefined later. */
12
07be1b83 13
58e23c8d 14/* Convert branch sequences to more efficient trie ops? */
07be1b83 15#define PERL_ENABLE_TRIE_OPTIMISATION 1
58e23c8d
YO
16
17/* Be really agressive about optimising patterns with trie sequences? */
07be1b83 18#define PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION 1
58e23c8d
YO
19
20/* Should the optimiser take positive assertions into account? */
1de06328 21#define PERL_ENABLE_POSITIVE_ASSERTION_STUDY 1
58e23c8d
YO
22
23/* Not for production use: */
07be1b83 24#define PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS 0
58e23c8d 25
b9b4dddf
YO
26/* Unless the next line is uncommented it is illegal to combine lazy
27 matching with possessive matching. Frankly it doesn't make much sense
28 to allow it as X*?+ matches nothing, X+?+ matches a single char only,
29 and X{min,max}?+ matches min times only.
30 */
31/* #define REG_ALLOW_MINMOD_SUSPEND */
07be1b83 32
a687059c
LW
33/*
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:
37 *
bd61b366 38 * regstart sv that must begin a match; NULL if none obvious
a687059c
LW
39 * reganch is the match anchored (at beginning-of-line only)?
40 * regmust string (pointer into program) that match must include, or NULL
79072805 41 * [regmust changed to SV* for bminstr()--law]
a687059c
LW
42 * regmlen length of regmust string
43 * [regmlen not used currently]
44 *
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
e50aee73 48 * that pregcomp() supplies a regmust only if the r.e. contains something
a687059c
LW
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
e50aee73 51 * supplied because the test in pregexec() needs it and pregcomp() is computing
a687059c
LW
52 * it anyway.
53 * [regmust is now supplied always. The tests that use regmust have a
54 * heuristic that disables the test if it usually matches.]
55 *
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.]
61 */
62
63/*
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
a3621e74
YO
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.
a687059c
LW
78 */
79
a687059c
LW
80/*
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.)
86 *
87 * Using two bytes for the "next" pointer is vast overkill for most things,
88 * but allows patterns to get big without disasters.
89 *
e91177ed 90 * [The "next" pointer is always aligned on an even
a687059c
LW
91 * boundary, and reads the offset directly as a short. Also, there is no
92 * special test to reverse the sign of BACK pointers since the offset is
93 * stored negative.]
94 */
6d5c990f
RGS
95typedef struct regexp_internal {
96 regexp_paren_ofs *swap; /* Swap copy of *startp / *endp */
97 U32 *offsets; /* offset annotations 20001228 MJD
98 data about mapping the program to the
99 string*/
100 regnode *regstclass; /* Optional startclass as identified or constructed
101 by the optimiser */
102 struct reg_data *data; /* Additional miscellaneous data used by the program.
103 Used to make it easier to clone and free arbitrary
104 data that the regops need. Often the ARG field of
105 a regop is an index into this structure */
106 regnode program[1]; /* Unwarranted chumminess with compiler. */
107} regexp_internal;
108
109#define RXi_SET(x,y) (x)->pprivate = (void*)(y)
110#define RXi_GET(x) ((regexp_internal *)((x)->pprivate))
111#define RXi_GET_DECL(r,ri) regexp_internal *ri = RXi_GET(r)
a687059c 112
c277df42 113struct regnode_string {
cd439c50 114 U8 str_len;
c277df42
IZ
115 U8 type;
116 U16 next_off;
cd439c50 117 char string[1];
c277df42 118};
a687059c 119
6bda09f9
YO
120/* Argument bearing node - workhorse,
121 arg1 is often for the data field */
c277df42
IZ
122struct regnode_1 {
123 U8 flags;
124 U8 type;
125 U16 next_off;
126 U32 arg1;
127};
128
6bda09f9
YO
129/* Similar to a regnode_1 but with an extra signed argument */
130struct regnode_2L {
131 U8 flags;
132 U8 type;
133 U16 next_off;
134 U32 arg1;
135 I32 arg2;
136};
137
138/* 'Two field' -- Two 16 bit unsigned args */
c277df42
IZ
139struct regnode_2 {
140 U8 flags;
141 U8 type;
142 U16 next_off;
143 U16 arg1;
144 U16 arg2;
145};
146
6bda09f9 147
936ed897 148#define ANYOF_BITMAP_SIZE 32 /* 256 b/(8 b/B) */
ffc61ed2 149#define ANYOF_CLASSBITMAP_SIZE 4 /* up to 32 (8*4) named classes */
936ed897 150
786e8c11 151/* also used by trie */
936ed897
IZ
152struct regnode_charclass {
153 U8 flags;
154 U8 type;
155 U16 next_off;
ffc61ed2 156 U32 arg1;
19103b91 157 char bitmap[ANYOF_BITMAP_SIZE]; /* only compile-time */
936ed897
IZ
158};
159
19103b91
JH
160struct regnode_charclass_class { /* has [[:blah:]] classes */
161 U8 flags; /* should have ANYOF_CLASS here */
936ed897
IZ
162 U8 type;
163 U16 next_off;
ffc61ed2 164 U32 arg1;
19103b91
JH
165 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time */
166 char classflags[ANYOF_CLASSBITMAP_SIZE]; /* and run-time */
936ed897
IZ
167};
168
9c7e81e8
AD
169/* XXX fix this description.
