diff options
Diffstat (limited to 'src/regex/regex_internal.c')
-rw-r--r-- | src/regex/regex_internal.c | 3706 |
1 files changed, 3706 insertions, 0 deletions
diff --git a/src/regex/regex_internal.c b/src/regex/regex_internal.c new file mode 100644 index 000000000..944ca9bb9 --- /dev/null +++ b/src/regex/regex_internal.c | |||
@@ -0,0 +1,3706 @@ | |||
1 | /* | ||
2 | This file is part of GNUnet | ||
3 | Copyright (C) 2012 GNUnet e.V. | ||
4 | |||
5 | GNUnet is free software: you can redistribute it and/or modify it | ||
6 | under the terms of the GNU Affero General Public License as published | ||
7 | by the Free Software Foundation, either version 3 of the License, | ||
8 | or (at your option) any later version. | ||
9 | |||
10 | GNUnet is distributed in the hope that it will be useful, but | ||
11 | WITHOUT ANY WARRANTY; without even the implied warranty of | ||
12 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | ||
13 | Affero General Public License for more details. | ||
14 | |||
15 | You should have received a copy of the GNU Affero General Public License | ||
16 | along with this program. If not, see <http://www.gnu.org/licenses/>. | ||
17 | */ | ||
18 | /** | ||
19 | * @file src/regex/regex_internal.c | ||
20 | * @brief library to create Deterministic Finite Automatons (DFAs) from regular | ||
21 | * expressions (regexes). | ||
22 | * @author Maximilian Szengel | ||
23 | */ | ||
24 | #include "platform.h" | ||
25 | #include "gnunet_util_lib.h" | ||
26 | #include "gnunet_regex_service.h" | ||
27 | #include "regex_internal_lib.h" | ||
28 | #include "regex_internal.h" | ||
29 | |||
30 | |||
31 | /** | ||
32 | * Set this to #GNUNET_YES to enable state naming. Used to debug NFA->DFA | ||
33 | * creation. Disabled by default for better performance. | ||
34 | */ | ||
35 | #define REGEX_DEBUG_DFA GNUNET_NO | ||
36 | |||
37 | /** | ||
38 | * Set of states using MDLL API. | ||
39 | */ | ||
40 | struct REGEX_INTERNAL_StateSet_MDLL | ||
41 | { | ||
42 | /** | ||
43 | * MDLL of states. | ||
44 | */ | ||
45 | struct REGEX_INTERNAL_State *head; | ||
46 | |||
47 | /** | ||
48 | * MDLL of states. | ||
49 | */ | ||
50 | struct REGEX_INTERNAL_State *tail; | ||
51 | |||
52 | /** | ||
53 | * Length of the MDLL. | ||
54 | */ | ||
55 | unsigned int len; | ||
56 | }; | ||
57 | |||
58 | |||
59 | /** | ||
60 | * Append state to the given StateSet. | ||
61 | * | ||
62 | * @param set set to be modified | ||
63 | * @param state state to be appended | ||
64 | */ | ||
65 | static void | ||
66 | state_set_append (struct REGEX_INTERNAL_StateSet *set, | ||
67 | struct REGEX_INTERNAL_State *state) | ||
68 | { | ||
69 | if (set->off == set->size) | ||
70 | GNUNET_array_grow (set->states, set->size, set->size * 2 + 4); | ||
71 | set->states[set->off++] = state; | ||
72 | } | ||
73 | |||
74 | |||
75 | /** | ||
76 | * Compare two strings for equality. If either is NULL they are not equal. | ||
77 | * | ||
78 | * @param str1 first string for comparison. | ||
79 | * @param str2 second string for comparison. | ||
80 | * | ||
81 | * @return 0 if the strings are the same or both NULL, 1 or -1 if not. | ||
82 | */ | ||
83 | static int | ||
84 | nullstrcmp (const char *str1, const char *str2) | ||
85 | { | ||
86 | if ((NULL == str1) != (NULL == str2)) | ||
87 | return -1; | ||
88 | if ((NULL == str1) && (NULL == str2)) | ||
89 | return 0; | ||
90 | |||
91 | return strcmp (str1, str2); | ||
92 | } | ||
93 | |||
94 | |||
95 | /** | ||
96 | * Adds a transition from one state to another on @a label. Does not add | ||
97 | * duplicate states. | ||
98 | * | ||
99 | * @param ctx context | ||
100 | * @param from_state starting state for the transition | ||
101 | * @param label transition label | ||
102 | * @param to_state state to where the transition should point to | ||
103 | */ | ||
104 | static void | ||
105 | state_add_transition (struct REGEX_INTERNAL_Context *ctx, | ||
106 | struct REGEX_INTERNAL_State *from_state, | ||
107 | const char *label, | ||
108 | struct REGEX_INTERNAL_State *to_state) | ||
109 | { | ||
110 | struct REGEX_INTERNAL_Transition *t; | ||
111 | struct REGEX_INTERNAL_Transition *oth; | ||
112 | |||
113 | if (NULL == from_state) | ||
114 | { | ||
115 | GNUNET_log (GNUNET_ERROR_TYPE_ERROR, | ||
116 | "Could not create Transition.\n"); | ||
117 | return; | ||
118 | } | ||
119 | |||
120 | /* Do not add duplicate state transitions */ | ||
121 | for (t = from_state->transitions_head; NULL != t; t = t->next) | ||
122 | { | ||
123 | if (t->to_state == to_state && 0 == nullstrcmp (t->label, label) && | ||
124 | t->from_state == from_state) | ||
125 | return; | ||
126 | } | ||
127 | |||
128 | /* sort transitions by label */ | ||
129 | for (oth = from_state->transitions_head; NULL != oth; oth = oth->next) | ||
130 | { | ||
131 | if (0 < nullstrcmp (oth->label, label)) | ||
132 | break; | ||
133 | } | ||
134 | |||
135 | t = GNUNET_new (struct REGEX_INTERNAL_Transition); | ||
136 | if (NULL != ctx) | ||
137 | t->id = ctx->transition_id++; | ||
138 | if (NULL != label) | ||
139 | t->label = GNUNET_strdup (label); | ||
140 | else | ||
141 | t->label = NULL; | ||
142 | t->to_state = to_state; | ||
143 | t->from_state = from_state; | ||
144 | |||
145 | /* Add outgoing transition to 'from_state' */ | ||
146 | from_state->transition_count++; | ||
147 | GNUNET_CONTAINER_DLL_insert_before (from_state->transitions_head, | ||
148 | from_state->transitions_tail, oth, t); | ||
149 | } | ||
150 | |||
151 | |||
152 | /** | ||
153 | * Remove a 'transition' from 'state'. | ||
154 | * | ||
155 | * @param state state from which the to-be-removed transition originates. | ||
156 | * @param transition transition that should be removed from state 'state'. | ||
157 | */ | ||
158 | static void | ||
159 | state_remove_transition (struct REGEX_INTERNAL_State *state, | ||
160 | struct REGEX_INTERNAL_Transition *transition) | ||
161 | { | ||
162 | if (NULL == state || NULL == transition) | ||
163 | return; | ||
164 | |||
165 | if (transition->from_state != state) | ||
166 | return; | ||
167 | |||
168 | GNUNET_free_non_null (transition->label); | ||
169 | |||
170 | state->transition_count--; | ||
171 | GNUNET_CONTAINER_DLL_remove (state->transitions_head, state->transitions_tail, | ||
172 | transition); | ||
173 | |||
174 | GNUNET_free (transition); | ||
175 | } | ||
176 | |||
177 | |||
178 | /** | ||
179 | * Compare two states. Used for sorting. | ||
180 | * | ||
181 | * @param a first state | ||
182 | * @param b second state | ||
183 | * | ||
184 | * @return an integer less than, equal to, or greater than zero | ||
185 | * if the first argument is considered to be respectively | ||
186 | * less than, equal to, or greater than the second. | ||
187 | */ | ||
188 | static int | ||
189 | state_compare (const void *a, const void *b) | ||
190 | { | ||
191 | struct REGEX_INTERNAL_State **s1 = (struct REGEX_INTERNAL_State **) a; | ||
192 | struct REGEX_INTERNAL_State **s2 = (struct REGEX_INTERNAL_State **) b; | ||
193 | |||
194 | return (*s1)->id - (*s2)->id; | ||
195 | } | ||
196 | |||
197 | |||
198 | /** | ||
199 | * Get all edges leaving state @a s. | ||
200 | * | ||
201 | * @param s state. | ||
202 | * @param edges all edges leaving @a s, expected to be allocated and have enough | ||
203 | * space for `s->transitions_count` elements. | ||
204 | * | ||
205 | * @return number of edges. | ||
206 | */ | ||
207 | static unsigned int | ||
208 | state_get_edges (struct REGEX_INTERNAL_State *s, | ||
209 | struct REGEX_BLOCK_Edge *edges) | ||
210 | { | ||
211 | struct REGEX_INTERNAL_Transition *t; | ||
212 | unsigned int count; | ||
213 | |||
214 | if (NULL == s) | ||
215 | return 0; | ||
216 | |||
217 | count = 0; | ||
218 | |||
219 | for (t = s->transitions_head; NULL != t; t = t->next) | ||
220 | { | ||
221 | if (NULL != t->to_state) | ||
222 | { | ||
223 | edges[count].label = t->label; | ||
224 | edges[count].destination = t->to_state->hash; | ||
225 | count++; | ||
226 | } | ||
227 | } | ||
228 | return count; | ||
229 | } | ||
230 | |||
231 | |||
232 | /** | ||
233 | * Compare to state sets by comparing the id's of the states that are contained | ||
234 | * in each set. Both sets are expected to be sorted by id! | ||
235 | * | ||
236 | * @param sset1 first state set | ||
237 | * @param sset2 second state set | ||
238 | * @return 0 if the sets are equal, otherwise non-zero | ||
239 | */ | ||
240 | static int | ||
241 | state_set_compare (struct REGEX_INTERNAL_StateSet *sset1, | ||
242 | struct REGEX_INTERNAL_StateSet *sset2) | ||
243 | { | ||
244 | int result; | ||
245 | unsigned int i; | ||
246 | |||
247 | if (NULL == sset1 || NULL == sset2) | ||
248 | return 1; | ||
249 | |||
250 | result = sset1->off - sset2->off; | ||
251 | if (result < 0) | ||
252 | return -1; | ||
253 | if (result > 0) | ||
254 | return 1; | ||
255 | for (i = 0; i < sset1->off; i++) | ||
256 | if (0 != (result = state_compare (&sset1->states[i], &sset2->states[i]))) | ||
257 | break; | ||
258 | return result; | ||
259 | } | ||
260 | |||
261 | |||
262 | /** | ||
263 | * Clears the given StateSet 'set' | ||
264 | * | ||
265 | * @param set set to be cleared | ||
266 | */ | ||
267 | static void | ||
268 | state_set_clear (struct REGEX_INTERNAL_StateSet *set) | ||
269 | { | ||
270 | GNUNET_array_grow (set->states, set->size, 0); | ||
271 | set->off = 0; | ||
272 | } | ||
273 | |||
274 | |||
275 | /** | ||
276 | * Clears an automaton fragment. Does not destroy the states inside the | ||
277 | * automaton. | ||
278 | * | ||
279 | * @param a automaton to be cleared | ||
280 | */ | ||
281 | static void | ||
282 | automaton_fragment_clear (struct REGEX_INTERNAL_Automaton *a) | ||
283 | { | ||
284 | if (NULL == a) | ||
285 | return; | ||
286 | |||
287 | a->start = NULL; | ||
288 | a->end = NULL; | ||
289 | a->states_head = NULL; | ||
290 | a->states_tail = NULL; | ||
291 | a->state_count = 0; | ||
292 | GNUNET_free (a); | ||
293 | } | ||
294 | |||
295 | |||
296 | /** | ||
297 | * Frees the memory used by State @a s | ||
298 | * | ||
299 | * @param s state that should be destroyed | ||
300 | */ | ||
301 | static void | ||
302 | automaton_destroy_state (struct REGEX_INTERNAL_State *s) | ||
303 | { | ||
304 | struct REGEX_INTERNAL_Transition *t; | ||
305 | struct REGEX_INTERNAL_Transition *next_t; | ||
306 | |||
307 | if (NULL == s) | ||
308 | return; | ||
309 | |||
310 | GNUNET_free_non_null (s->name); | ||
311 | GNUNET_free_non_null (s->proof); | ||
312 | state_set_clear (&s->nfa_set); | ||
313 | for (t = s->transitions_head; NULL != t; t = next_t) | ||
314 | { | ||
315 | next_t = t->next; | ||
316 | state_remove_transition (s, t); | ||
317 | } | ||
318 | |||
319 | GNUNET_free (s); | ||
320 | } | ||
321 | |||
322 | |||
323 | /** | ||
324 | * Remove a state from the given automaton 'a'. Always use this function when | ||
325 | * altering the states of an automaton. Will also remove all transitions leading | ||
326 | * to this state, before destroying it. | ||
327 | * | ||
328 | * @param a automaton | ||
329 | * @param s state to remove | ||
330 | */ | ||
331 | static void | ||
332 | automaton_remove_state (struct REGEX_INTERNAL_Automaton *a, | ||
333 | struct REGEX_INTERNAL_State *s) | ||
334 | { | ||
335 | struct REGEX_INTERNAL_State *s_check; | ||
336 | struct REGEX_INTERNAL_Transition *t_check; | ||
337 | struct REGEX_INTERNAL_Transition *t_check_next; | ||
338 | |||
339 | if (NULL == a || NULL == s) | ||
340 | return; | ||
341 | |||
342 | /* remove all transitions leading to this state */ | ||
343 | for (s_check = a->states_head; NULL != s_check; s_check = s_check->next) | ||
344 | { | ||
345 | for (t_check = s_check->transitions_head; NULL != t_check; | ||
346 | t_check = t_check_next) | ||
347 | { | ||
348 | t_check_next = t_check->next; | ||
349 | if (t_check->to_state == s) | ||
350 | state_remove_transition (s_check, t_check); | ||
351 | } | ||
352 | } | ||
353 | |||
354 | /* remove state */ | ||
355 | GNUNET_CONTAINER_DLL_remove (a->states_head, a->states_tail, s); | ||
356 | a->state_count--; | ||
357 | |||
358 | automaton_destroy_state (s); | ||
359 | } | ||
360 | |||
361 | |||
362 | /** | ||
363 | * Merge two states into one. Will merge 's1' and 's2' into 's1' and destroy | ||
364 | * 's2'. 's1' will contain all (non-duplicate) outgoing transitions of 's2'. | ||
365 | * | ||
366 | * @param ctx context | ||
367 | * @param a automaton | ||
368 | * @param s1 first state | ||
369 | * @param s2 second state, will be destroyed | ||
370 | */ | ||
371 | static void | ||
372 | automaton_merge_states (struct REGEX_INTERNAL_Context *ctx, | ||
373 | struct REGEX_INTERNAL_Automaton *a, | ||
374 | struct REGEX_INTERNAL_State *s1, | ||
375 | struct REGEX_INTERNAL_State *s2) | ||
376 | { | ||
377 | struct REGEX_INTERNAL_State *s_check; | ||
378 | struct REGEX_INTERNAL_Transition *t_check; | ||
379 | struct REGEX_INTERNAL_Transition *t; | ||
380 | struct REGEX_INTERNAL_Transition *t_next; | ||
381 | int is_dup; | ||
382 | |||
383 | if (s1 == s2) | ||
384 | return; | ||
385 | |||
386 | /* 1. Make all transitions pointing to s2 point to s1, unless this transition | ||
387 | * does not already exists, if it already exists remove transition. */ | ||
388 | for (s_check = a->states_head; NULL != s_check; s_check = s_check->next) | ||
389 | { | ||
390 | for (t_check = s_check->transitions_head; NULL != t_check; t_check = t_next) | ||
391 | { | ||
392 | t_next = t_check->next; | ||
393 | |||
394 | if (s2 == t_check->to_state) | ||
395 | { | ||
396 | is_dup = GNUNET_NO; | ||
397 | for (t = t_check->from_state->transitions_head; NULL != t; t = t->next) | ||
398 | { | ||
399 | if (t->to_state == s1 && 0 == strcmp (t_check->label, t->label)) | ||
400 | is_dup = GNUNET_YES; | ||
401 | } | ||
402 | if (GNUNET_NO == is_dup) | ||
403 | t_check->to_state = s1; | ||
404 | else | ||
405 | state_remove_transition (t_check->from_state, t_check); | ||
406 | } | ||
407 | } | ||
408 | } | ||
409 | |||
410 | /* 2. Add all transitions from s2 to sX to s1 */ | ||
411 | for (t_check = s2->transitions_head; NULL != t_check; t_check = t_check->next) | ||
412 | { | ||
413 | if (t_check->to_state != s1) | ||
414 | state_add_transition (ctx, s1, t_check->label, t_check->to_state); | ||
415 | } | ||
416 | |||
417 | /* 3. Rename s1 to {s1,s2} */ | ||
418 | #if REGEX_DEBUG_DFA | ||
419 | char *new_name; | ||
420 | |||
421 | new_name = s1->name; | ||
422 | GNUNET_asprintf (&s1->name, "{%s,%s}", new_name, s2->name); | ||
423 | GNUNET_free (new_name); | ||
424 | #endif | ||
425 | |||
426 | /* remove state */ | ||
427 | GNUNET_CONTAINER_DLL_remove (a->states_head, a->states_tail, s2); | ||
428 | a->state_count--; | ||
429 | automaton_destroy_state (s2); | ||
430 | } | ||
431 | |||
432 | |||
433 | /** | ||
434 | * Add a state to the automaton 'a', always use this function to alter the | ||
435 | * states DLL of the automaton. | ||
436 | * | ||
437 | * @param a automaton to add the state to | ||
438 | * @param s state that should be added | ||
439 | */ | ||
440 | static void | ||
441 | automaton_add_state (struct REGEX_INTERNAL_Automaton *a, | ||
442 | struct REGEX_INTERNAL_State *s) | ||
443 | { | ||
444 | GNUNET_CONTAINER_DLL_insert (a->states_head, a->states_tail, s); | ||
445 | a->state_count++; | ||
446 | } | ||
447 | |||
448 | |||
449 | /** | ||
450 | * Depth-first traversal (DFS) of all states that are reachable from state | ||
451 | * 's'. Performs 'action' on each visited state. | ||
452 | * | ||
453 | * @param s start state. | ||
454 | * @param marks an array of size a->state_count to remember which state was | ||
455 | * already visited. | ||
456 | * @param count current count of the state. | ||
457 | * @param check function that is checked before advancing on each transition | ||
458 | * in the DFS. | ||
459 | * @param check_cls closure for check. | ||
460 | * @param action action to be performed on each state. | ||
461 | * @param action_cls closure for action. | ||
462 | */ | ||
463 | static void | ||
464 | automaton_state_traverse (struct REGEX_INTERNAL_State *s, int *marks, | ||
465 | unsigned int *count, | ||
466 | REGEX_INTERNAL_traverse_check check, void *check_cls, | ||
467 | REGEX_INTERNAL_traverse_action action, void *action_cls) | ||
468 | { | ||
469 | struct REGEX_INTERNAL_Transition *t; | ||
470 | |||
