2 This file is part of GNUnet
3 (C) 2012 Christian Grothoff (and other contributing authors)
5 GNUnet is free software; you can redistribute it and/or modify
6 it under the terms of the GNU General Public License as published
7 by the Free Software Foundation; either version 3, or (at your
8 option) any later version.
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 General Public License for more details.
15 You should have received a copy of the GNU General Public License
16 along with GNUnet; see the file COPYING. If not, write to the
17 Free Software Foundation, Inc., 59 Temple Place - Suite 330,
18 Boston, MA 02111-1307, USA.
21 * @file src/regex/regex.c
22 * @brief library to create automatons from regular expressions
23 * @author Maximilian Szengel
26 #include "gnunet_container_lib.h"
27 #include "gnunet_crypto_lib.h"
28 #include "gnunet_regex_lib.h"
29 #include "regex_internal.h"
33 * Constant for how many bits the initial string regex should have.
35 #define INITIAL_BITS 8
41 struct GNUNET_REGEX_StateSet
46 struct GNUNET_REGEX_State **states;
49 * Length of the 'states' array.
56 * Compare two strings for equality. If either is NULL they are not equal.
58 * @param str1 first string for comparison.
59 * @param str2 second string for comparison.
61 * @return 0 if the strings are the same or both NULL, 1 or -1 if not.
64 nullstrcmp (const char *str1, const char *str2)
66 if ((NULL == str1) != (NULL == str2))
68 if ((NULL == str1) && (NULL == str2))
71 return strcmp (str1, str2);
76 * Adds a transition from one state to another on 'label'. Does not add
80 * @param from_state starting state for the transition
81 * @param label transition label
82 * @param to_state state to where the transition should point to
85 state_add_transition (struct GNUNET_REGEX_Context *ctx,
86 struct GNUNET_REGEX_State *from_state, const char *label,
87 struct GNUNET_REGEX_State *to_state)
90 struct GNUNET_REGEX_Transition *t;
91 struct GNUNET_REGEX_Transition *oth;
93 if (NULL == from_state)
95 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not create Transition.\n");
99 // Do not add duplicate state transitions
101 for (t = from_state->transitions_head; NULL != t; t = t->next)
103 if (t->to_state == to_state && 0 == nullstrcmp (t->label, label) &&
104 t->from_state == from_state)
111 if (GNUNET_YES == is_dup)
114 // sort transitions by label
115 for (oth = from_state->transitions_head; NULL != oth; oth = oth->next)
117 if (0 < nullstrcmp (oth->label, label))
121 t = GNUNET_malloc (sizeof (struct GNUNET_REGEX_Transition));
123 t->id = ctx->transition_id++;
125 t->label = GNUNET_strdup (label);
128 t->to_state = to_state;
129 t->from_state = from_state;
131 // Add outgoing transition to 'from_state'
132 from_state->transition_count++;
133 GNUNET_CONTAINER_DLL_insert_before (from_state->transitions_head,
134 from_state->transitions_tail, oth, t);
139 * Remove a 'transition' from 'state'.
141 * @param state state from which the to-be-removed transition originates.
142 * @param transition transition that should be removed from state 'state'.
145 state_remove_transition (struct GNUNET_REGEX_State *state,
146 struct GNUNET_REGEX_Transition *transition)
148 if (NULL == state || NULL == transition)
151 if (transition->from_state != state)
154 state->transition_count--;
155 GNUNET_CONTAINER_DLL_remove (state->transitions_head, state->transitions_tail,
157 GNUNET_free_non_null (transition->label);
158 GNUNET_free (transition);
163 * Compare two states. Used for sorting.
165 * @param a first state
166 * @param b second state
168 * @return an integer less than, equal to, or greater than zero
169 * if the first argument is considered to be respectively
170 * less than, equal to, or greater than the second.
173 state_compare (const void *a, const void *b)
175 struct GNUNET_REGEX_State **s1;
176 struct GNUNET_REGEX_State **s2;
178 s1 = (struct GNUNET_REGEX_State **) a;
179 s2 = (struct GNUNET_REGEX_State **) b;
181 return (*s1)->id - (*s2)->id;
186 * Get all edges leaving state 's'.
189 * @param edges all edges leaving 's', expected to be allocated and have enough
190 * space for s->transitions_count elements.
192 * @return number of edges.
195 state_get_edges (struct GNUNET_REGEX_State *s, struct GNUNET_REGEX_Edge *edges)
197 struct GNUNET_REGEX_Transition *t;
205 for (t = s->transitions_head; NULL != t; t = t->next)
207 if (NULL != t->to_state)
209 edges[count].label = t->label;
210 edges[count].destination = t->to_state->hash;
219 * Compare to state sets by comparing the id's of the states that are contained
220 * in each set. Both sets are expected to be sorted by id!
222 * @param sset1 first state set
223 * @param sset2 second state set
225 * @return an integer less than, equal to, or greater than zero
226 * if the first argument is considered to be respectively
227 * less than, equal to, or greater than the second.
230 state_set_compare (struct GNUNET_REGEX_StateSet *sset1,
231 struct GNUNET_REGEX_StateSet *sset2)
236 if (NULL == sset1 || NULL == sset2)
239 result = sset1->len - sset2->len;
241 for (i = 0; i < sset1->len; i++)
246 result = state_compare (&sset1->states[i], &sset2->states[i]);
253 * Clears the given StateSet 'set'
255 * @param set set to be cleared
258 state_set_clear (struct GNUNET_REGEX_StateSet *set)
262 GNUNET_free_non_null (set->states);
269 * Clears an automaton fragment. Does not destroy the states inside the
272 * @param a automaton to be cleared
275 automaton_fragment_clear (struct GNUNET_REGEX_Automaton *a)
282 a->states_head = NULL;
283 a->states_tail = NULL;
290 * Frees the memory used by State 's'
292 * @param s state that should be destroyed
295 automaton_destroy_state (struct GNUNET_REGEX_State *s)
297 struct GNUNET_REGEX_Transition *t;
298 struct GNUNET_REGEX_Transition *next_t;
303 GNUNET_free_non_null (s->name);
304 GNUNET_free_non_null (s->proof);
306 for (t = s->transitions_head; NULL != t; t = next_t)
309 GNUNET_CONTAINER_DLL_remove (s->transitions_head, s->transitions_tail, t);
310 GNUNET_free_non_null (t->label);
314 state_set_clear (s->nfa_set);
321 * Remove a state from the given automaton 'a'. Always use this function when
322 * altering the states of an automaton. Will also remove all transitions leading
323 * to this state, before destroying it.
326 * @param s state to remove
329 automaton_remove_state (struct GNUNET_REGEX_Automaton *a,
330 struct GNUNET_REGEX_State *s)
332 struct GNUNET_REGEX_State *ss;
333 struct GNUNET_REGEX_State *s_check;
334 struct GNUNET_REGEX_Transition *t_check;
336 if (NULL == a || NULL == s)
341 GNUNET_CONTAINER_DLL_remove (a->states_head, a->states_tail, s);
344 // remove all transitions leading to this state
345 for (s_check = a->states_head; NULL != s_check; s_check = s_check->next)
347 for (t_check = s_check->transitions_head; NULL != t_check;
348 t_check = t_check->next)
350 if (t_check->to_state == ss)
352 GNUNET_CONTAINER_DLL_remove (s_check->transitions_head,
353 s_check->transitions_tail, t_check);
354 s_check->transition_count--;
359 automaton_destroy_state (ss);
364 * Merge two states into one. Will merge 's1' and 's2' into 's1' and destroy
369 * @param s1 first state
370 * @param s2 second state, will be destroyed
373 automaton_merge_states (struct GNUNET_REGEX_Context *ctx,
374 struct GNUNET_REGEX_Automaton *a,
375 struct GNUNET_REGEX_State *s1,
376 struct GNUNET_REGEX_State *s2)
378 struct GNUNET_REGEX_State *s_check;
379 struct GNUNET_REGEX_Transition *t_check;
380 struct GNUNET_REGEX_Transition *t;
381 struct GNUNET_REGEX_Transition *t_next;
385 GNUNET_assert (NULL != ctx && NULL != a && NULL != s1 && NULL != s2);
390 // 1. Make all transitions pointing to s2 point to s1, unless this transition
391 // does not already exists, if it already exists remove transition.
392 for (s_check = a->states_head; NULL != s_check; s_check = s_check->next)
394 for (t_check = s_check->transitions_head; NULL != t_check; t_check = t_next)
396 t_next = t_check->next;
398 if (s2 == t_check->to_state)
401 for (t = t_check->from_state->transitions_head; NULL != t; t = t->next)
403 if (t->to_state == s1 && 0 == strcmp (t_check->label, t->label))
406 if (GNUNET_NO == is_dup)
407 t_check->to_state = s1;
409 state_remove_transition (t_check->from_state, t_check);
414 // 2. Add all transitions from s2 to sX to s1
415 for (t_check = s2->transitions_head; NULL != t_check; t_check = t_check->next)
417 if (t_check->to_state != s1)
418 state_add_transition (ctx, s1, t_check->label, t_check->to_state);
421 // 3. Rename s1 to {s1,s2}
423 GNUNET_asprintf (&s1->name, "{%s,%s}", new_name, s2->name);
424 GNUNET_free (new_name);
427 GNUNET_CONTAINER_DLL_remove (a->states_head, a->states_tail, s2);
429 automaton_destroy_state (s2);
434 * Add a state to the automaton 'a', always use this function to alter the
435 * states DLL of the automaton.
