1 /* dfa.h - declarations for GNU deterministic regexp compiler
2 Copyright (C) 1988, 1998 Free Software Foundation, Inc.
4 This program is free software; you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation; either version 2, or (at your option)
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program; if not, write to the Free Software
16 Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA */
18 /* Written June, 1988 by Mike Haertel */
20 /* $FreeBSD: src/gnu/usr.bin/grep/dfa.h,v 1.6 2000/01/04 03:25:39 obrien Exp $ */
23 2. We should not export so much of the DFA internals.
24 In addition to clobbering modularity, we eat up valuable
42 /* Number of bits in an unsigned char. */
47 /* First integer value that is greater than any character code. */
48 #define NOTCHAR (1 << CHARBITS)
50 /* INTBITS need not be exact, just a lower bound. */
52 #define INTBITS (CHARBITS * sizeof (int))
55 /* Number of ints required to hold a bit for every character. */
56 #define CHARCLASS_INTS ((NOTCHAR + INTBITS - 1) / INTBITS)
58 /* Sets of unsigned characters are stored as bit vectors in arrays of ints. */
59 typedef int charclass[CHARCLASS_INTS];
61 /* The regexp is parsed into an array of tokens in postfix form. Some tokens
62 are operators and others are terminal symbols. Most (but not all) of these
63 codes are returned by the lexical analyzer. */
67 END = -1, /* END is a terminal symbol that matches the
68 end of input; any value of END or less in
69 the parse tree is such a symbol. Accepting
70 states of the DFA are those that would have
71 a transition on END. */
73 /* Ordinary character values are terminal symbols that match themselves. */
75 EMPTY = NOTCHAR, /* EMPTY is a terminal symbol that matches
78 BACKREF, /* BACKREF is generated by \<digit>; it
79 it not completely handled. If the scanner
80 detects a transition on backref, it returns
81 a kind of "semi-success" indicating that
82 the match will have to be verified with
83 a backtracking matcher. */
85 BEGLINE, /* BEGLINE is a terminal symbol that matches
86 the empty string if it is at the beginning
89 ENDLINE, /* ENDLINE is a terminal symbol that matches
90 the empty string if it is at the end of
93 BEGWORD, /* BEGWORD is a terminal symbol that matches
94 the empty string if it is at the beginning
97 ENDWORD, /* ENDWORD is a terminal symbol that matches
98 the empty string if it is at the end of
101 LIMWORD, /* LIMWORD is a terminal symbol that matches
102 the empty string if it is at the beginning
103 or the end of a word. */
105 NOTLIMWORD, /* NOTLIMWORD is a terminal symbol that
106 matches the empty string if it is not at
107 the beginning or end of a word. */
109 QMARK, /* QMARK is an operator of one argument that
110 matches zero or one occurences of its
113 STAR, /* STAR is an operator of one argument that
114 matches the Kleene closure (zero or more
115 occurrences) of its argument. */
117 PLUS, /* PLUS is an operator of one argument that
118 matches the positive closure (one or more
119 occurrences) of its argument. */
121 REPMN, /* REPMN is a lexical token corresponding
122 to the {m,n} construct. REPMN never
123 appears in the compiled token vector. */
125 CAT, /* CAT is an operator of two arguments that
126 matches the concatenation of its
127 arguments. CAT is never returned by the
130 OR, /* OR is an operator of two arguments that
131 matches either of its arguments. */
133 ORTOP, /* OR at the toplevel in the parse tree.
134 This is used for a boyer-moore heuristic. */
136 LPAREN, /* LPAREN never appears in the parse tree,
137 it is only a lexeme. */
139 RPAREN, /* RPAREN never appears in the parse tree. */
141 CSET /* CSET and (and any value greater) is a
142 terminal symbol that matches any of a
143 class of characters. */
146 /* Sets are stored in an array in the compiled dfa; the index of the
147 array corresponding to a given set token is given by SET_INDEX(t). */
148 #define SET_INDEX(t) ((t) - CSET)
150 /* Sometimes characters can only be matched depending on the surrounding
151 context. Such context decisions depend on what the previous character
152 was, and the value of the current (lookahead) character. Context
153 dependent constraints are encoded as 8 bit integers. Each bit that
154 is set indicates that the constraint succeeds in the corresponding
157 bit 7 - previous and current are newlines
158 bit 6 - previous was newline, current isn't
159 bit 5 - previous wasn't newline, current is
160 bit 4 - neither previous nor current is a newline
161 bit 3 - previous and current are word-constituents
162 bit 2 - previous was word-constituent, current isn't
163 bit 1 - previous wasn't word-constituent, current is
164 bit 0 - neither previous nor current is word-constituent
166 Word-constituent characters are those that satisfy isalnum().
168 The macro SUCCEEDS_IN_CONTEXT determines whether a a given constraint
169 succeeds in a particular context. Prevn is true if the previous character
170 was a newline, currn is true if the lookahead character is a newline.