471 | if (GNUNET_YES == marks[s->traversal_id]) | ||
472 | return; | ||
473 | |||
474 | marks[s->traversal_id] = GNUNET_YES; | ||
475 | |||
476 | if (NULL != action) | ||
477 | action (action_cls, *count, s); | ||
478 | |||
479 | (*count)++; | ||
480 | |||
481 | for (t = s->transitions_head; NULL != t; t = t->next) | ||
482 | { | ||
483 | if (NULL == check || | ||
484 | (NULL != check && GNUNET_YES == check (check_cls, s, t))) | ||
485 | { | ||
486 | automaton_state_traverse (t->to_state, marks, count, check, check_cls, | ||
487 | action, action_cls); | ||
488 | } | ||
489 | } | ||
490 | } | ||
491 | |||
492 | |||
493 | /** | ||
494 | * Traverses the given automaton using depth-first-search (DFS) from it's start | ||
495 | * state, visiting all reachable states and calling 'action' on each one of | ||
496 | * them. | ||
497 | * | ||
498 | * @param a automaton to be traversed. | ||
499 | * @param start start state, pass a->start or NULL to traverse the whole automaton. | ||
500 | * @param check function that is checked before advancing on each transition | ||
501 | * in the DFS. | ||
502 | * @param check_cls closure for @a check. | ||
503 | * @param action action to be performed on each state. | ||
504 | * @param action_cls closure for @a action | ||
505 | */ | ||
506 | void | ||
507 | REGEX_INTERNAL_automaton_traverse (const struct REGEX_INTERNAL_Automaton *a, | ||
508 | struct REGEX_INTERNAL_State *start, | ||
509 | REGEX_INTERNAL_traverse_check check, | ||
510 | void *check_cls, | ||
511 | REGEX_INTERNAL_traverse_action action, | ||
512 | void *action_cls) | ||
513 | { | ||
514 | unsigned int count; | ||
515 | struct REGEX_INTERNAL_State *s; | ||
516 | |||
517 | if (NULL == a || 0 == a->state_count) | ||
518 | return; | ||
519 | |||
520 | int marks[a->state_count]; | ||
521 | |||
522 | for (count = 0, s = a->states_head; NULL != s && count < a->state_count; | ||
523 | s = s->next, count++) | ||
524 | { | ||
525 | s->traversal_id = count; | ||
526 | marks[s->traversal_id] = GNUNET_NO; | ||
527 | } | ||
528 | |||
529 | count = 0; | ||
530 | |||
531 | if (NULL == start) | ||
532 | s = a->start; | ||
533 | else | ||
534 | s = start; | ||
535 | |||
536 | automaton_state_traverse (s, marks, &count, | ||
537 | check, check_cls, | ||
538 | action, action_cls); | ||
539 | } | ||
540 | |||
541 | |||
542 | /** | ||
543 | * String container for faster string operations. | ||
544 | */ | ||
545 | struct StringBuffer | ||
546 | { | ||
547 | /** | ||
548 | * Buffer holding the string (may start in the middle!); | ||
549 | * NOT 0-terminated! | ||
550 | */ | ||
551 | char *sbuf; | ||
552 | |||
553 | /** | ||
554 | * Allocated buffer. | ||
555 | */ | ||
556 | char *abuf; | ||
557 | |||
558 | /** | ||
559 | * Length of the string in the buffer. | ||
560 | */ | ||
561 | size_t slen; | ||
562 | |||
563 | /** | ||
564 | * Number of bytes allocated for @e sbuf | ||
565 | */ | ||
566 | unsigned int blen; | ||
567 | |||
568 | /** | ||
569 | * Buffer currently represents "NULL" (not the empty string!) | ||
570 | */ | ||
571 | int16_t null_flag; | ||
572 | |||
573 | /** | ||
574 | * If this entry is part of the last/current generation array, | ||
575 | * this flag is #GNUNET_YES if the last and current generation are | ||
576 | * identical (and thus copying is unnecessary if the value didn't | ||
577 | * change). This is used in an optimization that improves | ||
578 | * performance by about 1% --- if we use int16_t here. With just | ||
579 | * "int" for both flags, performance drops (on my system) significantly, | ||
580 | * most likely due to increased cache misses. | ||
581 | */ | ||
582 | int16_t synced; | ||
583 | |||
584 | }; | ||
585 | |||
586 | |||
587 | /** | ||
588 | * Compare two strings for equality. If either is NULL they are not equal. | ||
589 | * | ||
590 | * @param s1 first string for comparison. | ||
591 | * @param s2 second string for comparison. | ||
592 | * | ||
593 | * @return 0 if the strings are the same or both NULL, 1 or -1 if not. | ||
594 | */ | ||
595 | static int | ||
596 | sb_nullstrcmp (const struct StringBuffer *s1, | ||
597 | const struct StringBuffer *s2) | ||
598 | { | ||
599 | if ( (GNUNET_YES == s1->null_flag) && | ||
600 | (GNUNET_YES == s2->null_flag) ) | ||
601 | return 0; | ||
602 | if ( (GNUNET_YES == s1->null_flag) || | ||
603 | (GNUNET_YES == s2->null_flag) ) | ||
604 | return -1; | ||
605 | if (s1->slen != s2->slen) | ||
606 | return -1; | ||
607 | if (0 == s1->slen) | ||
608 | return 0; | ||
609 | return memcmp (s1->sbuf, s2->sbuf, s1->slen); | ||
610 | } | ||
611 | |||
612 | |||
613 | /** | ||
614 | * Compare two strings for equality. | ||
615 | * | ||
616 | * @param s1 first string for comparison. | ||
617 | * @param s2 second string for comparison. | ||
618 | * | ||
619 | * @return 0 if the strings are the same, 1 or -1 if not. | ||
620 | */ | ||
621 | static int | ||
622 | sb_strcmp (const struct StringBuffer *s1, | ||
623 | const struct StringBuffer *s2) | ||
624 | { | ||
625 | if (s1->slen != s2->slen) | ||
626 | return -1; | ||
627 | if (0 == s1->slen) | ||
628 | return 0; | ||
629 | return memcmp (s1->sbuf, s2->sbuf, s1->slen); | ||
630 | } | ||
631 | |||
632 | |||
633 | /** | ||
634 | * Reallocate the buffer of 'ret' to fit 'nlen' characters; | ||
635 | * move the existing string to the beginning of the new buffer. | ||
636 | * | ||
637 | * @param ret current buffer, to be updated | ||
638 | * @param nlen target length for the buffer, must be at least ret->slen | ||
639 | */ | ||
640 | static void | ||
641 | sb_realloc (struct StringBuffer *ret, | ||
642 | size_t nlen) | ||
643 | { | ||
644 | char *old; | ||
645 | |||
646 | GNUNET_assert (nlen >= ret->slen); | ||
647 | old = ret->abuf; | ||
648 | ret->abuf = GNUNET_malloc (nlen); | ||
649 | ret->blen = nlen; | ||
650 | GNUNET_memcpy (ret->abuf, | ||
651 | ret->sbuf, | ||
652 | ret->slen); | ||
653 | ret->sbuf = ret->abuf; | ||
654 | GNUNET_free_non_null (old); | ||
655 | } | ||
656 | |||
657 | |||
658 | /** | ||
659 | * Append a string. | ||
660 | * | ||
661 | * @param ret where to write the result | ||
662 | * @param sarg string to append | ||
663 | */ | ||
664 | static void | ||
665 | sb_append (struct StringBuffer *ret, | ||
666 | const struct StringBuffer *sarg) | ||
667 | { | ||
668 | if (GNUNET_YES == ret->null_flag) | ||
669 | ret->slen = 0; | ||
670 | ret->null_flag = GNUNET_NO; | ||
671 | if (ret->blen < sarg->slen + ret->slen) | ||
672 | sb_realloc (ret, ret->blen + sarg->slen + 128); | ||
673 | GNUNET_memcpy (&ret->sbuf[ret->slen], | ||
674 | sarg->sbuf, | ||
675 | sarg->slen); | ||
676 | ret->slen += sarg->slen; | ||
677 | } | ||
678 | |||
679 | |||
680 | /** | ||
681 | * Append a C string. | ||
682 | * | ||
683 | * @param ret where to write the result | ||
684 | * @param cstr string to append | ||
685 | */ | ||
686 | static void | ||
687 | sb_append_cstr (struct StringBuffer *ret, | ||
688 | const char *cstr) | ||
689 | { | ||
690 | size_t cstr_len = strlen (cstr); | ||
691 | |||
692 | if (GNUNET_YES == ret->null_flag) | ||
693 | ret->slen = 0; | ||
694 | ret->null_flag = GNUNET_NO; | ||
695 | if (ret->blen < cstr_len + ret->slen) | ||
696 | sb_realloc (ret, ret->blen + cstr_len + 128); | ||
697 | GNUNET_memcpy (&ret->sbuf[ret->slen], | ||
698 | cstr, | ||
699 | cstr_len); | ||
700 | ret->slen += cstr_len; | ||
701 | } | ||
702 | |||
703 | |||
704 | /** | ||
705 | * Wrap a string buffer, that is, set ret to the format string | ||
706 | * which contains an "%s" which is to be replaced with the original | ||
707 | * content of 'ret'. Note that optimizing this function is not | ||
708 | * really worth it, it is rarely called. | ||
709 | * | ||
710 | * @param ret where to write the result and take the input for %.*s from | ||
711 | * @param format format string, fprintf-style, with exactly one "%.*s" | ||
712 | * @param extra_chars how long will the result be, in addition to 'sarg' length | ||
713 | */ | ||
714 | static void | ||
715 | sb_wrap (struct StringBuffer *ret, | ||
716 | const char *format, | ||
717 | size_t extra_chars) | ||
718 | { | ||
719 | char *temp; | ||
720 | |||
721 | if (GNUNET_YES == ret->null_flag) | ||
722 | ret->slen = 0; | ||
723 | ret->null_flag = GNUNET_NO; | ||
724 | temp = GNUNET_malloc (ret->slen + extra_chars + 1); | ||
725 | GNUNET_snprintf (temp, | ||
726 | ret->slen + extra_chars + 1, | ||
727 | format, | ||
728 | (int) ret->slen, | ||
729 | ret->sbuf); | ||
730 | GNUNET_free_non_null (ret->abuf); | ||
731 | ret->abuf = temp; | ||
732 | ret->sbuf = temp; | ||
733 | ret->blen = ret->slen + extra_chars + 1; | ||
734 | ret->slen = ret->slen + extra_chars; | ||
735 | } | ||
736 | |||
737 | |||
738 | /** | ||
739 | * Format a string buffer. Note that optimizing this function is not | ||
740 | * really worth it, it is rarely called. | ||
741 | * | ||
742 | * @param ret where to write the result | ||
743 | * @param format format string, fprintf-style, with exactly one "%.*s" | ||
744 | * @param extra_chars how long will the result be, in addition to 'sarg' length | ||
745 | * @param sarg string to print into the format | ||
746 | */ | ||
747 | static void | ||
748 | sb_printf1 (struct StringBuffer *ret, | ||
749 | const char *format, | ||
750 | size_t extra_chars, | ||
751 | const struct StringBuffer *sarg) | ||
752 | { | ||
753 | if (ret->blen < sarg->slen + extra_chars + 1) | ||
754 | sb_realloc (ret, | ||
755 | sarg->slen + extra_chars + 1); | ||
756 | ret->null_flag = GNUNET_NO; | ||
757 | ret->sbuf = ret->abuf; | ||
758 | ret->slen = sarg->slen + extra_chars; | ||
759 | GNUNET_snprintf (ret->sbuf, | ||
760 | ret->blen, | ||
761 | format, | ||
762 | (int) sarg->slen, | ||
763 | sarg->sbuf); | ||
764 | } | ||
765 | |||
766 | |||
767 | /** | ||
768 | * Format a string buffer. | ||
769 | * | ||
770 | * @param ret where to write the result | ||
771 | * @param format format string, fprintf-style, with exactly two "%.*s" | ||
772 | * @param extra_chars how long will the result be, in addition to 'sarg1/2' length | ||
773 | * @param sarg1 first string to print into the format | ||
774 | * @param sarg2 second string to print into the format | ||
775 | */ | ||
776 | static void | ||
777 | sb_printf2 (struct StringBuffer *ret, | ||
778 | const char *format, | ||
779 | size_t extra_chars, | ||
780 | const struct StringBuffer *sarg1, | ||
781 | const struct StringBuffer *sarg2) | ||
782 | { | ||
783 | if (ret->blen < sarg1->slen + sarg2->slen + extra_chars + 1) | ||
784 | sb_realloc (ret, | ||
785 | sarg1->slen + sarg2->slen + extra_chars + 1); | ||
786 | ret->null_flag = GNUNET_NO; | ||
787 | ret->slen = sarg1->slen + sarg2->slen + extra_chars; | ||
788 | ret->sbuf = ret->abuf; | ||
789 | GNUNET_snprintf (ret->sbuf, | ||
790 | ret->blen, | ||
791 | format, | ||
792 | (int) sarg1->slen, | ||
793 | sarg1->sbuf, | ||
794 | (int) sarg2->slen, | ||
795 | sarg2->sbuf); | ||
796 | } | ||
797 | |||
798 | |||
799 | /** | ||
800 | * Format a string buffer. Note that optimizing this function is not | ||
801 | * really worth it, it is rarely called. | ||
802 | * | ||
803 | * @param ret where to write the result | ||
804 | * @param format format string, fprintf-style, with exactly three "%.*s" | ||
805 | * @param extra_chars how long will the result be, in addition to 'sarg1/2/3' length | ||
806 | * @param sarg1 first string to print into the format | ||
807 | * @param sarg2 second string to print into the format | ||
808 | * @param sarg3 third string to print into the format | ||
809 | */ | ||
810 | static void | ||
811 | sb_printf3 (struct StringBuffer *ret, | ||
812 | const char *format, | ||
813 | size_t extra_chars, | ||
814 | const struct StringBuffer *sarg1, | ||
815 | const struct StringBuffer *sarg2, | ||
816 | const struct StringBuffer *sarg3) | ||
817 | { | ||
818 | if (ret->blen < sarg1->slen + sarg2->slen + sarg3->slen + extra_chars + 1) | ||
819 | sb_realloc (ret, | ||
820 | sarg1->slen + sarg2->slen + sarg3->slen + extra_chars + 1); | ||
821 | ret->null_flag = GNUNET_NO; | ||
822 | ret->slen = sarg1->slen + sarg2->slen + sarg3->slen + extra_chars; | ||
823 | ret->sbuf = ret->abuf; | ||
824 | GNUNET_snprintf (ret->sbuf, | ||
825 | ret->blen, | ||
826 | format, | ||
827 | (int) sarg1->slen, | ||
828 | sarg1->sbuf, | ||
829 | (int) sarg2->slen, | ||
830 | sarg2->sbuf, | ||
831 | (int) sarg3->slen, | ||
832 | sarg3->sbuf); | ||
833 | } | ||
834 | |||
835 | |||
836 | /** | ||
837 | * Free resources of the given string buffer. | ||
838 | * | ||
839 | * @param sb buffer to free (actual pointer is not freed, as they | ||
840 | * should not be individually allocated) | ||
841 | */ | ||
842 | static void | ||
843 | sb_free (struct StringBuffer *sb) | ||
844 | { | ||
845 | GNUNET_array_grow (sb->abuf, | ||
846 | sb->blen, | ||
847 | 0); | ||
848 | sb->slen = 0; | ||
849 | sb->sbuf = NULL; | ||
850 | sb->null_flag= GNUNET_YES; | ||
851 | } | ||
852 | |||
853 | |||
854 | /** | ||
855 | * Copy the given string buffer from 'in' to 'out'. | ||
856 | * | ||
857 | * @param in input string | ||
858 | * @param out output string | ||
859 | */ | ||
860 | static void | ||
861 | sb_strdup (struct StringBuffer *out, | ||
862 | const struct StringBuffer *in) | ||
863 | |||
864 | { | ||
865 | out->null_flag = in->null_flag; | ||
866 | if (GNUNET_YES == out->null_flag) | ||
867 | return; | ||
868 | if (out->blen < in->slen) | ||
869 | { | ||
870 | GNUNET_array_grow (out->abuf, | ||
871 | out->blen, | ||
872 | in->slen); | ||
873 | } | ||
874 | out->sbuf = out->abuf; | ||
875 | out->slen = in->slen; | ||
876 | GNUNET_memcpy (out->sbuf, in->sbuf, out->slen); | ||
877 | } | ||
878 | |||
879 | |||
880 | /** | ||
881 | * Copy the given string buffer from 'in' to 'out'. | ||
882 | * | ||
883 | * @param cstr input string | ||
884 | * @param out output string | ||
885 | */ | ||
886 | static void | ||
887 | sb_strdup_cstr (struct StringBuffer *out, | ||
888 | const char *cstr) | ||
889 | { | ||
890 | if (NULL == cstr) | ||
891 | { | ||
892 | out->null_flag = GNUNET_YES; | ||
893 | return; | ||
894 | } | ||
895 | out->null_flag = GNUNET_NO; | ||
896 | out->slen = strlen (cstr); | ||
897 | if (out->blen < out->slen) | ||
898 | { | ||
899 | GNUNET_array_grow (out->abuf, | ||
900 | out->blen, | ||
901 | out->slen); | ||
902 | } | ||
903 | out->sbuf = out->abuf; | ||
904 | GNUNET_memcpy (out->sbuf, cstr, out->slen); | ||
905 | } | ||
906 | |||
907 | |||
908 | /** | ||
909 | * Check if the given string @a str needs parentheses around it when | ||
910 | * using it to generate a regex. | ||
911 | * | ||
912 | * @param str string | ||
913 | * | ||
914 | * @return #GNUNET_YES if parentheses are needed, #GNUNET_NO otherwise | ||
915 | */ | ||
916 | static int | ||
917 | needs_parentheses (const struct StringBuffer *str) | ||
918 | { | ||
919 | size_t slen; | ||
920 | const char *op; | ||
921 | const char *cl; | ||
922 | const char *pos; | ||
923 | const char *end; | ||
924 | unsigned int cnt; | ||
925 | |||
926 | if ((GNUNET_YES == str->null_flag) || ((slen = str->slen) < 2)) | ||
927 | return GNUNET_NO; | ||
928 | pos = str->sbuf; | ||
929 | if ('(' != pos[0]) | ||
930 | return GNUNET_YES; | ||
931 | end = str->sbuf + slen; | ||
932 | cnt = 1; | ||
933 | pos++; | ||
934 | while (cnt > 0) | ||
935 | { | ||
936 | cl = memchr (pos, ')', end - pos); | ||
937 | if (NULL == cl) | ||
938 | { | ||
939 | GNUNET_break (0); | ||
940 | return GNUNET_YES; | ||
941 | } | ||
942 | /* while '(' before ')', count opening parens */ | ||
943 | while ( (NULL != (op = memchr (pos, '(', end - pos))) && | ||
944 | (op < cl) ) | ||
945 | { | ||
946 | cnt++; | ||
947 | pos = op + 1; | ||
948 | } | ||
949 | /* got ')' first */ | ||
950 | cnt--; | ||
951 | pos = cl + 1; | ||
952 | } | ||
953 | return (*pos == '\0') ? GNUNET_NO : GNUNET_YES; | ||
954 | } | ||
955 | |||
956 | |||
957 | /** | ||
958 | * Remove parentheses surrounding string @a str. | ||
959 | * Example: "(a)" becomes "a", "(a|b)|(a|c)" stays the same. | ||
960 | * You need to #GNUNET_free() the returned string. | ||
961 | * | ||
962 | * @param str string, modified to contain a | ||
963 | * @return string without surrounding parentheses, string 'str' if no preceding | ||
964 | * epsilon could be found, NULL if 'str' was NULL | ||
965 | */ | ||
966 | static void | ||