437 * @param a automaton to add the state to
438 * @param s state that should be added
441 automaton_add_state (struct GNUNET_REGEX_Automaton *a,
442 struct GNUNET_REGEX_State *s)
444 GNUNET_CONTAINER_DLL_insert (a->states_head, a->states_tail, s);
450 * Depth-first traversal (DFS) of all states that are reachable from state
451 * 's'. Performs 'action' on each visited state.
453 * @param s start state.
454 * @param marks an array of size a->state_count to remember which state was
456 * @param count current count of the state.
457 * @param check function that is checked before advancing on each transition
459 * @param check_cls closure for check.
460 * @param action action to be performed on each state.
461 * @param action_cls closure for action.
464 automaton_state_traverse (struct GNUNET_REGEX_State *s, int *marks,
466 GNUNET_REGEX_traverse_check check, void *check_cls,
467 GNUNET_REGEX_traverse_action action, void *action_cls)
469 struct GNUNET_REGEX_Transition *t;
471 if (GNUNET_YES == marks[s->traversal_id])
474 marks[s->traversal_id] = GNUNET_YES;
477 action (action_cls, *count, s);
481 for (t = s->transitions_head; NULL != t; t = t->next)
484 (NULL != check && GNUNET_YES == check (check_cls, s, t)))
486 automaton_state_traverse (t->to_state, marks, count, check, check_cls,
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
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
502 * @param check_cls closure for check.
503 * @param action action to be performed on each state.
504 * @param action_cls closure for action
507 GNUNET_REGEX_automaton_traverse (const struct GNUNET_REGEX_Automaton *a,
508 struct GNUNET_REGEX_State *start,
509 GNUNET_REGEX_traverse_check check,
511 GNUNET_REGEX_traverse_action action,
515 struct GNUNET_REGEX_State *s;
517 if (NULL == a || 0 == a->state_count)
520 int marks[a->state_count];
522 for (count = 0, s = a->states_head; NULL != s && count < a->state_count;
523 s = s->next, count++)
525 s->traversal_id = count;
526 marks[s->traversal_id] = GNUNET_NO;
536 automaton_state_traverse (s, marks, &count, check, check_cls, action,
542 * Context for adding strided transitions to a DFA.
544 struct GNUNET_REGEX_Strided_Context
547 * Length of the strides.
549 const unsigned int stride;
552 * Strided transitions DLL. New strided transitions will be stored in this DLL
553 * and afterwards added to the DFA.
555 struct GNUNET_REGEX_Transition *transitions_head;
558 * Strided transitions DLL.
560 struct GNUNET_REGEX_Transition *transitions_tail;
565 * Recursive helper function to add strides to a DFA.
567 * @param cls context, contains stride length and strided transitions DLL.
568 * @param depth current depth of the depth-first traversal of the graph.
569 * @param label current label, string that contains all labels on the path from
571 * @param start start state for the depth-first traversal of the graph.
572 * @param s current state in the depth-first traversal
575 add_multi_strides_to_dfa_helper (void *cls, const unsigned int depth,
576 char *label, struct GNUNET_REGEX_State *start,
577 struct GNUNET_REGEX_State *s)
579 struct GNUNET_REGEX_Strided_Context *ctx = cls;
580 struct GNUNET_REGEX_Transition *t;
583 if (depth == ctx->stride)
585 t = GNUNET_malloc (sizeof (struct GNUNET_REGEX_Transition));
586 t->label = GNUNET_strdup (label);
588 t->from_state = start;
589 GNUNET_CONTAINER_DLL_insert (ctx->transitions_head, ctx->transitions_tail,
594 for (t = s->transitions_head; NULL != t; t = t->next)
596 /* Do not consider self-loops, because it end's up in too many
598 if (t->to_state == t->from_state)
603 GNUNET_asprintf (&new_label, "%s%s", label, t->label);
606 new_label = GNUNET_strdup (t->label);
608 add_multi_strides_to_dfa_helper (cls, (depth + 1), new_label, start,
612 GNUNET_free_non_null (label);
617 * Function called for each state in the DFA. Starts a traversal of depth set in
618 * context starting from state 's'.
620 * @param cls context.
621 * @param count not used.
622 * @param s current state.
625 add_multi_strides_to_dfa (void *cls, const unsigned int count,
626 struct GNUNET_REGEX_State *s)
628 add_multi_strides_to_dfa_helper (cls, 0, NULL, s, s);
633 * Adds multi-strided transitions to the given 'dfa'.
635 * @param regex_ctx regex context needed to add transitions to the automaton.
636 * @param dfa DFA to which the multi strided transitions should be added.
637 * @param stride_len length of the strides.
640 GNUNET_REGEX_add_multi_strides_to_dfa (struct GNUNET_REGEX_Context *regex_ctx,
641 struct GNUNET_REGEX_Automaton *dfa,
642 const unsigned int stride_len)
644 struct GNUNET_REGEX_Strided_Context ctx = { stride_len, NULL, NULL };
645 struct GNUNET_REGEX_State *s;
646 struct GNUNET_REGEX_State *s_next;
647 struct GNUNET_REGEX_Transition *t;
648 struct GNUNET_REGEX_Transition *t_next;
650 if (1 > stride_len || GNUNET_YES == dfa->is_multistrided)
654 for (s = dfa->states_head; NULL != s; s = s->next)
655 s->marked = GNUNET_NO;
657 // Compute the new transitions of given stride_len
658 GNUNET_REGEX_automaton_traverse (dfa, dfa->start, NULL, NULL,
659 &add_multi_strides_to_dfa, &ctx);
661 // Add all the new transitions to the automaton.
662 for (t = ctx.transitions_head; NULL != t; t = t_next)
665 state_add_transition (regex_ctx, t->from_state, t->label, t->to_state);
666 GNUNET_CONTAINER_DLL_remove (ctx.transitions_head, ctx.transitions_tail, t);
667 GNUNET_free_non_null (t->label);
671 // Remove marked states (including their incoming and outgoing transitions)
672 for (s = dfa->states_head; NULL != s; s = s_next)
675 if (GNUNET_YES == s->marked)
676 automaton_remove_state (dfa, s);
679 // Mark this automaton as multistrided
680 dfa->is_multistrided = GNUNET_YES;
686 * Check if the given string 'str' needs parentheses around it when
687 * using it to generate a regex.
691 * @return GNUNET_YES if parentheses are needed, GNUNET_NO otherwise
694 needs_parentheses (const char *str)
702 if ((NULL == str) || ((slen = strlen (str)) < 2))
711 cl = strchr (pos, ')');
717 op = strchr (pos, '(');
718 if ((NULL != op) && (op < cl))
728 return (*pos == '\0') ? GNUNET_NO : GNUNET_YES;
733 * Remove parentheses surrounding string 'str'.
734 * Example: "(a)" becomes "a", "(a|b)|(a|c)" stays the same.
735 * You need to GNUNET_free the returned string.
737 * @param str string, free'd or re-used by this function, can be NULL
739 * @return string without surrounding parentheses, string 'str' if no preceding
740 * epsilon could be found, NULL if 'str' was NULL
743 remove_parentheses (char *str)
748 if ((NULL == str) || ('(' != str[0]) ||
749 (str[(slen = strlen (str)) - 1] != ')'))
752 pos = strchr (&str[1], ')');
753 if (pos == &str[slen - 1])
755 memmove (str, &str[1], slen - 2);
756 str[slen - 2] = '\0';
763 * Check if the string 'str' starts with an epsilon (empty string).
764 * Example: "(|a)" is starting with an epsilon.
766 * @param str string to test
768 * @return 0 if str has no epsilon, 1 if str starts with '(|' and ends with ')'
771 has_epsilon (const char *str)
773 return (NULL != str) && ('(' == str[0]) && ('|' == str[1]) &&
774 (')' == str[strlen (str) - 1]);
779 * Remove an epsilon from the string str. Where epsilon is an empty string
780 * Example: str = "(|a|b|c)", result: "a|b|c"
781 * The returned string needs to be freed.
785 * @return string without preceding epsilon, string 'str' if no preceding
786 * epsilon could be found, NULL if 'str' was NULL
789 remove_epsilon (const char *str)
795 if (('(' == str[0]) && ('|' == str[1]))
798 if (')' == str[len - 1])
799 return GNUNET_strndup (&str[2], len - 3);
801 return GNUNET_strdup (str);
806 * Compare 'str1', starting from position 'k', with whole 'str2'
808 * @param str1 first string to compare, starting from position 'k'
809 * @param str2 second string for comparison
810 * @param k starting position in 'str1'
812 * @return -1 if any of the strings is NULL, 0 if equal, non 0 otherwise
815 strkcmp (const char *str1, const char *str2, size_t k)
817 if ((NULL == str1) || (NULL == str2) || (strlen (str1) < k))
819 return strcmp (&str1[k], str2);
824 * Helper function used as 'action' in 'GNUNET_REGEX_automaton_traverse'
825 * function to create the depth-first numbering of the states.