170 Impose a limit of REG_INFTY on various pattern matching operations
171 to limit stack growth and to avoid "infinite" recursions.
172*/
173/* The default size for REG_INFTY is I16_MAX, which is the same as
174 SHORT_MAX (see perl.h). Unfortunately I16 isn't necessarily 16 bits
175 (see handy.h). On the Cray C90, sizeof(short)==4 and hence I16_MAX is
176 ((1<<31)-1), while on the Cray T90, sizeof(short)==8 and I16_MAX is
177 ((1<<63)-1). To limit stack growth to reasonable sizes, supply a
178 smaller default.
179 --Andy Dougherty 11 June 1998
180*/
181#if SHORTSIZE > 2
182# ifndef REG_INFTY
183# define REG_INFTY ((1<<15)-1)
184# endif
185#endif
186
187#ifndef REG_INFTY
83cfe48c
IZ
188# define REG_INFTY I16_MAX
189#endif
a687059c 190
e91177ed
IZ
191#define ARG_VALUE(arg) (arg)
192#define ARG__SET(arg,val) ((arg) = (val))
c277df42 193
e8e3b0fb
JH
194#undef ARG
195#undef ARG1
196#undef ARG2
197
c277df42
IZ
198#define ARG(p) ARG_VALUE(ARG_LOC(p))
199#define ARG1(p) ARG_VALUE(ARG1_LOC(p))
200#define ARG2(p) ARG_VALUE(ARG2_LOC(p))
6bda09f9 201#define ARG2L(p) ARG_VALUE(ARG2L_LOC(p))
786e8c11 202
c277df42
IZ
203#define ARG_SET(p, val) ARG__SET(ARG_LOC(p), (val))
204#define ARG1_SET(p, val) ARG__SET(ARG1_LOC(p), (val))
205#define ARG2_SET(p, val) ARG__SET(ARG2_LOC(p), (val))
6bda09f9 206#define ARG2L_SET(p, val) ARG__SET(ARG2L_LOC(p), (val))
c277df42 207
e8e3b0fb
JH
208#undef NEXT_OFF
209#undef NODE_ALIGN
210
075abff3
AL
211#define NEXT_OFF(p) ((p)->next_off)
212#define NODE_ALIGN(node)
213#define NODE_ALIGN_FILL(node) ((node)->flags = 0xde) /* deadbeef */
a687059c 214
c277df42
IZ
215#define SIZE_ALIGN NODE_ALIGN
216
e8e3b0fb
JH
217#undef OP
218#undef OPERAND
219#undef MASK
220#undef STRING
221
e91177ed
IZ
222#define OP(p) ((p)->type)
223#define OPERAND(p) (((struct regnode_string *)p)->string)
cd439c50
IZ
224#define MASK(p) ((char*)OPERAND(p))
225#define STR_LEN(p) (((struct regnode_string *)p)->str_len)
226#define STRING(p) (((struct regnode_string *)p)->string)
227#define STR_SZ(l) ((l + sizeof(regnode) - 1) / sizeof(regnode))
228#define NODE_SZ_STR(p) (STR_SZ(STR_LEN(p))+1)
229
e8e3b0fb
JH
230#undef NODE_ALIGN
231#undef ARG_LOC
232#undef NEXTOPER
233#undef PREVOPER
234
e91177ed
IZ
235#define NODE_ALIGN(node)
236#define ARG_LOC(p) (((struct regnode_1 *)p)->arg1)
237#define ARG1_LOC(p) (((struct regnode_2 *)p)->arg1)
238#define ARG2_LOC(p) (((struct regnode_2 *)p)->arg2)
6bda09f9 239#define ARG2L_LOC(p) (((struct regnode_2L *)p)->arg2)
786e8c11 240
e91177ed
IZ
241#define NODE_STEP_REGNODE 1 /* sizeof(regnode)/sizeof(regnode) */
242#define EXTRA_STEP_2ARGS EXTRA_SIZE(struct regnode_2)
243
244#define NODE_STEP_B 4
c277df42
IZ
245
246#define NEXTOPER(p) ((p) + NODE_STEP_REGNODE)
247#define PREVOPER(p) ((p) - NODE_STEP_REGNODE)
248
e91177ed 249#define FILL_ADVANCE_NODE(ptr, op) STMT_START { \
c277df42 250 (ptr)->type = op; (ptr)->next_off = 0; (ptr)++; } STMT_END
e91177ed 251#define FILL_ADVANCE_NODE_ARG(ptr, op, arg) STMT_START { \
c277df42 252 ARG_SET(ptr, arg); FILL_ADVANCE_NODE(ptr, op); (ptr) += 1; } STMT_END
a687059c 253
22c35a8c 254#define REG_MAGIC 0234