967 | remove_parentheses (struct StringBuffer *str) | ||
968 | { | ||
969 | size_t slen; | ||
970 | const char *pos; | ||
971 | const char *end; | ||
972 | const char *sbuf; | ||
973 | const char *op; | ||
974 | const char *cp; | ||
975 | unsigned int cnt; | ||
976 | |||
977 | if (0) | ||
978 | return; | ||
979 | sbuf = str->sbuf; | ||
980 | if ( (GNUNET_YES == str->null_flag) || | ||
981 | (1 >= (slen = str->slen)) || | ||
982 | ('(' != str->sbuf[0]) || | ||
983 | (')' != str->sbuf[slen - 1]) ) | ||
984 | return; | ||
985 | cnt = 0; | ||
986 | pos = &sbuf[1]; | ||
987 | end = &sbuf[slen - 1]; | ||
988 | op = memchr (pos, '(', end - pos); | ||
989 | cp = memchr (pos, ')', end - pos); | ||
990 | while (NULL != cp) | ||
991 | { | ||
992 | while ( (NULL != op) && | ||
993 | (op < cp) ) | ||
994 | { | ||
995 | cnt++; | ||
996 | pos = op + 1; | ||
997 | op = memchr (pos, '(', end - pos); | ||
998 | } | ||
999 | while ( (NULL != cp) && | ||
1000 | ( (NULL == op) || | ||
1001 | (cp < op) ) ) | ||
1002 | { | ||
1003 | if (0 == cnt) | ||
1004 | return; /* can't strip parens */ | ||
1005 | cnt--; | ||
1006 | pos = cp + 1; | ||
1007 | cp = memchr (pos, ')', end - pos); | ||
1008 | } | ||
1009 | } | ||
1010 | if (0 != cnt) | ||
1011 | { | ||
1012 | GNUNET_break (0); | ||
1013 | return; | ||
1014 | } | ||
1015 | str->sbuf++; | ||
1016 | str->slen -= 2; | ||
1017 | } | ||
1018 | |||
1019 | |||
1020 | /** | ||
1021 | * Check if the string 'str' starts with an epsilon (empty string). | ||
1022 | * Example: "(|a)" is starting with an epsilon. | ||
1023 | * | ||
1024 | * @param str string to test | ||
1025 | * | ||
1026 | * @return 0 if str has no epsilon, 1 if str starts with '(|' and ends with ')' | ||
1027 | */ | ||
1028 | static int | ||
1029 | has_epsilon (const struct StringBuffer *str) | ||
1030 | { | ||
1031 | return | ||
1032 | (GNUNET_YES != str->null_flag) && | ||
1033 | (0 < str->slen) && | ||
1034 | ('(' == str->sbuf[0]) && | ||
1035 | ('|' == str->sbuf[1]) && | ||
1036 | (')' == str->sbuf[str->slen - 1]); | ||
1037 | } | ||
1038 | |||
1039 | |||
1040 | /** | ||
1041 | * Remove an epsilon from the string str. Where epsilon is an empty string | ||
1042 | * Example: str = "(|a|b|c)", result: "a|b|c" | ||
1043 | * The returned string needs to be freed. | ||
1044 | * | ||
1045 | * @param str original string | ||
1046 | * @param ret where to return string without preceding epsilon, string 'str' if no preceding | ||
1047 | * epsilon could be found, NULL if 'str' was NULL | ||
1048 | */ | ||
1049 | static void | ||
1050 | remove_epsilon (const struct StringBuffer *str, | ||
1051 | struct StringBuffer *ret) | ||
1052 | { | ||
1053 | if (GNUNET_YES == str->null_flag) | ||
1054 | { | ||
1055 | ret->null_flag = GNUNET_YES; | ||
1056 | return; | ||
1057 | } | ||
1058 | if ( (str->slen > 1) && | ||
1059 | ('(' == str->sbuf[0]) && | ||
1060 | ('|' == str->sbuf[1]) && | ||
1061 | (')' == str->sbuf[str->slen - 1]) ) | ||
1062 | { | ||
1063 | /* remove epsilon */ | ||
1064 | if (ret->blen < str->slen - 3) | ||
1065 | { | ||
1066 | GNUNET_array_grow (ret->abuf, | ||
1067 | ret->blen, | ||
1068 | str->slen - 3); | ||
1069 | } | ||
1070 | ret->sbuf = ret->abuf; | ||
1071 | ret->slen = str->slen - 3; | ||
1072 | GNUNET_memcpy (ret->sbuf, &str->sbuf[2], ret->slen); | ||
1073 | return; | ||
1074 | } | ||
1075 | sb_strdup (ret, str); | ||
1076 | } | ||
1077 | |||
1078 | |||
1079 | /** | ||
1080 | * Compare n bytes of 'str1' and 'str2' | ||
1081 | * | ||
1082 | * @param str1 first string to compare | ||
1083 | * @param str2 second string for comparison | ||
1084 | * @param n number of bytes to compare | ||
1085 | * | ||
1086 | * @return -1 if any of the strings is NULL, 0 if equal, non 0 otherwise | ||
1087 | */ | ||
1088 | static int | ||
1089 | sb_strncmp (const struct StringBuffer *str1, | ||
1090 | const struct StringBuffer *str2, size_t n) | ||
1091 | { | ||
1092 | size_t max; | ||
1093 | |||
1094 | if ( (str1->slen != str2->slen) && | ||
1095 | ( (str1->slen < n) || | ||
1096 | (str2->slen < n) ) ) | ||
1097 | return -1; | ||
1098 | max = GNUNET_MAX (str1->slen, str2->slen); | ||
1099 | if (max > n) | ||
1100 | max = n; | ||
1101 | return memcmp (str1->sbuf, str2->sbuf, max); | ||
1102 | } | ||
1103 | |||
1104 | |||
1105 | /** | ||
1106 | * Compare n bytes of 'str1' and 'str2' | ||
1107 | * | ||
1108 | * @param str1 first string to compare | ||
1109 | * @param str2 second C string for comparison | ||
1110 | * @param n number of bytes to compare (and length of str2) | ||
1111 | * | ||
1112 | * @return -1 if any of the strings is NULL, 0 if equal, non 0 otherwise | ||
1113 | */ | ||
1114 | static int | ||
1115 | sb_strncmp_cstr (const struct StringBuffer *str1, | ||
1116 | const char *str2, size_t n) | ||
1117 | { | ||
1118 | if (str1->slen < n) | ||
1119 | return -1; | ||
1120 | return memcmp (str1->sbuf, str2, n); | ||
1121 | } | ||
1122 | |||
1123 | |||
1124 | /** | ||
1125 | * Initialize string buffer for storing strings of up to n | ||
1126 | * characters. | ||
1127 | * | ||
1128 | * @param sb buffer to initialize | ||
1129 | * @param n desired target length | ||
1130 | */ | ||
1131 | static void | ||
1132 | sb_init (struct StringBuffer *sb, | ||
1133 | size_t n) | ||
1134 | { | ||
1135 | sb->null_flag = GNUNET_NO; | ||
1136 | sb->abuf = sb->sbuf = (0 == n) ? NULL : GNUNET_malloc (n); | ||
1137 | sb->blen = n; | ||
1138 | sb->slen = 0; | ||
1139 | } | ||
1140 | |||
1141 | |||
1142 | /** | ||
1143 | * Compare 'str1', starting from position 'k', with whole 'str2' | ||
1144 | * | ||
1145 | * @param str1 first string to compare, starting from position 'k' | ||
1146 | * @param str2 second string for comparison | ||
1147 | * @param k starting position in 'str1' | ||
1148 | * | ||
1149 | * @return -1 if any of the strings is NULL, 0 if equal, non 0 otherwise | ||
1150 | */ | ||
1151 | static int | ||
1152 | sb_strkcmp (const struct StringBuffer *str1, | ||
1153 | const struct StringBuffer *str2, size_t k) | ||
1154 | { | ||
1155 | if ( (GNUNET_YES == str1->null_flag) || | ||
1156 | (GNUNET_YES == str2->null_flag) || | ||
1157 | (k > str1->slen) || | ||
1158 | (str1->slen - k != str2->slen) ) | ||
1159 | return -1; | ||
1160 | return memcmp (&str1->sbuf[k], str2->sbuf, str2->slen); | ||
1161 | } | ||
1162 | |||
1163 | |||
1164 | /** | ||
1165 | * Helper function used as 'action' in 'REGEX_INTERNAL_automaton_traverse' | ||
1166 | * function to create the depth-first numbering of the states. | ||
1167 | * | ||
1168 | * @param cls states array. | ||
1169 | * @param count current state counter. | ||
1170 | * @param s current state. | ||
1171 | */ | ||
1172 | static void | ||
1173 | number_states (void *cls, const unsigned int count, | ||
1174 | struct REGEX_INTERNAL_State *s) | ||
1175 | { | ||
1176 | struct REGEX_INTERNAL_State **states = cls; | ||
1177 | |||
1178 | s->dfs_id = count; | ||
1179 | if (NULL != states) | ||
1180 | states[count] = s; | ||
1181 | } | ||
1182 | |||
1183 | |||
1184 | |||
1185 | #define PRIS(a) \ | ||
1186 | ((GNUNET_YES == a.null_flag) ? 6 : (int) a.slen), \ | ||
1187 | ((GNUNET_YES == a.null_flag) ? "(null)" : a.sbuf) | ||
1188 | |||
1189 | |||
1190 | /** | ||
1191 | * Construct the regular expression given the inductive step, | ||
1192 | * $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* | ||
1193 | * R^{(k-1)}_{kj}, and simplify the resulting expression saved in R_cur_ij. | ||
1194 | * | ||
1195 | * @param R_last_ij value of $R^{(k-1)_{ij}. | ||
1196 | * @param R_last_ik value of $R^{(k-1)_{ik}. | ||
1197 | * @param R_last_kk value of $R^{(k-1)_{kk}. | ||
1198 | * @param R_last_kj value of $R^{(k-1)_{kj}. | ||
1199 | * @param R_cur_ij result for this inductive step is saved in R_cur_ij, R_cur_ij | ||
1200 | * is expected to be NULL when called! | ||
1201 | * @param R_cur_l optimization -- kept between iterations to avoid realloc | ||
1202 | * @param R_cur_r optimization -- kept between iterations to avoid realloc | ||
1203 | */ | ||
1204 | static void | ||
1205 | automaton_create_proofs_simplify (const struct StringBuffer *R_last_ij, | ||
1206 | const struct StringBuffer *R_last_ik, | ||
1207 | const struct StringBuffer *R_last_kk, | ||
1208 | const struct StringBuffer *R_last_kj, | ||
1209 | struct StringBuffer *R_cur_ij, | ||
1210 | struct StringBuffer *R_cur_l, | ||
1211 | struct StringBuffer *R_cur_r) | ||
1212 | { | ||
1213 | struct StringBuffer R_temp_ij; | ||
1214 | struct StringBuffer R_temp_ik; | ||
1215 | struct StringBuffer R_temp_kj; | ||
1216 | struct StringBuffer R_temp_kk; | ||
1217 | int eps_check; | ||
1218 | int ij_ik_cmp; | ||
1219 | int ij_kj_cmp; | ||
1220 | int ik_kk_cmp; | ||
1221 | int kk_kj_cmp; | ||
1222 | int clean_ik_kk_cmp; | ||
1223 | int clean_kk_kj_cmp; | ||
1224 | size_t length; | ||
1225 | size_t length_l; | ||
1226 | size_t length_r; | ||
1227 | |||
1228 | /* | ||
1229 | * $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj} | ||
1230 | * R_last == R^{(k-1)}, R_cur == R^{(k)} | ||
1231 | * R_cur_ij = R_cur_l | R_cur_r | ||
1232 | * R_cur_l == R^{(k-1)}_{ij} | ||
1233 | * R_cur_r == R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj} | ||
1234 | */ | ||
1235 | |||
1236 | if ( (GNUNET_YES == R_last_ij->null_flag) && | ||
1237 | ( (GNUNET_YES == R_last_ik->null_flag) || | ||
1238 | (GNUNET_YES == R_last_kj->null_flag))) | ||
1239 | { | ||
1240 | /* R^{(k)}_{ij} = N | N */ | ||
1241 | R_cur_ij->null_flag = GNUNET_YES; | ||
1242 | R_cur_ij->synced = GNUNET_NO; | ||
1243 | return; | ||
1244 | } | ||
1245 | |||
1246 | if ( (GNUNET_YES == R_last_ik->null_flag) || | ||
1247 | (GNUNET_YES == R_last_kj->null_flag) ) | ||
1248 | { | ||
1249 | /* R^{(k)}_{ij} = R^{(k-1)}_{ij} | N */ | ||
1250 | if (GNUNET_YES == R_last_ij->synced) | ||
1251 | { | ||
1252 | R_cur_ij->synced = GNUNET_YES; | ||
1253 | R_cur_ij->null_flag = GNUNET_NO; | ||
1254 | return; | ||
1255 | } | ||
1256 | R_cur_ij->synced = GNUNET_YES; | ||
1257 | sb_strdup (R_cur_ij, R_last_ij); | ||
1258 | return; | ||
1259 | } | ||
1260 | R_cur_ij->synced = GNUNET_NO; | ||
1261 | |||
1262 | /* $R^{(k)}_{ij} = N | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj} OR | ||
1263 | * $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj} */ | ||
1264 | |||
1265 | R_cur_r->null_flag = GNUNET_YES; | ||
1266 | R_cur_r->slen = 0; | ||
1267 | R_cur_l->null_flag = GNUNET_YES; | ||
1268 | R_cur_l->slen = 0; | ||
1269 | |||
1270 | /* cache results from strcmp, we might need these many times */ | ||
1271 | ij_kj_cmp = sb_nullstrcmp (R_last_ij, R_last_kj); | ||
1272 | ij_ik_cmp = sb_nullstrcmp (R_last_ij, R_last_ik); | ||
1273 | ik_kk_cmp = sb_nullstrcmp (R_last_ik, R_last_kk); | ||
1274 | kk_kj_cmp = sb_nullstrcmp (R_last_kk, R_last_kj); | ||
1275 | |||
1276 | /* Assign R_temp_(ik|kk|kj) to R_last[][] and remove epsilon as well | ||
1277 | * as parentheses, so we can better compare the contents */ | ||
1278 | |||
1279 | memset (&R_temp_ij, 0, sizeof (struct StringBuffer)); | ||
1280 | memset (&R_temp_ik, 0, sizeof (struct StringBuffer)); | ||
1281 | memset (&R_temp_kk, 0, sizeof (struct StringBuffer)); | ||
1282 | memset (&R_temp_kj, 0, sizeof (struct StringBuffer)); | ||
1283 | remove_epsilon (R_last_ik, &R_temp_ik); | ||
1284 | remove_epsilon (R_last_kk, &R_temp_kk); | ||
1285 | remove_epsilon (R_last_kj, &R_temp_kj); | ||
1286 | remove_parentheses (&R_temp_ik); | ||
1287 | remove_parentheses (&R_temp_kk); | ||
1288 | remove_parentheses (&R_temp_kj); | ||
1289 | clean_ik_kk_cmp = sb_nullstrcmp (R_last_ik, &R_temp_kk); | ||
1290 | clean_kk_kj_cmp = sb_nullstrcmp (&R_temp_kk, R_last_kj); | ||
1291 | |||
1292 | /* construct R_cur_l (and, if necessary R_cur_r) */ | ||
1293 | if (GNUNET_YES != R_last_ij->null_flag) | ||
1294 | { | ||
1295 | /* Assign R_temp_ij to R_last_ij and remove epsilon as well | ||
1296 | * as parentheses, so we can better compare the contents */ | ||
1297 | remove_epsilon (R_last_ij, &R_temp_ij); | ||
1298 | remove_parentheses (&R_temp_ij); | ||
1299 | |||
1300 | if ( (0 == sb_strcmp (&R_temp_ij, &R_temp_ik)) && | ||
1301 | (0 == sb_strcmp (&R_temp_ik, &R_temp_kk)) && | ||
1302 | (0 == sb_strcmp (&R_temp_kk, &R_temp_kj)) ) | ||
1303 | { | ||
1304 | if (0 == R_temp_ij.slen) | ||
1305 | { | ||
1306 | R_cur_r->null_flag = GNUNET_NO; | ||
1307 | } | ||
1308 | else if ((0 == sb_strncmp_cstr (R_last_ij, "(|", 2)) || | ||
1309 | (0 == sb_strncmp_cstr (R_last_ik, "(|", 2) && | ||
1310 | 0 == sb_strncmp_cstr (R_last_kj, "(|", 2))) | ||
1311 | { | ||
1312 | /* | ||
1313 | * a|(e|a)a*(e|a) = a* | ||
1314 | * a|(e|a)(e|a)*(e|a) = a* | ||
1315 | * (e|a)|aa*a = a* | ||
1316 | * (e|a)|aa*(e|a) = a* | ||
1317 | * (e|a)|(e|a)a*a = a* | ||
1318 | * (e|a)|(e|a)a*(e|a) = a* | ||
1319 | * (e|a)|(e|a)(e|a)*(e|a) = a* | ||
1320 | */ | ||
1321 | if (GNUNET_YES == needs_parentheses (&R_temp_ij)) | ||
1322 | sb_printf1 (R_cur_r, "(%.*s)*", 3, &R_temp_ij); | ||
1323 | else | ||
1324 | sb_printf1 (R_cur_r, "%.*s*", 1, &R_temp_ij); | ||
1325 | } | ||
1326 | else | ||
1327 | { | ||
1328 | /* | ||
1329 | * a|aa*a = a+ | ||
1330 | * a|(e|a)a*a = a+ | ||
1331 | * a|aa*(e|a) = a+ | ||
1332 | * a|(e|a)(e|a)*a = a+ | ||
1333 | * a|a(e|a)*(e|a) = a+ | ||
1334 | */ | ||
1335 | if (GNUNET_YES == needs_parentheses (&R_temp_ij)) | ||
1336 | sb_printf1 (R_cur_r, "(%.*s)+", 3, &R_temp_ij); | ||
1337 | else | ||
1338 | sb_printf1 (R_cur_r, "%.*s+", 1, &R_temp_ij); | ||
1339 | } | ||
1340 | } | ||
1341 | else if ( (0 == ij_ik_cmp) && (0 == clean_kk_kj_cmp) && (0 != clean_ik_kk_cmp) ) | ||
1342 | { | ||
1343 | /* a|ab*b = ab* */ | ||
1344 | if (0 == R_last_kk->slen) | ||
1345 | sb_strdup (R_cur_r, R_last_ij); | ||
1346 | else if (GNUNET_YES == needs_parentheses (&R_temp_kk)) | ||
1347 | sb_printf2 (R_cur_r, "%.*s(%.*s)*", 3, R_last_ij, &R_temp_kk); | ||
1348 | else | ||
1349 | sb_printf2 (R_cur_r, "%.*s%.*s*", 1, R_last_ij, R_last_kk); | ||
1350 | R_cur_l->null_flag = GNUNET_YES; | ||
1351 | } | ||
1352 | else if ( (0 == ij_kj_cmp) && (0 == clean_ik_kk_cmp) && (0 != clean_kk_kj_cmp)) | ||
1353 | { | ||
1354 | /* a|bb*a = b*a */ | ||
1355 | if (R_last_kk->slen < 1) | ||
1356 | { | ||
1357 | sb_strdup (R_cur_r, R_last_kj); | ||
1358 | } | ||
1359 | else if (GNUNET_YES == needs_parentheses (&R_temp_kk)) | ||
1360 | sb_printf2 (R_cur_r, "(%.*s)*%.*s", 3, &R_temp_kk, R_last_kj); | ||
1361 | else | ||
1362 | sb_printf2 (R_cur_r, "%.*s*%.*s", 1, &R_temp_kk, R_last_kj); | ||
1363 | |||
1364 | R_cur_l->null_flag = GNUNET_YES; | ||
1365 | } | ||
1366 | else if ( (0 == ij_ik_cmp) && (0 == kk_kj_cmp) && (! has_epsilon (R_last_ij)) && | ||
1367 | has_epsilon (R_last_kk)) | ||
1368 | { | ||
1369 | /* a|a(e|b)*(e|b) = a|ab* = a|a|ab|abb|abbb|... = ab* */ | ||
1370 | if (needs_parentheses (&R_temp_kk)) | ||
1371 | sb_printf2 (R_cur_r, "%.*s(%.*s)*", 3, R_last_ij, &R_temp_kk); | ||
1372 | else | ||
1373 | sb_printf2 (R_cur_r, "%.*s%.*s*", 1, R_last_ij, &R_temp_kk); | ||
1374 | R_cur_l->null_flag = GNUNET_YES; | ||
1375 | } | ||
1376 | else if ( (0 == ij_kj_cmp) && (0 == ik_kk_cmp) && (! has_epsilon (R_last_ij)) && | ||
1377 | has_epsilon (R_last_kk)) | ||
1378 | { | ||
1379 | /* a|(e|b)(e|b)*a = a|b*a = a|a|ba|bba|bbba|... = b*a */ | ||
1380 | if (needs_parentheses (&R_temp_kk)) | ||
1381 | sb_printf2 (R_cur_r, "(%.*s)*%.*s", 3, &R_temp_kk, R_last_ij); | ||
1382 | else | ||
1383 | sb_printf2 (R_cur_r, "%.*s*%.*s", 1, &R_temp_kk, R_last_ij); | ||
1384 | R_cur_l->null_flag = GNUNET_YES; | ||
1385 | } | ||
1386 | else | ||
1387 | { | ||
1388 | sb_strdup (R_cur_l, R_last_ij); | ||
1389 | remove_parentheses (R_cur_l); | ||
1390 | } | ||
1391 | } | ||
1392 | else | ||
1393 | { | ||
1394 | /* we have no left side */ | ||
1395 | R_cur_l->null_flag = GNUNET_YES; | ||
1396 | } | ||
1397 | |||
1398 | /* construct R_cur_r, if not already constructed */ | ||
1399 | if (GNUNET_YES == R_cur_r->null_flag) | ||
1400 | { | ||
1401 | length = R_temp_kk.slen - R_last_ik->slen; | ||
1402 | |||
1403 | /* a(ba)*bx = (ab)+x */ | ||
1404 | if ( (length > 0) && | ||
1405 | (GNUNET_YES != R_last_kk->null_flag) && | ||
1406 | (0 < R_last_kk->slen) && | ||
1407 | (GNUNET_YES != R_last_kj->null_flag) && | ||
1408 | (0 < R_last_kj->slen) && | ||
1409 | (GNUNET_YES != R_last_ik->null_flag) && | ||
1410 | (0 < R_last_ik->slen) && | ||
1411 | (0 == sb_strkcmp (&R_temp_kk, R_last_ik, length)) && | ||