171 Prevl and currl similarly depend upon whether the previous and current
172 characters are word-constituent letters. */
173 #define MATCHES_NEWLINE_CONTEXT(constraint, prevn, currn) \
174 ((constraint) & 1 << (((prevn) ? 2 : 0) + ((currn) ? 1 : 0) + 4))
175 #define MATCHES_LETTER_CONTEXT(constraint, prevl, currl) \
176 ((constraint) & 1 << (((prevl) ? 2 : 0) + ((currl) ? 1 : 0)))
177 #define SUCCEEDS_IN_CONTEXT(constraint, prevn, currn, prevl, currl) \
178 (MATCHES_NEWLINE_CONTEXT(constraint, prevn, currn) \
179 && MATCHES_LETTER_CONTEXT(constraint, prevl, currl))
181 /* The following macros give information about what a constraint depends on. */
182 #define PREV_NEWLINE_DEPENDENT(constraint) \
183 (((constraint) & 0xc0) >> 2 != ((constraint) & 0x30))
184 #define PREV_LETTER_DEPENDENT(constraint) \
185 (((constraint) & 0x0c) >> 2 != ((constraint) & 0x03))
187 /* Tokens that match the empty string subject to some constraint actually
188 work by applying that constraint to determine what may follow them,
189 taking into account what has gone before. The following values are
190 the constraints corresponding to the special tokens previously defined. */
191 #define NO_CONSTRAINT 0xff
192 #define BEGLINE_CONSTRAINT 0xcf
193 #define ENDLINE_CONSTRAINT 0xaf
194 #define BEGWORD_CONSTRAINT 0xf2
195 #define ENDWORD_CONSTRAINT 0xf4
196 #define LIMWORD_CONSTRAINT 0xf6
197 #define NOTLIMWORD_CONSTRAINT 0xf9
199 /* States of the recognizer correspond to sets of positions in the parse
200 tree, together with the constraints under which they may be matched.
201 So a position is encoded as an index into the parse tree together with
205 unsigned index; /* Index into the parse array. */
206 unsigned constraint; /* Constraint for matching this position. */
209 /* Sets of positions are stored as arrays. */
212 position *elems; /* Elements of this position set. */
213 int nelem; /* Number of elements in this set. */
216 /* A state of the dfa consists of a set of positions, some flags,
217 and the token value of the lowest-numbered position of the state that
218 contains an END token. */
221 int hash; /* Hash of the positions of this state. */
222 position_set elems; /* Positions this state could match. */
223 char newline; /* True if previous state matched newline. */
224 char letter; /* True if previous state matched a letter. */
225 char backref; /* True if this state matches a \<digit>. */
226 unsigned char constraint; /* Constraint for this state to accept. */
227 int first_end; /* Token value of the first END in elems. */
230 /* Element of a list of strings, at least one of which is known to
231 appear in any R.E. matching the DFA. */
236 struct dfamust *next;
239 /* A compiled regular expression. */
242 /* Stuff built by the scanner. */
243 charclass *charclasses; /* Array of character sets for CSET tokens. */
244 int cindex; /* Index for adding new charclasses. */
245 int calloc; /* Number of charclasses currently allocated. */
247 /* Stuff built by the parser. */
248 token *tokens; /* Postfix parse array. */
249 int tindex; /* Index for adding new tokens. */
250 int talloc; /* Number of tokens currently allocated. */
251 int depth; /* Depth required of an evaluation stack
252 used for depth-first traversal of the
254 int nleaves; /* Number of leaves on the parse tree. */
255 int nregexps; /* Count of parallel regexps being built
258 /* Stuff owned by the state builder. */
259 dfa_state *states; /* States of the dfa. */
260 int sindex; /* Index for adding new states. */
261 int salloc; /* Number of states currently allocated. */
263 /* Stuff built by the structure analyzer. */
264 position_set *follows; /* Array of follow sets, indexed by position
265 index. The follow of a position is the set
266 of positions containing characters that
267 could conceivably follow a character
268 matching the given position in a string
269 matching the regexp. Allocated to the
270 maximum possible position index. */
271 int searchflag; /* True if we are supposed to build a searching
272 as opposed to an exact matcher. A searching
273 matcher finds the first and shortest string
274 matching a regexp anywhere in the buffer,
275 whereas an exact matcher finds the longest
276 string matching, but anchored to the
277 beginning of the buffer. */
279 /* Stuff owned by the executor. */
280 int tralloc; /* Number of transition tables that have