827 * @param cls states array.
828 * @param count current state counter.
829 * @param s current state.
832 number_states (void *cls, const unsigned int count,
833 struct GNUNET_REGEX_State *s)
835 struct GNUNET_REGEX_State **states = cls;
844 * Construct the regular expression given the inductive step,
845 * $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^*
846 * R^{(k-1)}_{kj}, and simplify the resulting expression saved in R_cur_ij.
848 * @param R_last_ij value of $R^{(k-1)_{ij}.
849 * @param R_last_ik value of $R^{(k-1)_{ik}.
850 * @param R_last_kk value of $R^{(k-1)_{kk}.
851 * @param R_last_kj value of $R^{(k-1)_{kj}.
852 * @param R_cur_ij result for this inductive step is saved in R_cur_ij, R_cur_ij
853 * is expected to be NULL when called!
856 automaton_create_proofs_simplify (char *R_last_ij, char *R_last_ik,
857 char *R_last_kk, char *R_last_kj,
883 GNUNET_assert (NULL == *R_cur_ij && NULL != R_cur_ij);
885 // $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj}
886 // R_last == R^{(k-1)}, R_cur == R^{(k)}
887 // R_cur_ij = R_cur_l | R_cur_r
888 // R_cur_l == R^{(k-1)}_{ij}
889 // R_cur_r == R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj}
891 if ((NULL == R_last_ij) && ((NULL == R_last_ik) || (NULL == R_last_kk) || /* technically cannot happen, but looks saner */
892 (NULL == R_last_kj)))
894 /* R^{(k)}_{ij} = N | N */
899 if ((NULL == R_last_ik) || (NULL == R_last_kk) || /* technically cannot happen, but looks saner */
902 /* R^{(k)}_{ij} = R^{(k-1)}_{ij} | N */
903 *R_cur_ij = GNUNET_strdup (R_last_ij);
907 // $R^{(k)}_{ij} = N | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj} OR
908 // $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj}
913 // cache results from strcmp, we might need these many times
914 ij_kj_cmp = nullstrcmp (R_last_ij, R_last_kj);
915 ij_ik_cmp = nullstrcmp (R_last_ij, R_last_ik);
916 ik_kk_cmp = nullstrcmp (R_last_ik, R_last_kk);
917 kk_kj_cmp = nullstrcmp (R_last_kk, R_last_kj);
919 // Assign R_temp_(ik|kk|kj) to R_last[][] and remove epsilon as well
920 // as parentheses, so we can better compare the contents
921 R_temp_ik = remove_parentheses (remove_epsilon (R_last_ik));
922 R_temp_kk = remove_parentheses (remove_epsilon (R_last_kk));
923 R_temp_kj = remove_parentheses (remove_epsilon (R_last_kj));
925 clean_ik_kk_cmp = nullstrcmp (R_last_ik, R_temp_kk);
926 clean_kk_kj_cmp = nullstrcmp (R_temp_kk, R_last_kj);
928 // construct R_cur_l (and, if necessary R_cur_r)
929 if (NULL != R_last_ij)
931 // Assign R_temp_ij to R_last_ij and remove epsilon as well
932 // as parentheses, so we can better compare the contents
933 R_temp_ij = remove_parentheses (remove_epsilon (R_last_ij));
935 if (0 == strcmp (R_temp_ij, R_temp_ik) && 0 == strcmp (R_temp_ik, R_temp_kk)
936 && 0 == strcmp (R_temp_kk, R_temp_kj))
938 if (0 == strlen (R_temp_ij))
940 R_cur_r = GNUNET_strdup ("");
942 else if ((0 == strncmp (R_last_ij, "(|", 2)) ||
943 (0 == strncmp (R_last_ik, "(|", 2) &&
944 0 == strncmp (R_last_kj, "(|", 2)))
946 // a|(e|a)a*(e|a) = a*
947 // a|(e|a)(e|a)*(e|a) = a*
949 // (e|a)|aa*(e|a) = a*
950 // (e|a)|(e|a)a*a = a*
951 // (e|a)|(e|a)a*(e|a) = a*
952 // (e|a)|(e|a)(e|a)*(e|a) = a*
953 if (GNUNET_YES == needs_parentheses (R_temp_ij))
954 GNUNET_asprintf (&R_cur_r, "(%s)*", R_temp_ij);
956 GNUNET_asprintf (&R_cur_r, "%s*", R_temp_ij);
963 // a|(e|a)(e|a)*a = a+
964 // a|a(e|a)*(e|a) = a+
965 if (GNUNET_YES == needs_parentheses (R_temp_ij))
966 GNUNET_asprintf (&R_cur_r, "(%s)+", R_temp_ij);
968 GNUNET_asprintf (&R_cur_r, "%s+", R_temp_ij);
971 else if (0 == ij_ik_cmp && 0 == clean_kk_kj_cmp && 0 != clean_ik_kk_cmp)
974 if (strlen (R_last_kk) < 1)
975 R_cur_r = GNUNET_strdup (R_last_ij);
976 else if (GNUNET_YES == needs_parentheses (R_temp_kk))
977 GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last_ij, R_temp_kk);
979 GNUNET_asprintf (&R_cur_r, "%s%s*", R_last_ij, R_last_kk);
983 else if (0 == ij_kj_cmp && 0 == clean_ik_kk_cmp && 0 != clean_kk_kj_cmp)
986 if (strlen (R_last_kk) < 1)
987 R_cur_r = GNUNET_strdup (R_last_kj);
988 else if (GNUNET_YES == needs_parentheses (R_temp_kk))
989 GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk, R_last_kj);
991 GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last_kj);
995 else if (0 == ij_ik_cmp && 0 == kk_kj_cmp && !has_epsilon (R_last_ij) &&
996 has_epsilon (R_last_kk))
998 // a|a(e|b)*(e|b) = a|ab* = a|a|ab|abb|abbb|... = ab*
999 if (needs_parentheses (R_temp_kk))
1000 GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last_ij, R_temp_kk);
1002 GNUNET_asprintf (&R_cur_r, "%s%s*", R_last_ij, R_temp_kk);
1006 else if (0 == ij_kj_cmp && 0 == ik_kk_cmp && !has_epsilon (R_last_ij) &&
1007 has_epsilon (R_last_kk))
1009 // a|(e|b)(e|b)*a = a|b*a = a|a|ba|bba|bbba|... = b*a
1010 if (needs_parentheses (R_temp_kk))
1011 GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk, R_last_ij);
1013 GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last_ij);
1019 temp_a = (NULL == R_last_ij) ? NULL : GNUNET_strdup (R_last_ij);
1020 temp_a = remove_parentheses (temp_a);
1024 GNUNET_free_non_null (R_temp_ij);
1028 // we have no left side
1032 // construct R_cur_r, if not already constructed
1033 if (NULL == R_cur_r)
1035 length = strlen (R_temp_kk) - strlen (R_last_ik);
1037 // a(ba)*bx = (ab)+x
1038 if (length > 0 && NULL != R_last_kk && 0 < strlen (R_last_kk) &&
1039 NULL != R_last_kj && 0 < strlen (R_last_kj) && NULL != R_last_ik &&
1040 0 < strlen (R_last_ik) && 0 == strkcmp (R_temp_kk, R_last_ik, length) &&
1041 0 == strncmp (R_temp_kk, R_last_kj, length))
1043 temp_a = GNUNET_malloc (length + 1);
1044 temp_b = GNUNET_malloc ((strlen (R_last_kj) - length) + 1);
1049 for (cnt = 0; cnt < strlen (R_last_kj); cnt++)
1053 temp_a[length_l] = R_last_kj[cnt];
1058 temp_b[length_r] = R_last_kj[cnt];
1062 temp_a[length_l] = '\0';
1063 temp_b[length_r] = '\0';
1066 if (NULL != R_cur_l && 0 == strlen (R_cur_l) && 0 == strlen (temp_b))
1068 GNUNET_asprintf (&R_cur_r, "(%s%s)*", R_last_ik, temp_a);
1069 GNUNET_free (R_cur_l);
1074 GNUNET_asprintf (&R_cur_r, "(%s%s)+%s", R_last_ik, temp_a, temp_b);
1076 GNUNET_free (temp_a);
1077 GNUNET_free (temp_b);
1079 else if (0 == strcmp (R_temp_ik, R_temp_kk) &&
1080 0 == strcmp (R_temp_kk, R_temp_kj))
1082 // (e|a)a*(e|a) = a*
1083 // (e|a)(e|a)*(e|a) = a*
1084 if (has_epsilon (R_last_ik) && has_epsilon (R_last_kj))
1086 if (needs_parentheses (R_temp_kk))
1087 GNUNET_asprintf (&R_cur_r, "(%s)*", R_temp_kk);
1089 GNUNET_asprintf (&R_cur_r, "%s*", R_temp_kk);
1092 else if (0 == clean_ik_kk_cmp && 0 == clean_kk_kj_cmp &&
1093 !has_epsilon (R_last_ik))
1095 if (needs_parentheses (R_temp_kk))
1096 GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_temp_kk, R_temp_kk);
1098 GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_temp_kk, R_temp_kk);
1102 // a(e|a)*(e|a) = a+
1107 (has_epsilon (R_last_ik) + has_epsilon (R_last_kk) +
1108 has_epsilon (R_last_kj));