1412 | (0 == sb_strncmp (&R_temp_kk, R_last_kj, length)) ) | ||
1413 | { | ||
1414 | struct StringBuffer temp_a; | ||
1415 | struct StringBuffer temp_b; | ||
1416 | |||
1417 | sb_init (&temp_a, length); | ||
1418 | sb_init (&temp_b, R_last_kj->slen - length); | ||
1419 | |||
1420 | length_l = length; | ||
1421 | temp_a.sbuf = temp_a.abuf; | ||
1422 | GNUNET_memcpy (temp_a.sbuf, R_last_kj->sbuf, length_l); | ||
1423 | temp_a.slen = length_l; | ||
1424 | |||
1425 | length_r = R_last_kj->slen - length; | ||
1426 | temp_b.sbuf = temp_b.abuf; | ||
1427 | GNUNET_memcpy (temp_b.sbuf, &R_last_kj->sbuf[length], length_r); | ||
1428 | temp_b.slen = length_r; | ||
1429 | |||
1430 | /* e|(ab)+ = (ab)* */ | ||
1431 | if ( (GNUNET_YES != R_cur_l->null_flag) && | ||
1432 | (0 == R_cur_l->slen) && | ||
1433 | (0 == temp_b.slen) ) | ||
1434 | { | ||
1435 | sb_printf2 (R_cur_r, "(%.*s%.*s)*", 3, R_last_ik, &temp_a); | ||
1436 | sb_free (R_cur_l); | ||
1437 | R_cur_l->null_flag = GNUNET_YES; | ||
1438 | } | ||
1439 | else | ||
1440 | { | ||
1441 | sb_printf3 (R_cur_r, "(%.*s%.*s)+%.*s", 3, R_last_ik, &temp_a, &temp_b); | ||
1442 | } | ||
1443 | sb_free (&temp_a); | ||
1444 | sb_free (&temp_b); | ||
1445 | } | ||
1446 | else if (0 == sb_strcmp (&R_temp_ik, &R_temp_kk) && | ||
1447 | 0 == sb_strcmp (&R_temp_kk, &R_temp_kj)) | ||
1448 | { | ||
1449 | /* | ||
1450 | * (e|a)a*(e|a) = a* | ||
1451 | * (e|a)(e|a)*(e|a) = a* | ||
1452 | */ | ||
1453 | if (has_epsilon (R_last_ik) && has_epsilon (R_last_kj)) | ||
1454 | { | ||
1455 | if (needs_parentheses (&R_temp_kk)) | ||
1456 | sb_printf1 (R_cur_r, "(%.*s)*", 3, &R_temp_kk); | ||
1457 | else | ||
1458 | sb_printf1 (R_cur_r, "%.*s*", 1, &R_temp_kk); | ||
1459 | } | ||
1460 | /* aa*a = a+a */ | ||
1461 | else if ( (0 == clean_ik_kk_cmp) && | ||
1462 | (0 == clean_kk_kj_cmp) && | ||
1463 | (! has_epsilon (R_last_ik)) ) | ||
1464 | { | ||
1465 | if (needs_parentheses (&R_temp_kk)) | ||
1466 | sb_printf2 (R_cur_r, "(%.*s)+%.*s", 3, &R_temp_kk, &R_temp_kk); | ||
1467 | else | ||
1468 | sb_printf2 (R_cur_r, "%.*s+%.*s", 1, &R_temp_kk, &R_temp_kk); | ||
1469 | } | ||
1470 | /* | ||
1471 | * (e|a)a*a = a+ | ||
1472 | * aa*(e|a) = a+ | ||
1473 | * a(e|a)*(e|a) = a+ | ||
1474 | * (e|a)a*a = a+ | ||
1475 | */ | ||
1476 | else | ||
1477 | { | ||
1478 | eps_check = | ||
1479 | (has_epsilon (R_last_ik) + has_epsilon (R_last_kk) + | ||
1480 | has_epsilon (R_last_kj)); | ||
1481 | |||
1482 | if (1 == eps_check) | ||
1483 | { | ||
1484 | if (needs_parentheses (&R_temp_kk)) | ||
1485 | sb_printf1 (R_cur_r, "(%.*s)+", 3, &R_temp_kk); | ||
1486 | else | ||
1487 | sb_printf1 (R_cur_r, "%.*s+", 1, &R_temp_kk); | ||
1488 | } | ||
1489 | } | ||
1490 | } | ||
1491 | /* | ||
1492 | * aa*b = a+b | ||
1493 | * (e|a)(e|a)*b = a*b | ||
1494 | */ | ||
1495 | else if (0 == sb_strcmp (&R_temp_ik, &R_temp_kk)) | ||
1496 | { | ||
1497 | if (has_epsilon (R_last_ik)) | ||
1498 | { | ||
1499 | if (needs_parentheses (&R_temp_kk)) | ||
1500 | sb_printf2 (R_cur_r, "(%.*s)*%.*s", 3, &R_temp_kk, R_last_kj); | ||
1501 | else | ||
1502 | sb_printf2 (R_cur_r, "%.*s*%.*s", 1, &R_temp_kk, R_last_kj); | ||
1503 | } | ||
1504 | else | ||
1505 | { | ||
1506 | if (needs_parentheses (&R_temp_kk)) | ||
1507 | sb_printf2 (R_cur_r, "(%.*s)+%.*s", 3, &R_temp_kk, R_last_kj); | ||
1508 | else | ||
1509 | sb_printf2 (R_cur_r, "%.*s+%.*s", 1, &R_temp_kk, R_last_kj); | ||
1510 | } | ||
1511 | } | ||
1512 | /* | ||
1513 | * ba*a = ba+ | ||
1514 | * b(e|a)*(e|a) = ba* | ||
1515 | */ | ||
1516 | else if (0 == sb_strcmp (&R_temp_kk, &R_temp_kj)) | ||
1517 | { | ||
1518 | if (has_epsilon (R_last_kj)) | ||
1519 | { | ||
1520 | if (needs_parentheses (&R_temp_kk)) | ||
1521 | sb_printf2 (R_cur_r, "%.*s(%.*s)*", 3, R_last_ik, &R_temp_kk); | ||
1522 | else | ||
1523 | sb_printf2 (R_cur_r, "%.*s%.*s*", 1, R_last_ik, &R_temp_kk); | ||
1524 | } | ||
1525 | else | ||
1526 | { | ||
1527 | if (needs_parentheses (&R_temp_kk)) | ||
1528 | sb_printf2 (R_cur_r, "(%.*s)+%.*s", 3, R_last_ik, &R_temp_kk); | ||
1529 | else | ||
1530 | sb_printf2 (R_cur_r, "%.*s+%.*s", 1, R_last_ik, &R_temp_kk); | ||
1531 | } | ||
1532 | } | ||
1533 | else | ||
1534 | { | ||
1535 | if (0 < R_temp_kk.slen) | ||
1536 | { | ||
1537 | if (needs_parentheses (&R_temp_kk)) | ||
1538 | { | ||
1539 | sb_printf3 (R_cur_r, "%.*s(%.*s)*%.*s", 3, R_last_ik, &R_temp_kk, | ||
1540 | R_last_kj); | ||
1541 | } | ||
1542 | else | ||
1543 | { | ||
1544 | sb_printf3 (R_cur_r, "%.*s%.*s*%.*s", 1, R_last_ik, &R_temp_kk, | ||
1545 | R_last_kj); | ||
1546 | } | ||
1547 | } | ||
1548 | else | ||
1549 | { | ||
1550 | sb_printf2 (R_cur_r, "%.*s%.*s", 0, R_last_ik, R_last_kj); | ||
1551 | } | ||
1552 | } | ||
1553 | } | ||
1554 | sb_free (&R_temp_ij); | ||
1555 | sb_free (&R_temp_ik); | ||
1556 | sb_free (&R_temp_kk); | ||
1557 | sb_free (&R_temp_kj); | ||
1558 | |||
1559 | if ( (GNUNET_YES == R_cur_l->null_flag) && | ||
1560 | (GNUNET_YES == R_cur_r->null_flag) ) | ||
1561 | { | ||
1562 | R_cur_ij->null_flag = GNUNET_YES; | ||
1563 | return; | ||
1564 | } | ||
1565 | |||
1566 | if ( (GNUNET_YES != R_cur_l->null_flag) && | ||
1567 | (GNUNET_YES == R_cur_r->null_flag) ) | ||
1568 | { | ||
1569 | struct StringBuffer tmp; | ||
1570 | |||
1571 | tmp = *R_cur_ij; | ||
1572 | *R_cur_ij = *R_cur_l; | ||
1573 | *R_cur_l = tmp; | ||
1574 | return; | ||
1575 | } | ||
1576 | |||
1577 | if ( (GNUNET_YES == R_cur_l->null_flag) && | ||
1578 | (GNUNET_YES != R_cur_r->null_flag) ) | ||
1579 | { | ||
1580 | struct StringBuffer tmp; | ||
1581 | |||
1582 | tmp = *R_cur_ij; | ||
1583 | *R_cur_ij = *R_cur_r; | ||
1584 | *R_cur_r = tmp; | ||
1585 | return; | ||
1586 | } | ||
1587 | |||
1588 | if (0 == sb_nullstrcmp (R_cur_l, R_cur_r)) | ||
1589 | { | ||
1590 | struct StringBuffer tmp; | ||
1591 | |||
1592 | tmp = *R_cur_ij; | ||
1593 | *R_cur_ij = *R_cur_l; | ||
1594 | *R_cur_l = tmp; | ||
1595 | return; | ||
1596 | } | ||
1597 | sb_printf2 (R_cur_ij, "(%.*s|%.*s)", 3, R_cur_l, R_cur_r); | ||
1598 | } | ||
1599 | |||
1600 | |||
1601 | /** | ||
1602 | * Create proofs for all states in the given automaton. Implementation of the | ||
1603 | * algorithm descriped in chapter 3.2.1 of "Automata Theory, Languages, and | ||
1604 | * Computation 3rd Edition" by Hopcroft, Motwani and Ullman. | ||
1605 | * | ||
1606 | * Each state in the automaton gets assigned 'proof' and 'hash' (hash of the | ||
1607 | * proof) fields. The starting state will only have a valid proof/hash if it has | ||
1608 | * any incoming transitions. | ||
1609 | * | ||
1610 | * @param a automaton for which to assign proofs and hashes, must not be NULL | ||
1611 | */ | ||
1612 | static int | ||
1613 | automaton_create_proofs (struct REGEX_INTERNAL_Automaton *a) | ||
1614 | { | ||
1615 | unsigned int n = a->state_count; | ||
1616 | struct REGEX_INTERNAL_State *states[n]; | ||
1617 | struct StringBuffer *R_last; | ||
1618 | struct StringBuffer *R_cur; | ||
1619 | struct StringBuffer R_cur_r; | ||
1620 | struct StringBuffer R_cur_l; | ||
1621 | struct StringBuffer *R_swap; | ||
1622 | struct REGEX_INTERNAL_Transition *t; | ||
1623 | struct StringBuffer complete_regex; | ||
1624 | unsigned int i; | ||
1625 | unsigned int j; | ||
1626 | unsigned int k; | ||
1627 | |||
1628 | R_last = GNUNET_malloc_large (sizeof (struct StringBuffer) * n * n); | ||
1629 | R_cur = GNUNET_malloc_large (sizeof (struct StringBuffer) * n * n); | ||
1630 | if ( (NULL == R_last) || | ||
1631 | (NULL == R_cur) ) | ||
1632 | { | ||
1633 | GNUNET_log_strerror (GNUNET_ERROR_TYPE_ERROR, "malloc"); | ||
1634 | GNUNET_free_non_null (R_cur); | ||
1635 | GNUNET_free_non_null (R_last); | ||
1636 | return GNUNET_SYSERR; | ||
1637 | } | ||
1638 | |||
1639 | /* create depth-first numbering of the states, initializes 'state' */ | ||
1640 | REGEX_INTERNAL_automaton_traverse (a, a->start, NULL, NULL, &number_states, | ||
1641 | states); | ||
1642 | |||
1643 | for (i = 0; i < n; i++) | ||
1644 | GNUNET_assert (NULL != states[i]); | ||
1645 | for (i = 0; i < n; i++) | ||
1646 | for (j = 0; j < n; j++) | ||
1647 | R_last[i *n + j].null_flag = GNUNET_YES; | ||
1648 | |||
1649 | /* Compute regular expressions of length "1" between each pair of states */ | ||
1650 | for (i = 0; i < n; i++) | ||
1651 | { | ||
1652 | for (t = states[i]->transitions_head; NULL != t; t = t->next) | ||
1653 | { | ||
1654 | j = t->to_state->dfs_id; | ||
1655 | if (GNUNET_YES == R_last[i * n + j].null_flag) | ||
1656 | { | ||
1657 | sb_strdup_cstr (&R_last[i * n + j], t->label); | ||
1658 | } | ||
1659 | else | ||
1660 | { | ||
1661 | sb_append_cstr (&R_last[i * n + j], "|"); | ||
1662 | sb_append_cstr (&R_last[i * n + j], t->label); | ||
1663 | } | ||
1664 | } | ||
1665 | /* add self-loop: i is reachable from i via epsilon-transition */ | ||
1666 | if (GNUNET_YES == R_last[i * n + i].null_flag) | ||
1667 | { | ||
1668 | R_last[i * n + i].slen = 0; | ||
1669 | R_last[i * n + i].null_flag = GNUNET_NO; | ||
1670 | } | ||
1671 | else | ||
1672 | { | ||
1673 | sb_wrap (&R_last[i * n + i], "(|%.*s)", 3); | ||
1674 | } | ||
1675 | } | ||
1676 | for (i = 0; i < n; i++) | ||
1677 | for (j = 0; j < n; j++) | ||
1678 | if (needs_parentheses (&R_last[i * n + j])) | ||
1679 | sb_wrap (&R_last[i * n + j], "(%.*s)", 2); | ||
1680 | /* Compute regular expressions of length "k" between each pair of states per | ||
1681 | * induction */ | ||
1682 | memset (&R_cur_l, 0, sizeof (struct StringBuffer)); | ||
1683 | memset (&R_cur_r, 0, sizeof (struct StringBuffer)); | ||
1684 | for (k = 0; k < n; k++) | ||
1685 | { | ||
1686 | for (i = 0; i < n; i++) | ||
1687 | { | ||
1688 | for (j = 0; j < n; j++) | ||
1689 | { | ||
1690 | /* Basis for the recursion: | ||
1691 | * $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj} | ||
1692 | * R_last == R^{(k-1)}, R_cur == R^{(k)} | ||
1693 | */ | ||
1694 | |||
1695 | /* Create R_cur[i][j] and simplify the expression */ | ||
1696 | automaton_create_proofs_simplify (&R_last[i * n + j], &R_last[i * n + k], | ||
1697 | &R_last[k * n + k], &R_last[k * n + j], | ||
1698 | &R_cur[i * n + j], | ||
1699 | &R_cur_l, &R_cur_r); | ||
1700 | } | ||
1701 | } | ||
1702 | /* set R_last = R_cur */ | ||
1703 | R_swap = R_last; | ||
1704 | R_last = R_cur; | ||
1705 | R_cur = R_swap; | ||
1706 | /* clear 'R_cur' for next iteration */ | ||
1707 | for (i = 0; i < n; i++) | ||
1708 | for (j = 0; j < n; j++) | ||
1709 | R_cur[i * n + j].null_flag = GNUNET_YES; | ||
1710 | } | ||
1711 | sb_free (&R_cur_l); | ||
1712 | sb_free (&R_cur_r); | ||
1713 | /* assign proofs and hashes */ | ||
1714 | for (i = 0; i < n; i++) | ||
1715 | { | ||
1716 | if (GNUNET_YES != R_last[a->start->dfs_id * n + i].null_flag) | ||
1717 | { | ||
1718 | states[i]->proof = GNUNET_strndup (R_last[a->start->dfs_id * n + i].sbuf, | ||
1719 | R_last[a->start->dfs_id * n + i].slen); | ||
1720 | GNUNET_CRYPTO_hash (states[i]->proof, strlen (states[i]->proof), | ||
1721 | &states[i]->hash); | ||
1722 | } | ||
1723 | } | ||
1724 | |||
1725 | /* complete regex for whole DFA: union of all pairs (start state/accepting | ||
1726 | * state(s)). */ | ||
1727 | sb_init (&complete_regex, 16 * n); | ||
1728 | for (i = 0; i < n; i++) | ||
1729 | { | ||
1730 | if (states[i]->accepting) | ||
1731 | { | ||
1732 | if ( (0 == complete_regex.slen) && | ||
1733 | (0 < R_last[a->start->dfs_id * n + i].slen) ) | ||
1734 | { | ||
1735 | sb_append (&complete_regex, | ||
1736 | &R_last[a->start->dfs_id * n + i]); | ||
1737 | } | ||
1738 | else if ( (GNUNET_YES != R_last[a->start->dfs_id * n + i].null_flag) && | ||
1739 | (0 < R_last[a->start->dfs_id * n + i].slen) ) | ||
1740 | { | ||
1741 | sb_append_cstr (&complete_regex, "|"); | ||
1742 | sb_append (&complete_regex, | ||
1743 | &R_last[a->start->dfs_id * n + i]); | ||
1744 | } | ||
1745 | } | ||
1746 | } | ||
1747 | a->canonical_regex = GNUNET_strndup (complete_regex.sbuf, complete_regex.slen); | ||
1748 | |||
1749 | /* cleanup */ | ||
1750 | sb_free (&complete_regex); | ||
1751 | for (i = 0; i < n; i++) | ||
1752 | for (j = 0; j < n; j++) | ||
1753 | { | ||
1754 | sb_free (&R_cur[i * n + j]); | ||
1755 | sb_free (&R_last[i * n + j]); | ||
1756 | } | ||
1757 | GNUNET_free (R_cur); | ||
1758 | GNUNET_free (R_last); | ||
1759 | return GNUNET_OK; | ||
1760 | } | ||
1761 | |||
1762 | |||
1763 | /** | ||
1764 | * Creates a new DFA state based on a set of NFA states. Needs to be freed using | ||
1765 | * automaton_destroy_state. | ||
1766 | * | ||
1767 | * @param ctx context | ||
1768 | * @param nfa_states set of NFA states on which the DFA should be based on | ||
1769 | * | ||
1770 | * @return new DFA state | ||
1771 | */ | ||
1772 | static struct REGEX_INTERNAL_State * | ||
1773 | dfa_state_create (struct REGEX_INTERNAL_Context *ctx, | ||
1774 | struct REGEX_INTERNAL_StateSet *nfa_states) | ||
1775 | { | ||
1776 | struct REGEX_INTERNAL_State *s; | ||
1777 | char *pos; | ||
1778 | size_t len; | ||
1779 | struct REGEX_INTERNAL_State *cstate; | ||
1780 | struct REGEX_INTERNAL_Transition *ctran; | ||
1781 | unsigned int i; | ||
1782 | |||
1783 | s = GNUNET_new (struct REGEX_INTERNAL_State); | ||
1784 | s->id = ctx->state_id++; | ||
1785 | s->index = -1; | ||
1786 | s->lowlink = -1; | ||
1787 | |||
1788 | if (NULL == nfa_states) | ||
1789 | { | ||
1790 | GNUNET_asprintf (&s->name, "s%i", s->id); | ||
1791 | return s; | ||
1792 | } | ||
1793 | |||
1794 | s->nfa_set = *nfa_states; | ||
1795 | |||
1796 | if (nfa_states->off < 1) | ||
1797 | return s; | ||
1798 | |||
1799 | /* Create a name based on 'nfa_states' */ | ||
1800 | len = nfa_states->off * 14 + 4; | ||
1801 | s->name = GNUNET_malloc (len); | ||
1802 | strcat (s->name, "{"); | ||
1803 | pos = s->name + 1; | ||
1804 | |||
1805 | for (i = 0; i < nfa_states->off; i++) | ||
1806 | { | ||
1807 | cstate = nfa_states->states[i]; | ||
1808 | GNUNET_snprintf (pos, | ||
1809 | pos - s->name + len, | ||
1810 | "%i,", | ||
1811 | cstate->id); | ||
1812 | pos += strlen (pos); | ||
1813 | |||
1814 | /* Add a transition for each distinct label to NULL state */ | ||
1815 | for (ctran = cstate->transitions_head; NULL != ctran; ctran = ctran->next) | ||
1816 | if (NULL != ctran->label) | ||
1817 | state_add_transition (ctx, s, ctran->label, NULL); | ||
1818 | |||
1819 | /* If the nfa_states contain an accepting state, the new dfa state is also | ||
1820 | * accepting. */ | ||
1821 | if (cstate->accepting) | ||
1822 | s->accepting = 1; | ||
1823 | } | ||
1824 | pos[-1] = '}'; | ||
1825 | s->name = GNUNET_realloc (s->name, strlen (s->name) + 1); | ||
1826 | |||
1827 | memset (nfa_states, 0, sizeof (struct REGEX_INTERNAL_StateSet)); | ||
1828 | return s; | ||
1829 | } | ||
1830 | |||
1831 | |||
1832 | /** | ||
1833 | * Move from the given state 's' to the next state on transition 'str'. Consumes | ||
1834 | * as much of the given 'str' as possible (usefull for strided DFAs). On return | ||
1835 | * 's' will point to the next state, and the length of the substring used for | ||
1836 | * this transition will be returned. If no transition possible 0 is returned and | ||
1837 | * 's' points to NULL. | ||
1838 | * | ||
1839 | * @param s starting state, will point to the next state or NULL (if no | ||
1840 | * transition possible) | ||
1841 | * @param str edge label to follow (will match longest common prefix) | ||
1842 | * | ||
1843 | * @return length of the substring comsumed from 'str' | ||
1844 | */ | ||
1845 | static unsigned int | ||
1846 | dfa_move (struct REGEX_INTERNAL_State **s, const char *str) | ||
1847 | { | ||
1848 | struct REGEX_INTERNAL_Transition *t; | ||
1849 | struct REGEX_INTERNAL_State *new_s; | ||
1850 | unsigned int len; | ||
1851 | unsigned int max_len; | ||
1852 | |||
1853 | if (NULL == s) | ||
1854 | return 0; | ||
1855 | |||
1856 | new_s = NULL; | ||
1857 | max_len = 0; | ||
1858 | for (t = (*s)->transitions_head; NULL != t; t = t->next) | ||
1859 | { | ||
1860 | len = strlen (t->label); | ||
1861 | |||
1862 | if (0 == strncmp (t->label, str, len)) | ||
1863 | { | ||
1864 | if (len >= max_len) | ||
1865 | { | ||
1866 | max_len = len; | ||
1867 | new_s = t->to_state; | ||
1868 | } | ||
1869 | } | ||
1870 | } | ||
1871 | |||
1872 | *s = new_s; | ||
1873 | return max_len; | ||
1874 | } | ||
1875 | |||
1876 | |||
1877 | /** | ||
1878 | * Set the given state 'marked' to #GNUNET_YES. Used by the | ||