a687059c 255
830247a4 256#define SIZE_ONLY (RExC_emit == &PL_regdummy)
c277df42 257
936ed897 258/* Flags for node->flags of ANYOF */
b8c5462f 259
19103b91 260#define ANYOF_CLASS 0x08 /* has [[:blah:]] classes */
ffc61ed2
JH
261#define ANYOF_INVERT 0x04
262#define ANYOF_FOLD 0x02
263#define ANYOF_LOCALE 0x01
b8c5462f 264
653099ff 265/* Used for regstclass only */
ffc61ed2
JH
266#define ANYOF_EOS 0x10 /* Can match an empty string too */
267
268/* There is a character or a range past 0xff */
269#define ANYOF_UNICODE 0x20
1aa99e6b 270#define ANYOF_UNICODE_ALL 0x40 /* Can match any char past 0xff */
ffc61ed2 271
4f66b38d
HS
272/* size of node is large (includes class pointer) */
273#define ANYOF_LARGE 0x80
274
ffc61ed2
JH
275/* Are there any runtime flags on in this node? */
276#define ANYOF_RUNTIME(s) (ANYOF_FLAGS(s) & 0x0f)
277
278#define ANYOF_FLAGS_ALL 0xff
653099ff 279
936ed897 280/* Character classes for node->classflags of ANYOF */
653099ff
GS
281/* Should be synchronized with a table in regprop() */
282/* 2n should pair with 2n+1 */
b8c5462f 283
289d4f09 284#define ANYOF_ALNUM 0 /* \w, PL_utf8_alnum, utf8::IsWord, ALNUM */
b8c5462f 285#define ANYOF_NALNUM 1
aaa51d5e 286#define ANYOF_SPACE 2 /* \s */
b8c5462f
JH
287#define ANYOF_NSPACE 3
288#define ANYOF_DIGIT 4
289#define ANYOF_NDIGIT 5
289d4f09 290#define ANYOF_ALNUMC 6 /* isalnum(3), utf8::IsAlnum, ALNUMC */
b8c5462f
JH
291#define ANYOF_NALNUMC 7
292#define ANYOF_ALPHA 8
293#define ANYOF_NALPHA 9
294#define ANYOF_ASCII 10
295#define ANYOF_NASCII 11
296#define ANYOF_CNTRL 12
297#define ANYOF_NCNTRL 13
298#define ANYOF_GRAPH 14
299#define ANYOF_NGRAPH 15
300#define ANYOF_LOWER 16
301#define ANYOF_NLOWER 17
302#define ANYOF_PRINT 18
303#define ANYOF_NPRINT 19
304#define ANYOF_PUNCT 20
305#define ANYOF_NPUNCT 21
306#define ANYOF_UPPER 22
307#define ANYOF_NUPPER 23
308#define ANYOF_XDIGIT 24
309#define ANYOF_NXDIGIT 25
aaa51d5e
JF
310#define ANYOF_PSXSPC 26 /* POSIX space: \s plus the vertical tab */
311#define ANYOF_NPSXSPC 27
ffc61ed2 312#define ANYOF_BLANK 28 /* GNU extension: space and tab: non-vertical space */
289d4f09 313#define ANYOF_NBLANK 29
b8c5462f 314
aaa51d5e 315#define ANYOF_MAX 32
b8c5462f
JH
316
317/* Backward source code compatibility. */
318
319#define ANYOF_ALNUML ANYOF_ALNUM
320#define ANYOF_NALNUML ANYOF_NALNUM
321#define ANYOF_SPACEL ANYOF_SPACE
322#define ANYOF_NSPACEL ANYOF_NSPACE
323
324/* Utility macros for the bitmap and classes of ANYOF */
325
936ed897
IZ
326#define ANYOF_SIZE (sizeof(struct regnode_charclass))
327#define ANYOF_CLASS_SIZE (sizeof(struct regnode_charclass_class))
b8c5462f 328
936ed897 329#define ANYOF_FLAGS(p) ((p)->flags)
b8c5462f
JH
330
331#define ANYOF_BIT(c) (1 << ((c) & 7))
332
936ed897 333#define ANYOF_CLASS_BYTE(p, c) (((struct regnode_charclass_class*)(p))->classflags[((c) >> 3) & 3])
b8c5462f
JH
334#define ANYOF_CLASS_SET(p, c) (ANYOF_CLASS_BYTE(p, c) |= ANYOF_BIT(c))
335#define ANYOF_CLASS_CLEAR(p, c) (ANYOF_CLASS_BYTE(p, c) &= ~ANYOF_BIT(c))
336#define ANYOF_CLASS_TEST(p, c) (ANYOF_CLASS_BYTE(p, c) & ANYOF_BIT(c))
337
936ed897
IZ