1112 if (needs_parentheses (R_temp_kk))
1113 GNUNET_asprintf (&R_cur_r, "(%s)+", R_temp_kk);
1115 GNUNET_asprintf (&R_cur_r, "%s+", R_temp_kk);
1120 // (e|a)(e|a)*b = a*b
1121 else if (0 == strcmp (R_temp_ik, R_temp_kk))
1123 if (has_epsilon (R_last_ik))
1125 if (needs_parentheses (R_temp_kk))
1126 GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk, R_last_kj);
1128 GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last_kj);
1132 if (needs_parentheses (R_temp_kk))
1133 GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_temp_kk, R_last_kj);
1135 GNUNET_asprintf (&R_cur_r, "%s+%s", R_temp_kk, R_last_kj);
1139 // b(e|a)*(e|a) = ba*
1140 else if (0 == strcmp (R_temp_kk, R_temp_kj))
1142 if (has_epsilon (R_last_kj))
1144 if (needs_parentheses (R_temp_kk))
1145 GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last_ik, R_temp_kk);
1147 GNUNET_asprintf (&R_cur_r, "%s%s*", R_last_ik, R_temp_kk);
1151 if (needs_parentheses (R_temp_kk))
1152 GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_last_ik, R_temp_kk);
1154 GNUNET_asprintf (&R_cur_r, "%s+%s", R_last_ik, R_temp_kk);
1159 if (strlen (R_temp_kk) > 0)
1161 if (needs_parentheses (R_temp_kk))
1163 GNUNET_asprintf (&R_cur_r, "%s(%s)*%s", R_last_ik, R_temp_kk,
1168 GNUNET_asprintf (&R_cur_r, "%s%s*%s", R_last_ik, R_temp_kk,
1174 GNUNET_asprintf (&R_cur_r, "%s%s", R_last_ik, R_last_kj);
1179 GNUNET_free_non_null (R_temp_ik);
1180 GNUNET_free_non_null (R_temp_kk);
1181 GNUNET_free_non_null (R_temp_kj);
1183 if (NULL == R_cur_l && NULL == R_cur_r)
1189 if (NULL != R_cur_l && NULL == R_cur_r)
1191 *R_cur_ij = R_cur_l;
1195 if (NULL == R_cur_l && NULL != R_cur_r)
1197 *R_cur_ij = R_cur_r;
1201 if (0 == nullstrcmp (R_cur_l, R_cur_r))
1203 *R_cur_ij = R_cur_l;
1204 GNUNET_free (R_cur_r);
1208 GNUNET_asprintf (R_cur_ij, "(%s|%s)", R_cur_l, R_cur_r);
1210 GNUNET_free (R_cur_l);
1211 GNUNET_free (R_cur_r);
1216 * create proofs for all states in the given automaton. Implementation of the
1217 * algorithm descriped in chapter 3.2.1 of "Automata Theory, Languages, and
1218 * Computation 3rd Edition" by Hopcroft, Motwani and Ullman.
1220 * @param a automaton.
1223 automaton_create_proofs (struct GNUNET_REGEX_Automaton *a)
1227 GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
1228 "Could not create proofs, automaton was NULL\n");
1232 unsigned int n = a->state_count;
1233 struct GNUNET_REGEX_State *states[n];
1237 struct GNUNET_REGEX_Transition *t;
1238 char *complete_regex;
1243 /* create depth-first numbering of the states, initializes 'state' */
1244 GNUNET_REGEX_automaton_traverse (a, a->start, NULL, NULL, &number_states,
1247 for (i = 0; i < n; i++)
1248 GNUNET_assert (NULL != states[i]);
1250 /* Compute regular expressions of length "1" between each pair of states */
1251 for (i = 0; i < n; i++)
1253 for (j = 0; j < n; j++)
1256 R_last[i][j] = NULL;
1258 for (t = states[i]->transitions_head; NULL != t; t = t->next)
1260 j = t->to_state->dfs_id;
1261 if (NULL == R_last[i][j])
1262 GNUNET_asprintf (&R_last[i][j], "%s", t->label);
1265 temp = R_last[i][j];
1266 GNUNET_asprintf (&R_last[i][j], "%s|%s", R_last[i][j], t->label);
1270 if (NULL == R_last[i][i])
1271 GNUNET_asprintf (&R_last[i][i], "");
1274 temp = R_last[i][i];
1275 GNUNET_asprintf (&R_last[i][i], "(|%s)", R_last[i][i]);
1279 for (i = 0; i < n; i++)
1280 for (j = 0; j < n; j++)
1281 if (needs_parentheses (R_last[i][j]))
1283 temp = R_last[i][j];
1284 GNUNET_asprintf (&R_last[i][j], "(%s)", R_last[i][j]);
1288 /* Compute regular expressions of length "k" between each pair of states per
1290 for (k = 0; k < n; k++)
1292 for (i = 0; i < n; i++)
1294 for (j = 0; j < n; j++)
1296 // Basis for the recursion:
1297 // $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj}
1298 // R_last == R^{(k-1)}, R_cur == R^{(k)}
1300 // Create R_cur[i][j] and simplify the expression
1301 automaton_create_proofs_simplify (R_last[i][j], R_last[i][k],
1302 R_last[k][k], R_last[k][j],
1307 // set R_last = R_cur
1308 for (i = 0; i < n; i++)
1310 for (j = 0; j < n; j++)
1312 GNUNET_free_non_null (R_last[i][j]);
1313 R_last[i][j] = R_cur[i][j];
1319 // assign proofs and hashes
1320 for (i = 0; i < n; i++)
1322 if (NULL != R_last[a->start->dfs_id][i])
1324 states[i]->proof = GNUNET_strdup (R_last[a->start->dfs_id][i]);
1325 GNUNET_CRYPTO_hash (states[i]->proof, strlen (states[i]->proof),
1330 // complete regex for whole DFA: union of all pairs (start state/accepting
1332 complete_regex = NULL;
1333 for (i = 0; i < n; i++)
1335 if (states[i]->accepting)
1337 if (NULL == complete_regex && 0 < strlen (R_last[a->start->dfs_id][i]))
1339 GNUNET_asprintf (&complete_regex, "%s", R_last[a->start->dfs_id][i]);
1341 else if (NULL != R_last[a->start->dfs_id][i] &&
1342 0 < strlen (R_last[a->start->dfs_id][i]))
1344 temp = complete_regex;
1345 GNUNET_asprintf (&complete_regex, "%s|%s", complete_regex,
1346 R_last[a->start->dfs_id][i]);
1351 a->canonical_regex = complete_regex;
1354 for (i = 0; i < n; i++)
1356 for (j = 0; j < n; j++)
1357 GNUNET_free_non_null (R_last[i][j]);
1363 * Creates a new DFA state based on a set of NFA states. Needs to be freed using
1364 * automaton_destroy_state.
1366 * @param ctx context
1367 * @param nfa_states set of NFA states on which the DFA should be based on
1369 * @return new DFA state
1371 static struct GNUNET_REGEX_State *
1372 dfa_state_create (struct GNUNET_REGEX_Context *ctx,
1373 struct GNUNET_REGEX_StateSet *nfa_states)
1375 struct GNUNET_REGEX_State *s;
1378 struct GNUNET_REGEX_State *cstate;
1379 struct GNUNET_REGEX_Transition *ctran;
1382 s = GNUNET_malloc (sizeof (struct GNUNET_REGEX_State));
1383 s->id = ctx->state_id++;
1385 s->marked = GNUNET_NO;
1393 if (NULL == nfa_states)
1395 GNUNET_asprintf (&s->name, "s%i", s->id);
1399 s->nfa_set = nfa_states;
1401 if (nfa_states->len < 1)
1404 // Create a name based on 'nfa_states'
1405 s->name = GNUNET_malloc (sizeof (char) * 2);
1406 strcat (s->name, "{");
1409 for (i = 0; i < nfa_states->len; i++)
1411 cstate = nfa_states->states[i];
1412 GNUNET_asprintf (&name, "%i,", cstate->id);
1416 len = strlen (s->name) + strlen (name) + 1;
1417 s->name = GNUNET_realloc (s->name, len);
1418 strcat (s->name, name);
1423 // Add a transition for each distinct label to NULL state
1424 for (ctran = cstate->transitions_head; NULL != ctran; ctran = ctran->next)
1426 if (NULL != ctran->label)
1427 state_add_transition (ctx, s, ctran->label, NULL);
1430 // If the nfa_states contain an accepting state, the new dfa state is also
1432 if (cstate->accepting)
1436 s->name[strlen (s->name) - 1] = '}';
1443 * Move from the given state 's' to the next state on transition 'str'. Consumes
1444 * as much of the given 'str' as possible (usefull for strided DFAs). On return
1445 * 's' will point to the next state, and the length of the substring used for
1446 * this transition will be returned. If no transition possible 0 is returned and
1447 * 's' points to NULL.