1879 | * #dfa_remove_unreachable_states() function to detect unreachable states in the | ||
1880 | * automaton. | ||
1881 | * | ||
1882 | * @param cls closure, not used. | ||
1883 | * @param count count, not used. | ||
1884 | * @param s state where the marked attribute will be set to #GNUNET_YES. | ||
1885 | */ | ||
1886 | static void | ||
1887 | mark_states (void *cls, | ||
1888 | const unsigned int count, | ||
1889 | struct REGEX_INTERNAL_State *s) | ||
1890 | { | ||
1891 | s->marked = GNUNET_YES; | ||
1892 | } | ||
1893 | |||
1894 | |||
1895 | /** | ||
1896 | * Remove all unreachable states from DFA 'a'. Unreachable states are those | ||
1897 | * states that are not reachable from the starting state. | ||
1898 | * | ||
1899 | * @param a DFA automaton | ||
1900 | */ | ||
1901 | static void | ||
1902 | dfa_remove_unreachable_states (struct REGEX_INTERNAL_Automaton *a) | ||
1903 | { | ||
1904 | struct REGEX_INTERNAL_State *s; | ||
1905 | struct REGEX_INTERNAL_State *s_next; | ||
1906 | |||
1907 | /* 1. unmark all states */ | ||
1908 | for (s = a->states_head; NULL != s; s = s->next) | ||
1909 | s->marked = GNUNET_NO; | ||
1910 | |||
1911 | /* 2. traverse dfa from start state and mark all visited states */ | ||
1912 | REGEX_INTERNAL_automaton_traverse (a, a->start, NULL, NULL, &mark_states, NULL); | ||
1913 | |||
1914 | /* 3. delete all states that were not visited */ | ||
1915 | for (s = a->states_head; NULL != s; s = s_next) | ||
1916 | { | ||
1917 | s_next = s->next; | ||
1918 | if (GNUNET_NO == s->marked) | ||
1919 | automaton_remove_state (a, s); | ||
1920 | } | ||
1921 | } | ||
1922 | |||
1923 | |||
1924 | /** | ||
1925 | * Remove all dead states from the DFA 'a'. Dead states are those states that do | ||
1926 | * not transition to any other state but themselves. | ||
1927 | * | ||
1928 | * @param a DFA automaton | ||
1929 | */ | ||
1930 | static void | ||
1931 | dfa_remove_dead_states (struct REGEX_INTERNAL_Automaton *a) | ||
1932 | { | ||
1933 | struct REGEX_INTERNAL_State *s; | ||
1934 | struct REGEX_INTERNAL_State *s_next; | ||
1935 | struct REGEX_INTERNAL_Transition *t; | ||
1936 | int dead; | ||
1937 | |||
1938 | GNUNET_assert (DFA == a->type); | ||
1939 | |||
1940 | for (s = a->states_head; NULL != s; s = s_next) | ||
1941 | { | ||
1942 | s_next = s->next; | ||
1943 | |||
1944 | if (s->accepting) | ||
1945 | continue; | ||
1946 | |||
1947 | dead = 1; | ||
1948 | for (t = s->transitions_head; NULL != t; t = t->next) | ||
1949 | { | ||
1950 | if (NULL != t->to_state && t->to_state != s) | ||
1951 | { | ||
1952 | dead = 0; | ||
1953 | break; | ||
1954 | } | ||
1955 | } | ||
1956 | |||
1957 | if (0 == dead) | ||
1958 | continue; | ||
1959 | |||
1960 | /* state s is dead, remove it */ | ||
1961 | automaton_remove_state (a, s); | ||
1962 | } | ||
1963 | } | ||
1964 | |||
1965 | |||
1966 | /** | ||
1967 | * Merge all non distinguishable states in the DFA 'a' | ||
1968 | * | ||
1969 | * @param ctx context | ||
1970 | * @param a DFA automaton | ||
1971 | * @return #GNUNET_OK on success | ||
1972 | */ | ||
1973 | static int | ||
1974 | dfa_merge_nondistinguishable_states (struct REGEX_INTERNAL_Context *ctx, | ||
1975 | struct REGEX_INTERNAL_Automaton *a) | ||
1976 | { | ||
1977 | uint32_t *table; | ||
1978 | struct REGEX_INTERNAL_State *s1; | ||
1979 | struct REGEX_INTERNAL_State *s2; | ||
1980 | struct REGEX_INTERNAL_Transition *t1; | ||
1981 | struct REGEX_INTERNAL_Transition *t2; | ||
1982 | struct REGEX_INTERNAL_State *s1_next; | ||
1983 | struct REGEX_INTERNAL_State *s2_next; | ||
1984 | int change; | ||
1985 | unsigned int num_equal_edges; | ||
1986 | unsigned int i; | ||
1987 | unsigned int state_cnt; | ||
1988 | unsigned long long idx; | ||
1989 | unsigned long long idx1; | ||
1990 | |||
1991 | if ( (NULL == a) || (0 == a->state_count) ) | ||
1992 | { | ||
1993 | GNUNET_log (GNUNET_ERROR_TYPE_ERROR, | ||
1994 | "Could not merge nondistinguishable states, automaton was NULL.\n"); | ||
1995 | return GNUNET_SYSERR; | ||
1996 | } | ||
1997 | |||
1998 | state_cnt = a->state_count; | ||
1999 | table = GNUNET_malloc_large ((sizeof (uint32_t) * state_cnt * state_cnt / 32) + sizeof (uint32_t)); | ||
2000 | if (NULL == table) | ||
2001 | { | ||
2002 | GNUNET_log_strerror (GNUNET_ERROR_TYPE_ERROR, "malloc"); | ||
2003 | return GNUNET_SYSERR; | ||
2004 | } | ||
2005 | |||
2006 | for (i = 0, s1 = a->states_head; NULL != s1; s1 = s1->next) | ||
2007 | s1->marked = i++; | ||
2008 | |||
2009 | /* Mark all pairs of accepting/!accepting states */ | ||
2010 | for (s1 = a->states_head; NULL != s1; s1 = s1->next) | ||
2011 | for (s2 = a->states_head; NULL != s2; s2 = s2->next) | ||
2012 | if ( (s1->accepting && !s2->accepting) || | ||
2013 | (!s1->accepting && s2->accepting) ) | ||
2014 | { | ||
2015 | idx = (unsigned long long) s1->marked * state_cnt + s2->marked; | ||
2016 | table[idx / 32] |= (1U << (idx % 32)); | ||
2017 | } | ||
2018 | |||
2019 | /* Find all equal states */ | ||
2020 | change = 1; | ||
2021 | while (0 != change) | ||
2022 | { | ||
2023 | change = 0; | ||
2024 | for (s1 = a->states_head; NULL != s1; s1 = s1->next) | ||
2025 | { | ||
2026 | for (s2 = a->states_head; NULL != s2 && s1 != s2; s2 = s2->next) | ||
2027 | { | ||
2028 | idx = (unsigned long long) s1->marked * state_cnt + s2->marked; | ||
2029 | if (0 != (table[idx / 32] & (1U << (idx % 32)))) | ||
2030 | continue; | ||
2031 | num_equal_edges = 0; | ||
2032 | for (t1 = s1->transitions_head; NULL != t1; t1 = t1->next) | ||
2033 | { | ||
2034 | for (t2 = s2->transitions_head; NULL != t2; t2 = t2->next) | ||
2035 | { | ||
2036 | if (0 == strcmp (t1->label, t2->label)) | ||
2037 | { | ||
2038 | num_equal_edges++; | ||
2039 | /* same edge, but targets definitively different, so we're different | ||
2040 | as well */ | ||
2041 | if (t1->to_state->marked > t2->to_state->marked) | ||
2042 | idx1 = (unsigned long long) t1->to_state->marked * state_cnt + t2->to_state->marked; | ||
2043 | else | ||
2044 | idx1 = (unsigned long long) t2->to_state->marked * state_cnt + t1->to_state->marked; | ||
2045 | if (0 != (table[idx1 / 32] & (1U << (idx1 % 32)))) | ||
2046 | { | ||
2047 | table[idx / 32] |= (1U << (idx % 32)); | ||
2048 | change = 1; /* changed a marker, need to run again */ | ||
2049 | } | ||
2050 | } | ||
2051 | } | ||
2052 | } | ||
2053 | if ( (num_equal_edges != s1->transition_count) || | ||
2054 | (num_equal_edges != s2->transition_count) ) | ||
2055 | { | ||
2056 | /* Make sure ALL edges of possible equal states are the same */ | ||
2057 | table[idx / 32] |= (1U << (idx % 32)); | ||
2058 | change = 1; /* changed a marker, need to run again */ | ||
2059 | } | ||
2060 | } | ||
2061 | } | ||
2062 | } | ||
2063 | |||
2064 | /* Merge states that are equal */ | ||
2065 | for (s1 = a->states_head; NULL != s1; s1 = s1_next) | ||
2066 | { | ||
2067 | s1_next = s1->next; | ||
2068 | for (s2 = a->states_head; NULL != s2 && s1 != s2; s2 = s2_next) | ||
2069 | { | ||
2070 | s2_next = s2->next; | ||
2071 | idx = (unsigned long long) s1->marked * state_cnt + s2->marked; | ||
2072 | if (0 == (table[idx / 32] & (1U << (idx % 32)))) | ||
2073 | automaton_merge_states (ctx, a, s1, s2); | ||
2074 | } | ||
2075 | } | ||
2076 | |||
2077 | GNUNET_free (table); | ||
2078 | return GNUNET_OK; | ||
2079 | } | ||
2080 | |||
2081 | |||
2082 | /** | ||
2083 | * Minimize the given DFA 'a' by removing all unreachable states, removing all | ||
2084 | * dead states and merging all non distinguishable states | ||
2085 | * | ||
2086 | * @param ctx context | ||
2087 | * @param a DFA automaton | ||
2088 | * @return GNUNET_OK on success | ||
2089 | */ | ||
2090 | static int | ||
2091 | dfa_minimize (struct REGEX_INTERNAL_Context *ctx, | ||
2092 | struct REGEX_INTERNAL_Automaton *a) | ||
2093 | { | ||
2094 | if (NULL == a) | ||
2095 | return GNUNET_SYSERR; | ||
2096 | |||
2097 | GNUNET_assert (DFA == a->type); | ||
2098 | |||
2099 | /* 1. remove unreachable states */ | ||
2100 | dfa_remove_unreachable_states (a); | ||
2101 | |||
2102 | /* 2. remove dead states */ | ||
2103 | dfa_remove_dead_states (a); | ||
2104 | |||
2105 | /* 3. Merge nondistinguishable states */ | ||
2106 | if (GNUNET_OK != dfa_merge_nondistinguishable_states (ctx, a)) | ||
2107 | return GNUNET_SYSERR; | ||
2108 | return GNUNET_OK; | ||
2109 | } | ||
2110 | |||
2111 | |||
2112 | /** | ||
2113 | * Context for adding strided transitions to a DFA. | ||
2114 | */ | ||
2115 | struct REGEX_INTERNAL_Strided_Context | ||
2116 | { | ||
2117 | /** | ||
2118 | * Length of the strides. | ||
2119 | */ | ||
2120 | const unsigned int stride; | ||
2121 | |||
2122 | /** | ||
2123 | * Strided transitions DLL. New strided transitions will be stored in this DLL | ||
2124 | * and afterwards added to the DFA. | ||
2125 | */ | ||
2126 | struct REGEX_INTERNAL_Transition *transitions_head; | ||
2127 | |||
2128 | /** | ||
2129 | * Strided transitions DLL. | ||
2130 | */ | ||
2131 | struct REGEX_INTERNAL_Transition *transitions_tail; | ||
2132 | }; | ||
2133 | |||
2134 | |||
2135 | /** | ||
2136 | * Recursive helper function to add strides to a DFA. | ||
2137 | * | ||
2138 | * @param cls context, contains stride length and strided transitions DLL. | ||
2139 | * @param depth current depth of the depth-first traversal of the graph. | ||
2140 | * @param label current label, string that contains all labels on the path from | ||
2141 | * 'start' to 's'. | ||
2142 | * @param start start state for the depth-first traversal of the graph. | ||
2143 | * @param s current state in the depth-first traversal | ||
2144 | */ | ||
2145 | static void | ||
2146 | dfa_add_multi_strides_helper (void *cls, const unsigned int depth, char *label, | ||
2147 | struct REGEX_INTERNAL_State *start, | ||
2148 | struct REGEX_INTERNAL_State *s) | ||
2149 | { | ||
2150 | struct REGEX_INTERNAL_Strided_Context *ctx = cls; | ||
2151 | struct REGEX_INTERNAL_Transition *t; | ||
2152 | char *new_label; | ||
2153 | |||
2154 | if (depth == ctx->stride) | ||
2155 | { | ||
2156 | t = GNUNET_new (struct REGEX_INTERNAL_Transition); | ||
2157 | t->label = GNUNET_strdup (label); | ||
2158 | t->to_state = s; | ||
2159 | t->from_state = start; | ||
2160 | GNUNET_CONTAINER_DLL_insert (ctx->transitions_head, ctx->transitions_tail, | ||
2161 | t); | ||
2162 | } | ||
2163 | else | ||
2164 | { | ||
2165 | for (t = s->transitions_head; NULL != t; t = t->next) | ||
2166 | { | ||
2167 | /* Do not consider self-loops, because it end's up in too many | ||
2168 | * transitions */ | ||
2169 | if (t->to_state == t->from_state) | ||
2170 | continue; | ||
2171 | |||
2172 | if (NULL != label) | ||
2173 | { | ||
2174 | GNUNET_asprintf (&new_label, "%s%s", label, t->label); | ||
2175 | } | ||
2176 | else | ||
2177 | new_label = GNUNET_strdup (t->label); | ||
2178 | |||
2179 | dfa_add_multi_strides_helper (cls, (depth + 1), new_label, start, | ||
2180 | t->to_state); | ||
2181 | } | ||
2182 | } | ||
2183 | GNUNET_free_non_null (label); | ||
2184 | } | ||
2185 | |||
2186 | |||
2187 | /** | ||
2188 | * Function called for each state in the DFA. Starts a traversal of depth set in | ||
2189 | * context starting from state 's'. | ||
2190 | * | ||
2191 | * @param cls context. | ||
2192 | * @param count not used. | ||
2193 | * @param s current state. | ||
2194 | */ | ||
2195 | static void | ||
2196 | dfa_add_multi_strides (void *cls, const unsigned int count, | ||
2197 | struct REGEX_INTERNAL_State *s) | ||
2198 | { | ||
2199 | dfa_add_multi_strides_helper (cls, 0, NULL, s, s); | ||
2200 | } | ||
2201 | |||
2202 | |||
2203 | /** | ||
2204 | * Adds multi-strided transitions to the given 'dfa'. | ||
2205 | * | ||
2206 | * @param regex_ctx regex context needed to add transitions to the automaton. | ||
2207 | * @param dfa DFA to which the multi strided transitions should be added. | ||
2208 | * @param stride_len length of the strides. | ||
2209 | */ | ||
2210 | void | ||
2211 | REGEX_INTERNAL_dfa_add_multi_strides (struct REGEX_INTERNAL_Context *regex_ctx, | ||
2212 | struct REGEX_INTERNAL_Automaton *dfa, | ||
2213 | const unsigned int stride_len) | ||
2214 | { | ||
2215 | struct REGEX_INTERNAL_Strided_Context ctx = { stride_len, NULL, NULL }; | ||
2216 | struct REGEX_INTERNAL_Transition *t; | ||
2217 | struct REGEX_INTERNAL_Transition *t_next; | ||
2218 | |||
2219 | if (1 > stride_len || GNUNET_YES == dfa->is_multistrided) | ||
2220 | return; | ||
2221 | |||
2222 | /* Compute the new transitions of given stride_len */ | ||
2223 | REGEX_INTERNAL_automaton_traverse (dfa, dfa->start, NULL, NULL, | ||
2224 | &dfa_add_multi_strides, &ctx); | ||
2225 | |||
2226 | /* Add all the new transitions to the automaton. */ | ||
2227 | for (t = ctx.transitions_head; NULL != t; t = t_next) | ||
2228 | { | ||
2229 | t_next = t->next; | ||
2230 | state_add_transition (regex_ctx, t->from_state, t->label, t->to_state); | ||
2231 | GNUNET_CONTAINER_DLL_remove (ctx.transitions_head, ctx.transitions_tail, t); | ||
2232 | GNUNET_free_non_null (t->label); | ||
2233 | GNUNET_free (t); | ||
2234 | } | ||
2235 | |||
2236 | /* Mark this automaton as multistrided */ | ||
2237 | dfa->is_multistrided = GNUNET_YES; | ||
2238 | } | ||
2239 | |||
2240 | /** | ||
2241 | * Recursive Helper function for DFA path compression. Does DFS on the DFA graph | ||
2242 | * and adds new transitions to the given transitions DLL and marks states that | ||
2243 | * should be removed by setting state->contained to GNUNET_YES. | ||
2244 | * | ||
2245 | * @param dfa DFA for which the paths should be compressed. | ||
2246 | * @param start starting state for linear path search. | ||
2247 | * @param cur current state in the recursive DFS. | ||
2248 | * @param label current label (string of traversed labels). | ||
2249 | * @param max_len maximal path compression length. | ||
2250 | * @param transitions_head transitions DLL. | ||
2251 | * @param transitions_tail transitions DLL. | ||
2252 | */ | ||
2253 | void | ||
2254 | dfa_compress_paths_helper (struct REGEX_INTERNAL_Automaton *dfa, | ||
2255 | struct REGEX_INTERNAL_State *start, | ||
2256 | struct REGEX_INTERNAL_State *cur, char *label, | ||
2257 | unsigned int max_len, | ||
2258 | struct REGEX_INTERNAL_Transition **transitions_head, | ||
2259 | struct REGEX_INTERNAL_Transition **transitions_tail) | ||
2260 | { | ||
2261 | struct REGEX_INTERNAL_Transition *t; | ||
2262 | char *new_label; | ||
2263 | |||
2264 | |||
2265 | if (NULL != label && | ||
2266 | ((cur->incoming_transition_count > 1 || GNUNET_YES == cur->accepting || | ||
2267 | GNUNET_YES == cur->marked) || (start != dfa->start && max_len > 0 && | ||
2268 | max_len == strlen (label)) || | ||
2269 | (start == dfa->start && GNUNET_REGEX_INITIAL_BYTES == strlen (label)))) | ||
2270 | { | ||
2271 | t = GNUNET_new (struct REGEX_INTERNAL_Transition); | ||
2272 | t->label = GNUNET_strdup (label); | ||
2273 | t->to_state = cur; | ||
2274 | t->from_state = start; | ||
2275 | GNUNET_CONTAINER_DLL_insert (*transitions_head, *transitions_tail, t); | ||
2276 | |||
2277 | if (GNUNET_NO == cur->marked) | ||
2278 | { | ||
2279 | dfa_compress_paths_helper (dfa, cur, cur, NULL, max_len, transitions_head, | ||
2280 | transitions_tail); | ||
2281 | } | ||
2282 | return; | ||
2283 | } | ||
2284 | else if (cur != start) | ||
2285 | cur->contained = GNUNET_YES; | ||
2286 | |||
2287 | if (GNUNET_YES == cur->marked && cur != start) | ||
2288 | return; | ||
2289 | |||
2290 | cur->marked = GNUNET_YES; | ||
2291 | |||
2292 | |||
2293 | for (t = cur->transitions_head; NULL != t; t = t->next) | ||
2294 | { | ||
2295 | if (NULL != label) | ||
2296 | GNUNET_asprintf (&new_label, "%s%s", label, t->label); | ||
2297 | else | ||
2298 | new_label = GNUNET_strdup (t->label); | ||
2299 | |||
2300 | if (t->to_state != cur) | ||
2301 | { | ||
2302 | dfa_compress_paths_helper (dfa, start, t->to_state, new_label, max_len, | ||
2303 | transitions_head, transitions_tail); | ||
2304 | } | ||
2305 | GNUNET_free (new_label); | ||
2306 | } | ||
2307 | } | ||
2308 | |||
2309 | |||
2310 | /** | ||
2311 | * Compress paths in the given 'dfa'. Linear paths like 0->1->2->3 will be | ||
2312 | * compressed to 0->3 by combining transitions. | ||
2313 | * | ||
2314 | * @param regex_ctx context for adding new transitions. | ||
2315 | * @param dfa DFA representation, will directly modify the given DFA. | ||