338#define ANYOF_CLASS_ZERO(ret) Zero(((struct regnode_charclass_class*)(ret))->classflags, ANYOF_CLASSBITMAP_SIZE, char)
339#define ANYOF_BITMAP_ZERO(ret) Zero(((struct regnode_charclass*)(ret))->bitmap, ANYOF_BITMAP_SIZE, char)
340
341#define ANYOF_BITMAP(p) (((struct regnode_charclass*)(p))->bitmap)
786e8c11 342#define ANYOF_BITMAP_BYTE(p, c) (ANYOF_BITMAP(p)[(((U8)(c)) >> 3) & 31])
b8c5462f
JH
343#define ANYOF_BITMAP_SET(p, c) (ANYOF_BITMAP_BYTE(p, c) |= ANYOF_BIT(c))
344#define ANYOF_BITMAP_CLEAR(p,c) (ANYOF_BITMAP_BYTE(p, c) &= ~ANYOF_BIT(c))
345#define ANYOF_BITMAP_TEST(p, c) (ANYOF_BITMAP_BYTE(p, c) & ANYOF_BIT(c))
346
f8bef550
NC
347#define ANYOF_BITMAP_SETALL(p) \
348 memset (ANYOF_BITMAP(p), 255, ANYOF_BITMAP_SIZE)
349#define ANYOF_BITMAP_CLEARALL(p) \
350 Zero (ANYOF_BITMAP(p), ANYOF_BITMAP_SIZE)
351/* Check that all 256 bits are all set. Used in S_cl_is_anything() */
352#define ANYOF_BITMAP_TESTALLSET(p) \
353 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)
354
936ed897
IZ
355#define ANYOF_SKIP ((ANYOF_SIZE - 1)/sizeof(regnode))
356#define ANYOF_CLASS_SKIP ((ANYOF_CLASS_SIZE - 1)/sizeof(regnode))
357#define ANYOF_CLASS_ADD_SKIP (ANYOF_CLASS_SKIP - ANYOF_SKIP)
c277df42 358
786e8c11 359
a687059c
LW
360/*
361 * Utility definitions.
362 */
2304df62 363#ifndef CHARMASK
075abff3 364# define UCHARAT(p) ((int)*(const U8*)(p))
a687059c 365#else
075abff3 366# define UCHARAT(p) ((int)*(p)&CHARMASK)
a687059c 367#endif
a687059c 368
c277df42
IZ
369#define EXTRA_SIZE(guy) ((sizeof(guy)-1)/sizeof(struct regnode))
370
3dab1dad
YO
371#define REG_SEEN_ZERO_LEN 0x00000001
372#define REG_SEEN_LOOKBEHIND 0x00000002
373#define REG_SEEN_GPOS 0x00000004
374#define REG_SEEN_EVAL 0x00000008
375#define REG_SEEN_CANY 0x00000010
cce850e4 376#define REG_SEEN_SANY REG_SEEN_CANY /* src bckwrd cmpt */
6bda09f9 377#define REG_SEEN_RECURSE 0x00000020
eaf3ca90 378#define REG_TOP_LEVEL_BRANCHES 0x00000040
e2e6a0f1 379#define REG_SEEN_VERBARG 0x00000080
5d458dd8 380#define REG_SEEN_CUTGROUP 0x00000100
d09b2d29 381
73c4f7a1
GS
382START_EXTERN_C
383
be8e71aa
YO
384#ifdef PLUGGABLE_RE_EXTENSION
385#include "re_nodes.h"
386#else
d09b2d29 387#include "regnodes.h"
be8e71aa 388#endif
d09b2d29 389
4d61ec05 390/* The following have no fixed length. U8 so we can do strchr() on it. */
d09b2d29 391#ifndef DOINIT
4d61ec05 392EXTCONST U8 PL_varies[];
d09b2d29 393#else
4d61ec05 394EXTCONST U8 PL_varies[] = {
0111c4fd 395 BRANCH, BACK, STAR, PLUS, CURLY, CURLYX, REF, REFF, REFFL,
99ed40ff
YO
396 WHILEM, CURLYM, CURLYN, BRANCHJ, IFTHEN, SUSPEND, CLUMP,
397 NREF, NREFF, NREFFL,
398 0
c277df42 399};
d09b2d29 400#endif
c277df42 401
4d61ec05
GS
402/* The following always have a length of 1. U8 we can do strchr() on it. */
403/* (Note that length 1 means "one character" under UTF8, not "one octet".) */
d09b2d29 404#ifndef DOINIT
4d61ec05 405EXTCONST U8 PL_simple[];
d09b2d29 406#else
4d61ec05 407EXTCONST U8 PL_simple[] = {
f33976b4 408 REG_ANY, SANY, CANY,
ffc61ed2
JH
409 ANYOF,
410 ALNUM, ALNUML,
411 NALNUM, NALNUML,
412 SPACE, SPACEL,
413 NSPACE, NSPACEL,
414 DIGIT, NDIGIT,
415 0
c277df42
IZ
416};
417#endif
418
f9f4320a
YO