1449 * @param s starting state, will point to the next state or NULL (if no
1450 * transition possible)
1451 * @param str edge label to follow (will match longest common prefix)
1453 * @return length of the substring comsumed from 'str'
1456 dfa_move (struct GNUNET_REGEX_State **s, const char *str)
1458 struct GNUNET_REGEX_Transition *t;
1459 struct GNUNET_REGEX_State *new_s;
1461 unsigned int max_len;
1468 for (t = (*s)->transitions_head; NULL != t; t = t->next)
1470 len = strlen (t->label);
1472 if (0 == strncmp (t->label, str, len))
1477 new_s = t->to_state;
1487 * Set the given state 'marked' to GNUNET_YES. Used by the
1488 * 'dfa_remove_unreachable_states' function to detect unreachable states in the
1491 * @param cls closure, not used.
1492 * @param count count, not used.
1493 * @param s state where the marked attribute will be set to GNUNET_YES.
1496 mark_states (void *cls, const unsigned int count, struct GNUNET_REGEX_State *s)
1498 s->marked = GNUNET_YES;
1502 * Remove all unreachable states from DFA 'a'. Unreachable states are those
1503 * states that are not reachable from the starting state.
1505 * @param a DFA automaton
1508 dfa_remove_unreachable_states (struct GNUNET_REGEX_Automaton *a)
1510 struct GNUNET_REGEX_State *s;
1511 struct GNUNET_REGEX_State *s_next;
1513 // 1. unmark all states
1514 for (s = a->states_head; NULL != s; s = s->next)
1515 s->marked = GNUNET_NO;
1517 // 2. traverse dfa from start state and mark all visited states
1518 GNUNET_REGEX_automaton_traverse (a, a->start, NULL, NULL, &mark_states, NULL);
1520 // 3. delete all states that were not visited
1521 for (s = a->states_head; NULL != s; s = s_next)
1524 if (GNUNET_NO == s->marked)
1525 automaton_remove_state (a, s);
1531 * Remove all dead states from the DFA 'a'. Dead states are those states that do
1532 * not transition to any other state but themselves.
1534 * @param a DFA automaton
1537 dfa_remove_dead_states (struct GNUNET_REGEX_Automaton *a)
1539 struct GNUNET_REGEX_State *s;
1540 struct GNUNET_REGEX_Transition *t;
1543 GNUNET_assert (DFA == a->type);
1545 for (s = a->states_head; NULL != s; s = s->next)
1551 for (t = s->transitions_head; NULL != t; t = t->next)
1553 if (NULL != t->to_state && t->to_state != s)
1563 // state s is dead, remove it
1564 automaton_remove_state (a, s);
1570 * Merge all non distinguishable states in the DFA 'a'
1572 * @param ctx context
1573 * @param a DFA automaton
1576 dfa_merge_nondistinguishable_states (struct GNUNET_REGEX_Context *ctx,
1577 struct GNUNET_REGEX_Automaton *a)
1579 int table[a->state_count][a->state_count];
1580 struct GNUNET_REGEX_State *s1;
1581 struct GNUNET_REGEX_State *s2;
1582 struct GNUNET_REGEX_Transition *t1;
1583 struct GNUNET_REGEX_Transition *t2;
1584 struct GNUNET_REGEX_State *s1_next;
1585 struct GNUNET_REGEX_State *s2_next;
1587 unsigned int num_equal_edges;
1590 for (i = 0, s1 = a->states_head; i < a->state_count && NULL != s1;
1596 // Mark all pairs of accepting/!accepting states
1597 for (s1 = a->states_head; NULL != s1; s1 = s1->next)
1599 for (s2 = a->states_head; NULL != s2; s2 = s2->next)
1601 table[s1->marked][s2->marked] = 0;
1603 if ((s1->accepting && !s2->accepting) ||
1604 (!s1->accepting && s2->accepting))
1606 table[s1->marked][s2->marked] = 1;
1611 // Find all equal states
1616 for (s1 = a->states_head; NULL != s1; s1 = s1->next)
1618 for (s2 = a->states_head; NULL != s2 && s1 != s2; s2 = s2->next)
1620 if (0 != table[s1->marked][s2->marked])
1623 num_equal_edges = 0;
1624 for (t1 = s1->transitions_head; NULL != t1; t1 = t1->next)
1626 for (t2 = s2->transitions_head; NULL != t2; t2 = t2->next)
1628 if (0 == strcmp (t1->label, t2->label))
1631 if (0 != table[t1->to_state->marked][t2->to_state->marked] ||
1632 0 != table[t2->to_state->marked][t1->to_state->marked])
1634 table[s1->marked][s2->marked] = 1;
1640 if (num_equal_edges != s1->transition_count ||
1641 num_equal_edges != s2->transition_count)
1643 // Make sure ALL edges of possible equal states are the same
1644 table[s1->marked][s2->marked] = -2;
1650 // Merge states that are equal
1651 for (s1 = a->states_head; NULL != s1; s1 = s1_next)
1654 for (s2 = a->states_head; NULL != s2 && s1 != s2; s2 = s2_next)
1657 if (table[s1->marked][s2->marked] == 0)
1658 automaton_merge_states (ctx, a, s1, s2);
1665 * Minimize the given DFA 'a' by removing all unreachable states, removing all
1666 * dead states and merging all non distinguishable states
1668 * @param ctx context
1669 * @param a DFA automaton
1672 dfa_minimize (struct GNUNET_REGEX_Context *ctx,
1673 struct GNUNET_REGEX_Automaton *a)
1678 GNUNET_assert (DFA == a->type);
1680 // 1. remove unreachable states
1681 dfa_remove_unreachable_states (a);
1683 // 2. remove dead states
1684 dfa_remove_dead_states (a);
1686 // 3. Merge nondistinguishable states
1687 dfa_merge_nondistinguishable_states (ctx, a);
1692 * Creates a new NFA fragment. Needs to be cleared using
1693 * automaton_fragment_clear.
1695 * @param start starting state
1696 * @param end end state
1698 * @return new NFA fragment
1700 static struct GNUNET_REGEX_Automaton *
1701 nfa_fragment_create (struct GNUNET_REGEX_State *start,
1702 struct GNUNET_REGEX_State *end)
1704 struct GNUNET_REGEX_Automaton *n;
1706 n = GNUNET_malloc (sizeof (struct GNUNET_REGEX_Automaton));
1713 if (NULL == start || NULL == end)
1716 automaton_add_state (n, end);
1717 automaton_add_state (n, start);
1729 * Adds a list of states to the given automaton 'n'.
1731 * @param n automaton to which the states should be added
1732 * @param states_head head of the DLL of states
1733 * @param states_tail tail of the DLL of states
1736 nfa_add_states (struct GNUNET_REGEX_Automaton *n,
1737 struct GNUNET_REGEX_State *states_head,
1738 struct GNUNET_REGEX_State *states_tail)
1740 struct GNUNET_REGEX_State *s;
1742 if (NULL == n || NULL == states_head)
1744 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not add states\n");
1748 if (NULL == n->states_head)
1750 n->states_head = states_head;
1751 n->states_tail = states_tail;
1755 if (NULL != states_head)
1757 n->states_tail->next = states_head;
1758 n->states_tail = states_tail;
1761 for (s = states_head; NULL != s; s = s->next)
1767 * Creates a new NFA state. Needs to be freed using automaton_destroy_state.
1769 * @param ctx context
1770 * @param accepting is it an accepting state or not
1772 * @return new NFA state
1774 static struct GNUNET_REGEX_State *
1775 nfa_state_create (struct GNUNET_REGEX_Context *ctx, int accepting)
1777 struct GNUNET_REGEX_State *s;
1779 s = GNUNET_malloc (sizeof (struct GNUNET_REGEX_State));
1780 s->id = ctx->state_id++;
1781 s->accepting = accepting;
1782 s->marked = GNUNET_NO;
1788 GNUNET_asprintf (&s->name, "s%i", s->id);
1795 * Calculates the NFA closure set for the given state.