2316 | * @param max_len maximal length of the compressed paths. | ||
2317 | */ | ||
2318 | static void | ||
2319 | dfa_compress_paths (struct REGEX_INTERNAL_Context *regex_ctx, | ||
2320 | struct REGEX_INTERNAL_Automaton *dfa, unsigned int max_len) | ||
2321 | { | ||
2322 | struct REGEX_INTERNAL_State *s; | ||
2323 | struct REGEX_INTERNAL_State *s_next; | ||
2324 | struct REGEX_INTERNAL_Transition *t; | ||
2325 | struct REGEX_INTERNAL_Transition *t_next; | ||
2326 | struct REGEX_INTERNAL_Transition *transitions_head = NULL; | ||
2327 | struct REGEX_INTERNAL_Transition *transitions_tail = NULL; | ||
2328 | |||
2329 | if (NULL == dfa) | ||
2330 | return; | ||
2331 | |||
2332 | /* Count the incoming transitions on each state. */ | ||
2333 | for (s = dfa->states_head; NULL != s; s = s->next) | ||
2334 | { | ||
2335 | for (t = s->transitions_head; NULL != t; t = t->next) | ||
2336 | { | ||
2337 | if (NULL != t->to_state) | ||
2338 | t->to_state->incoming_transition_count++; | ||
2339 | } | ||
2340 | } | ||
2341 | |||
2342 | /* Unmark all states. */ | ||
2343 | for (s = dfa->states_head; NULL != s; s = s->next) | ||
2344 | { | ||
2345 | s->marked = GNUNET_NO; | ||
2346 | s->contained = GNUNET_NO; | ||
2347 | } | ||
2348 | |||
2349 | /* Add strides and mark states that can be deleted. */ | ||
2350 | dfa_compress_paths_helper (dfa, dfa->start, dfa->start, NULL, max_len, | ||
2351 | &transitions_head, &transitions_tail); | ||
2352 | |||
2353 | /* Add all the new transitions to the automaton. */ | ||
2354 | for (t = transitions_head; NULL != t; t = t_next) | ||
2355 | { | ||
2356 | t_next = t->next; | ||
2357 | state_add_transition (regex_ctx, t->from_state, t->label, t->to_state); | ||
2358 | GNUNET_CONTAINER_DLL_remove (transitions_head, transitions_tail, t); | ||
2359 | GNUNET_free_non_null (t->label); | ||
2360 | GNUNET_free (t); | ||
2361 | } | ||
2362 | |||
2363 | /* Remove marked states (including their incoming and outgoing transitions). */ | ||
2364 | for (s = dfa->states_head; NULL != s; s = s_next) | ||
2365 | { | ||
2366 | s_next = s->next; | ||
2367 | if (GNUNET_YES == s->contained) | ||
2368 | automaton_remove_state (dfa, s); | ||
2369 | } | ||
2370 | } | ||
2371 | |||
2372 | |||
2373 | /** | ||
2374 | * Creates a new NFA fragment. Needs to be cleared using | ||
2375 | * automaton_fragment_clear. | ||
2376 | * | ||
2377 | * @param start starting state | ||
2378 | * @param end end state | ||
2379 | * | ||
2380 | * @return new NFA fragment | ||
2381 | */ | ||
2382 | static struct REGEX_INTERNAL_Automaton * | ||
2383 | nfa_fragment_create (struct REGEX_INTERNAL_State *start, | ||
2384 | struct REGEX_INTERNAL_State *end) | ||
2385 | { | ||
2386 | struct REGEX_INTERNAL_Automaton *n; | ||
2387 | |||
2388 | n = GNUNET_new (struct REGEX_INTERNAL_Automaton); | ||
2389 | |||
2390 | n->type = NFA; | ||
2391 | n->start = NULL; | ||
2392 | n->end = NULL; | ||
2393 | n->state_count = 0; | ||
2394 | |||
2395 | if (NULL == start || NULL == end) | ||
2396 | return n; | ||
2397 | |||
2398 | automaton_add_state (n, end); | ||
2399 | automaton_add_state (n, start); | ||
2400 | |||
2401 | n->state_count = 2; | ||
2402 | |||
2403 | n->start = start; | ||
2404 | n->end = end; | ||
2405 | |||
2406 | return n; | ||
2407 | } | ||
2408 | |||
2409 | |||
2410 | /** | ||
2411 | * Adds a list of states to the given automaton 'n'. | ||
2412 | * | ||
2413 | * @param n automaton to which the states should be added | ||
2414 | * @param states_head head of the DLL of states | ||
2415 | * @param states_tail tail of the DLL of states | ||
2416 | */ | ||
2417 | static void | ||
2418 | nfa_add_states (struct REGEX_INTERNAL_Automaton *n, | ||
2419 | struct REGEX_INTERNAL_State *states_head, | ||
2420 | struct REGEX_INTERNAL_State *states_tail) | ||
2421 | { | ||
2422 | struct REGEX_INTERNAL_State *s; | ||
2423 | |||
2424 | if (NULL == n || NULL == states_head) | ||
2425 | { | ||
2426 | GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not add states\n"); | ||
2427 | return; | ||
2428 | } | ||
2429 | |||
2430 | if (NULL == n->states_head) | ||
2431 | { | ||
2432 | n->states_head = states_head; | ||
2433 | n->states_tail = states_tail; | ||
2434 | return; | ||
2435 | } | ||
2436 | |||
2437 | if (NULL != states_head) | ||
2438 | { | ||
2439 | n->states_tail->next = states_head; | ||
2440 | n->states_tail = states_tail; | ||
2441 | } | ||
2442 | |||
2443 | for (s = states_head; NULL != s; s = s->next) | ||
2444 | n->state_count++; | ||
2445 | } | ||
2446 | |||
2447 | |||
2448 | /** | ||
2449 | * Creates a new NFA state. Needs to be freed using automaton_destroy_state. | ||
2450 | * | ||
2451 | * @param ctx context | ||
2452 | * @param accepting is it an accepting state or not | ||
2453 | * | ||
2454 | * @return new NFA state | ||
2455 | */ | ||
2456 | static struct REGEX_INTERNAL_State * | ||
2457 | nfa_state_create (struct REGEX_INTERNAL_Context *ctx, int accepting) | ||
2458 | { | ||
2459 | struct REGEX_INTERNAL_State *s; | ||
2460 | |||
2461 | s = GNUNET_new (struct REGEX_INTERNAL_State); | ||
2462 | s->id = ctx->state_id++; | ||
2463 | s->accepting = accepting; | ||
2464 | s->marked = GNUNET_NO; | ||
2465 | s->contained = 0; | ||
2466 | s->index = -1; | ||
2467 | s->lowlink = -1; | ||
2468 | s->scc_id = 0; | ||
2469 | s->name = NULL; | ||
2470 | GNUNET_asprintf (&s->name, "s%i", s->id); | ||
2471 | |||
2472 | return s; | ||
2473 | } | ||
2474 | |||
2475 | |||
2476 | /** | ||
2477 | * Calculates the closure set for the given set of states. | ||
2478 | * | ||
2479 | * @param ret set to sorted nfa closure on 'label' (epsilon closure if 'label' is NULL) | ||
2480 | * @param nfa the NFA containing 's' | ||
2481 | * @param states list of states on which to base the closure on | ||
2482 | * @param label transitioning label for which to base the closure on, | ||
2483 | * pass NULL for epsilon transition | ||
2484 | */ | ||
2485 | static void | ||
2486 | nfa_closure_set_create (struct REGEX_INTERNAL_StateSet *ret, | ||
2487 | struct REGEX_INTERNAL_Automaton *nfa, | ||
2488 | struct REGEX_INTERNAL_StateSet *states, const char *label) | ||
2489 | { | ||
2490 | struct REGEX_INTERNAL_State *s; | ||
2491 | unsigned int i; | ||
2492 | struct REGEX_INTERNAL_StateSet_MDLL cls_stack; | ||
2493 | struct REGEX_INTERNAL_State *clsstate; | ||
2494 | struct REGEX_INTERNAL_State *currentstate; | ||
2495 | struct REGEX_INTERNAL_Transition *ctran; | ||
2496 | |||
2497 | memset (ret, 0, sizeof (struct REGEX_INTERNAL_StateSet)); | ||
2498 | if (NULL == states) | ||
2499 | return; | ||
2500 | |||
2501 | for (i = 0; i < states->off; i++) | ||
2502 | { | ||
2503 | s = states->states[i]; | ||
2504 | |||
2505 | /* Add start state to closure only for epsilon closure */ | ||
2506 | if (NULL == label) | ||
2507 | state_set_append (ret, s); | ||
2508 | |||
2509 | /* initialize work stack */ | ||
2510 | cls_stack.head = NULL; | ||
2511 | cls_stack.tail = NULL; | ||
2512 | GNUNET_CONTAINER_MDLL_insert (ST, cls_stack.head, cls_stack.tail, s); | ||
2513 | cls_stack.len = 1; | ||
2514 | |||
2515 | while (NULL != (currentstate = cls_stack.tail)) | ||
2516 | { | ||
2517 | GNUNET_CONTAINER_MDLL_remove (ST, cls_stack.head, cls_stack.tail, | ||
2518 | currentstate); | ||
2519 | cls_stack.len--; | ||
2520 | for (ctran = currentstate->transitions_head; NULL != ctran; | ||
2521 | ctran = ctran->next) | ||
2522 | { | ||
2523 | if (NULL == (clsstate = ctran->to_state)) | ||
2524 | continue; | ||
2525 | if (0 != clsstate->contained) | ||
2526 | continue; | ||
2527 | if (0 != nullstrcmp (label, ctran->label)) | ||
2528 | continue; | ||
2529 | state_set_append (ret, clsstate); | ||
2530 | GNUNET_CONTAINER_MDLL_insert_tail (ST, cls_stack.head, cls_stack.tail, | ||
2531 | clsstate); | ||
2532 | cls_stack.len++; | ||
2533 | clsstate->contained = 1; | ||
2534 | } | ||
2535 | } | ||
2536 | } | ||
2537 | for (i = 0; i < ret->off; i++) | ||
2538 | ret->states[i]->contained = 0; | ||
2539 | |||
2540 | if (ret->off > 1) | ||
2541 | qsort (ret->states, ret->off, sizeof (struct REGEX_INTERNAL_State *), | ||
2542 | &state_compare); | ||
2543 | } | ||
2544 | |||
2545 | |||
2546 | /** | ||
2547 | * Pops two NFA fragments (a, b) from the stack and concatenates them (ab) | ||
2548 | * | ||
2549 | * @param ctx context | ||
2550 | */ | ||
2551 | static void | ||
2552 | nfa_add_concatenation (struct REGEX_INTERNAL_Context *ctx) | ||
2553 | { | ||
2554 | struct REGEX_INTERNAL_Automaton *a; | ||
2555 | struct REGEX_INTERNAL_Automaton *b; | ||
2556 | struct REGEX_INTERNAL_Automaton *new_nfa; | ||
2557 | |||
2558 | b = ctx->stack_tail; | ||
2559 | GNUNET_assert (NULL != b); | ||
2560 | GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, b); | ||
2561 | a = ctx->stack_tail; | ||
2562 | GNUNET_assert (NULL != a); | ||
2563 | GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a); | ||
2564 | |||
2565 | state_add_transition (ctx, a->end, NULL, b->start); | ||
2566 | a->end->accepting = 0; | ||
2567 | b->end->accepting = 1; | ||
2568 | |||
2569 | new_nfa = nfa_fragment_create (NULL, NULL); | ||
2570 | nfa_add_states (new_nfa, a->states_head, a->states_tail); | ||
2571 | nfa_add_states (new_nfa, b->states_head, b->states_tail); | ||
2572 | new_nfa->start = a->start; | ||
2573 | new_nfa->end = b->end; | ||
2574 | new_nfa->state_count += a->state_count + b->state_count; | ||
2575 | automaton_fragment_clear (a); | ||
2576 | automaton_fragment_clear (b); | ||
2577 | |||
2578 | GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new_nfa); | ||
2579 | } | ||
2580 | |||
2581 | |||
2582 | /** | ||
2583 | * Pops a NFA fragment from the stack (a) and adds a new fragment (a*) | ||
2584 | * | ||
2585 | * @param ctx context | ||
2586 | */ | ||
2587 | static void | ||
2588 | nfa_add_star_op (struct REGEX_INTERNAL_Context *ctx) | ||
2589 | { | ||
2590 | struct REGEX_INTERNAL_Automaton *a; | ||
2591 | struct REGEX_INTERNAL_Automaton *new_nfa; | ||
2592 | struct REGEX_INTERNAL_State *start; | ||
2593 | struct REGEX_INTERNAL_State *end; | ||
2594 | |||
2595 | a = ctx->stack_tail; | ||
2596 | |||
2597 | if (NULL == a) | ||
2598 | { | ||
2599 | GNUNET_log (GNUNET_ERROR_TYPE_ERROR, | ||
2600 | "nfa_add_star_op failed, because there was no element on the stack"); | ||
2601 | return; | ||
2602 | } | ||
2603 | |||
2604 | GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a); | ||
2605 | |||
2606 | start = nfa_state_create (ctx, 0); | ||
2607 | end = nfa_state_create (ctx, 1); | ||
2608 | |||
2609 | state_add_transition (ctx, start, NULL, a->start); | ||
2610 | state_add_transition (ctx, start, NULL, end); | ||
2611 | state_add_transition (ctx, a->end, NULL, a->start); | ||
2612 | state_add_transition (ctx, a->end, NULL, end); | ||
2613 | |||
2614 | a->end->accepting = 0; | ||
2615 | end->accepting = 1; | ||
2616 | |||
2617 | new_nfa = nfa_fragment_create (start, end); | ||
2618 | nfa_add_states (new_nfa, a->states_head, a->states_tail); | ||
2619 | automaton_fragment_clear (a); | ||
2620 | |||
2621 | GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new_nfa); | ||
2622 | } | ||
2623 | |||
2624 | |||
2625 | /** | ||
2626 | * Pops an NFA fragment (a) from the stack and adds a new fragment (a+) | ||
2627 | * | ||
2628 | * @param ctx context | ||
2629 | */ | ||
2630 | static void | ||
2631 | nfa_add_plus_op (struct REGEX_INTERNAL_Context *ctx) | ||
2632 | { | ||
2633 | struct REGEX_INTERNAL_Automaton *a; | ||
2634 | |||
2635 | a = ctx->stack_tail; | ||
2636 | |||
2637 | if (NULL == a) | ||
2638 | { | ||
2639 | GNUNET_log (GNUNET_ERROR_TYPE_ERROR, | ||
2640 | "nfa_add_plus_op failed, because there was no element on the stack"); | ||
2641 | return; | ||
2642 | } | ||
2643 | |||
2644 | GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a); | ||
2645 | |||
2646 | state_add_transition (ctx, a->end, NULL, a->start); | ||
2647 | |||
2648 | GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, a); | ||
2649 | } | ||
2650 | |||
2651 | |||
2652 | /** | ||
2653 | * Pops an NFA fragment (a) from the stack and adds a new fragment (a?) | ||
2654 | * | ||
2655 | * @param ctx context | ||
2656 | */ | ||
2657 | static void | ||
2658 | nfa_add_question_op (struct REGEX_INTERNAL_Context *ctx) | ||
2659 | { | ||
2660 | struct REGEX_INTERNAL_Automaton *a; | ||
2661 | struct REGEX_INTERNAL_Automaton *new_nfa; | ||
2662 | struct REGEX_INTERNAL_State *start; | ||
2663 | struct REGEX_INTERNAL_State *end; | ||
2664 | |||
2665 | a = ctx->stack_tail; | ||
2666 | if (NULL == a) | ||
2667 | { | ||
2668 | GNUNET_log (GNUNET_ERROR_TYPE_ERROR, | ||
2669 | "nfa_add_question_op failed, because there was no element on the stack"); | ||
2670 | return; | ||
2671 | } | ||
2672 | |||
2673 | GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a); | ||
2674 | |||
2675 | start = nfa_state_create (ctx, 0); | ||
2676 | end = nfa_state_create (ctx, 1); | ||
2677 | |||
2678 | state_add_transition (ctx, start, NULL, a->start); | ||
2679 | state_add_transition (ctx, start, NULL, end); | ||
2680 | state_add_transition (ctx, a->end, NULL, end); | ||
2681 | |||
2682 | a->end->accepting = 0; | ||
2683 | |||
2684 | new_nfa = nfa_fragment_create (start, end); | ||
2685 | nfa_add_states (new_nfa, a->states_head, a->states_tail); | ||
2686 | GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new_nfa); | ||
2687 | automaton_fragment_clear (a); | ||
2688 | } | ||
2689 | |||
2690 | |||
2691 | /** | ||
2692 | * Pops two NFA fragments (a, b) from the stack and adds a new NFA fragment that | ||
2693 | * alternates between a and b (a|b) | ||
2694 | * | ||
2695 | * @param ctx context | ||
2696 | */ | ||
2697 | static void | ||
2698 | nfa_add_alternation (struct REGEX_INTERNAL_Context *ctx) | ||
2699 | { | ||
2700 | struct REGEX_INTERNAL_Automaton *a; | ||
2701 | struct REGEX_INTERNAL_Automaton *b; | ||
2702 | struct REGEX_INTERNAL_Automaton *new_nfa; | ||
2703 | struct REGEX_INTERNAL_State *start; | ||
2704 | struct REGEX_INTERNAL_State *end; | ||
2705 | |||
2706 | b = ctx->stack_tail; | ||
2707 | GNUNET_assert (NULL != b); | ||
2708 | GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, b); | ||
2709 | a = ctx->stack_tail; | ||
2710 | GNUNET_assert (NULL != a); | ||
2711 | GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a); | ||
2712 | |||
2713 | start = nfa_state_create (ctx, 0); | ||
2714 | end = nfa_state_create (ctx, 1); | ||
2715 | state_add_transition (ctx, start, NULL, a->start); | ||
2716 | state_add_transition (ctx, start, NULL, b->start); | ||
2717 | |||
2718 | state_add_transition (ctx, a->end, NULL, end); | ||
2719 | state_add_transition (ctx, b->end, NULL, end); | ||
2720 | |||
2721 | a->end->accepting = 0; | ||
2722 | b->end->accepting = 0; | ||
2723 | end->accepting = 1; | ||
2724 | |||
2725 | new_nfa = nfa_fragment_create (start, end); | ||
2726 | nfa_add_states (new_nfa, a->states_head, a->states_tail); | ||
2727 | nfa_add_states (new_nfa, b->states_head, b->states_tail); | ||
2728 | automaton_fragment_clear (a); | ||
2729 | automaton_fragment_clear (b); | ||
2730 | |||
2731 | GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new_nfa); | ||
2732 | } | ||
2733 | |||
2734 | |||
2735 | /** | ||
2736 | * Adds a new nfa fragment to the stack | ||
2737 | * | ||
2738 | * @param ctx context | ||
2739 | * @param label label for nfa transition | ||
2740 | */ | ||
2741 | static void | ||
2742 | nfa_add_label (struct REGEX_INTERNAL_Context *ctx, const char *label) | ||
2743 | { | ||
2744 | struct REGEX_INTERNAL_Automaton *n; | ||
2745 | struct REGEX_INTERNAL_State *start; | ||
2746 | struct REGEX_INTERNAL_State *end; | ||
2747 | |||
2748 | GNUNET_assert (NULL != ctx); | ||
2749 | |||
2750 | start = nfa_state_create (ctx, 0); | ||
2751 | end = nfa_state_create (ctx, 1); | ||
2752 | state_add_transition (ctx, start, label, end); | ||
2753 | n = nfa_fragment_create (start, end); | ||
2754 | GNUNET_assert (NULL != n); | ||
2755 | GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, n); | ||
2756 | } | ||
2757 | |||
2758 | |||
2759 | /** | ||
2760 | * Initialize a new context | ||
2761 | * | ||
2762 | * @param ctx context | ||
2763 | */ | ||
2764 | static void | ||
2765 | REGEX_INTERNAL_context_init (struct REGEX_INTERNAL_Context *ctx) | ||
2766 | { | ||
2767 | if (NULL == ctx) | ||
2768 | { | ||