419#ifndef PLUGGABLE_RE_EXTENSION
420#ifndef DOINIT
421EXTCONST regexp_engine PL_core_reg_engine;
422#else /* DOINIT */
423EXTCONST regexp_engine PL_core_reg_engine = {
3c6f95d1 424 Perl_re_compile,
f9f4320a
YO
425 Perl_regexec_flags,
426 Perl_re_intuit_start,
427 Perl_re_intuit_string,
f8149455 428 Perl_regfree_internal,
f9f4320a 429#if defined(USE_ITHREADS)
f8149455 430 Perl_regdupe_internal
f9f4320a
YO
431#endif
432};
433#endif /* DOINIT */
434#endif /* PLUGGABLE_RE_EXTENSION */
435
436
73c4f7a1 437END_EXTERN_C
cad2e5aa 438
cad2e5aa 439
261faec3
MJD
440/* .what is a character array with one character for each member of .data
441 * The character describes the function of the corresponding .data item:
0111c4fd 442 * f - start-class data for regstclass optimization
261faec3
MJD
443 * n - Root of op tree for (?{EVAL}) item
444 * o - Start op for (?{EVAL}) item
3dab1dad 445 * p - Pad for (?{EVAL}) item
9e55ce06
JH
446 * s - swash for unicode-style character class, and the multicharacter
447 * strings resulting from casefolding the single-character entries
448 * in the character class
a3621e74 449 * t - trie struct
55eed653 450 * u - trie struct's widecharmap (a HV, so can't share, must dup)
2b8b4781 451 * also used for revcharmap and words under DEBUGGING
07be1b83 452 * T - aho-trie struct
81714fb9 453 * S - sv for named capture lookup
261faec3 454 * 20010712 mjd@plover.com
c3613ed7 455 * (Remember to update re_dup() and pregfree() if you add any items.)
261faec3 456 */
cad2e5aa
JH
457struct reg_data {
458 U32 count;
459 U8 *what;
460 void* data[1];
461};
462
463struct reg_substr_datum {
464 I32 min_offset;
465 I32 max_offset;
33b8afdf
JH
466 SV *substr; /* non-utf8 variant */
467 SV *utf8_substr; /* utf8 variant */
1de06328 468 I32 end_shift;
cad2e5aa
JH
469};
470
471struct reg_substr_data {
472 struct reg_substr_datum data[3]; /* Actual array */
473};
474
475#define anchored_substr substrs->data[0].substr
33b8afdf 476#define anchored_utf8 substrs->data[0].utf8_substr
cad2e5aa 477#define anchored_offset substrs->data[0].min_offset
1de06328
YO
478#define anchored_end_shift substrs->data[0].end_shift
479
cad2e5aa 480#define float_substr substrs->data[1].substr
33b8afdf 481#define float_utf8 substrs->data[1].utf8_substr
cad2e5aa
JH
482#define float_min_offset substrs->data[1].min_offset
483#define float_max_offset substrs->data[1].max_offset
1de06328
YO
484#define float_end_shift substrs->data[1].end_shift
485
cad2e5aa 486#define check_substr substrs->data[2].substr
33b8afdf 487#define check_utf8 substrs->data[2].utf8_substr
cad2e5aa
JH
488#define check_offset_min substrs->data[2].min_offset
489#define check_offset_max substrs->data[2].max_offset
1de06328 490#define check_end_shift substrs->data[2].end_shift
a3621e74
YO
491
492
493
494/* trie related stuff */
5d9a96ca 495
a3621e74
YO
496/* a transition record for the state machine. the
497 check field determines which state "owns" the
498 transition. the char the transition is for is
499 determined by offset from the owning states base
500 field. the next field determines which state
501 is to be transitioned to if any.