1797 * @param nfa the NFA containing 's'
1798 * @param s starting point state
1799 * @param label transitioning label on which to base the closure on,
1800 * pass NULL for epsilon transition
1802 * @return sorted nfa closure on 'label' (epsilon closure if 'label' is NULL)
1804 static struct GNUNET_REGEX_StateSet *
1805 nfa_closure_create (struct GNUNET_REGEX_Automaton *nfa,
1806 struct GNUNET_REGEX_State *s, const char *label)
1808 struct GNUNET_REGEX_StateSet *cls;
1809 struct GNUNET_REGEX_StateSet *cls_check;
1810 struct GNUNET_REGEX_State *clsstate;
1811 struct GNUNET_REGEX_State *currentstate;
1812 struct GNUNET_REGEX_Transition *ctran;
1817 cls = GNUNET_malloc (sizeof (struct GNUNET_REGEX_StateSet));
1818 cls_check = GNUNET_malloc (sizeof (struct GNUNET_REGEX_StateSet));
1820 for (clsstate = nfa->states_head; NULL != clsstate; clsstate = clsstate->next)
1821 clsstate->contained = 0;
1823 // Add start state to closure only for epsilon closure
1825 GNUNET_array_append (cls->states, cls->len, s);
1827 GNUNET_array_append (cls_check->states, cls_check->len, s);
1828 while (cls_check->len > 0)
1830 currentstate = cls_check->states[cls_check->len - 1];
1831 GNUNET_array_grow (cls_check->states, cls_check->len, cls_check->len - 1);
1833 for (ctran = currentstate->transitions_head; NULL != ctran;
1834 ctran = ctran->next)
1836 if (NULL != ctran->to_state && 0 == nullstrcmp (label, ctran->label))
1838 clsstate = ctran->to_state;
1840 if (NULL != clsstate && 0 == clsstate->contained)
1842 GNUNET_array_append (cls->states, cls->len, clsstate);
1843 GNUNET_array_append (cls_check->states, cls_check->len, clsstate);
1844 clsstate->contained = 1;
1849 GNUNET_assert (0 == cls_check->len);
1850 GNUNET_free (cls_check);
1854 qsort (cls->states, cls->len, sizeof (struct GNUNET_REGEX_State *),
1862 * Calculates the closure set for the given set of states.
1864 * @param nfa the NFA containing 's'
1865 * @param states list of states on which to base the closure on
1866 * @param label transitioning label for which to base the closure on,
1867 * pass NULL for epsilon transition
1869 * @return sorted nfa closure on 'label' (epsilon closure if 'label' is NULL)
1871 static struct GNUNET_REGEX_StateSet *
1872 nfa_closure_set_create (struct GNUNET_REGEX_Automaton *nfa,
1873 struct GNUNET_REGEX_StateSet *states, const char *label)
1875 struct GNUNET_REGEX_State *s;
1876 struct GNUNET_REGEX_StateSet *sset;
1877 struct GNUNET_REGEX_StateSet *cls;
1881 unsigned int contains;
1886 cls = GNUNET_malloc (sizeof (struct GNUNET_REGEX_StateSet));
1888 for (i = 0; i < states->len; i++)
1890 s = states->states[i];
1891 sset = nfa_closure_create (nfa, s, label);
1893 for (j = 0; j < sset->len; j++)
1896 for (k = 0; k < cls->len; k++)
1898 if (sset->states[j]->id == cls->states[k]->id)
1905 GNUNET_array_append (cls->states, cls->len, sset->states[j]);
1907 state_set_clear (sset);
1911 qsort (cls->states, cls->len, sizeof (struct GNUNET_REGEX_State *),
1919 * Pops two NFA fragments (a, b) from the stack and concatenates them (ab)
1921 * @param ctx context
1924 nfa_add_concatenation (struct GNUNET_REGEX_Context *ctx)
1926 struct GNUNET_REGEX_Automaton *a;
1927 struct GNUNET_REGEX_Automaton *b;
1928 struct GNUNET_REGEX_Automaton *new_nfa;
1930 b = ctx->stack_tail;
1931 GNUNET_assert (NULL != b);
1932 GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, b);
1933 a = ctx->stack_tail;
1934 GNUNET_assert (NULL != a);
1935 GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a);
1937 state_add_transition (ctx, a->end, NULL, b->start);
1938 a->end->accepting = 0;
1939 b->end->accepting = 1;
1941 new_nfa = nfa_fragment_create (NULL, NULL);
1942 nfa_add_states (new_nfa, a->states_head, a->states_tail);
1943 nfa_add_states (new_nfa, b->states_head, b->states_tail);
1944 new_nfa->start = a->start;
1945 new_nfa->end = b->end;
1946 new_nfa->state_count += a->state_count + b->state_count;
1947 automaton_fragment_clear (a);
1948 automaton_fragment_clear (b);
1950 GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new_nfa);
1955 * Pops a NFA fragment from the stack (a) and adds a new fragment (a*)
1957 * @param ctx context
1960 nfa_add_star_op (struct GNUNET_REGEX_Context *ctx)
1962 struct GNUNET_REGEX_Automaton *a;
1963 struct GNUNET_REGEX_Automaton *new_nfa;
1964 struct GNUNET_REGEX_State *start;
1965 struct GNUNET_REGEX_State *end;
1967 a = ctx->stack_tail;
1971 GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
1972 "nfa_add_star_op failed, because there was no element on the stack");
1976 GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a);
1978 start = nfa_state_create (ctx, 0);
1979 end = nfa_state_create (ctx, 1);
1981 state_add_transition (ctx, start, NULL, a->start);
1982 state_add_transition (ctx, start, NULL, end);
1983 state_add_transition (ctx, a->end, NULL, a->start);
1984 state_add_transition (ctx, a->end, NULL, end);
1986 a->end->accepting = 0;
1989 new_nfa = nfa_fragment_create (start, end);
1990 nfa_add_states (new_nfa, a->states_head, a->states_tail);
1991 automaton_fragment_clear (a);
1993 GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new_nfa);
1998 * Pops an NFA fragment (a) from the stack and adds a new fragment (a+)
2000 * @param ctx context
2003 nfa_add_plus_op (struct GNUNET_REGEX_Context *ctx)
2005 struct GNUNET_REGEX_Automaton *a;
2007 a = ctx->stack_tail;
2008 GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a);
2010 state_add_transition (ctx, a->end, NULL, a->start);
2012 GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, a);
2017 * Pops an NFA fragment (a) from the stack and adds a new fragment (a?)
2019 * @param ctx context
2022 nfa_add_question_op (struct GNUNET_REGEX_Context *ctx)
2024 struct GNUNET_REGEX_Automaton *a;
2025 struct GNUNET_REGEX_Automaton *new_nfa;
2026 struct GNUNET_REGEX_State *start;
2027 struct GNUNET_REGEX_State *end;
2029 a = ctx->stack_tail;
2033 GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
2034 "nfa_add_question_op failed, because there was no element on the stack");
2038 GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a);
2040 start = nfa_state_create (ctx, 0);
2041 end = nfa_state_create (ctx, 1);
2043 state_add_transition (ctx, start, NULL, a->start);
2044 state_add_transition (ctx, start, NULL, end);
2045 state_add_transition (ctx, a->end, NULL, end);
2047 a->end->accepting = 0;
2049 new_nfa = nfa_fragment_create (start, end);
2050 nfa_add_states (new_nfa, a->states_head, a->states_tail);
2051 automaton_fragment_clear (a);
2053 GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new_nfa);
2058 * Pops two NFA fragments (a, b) from the stack and adds a new NFA fragment that
2059 * alternates between a and b (a|b)
2061 * @param ctx context
2064 nfa_add_alternation (struct GNUNET_REGEX_Context *ctx)
2066 struct GNUNET_REGEX_Automaton *a;
2067 struct GNUNET_REGEX_Automaton *b;
2068 struct GNUNET_REGEX_Automaton *new_nfa;
2069 struct GNUNET_REGEX_State *start;
2070 struct GNUNET_REGEX_State *end;
2072 b = ctx->stack_tail;
2073 GNUNET_assert (NULL != b);
2074 GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, b);
2075 a = ctx->stack_tail;
2076 GNUNET_assert (NULL != a);
2077 GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a);
2079 start = nfa_state_create (ctx, 0);
2080 end = nfa_state_create (ctx, 1);
2081 state_add_transition (ctx, start, NULL, a->start);
2082 state_add_transition (ctx, start, NULL, b->start);
2084 state_add_transition (ctx, a->end, NULL, end);
2085 state_add_transition (ctx, b->end, NULL, end);
2087 a->end->accepting = 0;
2088 b->end->accepting = 0;
2091 new_nfa = nfa_fragment_create (start, end);
2092 nfa_add_states (new_nfa, a->states_head, a->states_tail);
2093 nfa_add_states (new_nfa, b->states_head, b->states_tail);
2094 automaton_fragment_clear (a);
2095 automaton_fragment_clear (b);
2097 GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new_nfa);
2102 * Adds a new nfa fragment to the stack
2104 * @param ctx context
2105 * @param label label for nfa transition
2108 nfa_add_label (struct GNUNET_REGEX_Context *ctx, const char *label)
2110 struct GNUNET_REGEX_Automaton *n;
2111 struct GNUNET_REGEX_State *start;
2112 struct GNUNET_REGEX_State *end;
2114 GNUNET_assert (NULL != ctx);
2116 start = nfa_state_create (ctx, 0);
2117 end = nfa_state_create (ctx, 1);
2118 state_add_transition (ctx, start, label, end);
2119 n = nfa_fragment_create (start, end);
2120 GNUNET_assert (NULL != n);
2121 GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, n);
2126 * Initialize a new context
2128 * @param ctx context
2131 GNUNET_REGEX_context_init (struct GNUNET_REGEX_Context *ctx)
2135 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Context was NULL!");
2139 ctx->transition_id = 0;
2140 ctx->stack_head = NULL;
2141 ctx->stack_tail = NULL;
2146 * Construct an NFA by parsing the regex string of length 'len'.