282 int trcount; /* Number of transition tables that have
283 actually been built. */
284 int **trans; /* Transition tables for states that can
285 never accept. If the transitions for a
286 state have not yet been computed, or the
287 state could possibly accept, its entry in
288 this table is NULL. */
289 int **realtrans; /* Trans always points to realtrans + 1; this
290 is so trans[-1] can contain NULL. */
291 int **fails; /* Transition tables after failing to accept
292 on a state that potentially could do so. */
293 int *success; /* Table of acceptance conditions used in
294 dfaexec and computed in build_state. */
295 int *newlines; /* Transitions on newlines. The entry for a
296 newline in any transition table is always
297 -1 so we can count lines without wasting
298 too many cycles. The transition for a
299 newline is stored separately and handled
300 as a special case. Newline is also used
301 as a sentinel at the end of the buffer. */
302 struct dfamust *musts; /* List of strings, at least one of which
303 is known to appear in any r.e. matching
307 /* Some macros for user access to dfa internals. */
309 /* ACCEPTING returns true if s could possibly be an accepting state of r. */
310 #define ACCEPTING(s, r) ((r).states[s].constraint)
312 /* ACCEPTS_IN_CONTEXT returns true if the given state accepts in the
313 specified context. */
314 #define ACCEPTS_IN_CONTEXT(prevn, currn, prevl, currl, state, dfa) \
315 SUCCEEDS_IN_CONTEXT((dfa).states[state].constraint, \
316 prevn, currn, prevl, currl)
318 /* FIRST_MATCHING_REGEXP returns the index number of the first of parallel
319 regexps that a given state could accept. Parallel regexps are numbered
321 #define FIRST_MATCHING_REGEXP(state, dfa) (-(dfa).states[state].first_end)
325 /* dfasyntax() takes three arguments; the first sets the syntax bits described
326 earlier in this file, the second sets the case-folding flag, and the
327 third specifies the line terminator. */
328 extern void dfasyntax PARAMS ((reg_syntax_t, int, int));
330 /* Compile the given string of the given length into the given struct dfa.
331 Final argument is a flag specifying whether to build a searching or an
333 extern void dfacomp PARAMS ((char *, size_t, struct dfa *, int));
335 /* Execute the given struct dfa on the buffer of characters. The
336 first char * points to the beginning, and the second points to the
337 first character after the end of the buffer, which must be a writable
338 place so a sentinel end-of-buffer marker can be stored there. The
339 second-to-last argument is a flag telling whether to allow newlines to
340 be part of a string matching the regexp. The next-to-last argument,
341 if non-NULL, points to a place to increment every time we see a
342 newline. The final argument, if non-NULL, points to a flag that will
343 be set if further examination by a backtracking matcher is needed in
344 order to verify backreferencing; otherwise the flag will be cleared.
345 Returns NULL if no match is found, or a pointer to the first
346 character after the first & shortest matching string in the buffer. */
347 extern char *dfaexec PARAMS ((struct dfa *, char *, char *, int, int *, int *));
349 /* Free the storage held by the components of a struct dfa. */
350 extern void dfafree PARAMS ((struct dfa *));
352 /* Entry points for people who know what they're doing. */
354 /* Initialize the components of a struct dfa. */
355 extern void dfainit PARAMS ((struct dfa *));
357 /* Incrementally parse a string of given length into a struct dfa. */
358 extern void dfaparse PARAMS ((char *, size_t, struct dfa *));
360 /* Analyze a parsed regexp; second argument tells whether to build a searching
361 or an exact matcher. */
362 extern void dfaanalyze PARAMS ((struct dfa *, int));
364 /* Compute, for each possible character, the transitions out of a given
365 state, storing them in an array of integers. */
366 extern void dfastate PARAMS ((int, struct dfa *, int []));
368 /* Error handling. */
370 /* dfaerror() is called by the regexp routines whenever an error occurs. It
371 takes a single argument, a NUL-terminated string describing the error.
372 The default dfaerror() prints the error message to stderr and exits.
373 The user can provide a different dfafree() if so desired. */
374 extern void dfaerror PARAMS ((const char *));