2769 | GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Context was NULL!"); | ||
2770 | return; | ||
2771 | } | ||
2772 | ctx->state_id = 0; | ||
2773 | ctx->transition_id = 0; | ||
2774 | ctx->stack_head = NULL; | ||
2775 | ctx->stack_tail = NULL; | ||
2776 | } | ||
2777 | |||
2778 | |||
2779 | /** | ||
2780 | * Construct an NFA by parsing the regex string of length 'len'. | ||
2781 | * | ||
2782 | * @param regex regular expression string | ||
2783 | * @param len length of the string | ||
2784 | * | ||
2785 | * @return NFA, needs to be freed using REGEX_INTERNAL_destroy_automaton | ||
2786 | */ | ||
2787 | struct REGEX_INTERNAL_Automaton * | ||
2788 | REGEX_INTERNAL_construct_nfa (const char *regex, const size_t len) | ||
2789 | { | ||
2790 | struct REGEX_INTERNAL_Context ctx; | ||
2791 | struct REGEX_INTERNAL_Automaton *nfa; | ||
2792 | const char *regexp; | ||
2793 | char curlabel[2]; | ||
2794 | char *error_msg; | ||
2795 | unsigned int count; | ||
2796 | unsigned int altcount; | ||
2797 | unsigned int atomcount; | ||
2798 | unsigned int poff; | ||
2799 | unsigned int psize; | ||
2800 | struct | ||
2801 | { | ||
2802 | int altcount; | ||
2803 | int atomcount; | ||
2804 | } *p; | ||
2805 | |||
2806 | if (NULL == regex || 0 == strlen (regex) || 0 == len) | ||
2807 | { | ||
2808 | GNUNET_log (GNUNET_ERROR_TYPE_ERROR, | ||
2809 | "Could not parse regex. Empty regex string provided.\n"); | ||
2810 | |||
2811 | return NULL; | ||
2812 | } | ||
2813 | REGEX_INTERNAL_context_init (&ctx); | ||
2814 | |||
2815 | regexp = regex; | ||
2816 | curlabel[1] = '\0'; | ||
2817 | p = NULL; | ||
2818 | error_msg = NULL; | ||
2819 | altcount = 0; | ||
2820 | atomcount = 0; | ||
2821 | poff = 0; | ||
2822 | psize = 0; | ||
2823 | |||
2824 | for (count = 0; count < len && *regexp; count++, regexp++) | ||
2825 | { | ||
2826 | switch (*regexp) | ||
2827 | { | ||
2828 | case '(': | ||
2829 | if (atomcount > 1) | ||
2830 | { | ||
2831 | --atomcount; | ||
2832 | nfa_add_concatenation (&ctx); | ||
2833 | } | ||
2834 | if (poff == psize) | ||
2835 | GNUNET_array_grow (p, psize, psize * 2 + 4); /* FIXME why *2 +4? */ | ||
2836 | p[poff].altcount = altcount; | ||
2837 | p[poff].atomcount = atomcount; | ||
2838 | poff++; | ||
2839 | altcount = 0; | ||
2840 | atomcount = 0; | ||
2841 | break; | ||
2842 | case '|': | ||
2843 | if (0 == atomcount) | ||
2844 | { | ||
2845 | error_msg = "Cannot append '|' to nothing"; | ||
2846 | goto error; | ||
2847 | } | ||
2848 | while (--atomcount > 0) | ||
2849 | nfa_add_concatenation (&ctx); | ||
2850 | altcount++; | ||
2851 | break; | ||
2852 | case ')': | ||
2853 | if (0 == poff) | ||
2854 | { | ||
2855 | error_msg = "Missing opening '('"; | ||
2856 | goto error; | ||
2857 | } | ||
2858 | if (0 == atomcount) | ||
2859 | { | ||
2860 | /* Ignore this: "()" */ | ||
2861 | poff--; | ||
2862 | altcount = p[poff].altcount; | ||
2863 | atomcount = p[poff].atomcount; | ||
2864 | break; | ||
2865 | } | ||
2866 | while (--atomcount > 0) | ||
2867 | nfa_add_concatenation (&ctx); | ||
2868 | for (; altcount > 0; altcount--) | ||
2869 | nfa_add_alternation (&ctx); | ||
2870 | poff--; | ||
2871 | altcount = p[poff].altcount; | ||
2872 | atomcount = p[poff].atomcount; | ||
2873 | atomcount++; | ||
2874 | break; | ||
2875 | case '*': | ||
2876 | if (atomcount == 0) | ||
2877 | { | ||
2878 | error_msg = "Cannot append '*' to nothing"; | ||
2879 | goto error; | ||
2880 | } | ||
2881 | nfa_add_star_op (&ctx); | ||
2882 | break; | ||
2883 | case '+': | ||
2884 | if (atomcount == 0) | ||
2885 | { | ||
2886 | error_msg = "Cannot append '+' to nothing"; | ||
2887 | goto error; | ||
2888 | } | ||
2889 | nfa_add_plus_op (&ctx); | ||
2890 | break; | ||
2891 | case '?': | ||
2892 | if (atomcount == 0) | ||
2893 | { | ||
2894 | error_msg = "Cannot append '?' to nothing"; | ||
2895 | goto error; | ||
2896 | } | ||
2897 | nfa_add_question_op (&ctx); | ||
2898 | break; | ||
2899 | default: | ||
2900 | if (atomcount > 1) | ||
2901 | { | ||
2902 | --atomcount; | ||
2903 | nfa_add_concatenation (&ctx); | ||
2904 | } | ||
2905 | curlabel[0] = *regexp; | ||
2906 | nfa_add_label (&ctx, curlabel); | ||
2907 | atomcount++; | ||
2908 | break; | ||
2909 | } | ||
2910 | } | ||
2911 | if (0 != poff) | ||
2912 | { | ||
2913 | error_msg = "Unbalanced parenthesis"; | ||
2914 | goto error; | ||
2915 | } | ||
2916 | while (--atomcount > 0) | ||
2917 | nfa_add_concatenation (&ctx); | ||
2918 | for (; altcount > 0; altcount--) | ||
2919 | nfa_add_alternation (&ctx); | ||
2920 | |||
2921 | GNUNET_array_grow (p, psize, 0); | ||
2922 | |||
2923 | nfa = ctx.stack_tail; | ||
2924 | GNUNET_CONTAINER_DLL_remove (ctx.stack_head, ctx.stack_tail, nfa); | ||
2925 | |||
2926 | if (NULL != ctx.stack_head) | ||
2927 | { | ||
2928 | error_msg = "Creating the NFA failed. NFA stack was not empty!"; | ||
2929 | goto error; | ||
2930 | } | ||
2931 | |||
2932 | /* Remember the regex that was used to generate this NFA */ | ||
2933 | nfa->regex = GNUNET_strdup (regex); | ||
2934 | |||
2935 | /* create depth-first numbering of the states for pretty printing */ | ||
2936 | REGEX_INTERNAL_automaton_traverse (nfa, NULL, NULL, NULL, &number_states, NULL); | ||
2937 | |||
2938 | /* No multistriding added so far */ | ||
2939 | nfa->is_multistrided = GNUNET_NO; | ||
2940 | |||
2941 | return nfa; | ||
2942 | |||
2943 | error: | ||
2944 | GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not parse regex: `%s'\n", regex); | ||
2945 | if (NULL != error_msg) | ||
2946 | GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "%s\n", error_msg); | ||
2947 | |||
2948 | GNUNET_free_non_null (p); | ||
2949 | |||
2950 | while (NULL != (nfa = ctx.stack_head)) | ||
2951 | { | ||
2952 | GNUNET_CONTAINER_DLL_remove (ctx.stack_head, ctx.stack_tail, nfa); | ||
2953 | REGEX_INTERNAL_automaton_destroy (nfa); | ||
2954 | } | ||
2955 | |||
2956 | return NULL; | ||
2957 | } | ||
2958 | |||
2959 | |||
2960 | /** | ||
2961 | * Create DFA states based on given 'nfa' and starting with 'dfa_state'. | ||
2962 | * | ||
2963 | * @param ctx context. | ||
2964 | * @param nfa NFA automaton. | ||
2965 | * @param dfa DFA automaton. | ||
2966 | * @param dfa_state current dfa state, pass epsilon closure of first nfa state | ||
2967 | * for starting. | ||
2968 | */ | ||
2969 | static void | ||
2970 | construct_dfa_states (struct REGEX_INTERNAL_Context *ctx, | ||
2971 | struct REGEX_INTERNAL_Automaton *nfa, | ||
2972 | struct REGEX_INTERNAL_Automaton *dfa, | ||
2973 | struct REGEX_INTERNAL_State *dfa_state) | ||
2974 | { | ||
2975 | struct REGEX_INTERNAL_Transition *ctran; | ||
2976 | struct REGEX_INTERNAL_State *new_dfa_state; | ||
2977 | struct REGEX_INTERNAL_State *state_contains; | ||
2978 | struct REGEX_INTERNAL_State *state_iter; | ||
2979 | struct REGEX_INTERNAL_StateSet tmp; | ||
2980 | struct REGEX_INTERNAL_StateSet nfa_set; | ||
2981 | |||
2982 | for (ctran = dfa_state->transitions_head; NULL != ctran; ctran = ctran->next) | ||
2983 | { | ||
2984 | if (NULL == ctran->label || NULL != ctran->to_state) | ||
2985 | continue; | ||
2986 | |||
2987 | nfa_closure_set_create (&tmp, nfa, &dfa_state->nfa_set, ctran->label); | ||
2988 | nfa_closure_set_create (&nfa_set, nfa, &tmp, NULL); | ||
2989 | state_set_clear (&tmp); | ||
2990 | |||
2991 | state_contains = NULL; | ||
2992 | for (state_iter = dfa->states_head; NULL != state_iter; | ||
2993 | state_iter = state_iter->next) | ||
2994 | { | ||
2995 | if (0 == state_set_compare (&state_iter->nfa_set, &nfa_set)) | ||
2996 | { | ||
2997 | state_contains = state_iter; | ||
2998 | break; | ||
2999 | } | ||
3000 | } | ||
3001 | if (NULL == state_contains) | ||
3002 | { | ||
3003 | new_dfa_state = dfa_state_create (ctx, &nfa_set); | ||
3004 | automaton_add_state (dfa, new_dfa_state); | ||
3005 | ctran->to_state = new_dfa_state; | ||
3006 | construct_dfa_states (ctx, nfa, dfa, new_dfa_state); | ||
3007 | } | ||
3008 | else | ||
3009 | { | ||
3010 | ctran->to_state = state_contains; | ||
3011 | state_set_clear (&nfa_set); | ||
3012 | } | ||
3013 | } | ||
3014 | } | ||
3015 | |||
3016 | |||
3017 | /** | ||
3018 | * Construct DFA for the given 'regex' of length 'len'. | ||
3019 | * | ||
3020 | * Path compression means, that for example a DFA o -> a -> b -> c -> o will be | ||
3021 | * compressed to o -> abc -> o. Note that this parameter influences the | ||
3022 | * non-determinism of states of the resulting NFA in the DHT (number of outgoing | ||
3023 | * edges with the same label). For example for an application that stores IPv4 | ||
3024 | * addresses as bitstrings it could make sense to limit the path compression to | ||
3025 | * 4 or 8. | ||
3026 | * | ||
3027 | * @param regex regular expression string. | ||
3028 | * @param len length of the regular expression. | ||
3029 | * @param max_path_len limit the path compression length to the | ||
3030 | * given value. If set to 1, no path compression is applied. Set to 0 for | ||
3031 | * maximal possible path compression (generally not desireable). | ||
3032 | * @return DFA, needs to be freed using REGEX_INTERNAL_automaton_destroy. | ||
3033 | */ | ||
3034 | struct REGEX_INTERNAL_Automaton * | ||
3035 | REGEX_INTERNAL_construct_dfa (const char *regex, const size_t len, | ||
3036 | unsigned int max_path_len) | ||
3037 | { | ||
3038 | struct REGEX_INTERNAL_Context ctx; | ||
3039 | struct REGEX_INTERNAL_Automaton *dfa; | ||
3040 | struct REGEX_INTERNAL_Automaton *nfa; | ||
3041 | struct REGEX_INTERNAL_StateSet nfa_start_eps_cls; | ||
3042 | struct REGEX_INTERNAL_StateSet singleton_set; | ||
3043 | |||
3044 | REGEX_INTERNAL_context_init (&ctx); | ||
3045 | |||
3046 | /* Create NFA */ | ||
3047 | nfa = REGEX_INTERNAL_construct_nfa (regex, len); | ||
3048 | |||
3049 | if (NULL == nfa) | ||
3050 | { | ||
3051 | GNUNET_log (GNUNET_ERROR_TYPE_ERROR, | ||
3052 | "Could not create DFA, because NFA creation failed\n"); | ||
3053 | return NULL; | ||
3054 | } | ||
3055 | |||
3056 | dfa = GNUNET_new (struct REGEX_INTERNAL_Automaton); | ||
3057 | dfa->type = DFA; | ||
3058 | dfa->regex = GNUNET_strdup (regex); | ||
3059 | |||
3060 | /* Create DFA start state from epsilon closure */ | ||
3061 | memset (&singleton_set, 0, sizeof (struct REGEX_INTERNAL_StateSet)); | ||
3062 | state_set_append (&singleton_set, nfa->start); | ||
3063 | nfa_closure_set_create (&nfa_start_eps_cls, nfa, &singleton_set, NULL); | ||
3064 | state_set_clear (&singleton_set); | ||
3065 | dfa->start = dfa_state_create (&ctx, &nfa_start_eps_cls); | ||
3066 | automaton_add_state (dfa, dfa->start); | ||
3067 | |||
3068 | construct_dfa_states (&ctx, nfa, dfa, dfa->start); | ||
3069 | REGEX_INTERNAL_automaton_destroy (nfa); | ||
3070 | |||
3071 | /* Minimize DFA */ | ||
3072 | if (GNUNET_OK != dfa_minimize (&ctx, dfa)) | ||
3073 | { | ||
3074 | REGEX_INTERNAL_automaton_destroy (dfa); | ||
3075 | return NULL; | ||
3076 | } | ||
3077 | |||
3078 | /* Create proofs and hashes for all states */ | ||
3079 | if (GNUNET_OK != automaton_create_proofs (dfa)) | ||
3080 | { | ||
3081 | REGEX_INTERNAL_automaton_destroy (dfa); | ||
3082 | return NULL; | ||
3083 | } | ||
3084 | |||
3085 | /* Compress linear DFA paths */ | ||
3086 | if (1 != max_path_len) | ||
3087 | dfa_compress_paths (&ctx, dfa, max_path_len); | ||
3088 | |||
3089 | return dfa; | ||
3090 | } | ||
3091 | |||
3092 | |||
3093 | /** | ||
3094 | * Free the memory allocated by constructing the REGEX_INTERNAL_Automaton data | ||
3095 | * structure. | ||
3096 | * | ||
3097 | * @param a automaton to be destroyed | ||
3098 | */ | ||
3099 | void | ||
3100 | REGEX_INTERNAL_automaton_destroy (struct REGEX_INTERNAL_Automaton *a) | ||
3101 | { | ||
3102 | struct REGEX_INTERNAL_State *s; | ||
3103 | struct REGEX_INTERNAL_State *next_state; | ||
3104 | |||
3105 | if (NULL == a) | ||
3106 | return; | ||
3107 | |||
3108 | GNUNET_free_non_null (a->regex); | ||
3109 | GNUNET_free_non_null (a->canonical_regex); | ||
3110 | |||
3111 | for (s = a->states_head; NULL != s; s = next_state) | ||
3112 | { | ||
3113 | next_state = s->next; | ||
3114 | GNUNET_CONTAINER_DLL_remove (a->states_head, a->states_tail, s); | ||
3115 | automaton_destroy_state (s); | ||
3116 | } | ||
3117 | |||
3118 | GNUNET_free (a); | ||
3119 | } | ||
3120 | |||
3121 | |||
3122 | /** | ||
3123 | * Evaluates the given string using the given DFA automaton | ||
3124 | * | ||
3125 | * @param a automaton, type must be DFA | ||
3126 | * @param string string that should be evaluated | ||
3127 | * | ||
3128 | * @return 0 if string matches, non-0 otherwise | ||
3129 | */ | ||
3130 | static int | ||
3131 | evaluate_dfa (struct REGEX_INTERNAL_Automaton *a, | ||
3132 | const char *string) | ||
3133 | { | ||
3134 | const char *strp; | ||
3135 | struct REGEX_INTERNAL_State *s; | ||
3136 | unsigned int step_len; | ||
3137 | |||
3138 | if (DFA != a->type) | ||
3139 | { | ||
3140 | GNUNET_log (GNUNET_ERROR_TYPE_ERROR, | ||
3141 | "Tried to evaluate DFA, but NFA automaton given"); | ||
3142 | return -1; | ||
3143 | } | ||
3144 | |||
3145 | s = a->start; | ||
3146 | |||
3147 | /* If the string is empty but the starting state is accepting, we accept. */ | ||
3148 | if ((NULL == string || 0 == strlen (string)) && s->accepting) | ||
3149 | return 0; | ||
3150 | |||
3151 | for (strp = string; NULL != strp && *strp; strp += step_len) | ||
3152 | { | ||
3153 | step_len = dfa_move (&s, strp); | ||
3154 | |||
3155 | if (NULL == s) | ||
3156 | break; | ||
3157 | } | ||
3158 | |||
3159 | if (NULL != s && s->accepting) | ||
3160 | return 0; | ||
3161 | |||
3162 | return 1; | ||
3163 | } | ||
3164 | |||
3165 | |||
3166 | /** | ||
3167 | * Evaluates the given string using the given NFA automaton | ||
3168 | * | ||
3169 | * @param a automaton, type must be NFA | ||
3170 | * @param string string that should be evaluated | ||
3171 | * @return 0 if string matches, non-0 otherwise | ||
3172 | */ | ||
3173 | static int | ||
3174 | evaluate_nfa (struct REGEX_INTERNAL_Automaton *a, | ||
3175 | const char *string) | ||
3176 | { | ||
3177 | const char *strp; | ||
3178 | char str[2]; | ||
3179 | struct REGEX_INTERNAL_State *s; | ||
3180 | struct REGEX_INTERNAL_StateSet sset; | ||
3181 | struct REGEX_INTERNAL_StateSet new_sset; | ||
3182 | struct REGEX_INTERNAL_StateSet singleton_set; | ||
3183 | unsigned int i; | ||
3184 | int result; | ||
3185 | |||
3186 | if (NFA != a->type) | ||
3187 | { | ||
3188 | GNUNET_log (GNUNET_ERROR_TYPE_ERROR, | ||
3189 | "Tried to evaluate NFA, but DFA automaton given"); | ||
3190 | return -1; | ||
3191 | } | ||
3192 | |||
3193 | /* If the string is empty but the starting state is accepting, we accept. */ | ||
3194 | if ((NULL == string || 0 == strlen (string)) && a->start->accepting) | ||
3195 | return 0; | ||
3196 | |||
3197 | result = 1; | ||
3198 | memset (&singleton_set, 0, sizeof (struct REGEX_INTERNAL_StateSet)); | ||
3199 | state_set_append (&singleton_set, a->start); | ||
3200 | nfa_closure_set_create (&sset, a, &singleton_set, NULL); | ||
3201 | state_set_clear (&singleton_set); | ||
3202 | |||
3203 | str[1] = '\0'; | ||
3204 | for (strp = string; NULL != strp && *strp; strp++) | ||
3205 | { | ||
3206 | str[0] = *strp; | ||
3207 | nfa_closure_set_create (&new_sset, a, &sset, str); | ||
3208 | state_set_clear (&sset); | ||
3209 | nfa_closure_set_create (&sset, a, &new_sset, 0); | ||
3210 | state_set_clear (&new_sset); | ||
3211 | } | ||
3212 | |||
3213 | for (i = 0; i < sset.off; i++) | ||
3214 | { | ||
3215 | s = sset.states[i]; | ||
3216 | if ( (NULL != s) && (s->accepting) ) | ||
3217 | { | ||
3218 | result = 0; | ||
3219 | break; | ||
3220 | } | ||
3221 | } | ||
3222 | |||
3223 | state_set_clear (&sset); | ||
3224 | return result; | ||
3225 | } | ||
3226 | |||
3227 | |||
3228 | /** | ||
3229 | * Evaluates the given @a string against the given compiled regex @a a | ||
3230 | * | ||
3231 | * @param a automaton | ||
3232 | * @param string string to check | ||
3233 | * @return 0 if string matches, non-0 otherwise | ||
3234 | */ | ||
3235 | int | ||
3236 | REGEX_INTERNAL_eval (struct REGEX_INTERNAL_Automaton *a, | ||
3237 | const char *string) | ||
3238 | { | ||
3239 | int result; | ||
3240 | |||
3241 | switch (a->type) | ||
3242 | { | ||
3243 | case DFA: | ||
3244 | result = evaluate_dfa (a, string); | ||
3245 | break; | ||
3246 | case NFA: | ||