2148 * @param regex regular expression string
2149 * @param len length of the string
2151 * @return NFA, needs to be freed using GNUNET_REGEX_destroy_automaton
2153 struct GNUNET_REGEX_Automaton *
2154 GNUNET_REGEX_construct_nfa (const char *regex, const size_t len)
2156 struct GNUNET_REGEX_Context ctx;
2157 struct GNUNET_REGEX_Automaton *nfa;
2162 unsigned int altcount;
2163 unsigned int atomcount;
2164 unsigned int pcount;
2171 if (NULL == regex || 0 == strlen (regex) || 0 == len)
2173 GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
2174 "Could not parse regex. Empty regex string provided.\n");
2179 GNUNET_REGEX_context_init (&ctx);
2189 for (count = 0; count < len && *regexp; count++, regexp++)
2197 nfa_add_concatenation (&ctx);
2199 GNUNET_array_grow (p, pcount, pcount + 1);
2200 p[pcount - 1].altcount = altcount;
2201 p[pcount - 1].atomcount = atomcount;
2208 error_msg = "Cannot append '|' to nothing";
2211 while (--atomcount > 0)
2212 nfa_add_concatenation (&ctx);
2218 error_msg = "Missing opening '('";
2223 // Ignore this: "()"
2225 altcount = p[pcount].altcount;
2226 atomcount = p[pcount].atomcount;
2229 while (--atomcount > 0)
2230 nfa_add_concatenation (&ctx);
2231 for (; altcount > 0; altcount--)
2232 nfa_add_alternation (&ctx);
2234 altcount = p[pcount].altcount;
2235 atomcount = p[pcount].atomcount;
2241 error_msg = "Cannot append '*' to nothing";
2244 nfa_add_star_op (&ctx);
2249 error_msg = "Cannot append '+' to nothing";
2252 nfa_add_plus_op (&ctx);
2257 error_msg = "Cannot append '?' to nothing";
2260 nfa_add_question_op (&ctx);
2266 nfa_add_concatenation (&ctx);
2268 curlabel[0] = *regexp;
2269 nfa_add_label (&ctx, curlabel);
2276 error_msg = "Unbalanced parenthesis";
2279 while (--atomcount > 0)
2280 nfa_add_concatenation (&ctx);
2281 for (; altcount > 0; altcount--)
2282 nfa_add_alternation (&ctx);
2284 GNUNET_free_non_null (p);
2286 nfa = ctx.stack_tail;
2287 GNUNET_CONTAINER_DLL_remove (ctx.stack_head, ctx.stack_tail, nfa);
2289 if (NULL != ctx.stack_head)
2291 error_msg = "Creating the NFA failed. NFA stack was not empty!";
2295 /* Remember the regex that was used to generate this NFA */
2296 nfa->regex = GNUNET_strdup (regex);
2298 /* create depth-first numbering of the states for pretty printing */
2299 GNUNET_REGEX_automaton_traverse (nfa, NULL, NULL, NULL, &number_states, NULL);
2301 /* No multistriding added so far */
2302 nfa->is_multistrided = GNUNET_NO;
2307 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not parse regex: %s\n", regex);
2308 if (NULL != error_msg)
2309 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "%s\n", error_msg);
2311 GNUNET_free_non_null (p);
2313 while (NULL != (nfa = ctx.stack_head))
2315 GNUNET_CONTAINER_DLL_remove (ctx.stack_head, ctx.stack_tail, nfa);
2316 GNUNET_REGEX_automaton_destroy (nfa);
2324 * Create DFA states based on given 'nfa' and starting with 'dfa_state'.
2326 * @param ctx context.
2327 * @param nfa NFA automaton.
2328 * @param dfa DFA automaton.
2329 * @param dfa_state current dfa state, pass epsilon closure of first nfa state
2333 construct_dfa_states (struct GNUNET_REGEX_Context *ctx,
2334 struct GNUNET_REGEX_Automaton *nfa,
2335 struct GNUNET_REGEX_Automaton *dfa,
2336 struct GNUNET_REGEX_State *dfa_state)
2338 struct GNUNET_REGEX_Transition *ctran;
2339 struct GNUNET_REGEX_State *state_iter;
2340 struct GNUNET_REGEX_State *new_dfa_state;
2341 struct GNUNET_REGEX_State *state_contains;
2342 struct GNUNET_REGEX_StateSet *tmp;
2343 struct GNUNET_REGEX_StateSet *nfa_set;
2345 for (ctran = dfa_state->transitions_head; NULL != ctran; ctran = ctran->next)
2347 if (NULL == ctran->label || NULL != ctran->to_state)
2350 tmp = nfa_closure_set_create (nfa, dfa_state->nfa_set, ctran->label);
2351 nfa_set = nfa_closure_set_create (nfa, tmp, 0);
2352 state_set_clear (tmp);
2353 new_dfa_state = dfa_state_create (ctx, nfa_set);
2354 state_contains = NULL;
2355 for (state_iter = dfa->states_head; NULL != state_iter;
2356 state_iter = state_iter->next)
2358 if (0 == state_set_compare (state_iter->nfa_set, new_dfa_state->nfa_set))
2359 state_contains = state_iter;
2362 if (NULL == state_contains)
2364 automaton_add_state (dfa, new_dfa_state);
2365 ctran->to_state = new_dfa_state;
2366 construct_dfa_states (ctx, nfa, dfa, new_dfa_state);
2370 ctran->to_state = state_contains;
2371 automaton_destroy_state (new_dfa_state);
2378 * Construct DFA for the given 'regex' of length 'len'
2380 * @param regex regular expression string
2381 * @param len length of the regular expression
2383 * @return DFA, needs to be freed using GNUNET_REGEX_destroy_automaton
2385 struct GNUNET_REGEX_Automaton *
2386 GNUNET_REGEX_construct_dfa (const char *regex, const size_t len)
2388 struct GNUNET_REGEX_Context ctx;
2389 struct GNUNET_REGEX_Automaton *dfa;
2390 struct GNUNET_REGEX_Automaton *nfa;
2391 struct GNUNET_REGEX_StateSet *nfa_start_eps_cls;
2393 GNUNET_REGEX_context_init (&ctx);
2396 nfa = GNUNET_REGEX_construct_nfa (regex, len);
2400 GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
2401 "Could not create DFA, because NFA creation failed\n");
2405 dfa = GNUNET_malloc (sizeof (struct GNUNET_REGEX_Automaton));
2407 dfa->state_count = 0;
2408 dfa->states_head = NULL;
2409 dfa->states_tail = NULL;
2410 dfa->regex = GNUNET_strdup (regex);
2411 dfa->is_multistrided = GNUNET_NO;
2413 // Create DFA start state from epsilon closure
2414 nfa_start_eps_cls = nfa_closure_create (nfa, nfa->start, 0);
2415 dfa->start = dfa_state_create (&ctx, nfa_start_eps_cls);
2416 automaton_add_state (dfa, dfa->start);
2418 construct_dfa_states (&ctx, nfa, dfa, dfa->start);
2420 GNUNET_REGEX_automaton_destroy (nfa);
2423 dfa_minimize (&ctx, dfa);
2425 // Create proofs for all states
2426 automaton_create_proofs (dfa);
2428 // Add strides to DFA
2429 // GNUNET_REGEX_add_multi_strides_to_dfa (&ctx, dfa, 2);
2436 * Free the memory allocated by constructing the GNUNET_REGEX_Automaton data
2439 * @param a automaton to be destroyed
2442 GNUNET_REGEX_automaton_destroy (struct GNUNET_REGEX_Automaton *a)
2444 struct GNUNET_REGEX_State *s;
2445 struct GNUNET_REGEX_State *next_state;
2450 GNUNET_free_non_null (a->regex);
2451 GNUNET_free_non_null (a->canonical_regex);
2453 for (s = a->states_head; NULL != s;)
2455 next_state = s->next;
2456 automaton_destroy_state (s);
2465 * Evaluates the given string using the given DFA automaton
2467 * @param a automaton, type must be DFA
2468 * @param string string that should be evaluated
2470 * @return 0 if string matches, non 0 otherwise
2473 evaluate_dfa (struct GNUNET_REGEX_Automaton *a, const char *string)
2476 struct GNUNET_REGEX_State *s;
2477 unsigned int step_len;
2481 GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
2482 "Tried to evaluate DFA, but NFA automaton given");
2488 // If the string is empty but the starting state is accepting, we accept.