502*/
503struct _reg_trie_trans {
504 U32 next;
505 U32 check;
506};
507
508/* a transition list element for the list based representation */
509struct _reg_trie_trans_list_elem {
510 U16 forid;
511 U32 newstate;
512};
513typedef struct _reg_trie_trans_list_elem reg_trie_trans_le;
514
515/* a state for compressed nodes. base is an offset
516 into an array of reg_trie_trans array. If wordnum is
517 nonzero the state is accepting. if base is zero then
518 the state has no children (and will be accepting)
519*/
520struct _reg_trie_state {
521 U16 wordnum;
522 union {
523 U32 base;
524 reg_trie_trans_le* list;
525 } trans;
526};
527
528
529
a3621e74
YO
530typedef struct _reg_trie_state reg_trie_state;
531typedef struct _reg_trie_trans reg_trie_trans;
532
533
534/* anything in here that needs to be freed later
23eab42c
NC
535 should be dealt with in pregfree.
536 refcount is first in both this and _reg_ac_data to allow a space
537 optimisation in Perl_regdupe. */
a3621e74 538struct _reg_trie_data {
23eab42c 539 U32 refcount; /* number of times this trie is referenced */
786e8c11
YO
540 U16 uniquecharcount; /* unique chars in trie (width of trans table) */
541 U32 lasttrans; /* last valid transition element */
542 U16 *charmap; /* byte to charid lookup array */
786e8c11
YO
543 reg_trie_state *states; /* state data */
544 reg_trie_trans *trans; /* array of transition elements */
545 char *bitmap; /* stclass bitmap */
786e8c11
YO
546 U32 startstate; /* initial state - used for common prefix optimisation */
547 STRLEN minlen; /* minimum length of words in trie - build/opt only? */
548 STRLEN maxlen; /* maximum length of words in trie - build/opt only? */
549 U32 *wordlen; /* array of lengths of words */
550 U16 *jump; /* optional 1 indexed array of offsets before tail
551 for the node following a given word. */
552 U16 *nextword; /* optional 1 indexed array to support linked list
553 of duplicate wordnums */
1e2e3d02
YO
554 U32 statecount; /* Build only - number of states in the states array
555 (including the unused zero state) */
786e8c11 556 U32 wordcount; /* Build only */
a3621e74 557#ifdef DEBUGGING
786e8c11 558 STRLEN charcount; /* Build only */
a3621e74
YO
559#endif
560};
2b8b4781
NC
561/* There is one (3 under DEBUGGING) pointers that logically belong in this
562 structure, but are held outside as they need duplication on thread cloning,
563 whereas the rest of the structure can be read only:
564 HV *widecharmap; code points > 255 to charid
565#ifdef DEBUGGING
566 AV *words; Array of words contained in trie, for dumping
567 AV *revcharmap; Map of each charid back to its character representation
568#endif
569*/
570
571#define TRIE_WORDS_OFFSET 2
572
a3621e74
YO
573typedef struct _reg_trie_data reg_trie_data;
574
23eab42c
NC
575/* refcount is first in both this and _reg_trie_data to allow a space
576 optimisation in Perl_regdupe. */
07be1b83 577struct _reg_ac_data {
23eab42c 578 U32 refcount;
07be1b83
YO
579 U32 *fail;
580 reg_trie_state *states;
3251b653 581 U32 trie;
07be1b83
YO
582};
583typedef struct _reg_ac_data reg_ac_data;
584
3dab1dad
YO
585/* ANY_BIT doesnt use the structure, so we can borrow it here.
586 This is simpler than refactoring all of it as wed end up with
587 three different sets... */
588
589#define TRIE_BITMAP(p) (((reg_trie_data *)(p))->bitmap)
786e8c11 590#define TRIE_BITMAP_BYTE(p, c) (TRIE_BITMAP(p)[(((U8)(c)) >> 3) & 31])
5dfe062f
YO
591#define TRIE_BITMAP_SET(p, c) (TRIE_BITMAP_BYTE(p, c) |= ANYOF_BIT((U8)c))
592#define TRIE_BITMAP_CLEAR(p,c) (TRIE_BITMAP_BYTE(p, c) &= ~ANYOF_BIT((U8)c))
593#define TRIE_BITMAP_TEST(p, c) (TRIE_BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
3dab1dad 594
1de06328
YO
595#define IS_ANYOF_TRIE(op) ((op)==TRIEC || (op)==AHOCORASICKC)
596#define IS_TRIE_AC(op) ((op)>=AHOCORASICK)
597
598
786e8c11
YO
599#define BITMAP_BYTE(p, c) (((U8*)p)[(((U8)(c)) >> 3) & 31])
600#define BITMAP_TEST(p, c) (BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
3dab1dad 601
a3621e74
YO
602/* these defines assume uniquecharcount is the correct variable, and state may be evaluated twice */
603#define TRIE_NODENUM(state) (((state)-1)/(trie->uniquecharcount)+1)
604#define SAFE_TRIE_NODENUM(state) ((state) ? (((state)-1)/(trie->uniquecharcount)+1) : (state))
605#define TRIE_NODEIDX(state) ((state) ? (((state)-1)*(trie->uniquecharcount)+1) : (state))
606
3dab1dad 607#ifdef DEBUGGING
3dab1dad 608#define TRIE_CHARCOUNT(trie) ((trie)->charcount)
3dab1dad 609#else
3dab1dad 610#define TRIE_CHARCOUNT(trie) (trie_charcount)
3dab1dad 611#endif
a3621e74 612
0111c4fd
RGS
613#define RE_TRIE_MAXBUF_INIT 65536
614#define RE_TRIE_MAXBUF_NAME "\022E_TRIE_MAXBUF"
a3621e74
YO
615#define RE_DEBUG_FLAGS "\022E_DEBUG_FLAGS"
616
be8e71aa
YO
617/*
618
619RE_DEBUG_FLAGS is used to control what debug output is emitted
620its divided into three groups of options, some of which interact.