3247 | result = evaluate_nfa (a, string); | ||
3248 | break; | ||
3249 | default: | ||
3250 | GNUNET_log (GNUNET_ERROR_TYPE_ERROR, | ||
3251 | "Evaluating regex failed, automaton has no type!\n"); | ||
3252 | result = GNUNET_SYSERR; | ||
3253 | break; | ||
3254 | } | ||
3255 | |||
3256 | return result; | ||
3257 | } | ||
3258 | |||
3259 | |||
3260 | /** | ||
3261 | * Get the canonical regex of the given automaton. | ||
3262 | * When constructing the automaton a proof is computed for each state, | ||
3263 | * consisting of the regular expression leading to this state. A complete | ||
3264 | * regex for the automaton can be computed by combining these proofs. | ||
3265 | * As of now this function is only useful for testing. | ||
3266 | * | ||
3267 | * @param a automaton for which the canonical regex should be returned. | ||
3268 | * | ||
3269 | * @return | ||
3270 | */ | ||
3271 | const char * | ||
3272 | REGEX_INTERNAL_get_canonical_regex (struct REGEX_INTERNAL_Automaton *a) | ||
3273 | { | ||
3274 | if (NULL == a) | ||
3275 | return NULL; | ||
3276 | |||
3277 | return a->canonical_regex; | ||
3278 | } | ||
3279 | |||
3280 | |||
3281 | /** | ||
3282 | * Get the number of transitions that are contained in the given automaton. | ||
3283 | * | ||
3284 | * @param a automaton for which the number of transitions should be returned. | ||
3285 | * | ||
3286 | * @return number of transitions in the given automaton. | ||
3287 | */ | ||
3288 | unsigned int | ||
3289 | REGEX_INTERNAL_get_transition_count (struct REGEX_INTERNAL_Automaton *a) | ||
3290 | { | ||
3291 | unsigned int t_count; | ||
3292 | struct REGEX_INTERNAL_State *s; | ||
3293 | |||
3294 | if (NULL == a) | ||
3295 | return 0; | ||
3296 | |||
3297 | t_count = 0; | ||
3298 | for (s = a->states_head; NULL != s; s = s->next) | ||
3299 | t_count += s->transition_count; | ||
3300 | |||
3301 | return t_count; | ||
3302 | } | ||
3303 | |||
3304 | |||
3305 | /** | ||
3306 | * Get the first key for the given @a input_string. This hashes the first x bits | ||
3307 | * of the @a input_string. | ||
3308 | * | ||
3309 | * @param input_string string. | ||
3310 | * @param string_len length of the @a input_string. | ||
3311 | * @param key pointer to where to write the hash code. | ||
3312 | * @return number of bits of @a input_string that have been consumed | ||
3313 | * to construct the key | ||
3314 | */ | ||
3315 | size_t | ||
3316 | REGEX_INTERNAL_get_first_key (const char *input_string, | ||
3317 | size_t string_len, | ||
3318 | struct GNUNET_HashCode *key) | ||
3319 | { | ||
3320 | size_t size; | ||
3321 | |||
3322 | size = string_len < GNUNET_REGEX_INITIAL_BYTES ? string_len : | ||
3323 | GNUNET_REGEX_INITIAL_BYTES; | ||
3324 | if (NULL == input_string) | ||
3325 | { | ||
3326 | GNUNET_log (GNUNET_ERROR_TYPE_ERROR, | ||
3327 | "Given input string was NULL!\n"); | ||
3328 | return 0; | ||
3329 | } | ||
3330 | GNUNET_CRYPTO_hash (input_string, size, key); | ||
3331 | |||
3332 | return size; | ||
3333 | } | ||
3334 | |||
3335 | |||
3336 | /** | ||
3337 | * Recursive function that calls the iterator for each synthetic start state. | ||
3338 | * | ||
3339 | * @param min_len minimum length of the path in the graph. | ||
3340 | * @param max_len maximum length of the path in the graph. | ||
3341 | * @param consumed_string string consumed by traversing the graph till this state. | ||
3342 | * @param state current state of the automaton. | ||
3343 | * @param iterator iterator function called for each edge. | ||
3344 | * @param iterator_cls closure for the @a iterator function. | ||
3345 | */ | ||
3346 | static void | ||
3347 | iterate_initial_edge (unsigned int min_len, | ||
3348 | unsigned int max_len, | ||
3349 | char *consumed_string, | ||
3350 | struct REGEX_INTERNAL_State *state, | ||
3351 | REGEX_INTERNAL_KeyIterator iterator, | ||
3352 | void *iterator_cls) | ||
3353 | { | ||
3354 | char *temp; | ||
3355 | struct REGEX_INTERNAL_Transition *t; | ||
3356 | unsigned int num_edges = state->transition_count; | ||
3357 | struct REGEX_BLOCK_Edge edges[num_edges]; | ||
3358 | struct REGEX_BLOCK_Edge edge[1]; | ||
3359 | struct GNUNET_HashCode hash; | ||
3360 | struct GNUNET_HashCode hash_new; | ||
3361 | unsigned int cur_len; | ||
3362 | |||
3363 | if (NULL != consumed_string) | ||
3364 | cur_len = strlen (consumed_string); | ||
3365 | else | ||
3366 | cur_len = 0; | ||
3367 | |||
3368 | if ( ( (cur_len >= min_len) || | ||
3369 | (GNUNET_YES == state->accepting) ) && | ||
3370 | (cur_len > 0) && | ||
3371 | (NULL != consumed_string) ) | ||
3372 | { | ||
3373 | if (cur_len <= max_len) | ||
3374 | { | ||
3375 | if ( (NULL != state->proof) && | ||
3376 | (0 != strcmp (consumed_string, | ||
3377 | state->proof)) ) | ||
3378 | { | ||
3379 | (void) state_get_edges (state, edges); | ||
3380 | GNUNET_CRYPTO_hash (consumed_string, | ||
3381 | strlen (consumed_string), | ||
3382 | &hash); | ||
3383 | GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, | ||
3384 | "Start state for string `%s' is %s\n", | ||
3385 | consumed_string, | ||
3386 | GNUNET_h2s (&hash)); | ||
3387 | iterator (iterator_cls, | ||
3388 | &hash, | ||
3389 | consumed_string, | ||
3390 | state->accepting, | ||
3391 | num_edges, edges); | ||
3392 | } | ||
3393 | |||
3394 | if ( (GNUNET_YES == state->accepting) && | ||
3395 | (cur_len > 1) && | ||
3396 | (state->transition_count < 1) && | ||
3397 | (cur_len < max_len) ) | ||
3398 | { | ||
3399 | /* Special case for regex consisting of just a string that is shorter than | ||
3400 | * max_len */ | ||
3401 | edge[0].label = &consumed_string[cur_len - 1]; | ||
3402 | edge[0].destination = state->hash; | ||
3403 | temp = GNUNET_strdup (consumed_string); | ||
3404 | temp[cur_len - 1] = '\0'; | ||
3405 | GNUNET_CRYPTO_hash (temp, | ||
3406 | cur_len - 1, | ||
3407 | &hash_new); | ||
3408 | GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, | ||
3409 | "Start state for short string `%s' is %s\n", | ||
3410 | temp, | ||
3411 | GNUNET_h2s (&hash_new)); | ||
3412 | iterator (iterator_cls, | ||
3413 | &hash_new, | ||
3414 | temp, | ||
3415 | GNUNET_NO, 1, | ||
3416 | edge); | ||
3417 | GNUNET_free (temp); | ||
3418 | } | ||
3419 | } | ||
3420 | else /* cur_len > max_len */ | ||
3421 | { | ||
3422 | /* Case where the concatenated labels are longer than max_len, then split. */ | ||
3423 | edge[0].label = &consumed_string[max_len]; | ||
3424 | edge[0].destination = state->hash; | ||
3425 | temp = GNUNET_strdup (consumed_string); | ||
3426 | temp[max_len] = '\0'; | ||
3427 | GNUNET_CRYPTO_hash (temp, max_len, &hash); | ||
3428 | GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, | ||
3429 | "Start state at split edge `%s'-`%s` is %s\n", | ||
3430 | temp, | ||
3431 | edge[0].label, | ||
3432 | GNUNET_h2s (&hash_new)); | ||
3433 | iterator (iterator_cls, | ||
3434 | &hash, | ||
3435 | temp, | ||
3436 | GNUNET_NO, | ||
3437 | 1, | ||
3438 | edge); | ||
3439 | GNUNET_free (temp); | ||
3440 | } | ||
3441 | } | ||
3442 | |||
3443 | if (cur_len < max_len) | ||
3444 | { | ||
3445 | for (t = state->transitions_head; NULL != t; t = t->next) | ||
3446 | { | ||
3447 | if (NULL != strchr (t->label, | ||
3448 | (int) '.')) | ||
3449 | { | ||
3450 | /* Wildcards not allowed during starting states */ | ||
3451 | GNUNET_break (0); | ||
3452 | continue; | ||
3453 | } | ||
3454 | if (NULL != consumed_string) | ||
3455 | GNUNET_asprintf (&temp, | ||
3456 | "%s%s", | ||
3457 | consumed_string, | ||
3458 | t->label); | ||
3459 | else | ||
3460 | GNUNET_asprintf (&temp, | ||
3461 | "%s", | ||
3462 | t->label); | ||
3463 | iterate_initial_edge (min_len, | ||
3464 | max_len, | ||
3465 | temp, | ||
3466 | t->to_state, | ||
3467 | iterator, | ||
3468 | iterator_cls); | ||
3469 | GNUNET_free (temp); | ||
3470 | } | ||
3471 | } | ||
3472 | } | ||
3473 | |||
3474 | |||
3475 | /** | ||
3476 | * Iterate over all edges starting from start state of automaton 'a'. Calling | ||
3477 | * iterator for each edge. | ||
3478 | * | ||
3479 | * @param a automaton. | ||
3480 | * @param iterator iterator called for each edge. | ||
3481 | * @param iterator_cls closure. | ||
3482 | */ | ||
3483 | void | ||
3484 | REGEX_INTERNAL_iterate_all_edges (struct REGEX_INTERNAL_Automaton *a, | ||
3485 | REGEX_INTERNAL_KeyIterator iterator, | ||
3486 | void *iterator_cls) | ||
3487 | { | ||
3488 | struct REGEX_INTERNAL_State *s; | ||
3489 | |||
3490 | GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, | ||
3491 | "Iterating over starting edges\n"); | ||
3492 | iterate_initial_edge (GNUNET_REGEX_INITIAL_BYTES, | ||
3493 | GNUNET_REGEX_INITIAL_BYTES, | ||
3494 | NULL, a->start, | ||
3495 | iterator, iterator_cls); | ||
3496 | GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, | ||
3497 | "Iterating over DFA edges\n"); | ||
3498 | for (s = a->states_head; NULL != s; s = s->next) | ||
3499 | { | ||
3500 | struct REGEX_BLOCK_Edge edges[s->transition_count]; | ||
3501 | unsigned int num_edges; | ||
3502 | |||
3503 | num_edges = state_get_edges (s, edges); | ||
3504 | if ( ( (NULL != s->proof) && | ||
3505 | (0 < strlen (s->proof)) ) || s->accepting) | ||
3506 | { | ||
3507 | GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, | ||
3508 | "Creating DFA edges at `%s' under key %s\n", | ||
3509 | s->proof, | ||
3510 | GNUNET_h2s (&s->hash)); | ||
3511 | iterator (iterator_cls, &s->hash, s->proof, | ||
3512 | s->accepting, | ||
3513 | num_edges, edges); | ||
3514 | } | ||
3515 | s->marked = GNUNET_NO; | ||
3516 | } | ||
3517 | } | ||
3518 | |||
3519 | |||
3520 | /** | ||
3521 | * Struct to hold all the relevant state information in the HashMap. | ||
3522 | * | ||
3523 | * Contains the same info as the Regex Iterator parametes except the key, | ||
3524 | * which comes directly from the HashMap iterator. | ||
3525 | */ | ||
3526 | struct temporal_state_store { | ||
3527 | int reachable; | ||
3528 | char *proof; | ||
3529 | int accepting; | ||
3530 | int num_edges; | ||
3531 | struct REGEX_BLOCK_Edge *edges; | ||
3532 | }; | ||
3533 | |||
3534 | |||
3535 | /** | ||
3536 | * Store regex iterator and cls in one place to pass to the hashmap iterator. | ||
3537 | */ | ||
3538 | struct client_iterator { | ||
3539 | REGEX_INTERNAL_KeyIterator iterator; | ||
3540 | void *iterator_cls; | ||
3541 | }; | ||
3542 | |||
3543 | |||
3544 | /** | ||
3545 | * Iterator over all edges of a dfa. Stores all of them in a HashMap | ||
3546 | * for later reachability marking. | ||
3547 | * | ||
3548 | * @param cls Closure (HashMap) | ||
3549 | * @param key hash for current state. | ||
3550 | * @param proof proof for current state | ||
3551 | * @param accepting GNUNET_YES if this is an accepting state, GNUNET_NO if not. | ||
3552 | * @param num_edges number of edges leaving current state. | ||
3553 | * @param edges edges leaving current state. | ||
3554 | */ | ||
3555 | static void | ||
3556 | store_all_states (void *cls, | ||
3557 | const struct GNUNET_HashCode *key, | ||
3558 | const char *proof, | ||
3559 | int accepting, | ||
3560 | unsigned int num_edges, | ||
3561 | const struct REGEX_BLOCK_Edge *edges) | ||
3562 | { | ||
3563 | struct GNUNET_CONTAINER_MultiHashMap *hm = cls; | ||
3564 | struct temporal_state_store *tmp; | ||
3565 | size_t edges_size; | ||
3566 | |||
3567 | tmp = GNUNET_new (struct temporal_state_store); | ||
3568 | tmp->reachable = GNUNET_NO; | ||
3569 | tmp->proof = GNUNET_strdup (proof); | ||
3570 | tmp->accepting = accepting; | ||
3571 | tmp->num_edges = num_edges; | ||
3572 | edges_size = sizeof (struct REGEX_BLOCK_Edge) * num_edges; | ||
3573 | tmp->edges = GNUNET_malloc (edges_size); | ||
3574 | GNUNET_memcpy(tmp->edges, edges, edges_size); | ||
3575 | GNUNET_CONTAINER_multihashmap_put (hm, key, tmp, | ||
3576 | GNUNET_CONTAINER_MULTIHASHMAPOPTION_UNIQUE_FAST); | ||
3577 | } | ||
3578 | |||
3579 | |||
3580 | /** | ||
3581 | * Mark state as reachable and call recursively on all its edges. | ||
3582 | * | ||
3583 | * If already marked as reachable, do nothing. | ||
3584 | * | ||
3585 | * @param state State to mark as reachable. | ||
3586 | * @param hm HashMap which stores all the states indexed by key. | ||
3587 | */ | ||
3588 | static void | ||
3589 | mark_as_reachable (struct temporal_state_store *state, | ||
3590 | struct GNUNET_CONTAINER_MultiHashMap *hm) | ||
3591 | { | ||
3592 | struct temporal_state_store *child; | ||
3593 | unsigned int i; | ||
3594 | |||
3595 | if (GNUNET_YES == state->reachable) | ||
3596 | /* visited */ | ||
3597 | return; | ||
3598 | |||
3599 | state->reachable = GNUNET_YES; | ||
3600 | for (i = 0; i < state->num_edges; i++) | ||
3601 | { | ||
3602 | child = GNUNET_CONTAINER_multihashmap_get (hm, | ||
3603 | &state->edges[i].destination); | ||
3604 | if (NULL == child) | ||
3605 | { | ||
3606 | GNUNET_break (0); | ||
3607 | continue; | ||
3608 | } | ||
3609 | mark_as_reachable (child, hm); | ||
3610 | } | ||
3611 | } | ||
3612 | |||
3613 | |||
3614 | /** | ||
3615 | * Iterator over hash map entries to mark the ones that are reachable. | ||
3616 | * | ||
3617 | * @param cls closure | ||
3618 | * @param key current key code | ||
3619 | * @param value value in the hash map | ||
3620 | * @return #GNUNET_YES if we should continue to iterate, | ||
3621 | * #GNUNET_NO if not. | ||
3622 | */ | ||
3623 | static int | ||
3624 | reachability_iterator (void *cls, | ||
3625 | const struct GNUNET_HashCode *key, | ||
3626 | void *value) | ||
3627 | { | ||
3628 | struct GNUNET_CONTAINER_MultiHashMap *hm = cls; | ||
3629 | struct temporal_state_store *state = value; | ||
3630 | |||
3631 | if (GNUNET_YES == state->reachable) | ||
3632 | /* already visited and marked */ | ||
3633 | return GNUNET_YES; | ||
3634 | |||
3635 | if (GNUNET_REGEX_INITIAL_BYTES > strlen (state->proof) && | ||
3636 | GNUNET_NO == state->accepting) | ||
3637 | /* not directly reachable */ | ||
3638 | return GNUNET_YES; | ||
3639 | |||
3640 | mark_as_reachable (state, hm); | ||
3641 | return GNUNET_YES; | ||
3642 | } | ||
3643 | |||
3644 | |||
3645 | /** | ||
3646 | * Iterator over hash map entries. | ||
3647 | * Calling the callback on the ones marked as reachables. | ||
3648 | * | ||
3649 | * @param cls closure | ||
3650 | * @param key current key code | ||
3651 | * @param value value in the hash map | ||
3652 | * @return #GNUNET_YES if we should continue to iterate, | ||
3653 | * #GNUNET_NO if not. | ||
3654 | */ | ||
3655 | static int | ||
3656 | iterate_reachables (void *cls, | ||
3657 | const struct GNUNET_HashCode *key, | ||
3658 | void *value) | ||
3659 | { | ||
3660 | struct client_iterator *ci = cls; | ||
3661 | struct temporal_state_store *state = value; | ||
3662 | |||
3663 | if (GNUNET_YES == state->reachable) | ||
3664 | { | ||
3665 | ci->iterator (ci->iterator_cls, key, | ||
3666 | state->proof, state->accepting, | ||
3667 | state->num_edges, state->edges); | ||
3668 | } | ||
3669 | GNUNET_free (state->edges); | ||
3670 | GNUNET_free (state->proof); | ||
3671 | GNUNET_free (state); | ||
3672 | return GNUNET_YES; | ||
3673 | |||
3674 | } | ||
3675 | |||
3676 | /** | ||
3677 | * Iterate over all edges of automaton 'a' that are reachable from a state with | ||
3678 | * a proof of at least GNUNET_REGEX_INITIAL_BYTES characters. | ||
3679 | * | ||
3680 | * Call the iterator for each such edge. | ||
3681 | * | ||
3682 | * @param a automaton. | ||
3683 | * @param iterator iterator called for each reachable edge. | ||
3684 | * @param iterator_cls closure. | ||
3685 | */ | ||
3686 | void | ||
3687 | REGEX_INTERNAL_iterate_reachable_edges (struct REGEX_INTERNAL_Automaton *a, | ||
3688 | REGEX_INTERNAL_KeyIterator iterator, | ||
3689 | void *iterator_cls) | ||
3690 | { | ||
3691 | struct GNUNET_CONTAINER_MultiHashMap *hm; | ||
3692 | struct client_iterator ci; | ||
3693 | |||
3694 | hm = GNUNET_CONTAINER_multihashmap_create (a->state_count * 2, GNUNET_NO); | ||
3695 | ci.iterator = iterator; | ||
3696 | ci.iterator_cls = iterator_cls; | ||
3697 | |||
3698 | REGEX_INTERNAL_iterate_all_edges (a, &store_all_states, hm); | ||
3699 | GNUNET_CONTAINER_multihashmap_iterate (hm, &reachability_iterator, hm); | ||
3700 | GNUNET_CONTAINER_multihashmap_iterate (hm, &iterate_reachables, &ci); | ||
3701 | |||
3702 | GNUNET_CONTAINER_multihashmap_destroy (hm); | ||
3703 | } | ||
3704 | |||
3705 | |||
3706 | /* end of regex_internal.c */ | ||