2489 if ((NULL == string || 0 == strlen (string)) && s->accepting)
2492 for (strp = string; NULL != strp && *strp; strp += step_len)
2494 step_len = dfa_move (&s, strp);
2500 if (NULL != s && s->accepting)
2508 * Evaluates the given string using the given NFA automaton
2510 * @param a automaton, type must be NFA
2511 * @param string string that should be evaluated
2513 * @return 0 if string matches, non 0 otherwise
2516 evaluate_nfa (struct GNUNET_REGEX_Automaton *a, const char *string)
2520 struct GNUNET_REGEX_State *s;
2521 struct GNUNET_REGEX_StateSet *sset;
2522 struct GNUNET_REGEX_StateSet *new_sset;
2528 GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
2529 "Tried to evaluate NFA, but DFA automaton given");
2533 // If the string is empty but the starting state is accepting, we accept.
2534 if ((NULL == string || 0 == strlen (string)) && a->start->accepting)
2538 sset = nfa_closure_create (a, a->start, 0);
2541 for (strp = string; NULL != strp && *strp; strp++)
2544 new_sset = nfa_closure_set_create (a, sset, str);
2545 state_set_clear (sset);
2546 sset = nfa_closure_set_create (a, new_sset, 0);
2547 state_set_clear (new_sset);
2550 for (i = 0; i < sset->len; i++)
2552 s = sset->states[i];
2553 if (NULL != s && s->accepting)
2560 state_set_clear (sset);
2566 * Evaluates the given 'string' against the given compiled regex
2568 * @param a automaton
2569 * @param string string to check
2571 * @return 0 if string matches, non 0 otherwise
2574 GNUNET_REGEX_eval (struct GNUNET_REGEX_Automaton *a, const char *string)
2581 result = evaluate_dfa (a, string);
2584 result = evaluate_nfa (a, string);
2587 GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
2588 "Evaluating regex failed, automaton has no type!\n");
2589 result = GNUNET_SYSERR;
2598 * Get the canonical regex of the given automaton.
2599 * When constructing the automaton a proof is computed for each state,
2600 * consisting of the regular expression leading to this state. A complete
2601 * regex for the automaton can be computed by combining these proofs.
2602 * As of now this function is only useful for testing.
2604 * @param a automaton for which the canonical regex should be returned.
2609 GNUNET_REGEX_get_canonical_regex (struct GNUNET_REGEX_Automaton *a)
2614 return a->canonical_regex;
2619 * Get the number of transitions that are contained in the given automaton.
2621 * @param a automaton for which the number of transitions should be returned.
2623 * @return number of transitions in the given automaton.
2626 GNUNET_REGEX_get_transition_count (struct GNUNET_REGEX_Automaton *a)
2628 unsigned int t_count;
2629 struct GNUNET_REGEX_State *s;
2634 for (t_count = 0, s = a->states_head; NULL != s; s = s->next)
2636 t_count += s->transition_count;
2644 * Get the first key for the given 'input_string'. This hashes the first x bits
2645 * of the 'input_string'.
2647 * @param input_string string.
2648 * @param string_len length of the 'input_string'.
2649 * @param key pointer to where to write the hash code.
2651 * @return number of bits of 'input_string' that have been consumed
2652 * to construct the key
2655 GNUNET_REGEX_get_first_key (const char *input_string, size_t string_len,
2656 struct GNUNET_HashCode * key)
2660 size = string_len < INITIAL_BITS ? string_len : INITIAL_BITS;
2662 if (NULL == input_string)
2664 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Given input string was NULL!\n");
2668 GNUNET_CRYPTO_hash (input_string, size, key);
2675 * Check if the given 'proof' matches the given 'key'.
2677 * @param proof partial regex of a state.
2678 * @param key hash of a state.
2680 * @return GNUNET_OK if the proof is valid for the given key.
2683 GNUNET_REGEX_check_proof (const char *proof, const struct GNUNET_HashCode *key)
2685 struct GNUNET_HashCode key_check;
2687 if (NULL == proof || NULL == key)
2689 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Proof check failed, was NULL.\n");
2693 GNUNET_CRYPTO_hash (proof, strlen (proof), &key_check);
2695 GNUNET_CRYPTO_hash_cmp (key, &key_check)) ? GNUNET_OK : GNUNET_NO;
2700 * Recursive helper function for iterate_initial_edges. Will call iterator
2701 * function for each initial state.
2703 * @param min_len minimum length of the path in the graph.
2704 * @param max_len maximum length of the path in the graph.
2705 * @param cur_len current length of the path already traversed.
2706 * @param consumed_string string consumed by traversing the graph till this state.
2707 * @param state current state of the automaton.
2708 * @param iterator iterator function called for each edge.
2709 * @param iterator_cls closure for the iterator function.
2712 iterate_initial_edge (const unsigned int min_len, const unsigned int max_len,
2713 unsigned int cur_len, char *consumed_string,
2714 struct GNUNET_REGEX_State *state,
2715 GNUNET_REGEX_KeyIterator iterator, void *iterator_cls)
2719 struct GNUNET_REGEX_Transition *t;
2720 unsigned int num_edges = state->transition_count;
2721 struct GNUNET_REGEX_Edge edges[num_edges];
2722 struct GNUNET_HashCode hash;
2724 if (cur_len > min_len && NULL != consumed_string && cur_len <= max_len)
2726 for (i = 0, t = state->transitions_head; NULL != t; t = t->next, i++)
2728 edges[i].label = t->label;
2729 edges[i].destination = t->to_state->hash;
2732 GNUNET_CRYPTO_hash (consumed_string, strlen (consumed_string), &hash);
2733 iterator (iterator_cls, &hash, consumed_string, state->accepting, num_edges,
2737 if (cur_len < max_len)
2740 for (t = state->transitions_head; NULL != t; t = t->next)
2742 if (NULL != consumed_string)
2743 GNUNET_asprintf (&temp, "%s%s", consumed_string, t->label);
2745 GNUNET_asprintf (&temp, "%s", t->label);
2747 iterate_initial_edge (min_len, max_len, cur_len, temp, t->to_state,
2748 iterator, iterator_cls);
2756 * Iterate over all initial edges that aren't actually part of the automaton.
2757 * This is needed to find the initial states returned by
2758 * GNUNET_REGEX_get_first_key. Iteration will start at the first state that has
2759 * more than one outgoing edge, i.e. the state that branches the graph.
2760 * For example consider the following graph:
2761 * a -> b -> c -> d -> ...
2764 * This function will not iterate over the edges leading to "c", because these
2765 * will be covered by the iterate_edges function.
2767 * @param a the automaton for which the initial states should be computed.
2768 * @param initial_len length of the initial state string.
2769 * @param iterator iterator function called for each edge.
2770 * @param iterator_cls closure for the iterator function.
2773 iterate_initial_edges (struct GNUNET_REGEX_Automaton *a,
2774 const unsigned int initial_len,
2775 GNUNET_REGEX_KeyIterator iterator, void *iterator_cls)
2777 char *consumed_string;
2779 struct GNUNET_REGEX_State *s;
2780 unsigned int cur_len;
2782 if (1 > initial_len)
2785 consumed_string = NULL;
2789 if (1 == s->transition_count)
2793 if (NULL != consumed_string)
2795 temp = consumed_string;
2796 GNUNET_asprintf (&consumed_string, "%s%s", consumed_string,
2797 s->transitions_head->label);
2801 GNUNET_asprintf (&consumed_string, "%s", s->transitions_head->label);
2803 s = s->transitions_head->to_state;
2804 cur_len += strlen (s->transitions_head->label);
2806 while (cur_len < initial_len && 1 == s->transition_count);
2809 iterate_initial_edge (cur_len, initial_len, cur_len, consumed_string, s,
2810 iterator, iterator_cls);
2812 GNUNET_free_non_null (consumed_string);
2817 * Iterate over all edges helper function starting from state 's', calling
2818 * iterator function for each edge.
2821 * @param iterator iterator function called for each edge.
2822 * @param iterator_cls closure.
2825 iterate_edge (struct GNUNET_REGEX_State *s, GNUNET_REGEX_KeyIterator iterator,
2828 struct GNUNET_REGEX_Transition *t;
2829 struct GNUNET_REGEX_Edge edges[s->transition_count];
2830 unsigned int num_edges;
2832 if (GNUNET_YES != s->marked)
2834 s->marked = GNUNET_YES;
2836 num_edges = state_get_edges (s, edges);
2838 if ((NULL != s->proof && 0 < strlen (s->proof)) || s->accepting)
2839 iterator (iterator_cls, &s->hash, s->proof, s->accepting, num_edges,
2842 for (t = s->transitions_head; NULL != t; t = t->next)
2843 iterate_edge (t->to_state, iterator, iterator_cls);
2849 * Iterate over all edges starting from start state of automaton 'a'. Calling
2850 * iterator for each edge.
2852 * @param a automaton.
2853 * @param iterator iterator called for each edge.
2854 * @param iterator_cls closure.
2857 GNUNET_REGEX_iterate_all_edges (struct GNUNET_REGEX_Automaton *a,
2858 GNUNET_REGEX_KeyIterator iterator,
2861 struct GNUNET_REGEX_State *s;
2863 for (s = a->states_head; NULL != s; s = s->next)
2864 s->marked = GNUNET_NO;
2866 iterate_initial_edges (a, INITIAL_BITS, iterator, iterator_cls);
2867 iterate_edge (a->start, iterator, iterator_cls);