621The three groups are: Compile, Execute, Extra. There is room for a
622further group, as currently only the low three bytes are used.
623
624 Compile Options:
07be1b83 625
be8e71aa
YO
626 PARSE
627 PEEP
628 TRIE
629 PROGRAM
630 OFFSETS
a3621e74 631
be8e71aa 632 Execute Options:
a3621e74 633
be8e71aa
YO
634 INTUIT
635 MATCH
636 TRIE
a3621e74 637
be8e71aa 638 Extra Options
a3621e74 639
be8e71aa
YO
640 TRIE
641 OFFSETS
642
643If you modify any of these make sure you make corresponding changes to
644re.pm, especially to the documentation.
a3621e74 645
be8e71aa
YO
646*/
647
648
649/* Compile */
650#define RE_DEBUG_COMPILE_MASK 0x0000FF
651#define RE_DEBUG_COMPILE_PARSE 0x000001
652#define RE_DEBUG_COMPILE_OPTIMISE 0x000002
653#define RE_DEBUG_COMPILE_TRIE 0x000004
654#define RE_DEBUG_COMPILE_DUMP 0x000008
be8e71aa
YO
655
656/* Execute */
657#define RE_DEBUG_EXECUTE_MASK 0x00FF00
658#define RE_DEBUG_EXECUTE_INTUIT 0x000100
659#define RE_DEBUG_EXECUTE_MATCH 0x000200
660#define RE_DEBUG_EXECUTE_TRIE 0x000400
661
662/* Extra */
663#define RE_DEBUG_EXTRA_MASK 0xFF0000
664#define RE_DEBUG_EXTRA_TRIE 0x010000
665#define RE_DEBUG_EXTRA_OFFSETS 0x020000
a5ca303d
YO
666#define RE_DEBUG_EXTRA_OFFDEBUG 0x040000
667#define RE_DEBUG_EXTRA_STATE 0x080000
668#define RE_DEBUG_EXTRA_OPTIMISE 0x100000
24b23f37
YO
669/* combined */
670#define RE_DEBUG_EXTRA_STACK 0x280000
be8e71aa 671
e68ec53f 672#define RE_DEBUG_FLAG(x) (re_debug_flags & x)
be8e71aa
YO
673/* Compile */
674#define DEBUG_COMPILE_r(x) DEBUG_r( \
e68ec53f 675 if (re_debug_flags & RE_DEBUG_COMPILE_MASK) x )
be8e71aa 676#define DEBUG_PARSE_r(x) DEBUG_r( \
e68ec53f 677 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x )
be8e71aa 678#define DEBUG_OPTIMISE_r(x) DEBUG_r( \
e68ec53f 679 if (re_debug_flags & RE_DEBUG_COMPILE_OPTIMISE) x )
be8e71aa 680#define DEBUG_PARSE_r(x) DEBUG_r( \
e68ec53f 681 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x )
be8e71aa 682#define DEBUG_DUMP_r(x) DEBUG_r( \
e68ec53f 683 if (re_debug_flags & RE_DEBUG_COMPILE_DUMP) x )
be8e71aa 684#define DEBUG_TRIE_COMPILE_r(x) DEBUG_r( \
e68ec53f 685 if (re_debug_flags & RE_DEBUG_COMPILE_TRIE) x )
be8e71aa
YO
686
687/* Execute */
688#define DEBUG_EXECUTE_r(x) DEBUG_r( \
e68ec53f 689 if (re_debug_flags & RE_DEBUG_EXECUTE_MASK) x )
be8e71aa 690#define DEBUG_INTUIT_r(x) DEBUG_r( \
e68ec53f 691 if (re_debug_flags & RE_DEBUG_EXECUTE_INTUIT) x )
be8e71aa 692#define DEBUG_MATCH_r(x) DEBUG_r( \
e68ec53f 693 if (re_debug_flags & RE_DEBUG_EXECUTE_MATCH) x )
be8e71aa 694#define DEBUG_TRIE_EXECUTE_r(x) DEBUG_r( \
e68ec53f 695 if (re_debug_flags & RE_DEBUG_EXECUTE_TRIE) x )
be8e71aa
YO
696
697/* Extra */
698#define DEBUG_EXTRA_r(x) DEBUG_r( \
e68ec53f 699 if (re_debug_flags & RE_DEBUG_EXTRA_MASK) x )
a5ca303d
YO
700#define DEBUG_OFFSETS_r(x) DEBUG_r( \
701 if (re_debug_flags & RE_DEBUG_EXTRA_OFFSETS) x )
ab3bbdeb
YO
702#define DEBUG_STATE_r(x) DEBUG_r( \
703 if (re_debug_flags & RE_DEBUG_EXTRA_STATE) x )
24b23f37
YO
704#define DEBUG_STACK_r(x) DEBUG_r( \
705 if (re_debug_flags & RE_DEBUG_EXTRA_STACK) x )
a5ca303d
YO
706#define DEBUG_OPTIMISE_MORE_r(x) DEBUG_r( \
707 if ((RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE) == \
708 (re_debug_flags & (RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE)) ) x )
07be1b83 709#define MJD_OFFSET_DEBUG(x) DEBUG_r( \
a5ca303d 710 if (re_debug_flags & RE_DEBUG_EXTRA_OFFDEBUG) \
be8e71aa
YO
711 Perl_warn_nocontext x )
712#define DEBUG_TRIE_COMPILE_MORE_r(x) DEBUG_TRIE_COMPILE_r( \
e68ec53f 713 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x )
be8e71aa 714#define DEBUG_TRIE_EXECUTE_MORE_r(x) DEBUG_TRIE_EXECUTE_r( \
e68ec53f 715 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x )
be8e71aa
YO
716
717#define DEBUG_TRIE_r(x) DEBUG_r( \
e68ec53f 718 if (re_debug_flags & (RE_DEBUG_COMPILE_TRIE \
be8e71aa
YO
719 | RE_DEBUG_EXECUTE_TRIE )) x )
720
721/* initialization */
722/* get_sv() can return NULL during global destruction. */
e68ec53f
YO
723#define GET_RE_DEBUG_FLAGS DEBUG_r({ \
724 SV * re_debug_flags_sv = NULL; \
725 re_debug_flags_sv = get_sv(RE_DEBUG_FLAGS, 1); \
726 if (re_debug_flags_sv) { \
8e9a8a48 727 if (!SvIOK(re_debug_flags_sv)) \
e68ec53f 728 sv_setuv(re_debug_flags_sv, RE_DEBUG_COMPILE_DUMP | RE_DEBUG_EXECUTE_MASK ); \
8e9a8a48 729 re_debug_flags=SvIV(re_debug_flags_sv); \
e68ec53f
YO
730 }\
731})
a3621e74
YO
732
733#ifdef DEBUGGING
ab3bbdeb 734
8e9a8a48 735#define GET_RE_DEBUG_FLAGS_DECL IV re_debug_flags = 0; GET_RE_DEBUG_FLAGS;
ab3bbdeb
YO
736
737#define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) \
738 const char * const rpv = \
ddc5bc0f 739 pv_pretty((dsv), (pv), (l), (m), \
ab3bbdeb
YO
740 PL_colors[(c1)],PL_colors[(c2)], \
741 ((isuni) ? PERL_PV_ESCAPE_UNI : 0) ); \
0df25f3d 742 const int rlen = SvCUR(dsv)
ab3bbdeb
YO
743
744#define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m) \
745 const char * const rpv = \
ddc5bc0f 746 pv_pretty((dsv), (SvPV_nolen_const(sv)), (SvCUR(sv)), (m), \
ab3bbdeb
YO
747 PL_colors[(c1)],PL_colors[(c2)], \
748 ((isuni) ? PERL_PV_ESCAPE_UNI : 0) )
749
750#define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) \
751 const char * const rpv = \
ddc5bc0f 752 pv_pretty((dsv), (pv), (l), (m), \
ab3bbdeb
YO
753 PL_colors[0], PL_colors[1], \
754 ( PERL_PV_PRETTY_QUOTE | PERL_PV_PRETTY_ELIPSES | \
755 ((isuni) ? PERL_PV_ESCAPE_UNI : 0)) \
756 )
757
758#define RE_SV_DUMPLEN(ItEm) (SvCUR(ItEm) - (SvTAIL(ItEm)!=0))
759#define RE_SV_TAIL(ItEm) (SvTAIL(ItEm) ? "$" : "")
760
761#else /* if not DEBUGGING */
762
a3621e74 763#define GET_RE_DEBUG_FLAGS_DECL
ab3bbdeb
YO
764#define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2)
765#define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m)
766#define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m)
767#define RE_SV_DUMPLEN(ItEm)
768#define RE_SV_TAIL(ItEm)
769
770#endif /* DEBUG RELATED DEFINES */
a3621e74 771
99ed40ff 772