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"
31 #define INITIAL_BITS 10
34 * Context that contains an id counter for states and transitions as well as a
35 * DLL of automatons used as a stack for NFA construction.
37 struct GNUNET_REGEX_Context
42 unsigned int state_id;
45 * Unique transition id.
47 unsigned int transition_id;
50 * Unique SCC (Strongly Connected Component) id.
55 * DLL of GNUNET_REGEX_Automaton's used as a stack.
57 struct GNUNET_REGEX_Automaton *stack_head;
60 * DLL of GNUNET_REGEX_Automaton's used as a stack.
62 struct GNUNET_REGEX_Automaton *stack_tail;
66 * Type of an automaton.
68 enum GNUNET_REGEX_AutomatonType
75 * Automaton representation.
77 struct GNUNET_REGEX_Automaton
80 * This is a linked list.
82 struct GNUNET_REGEX_Automaton *prev;
85 * This is a linked list.
87 struct GNUNET_REGEX_Automaton *next;
90 * First state of the automaton. This is mainly used for constructing an NFA,
91 * where each NFA itself consists of one or more NFAs linked together.
93 struct GNUNET_REGEX_State *start;
96 * End state of the automaton.
98 struct GNUNET_REGEX_State *end;
101 * Number of states in the automaton.
103 unsigned int state_count;
108 struct GNUNET_REGEX_State *states_head;
113 struct GNUNET_REGEX_State *states_tail;
116 * Type of the automaton.
118 enum GNUNET_REGEX_AutomatonType type;
126 * Computed regex (result of RX->NFA->DFA->RX)
128 char *computed_regex;
132 * A state. Can be used in DFA and NFA automatons.
134 struct GNUNET_REGEX_State
137 * This is a linked list.
139 struct GNUNET_REGEX_State *prev;
142 * This is a linked list.
144 struct GNUNET_REGEX_State *next;
152 * If this is an accepting state or not.
157 * Marking of the state. This is used for marking all visited states when
158 * traversing all states of an automaton and for cases where the state id
159 * cannot be used (dfa minimization).
164 * Marking the state as contained. This is used for checking, if the state is
165 * contained in a set in constant time
170 * Marking the state as part of an SCC (Strongly Connected Component). All
171 * states with the same scc_id are part of the same SCC. scc_id is 0, if state
172 * is not a part of any SCC.
177 * Used for SCC detection.
182 * Used for SCC detection.
187 * Human readable name of the automaton. Used for debugging and graph
195 struct GNUNET_HashCode hash;
198 * State ID for proof creation.
200 unsigned int proof_id;
203 * Proof for this state.
208 * Number of transitions from this state to other states.
210 unsigned int transition_count;
213 * DLL of transitions.
215 struct Transition *transitions_head;
218 * DLL of transitions.
220 struct Transition *transitions_tail;
223 * Set of states on which this state is based on. Used when creating a DFA out
224 * of several NFA states.
226 struct GNUNET_REGEX_StateSet *nfa_set;
230 * Transition between two states. Each state can have 0-n transitions. If label
231 * is 0, this is considered to be an epsilon transition.
236 * This is a linked list.
238 struct Transition *prev;
241 * This is a linked list.
243 struct Transition *next;
246 * Unique id of this transition.
251 * Label for this transition. This is basically the edge label for the graph.
256 * State to which this transition leads.
258 struct GNUNET_REGEX_State *to_state;
261 * State from which this transition origins.
263 struct GNUNET_REGEX_State *from_state;
266 * Mark this transition. For example when reversing the automaton.
274 struct GNUNET_REGEX_StateSet
279 struct GNUNET_REGEX_State **states;
282 * Length of the 'states' array.
288 * Debug helper functions
291 debug_print_transitions (struct GNUNET_REGEX_State *);
294 debug_print_state (struct GNUNET_REGEX_State *s)
298 if (NULL == s->proof)
303 GNUNET_log (GNUNET_ERROR_TYPE_DEBUG,
304 "State %i: %s marked: %i accepting: %i scc_id: %i transitions: %i proof: %s\n",
305 s->id, s->name, s->marked, s->accepting, s->scc_id,
306 s->transition_count, proof);
308 GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Transitions:\n");
309 debug_print_transitions (s);
313 debug_print_states (struct GNUNET_REGEX_Automaton *a)
315 struct GNUNET_REGEX_State *s;
317 for (s = a->states_head; NULL != s; s = s->next)
318 debug_print_state (s);
322 debug_print_transition (struct Transition *t)
336 if (NULL == t->to_state)
339 to_state = t->to_state->name;
341 if (NULL == t->from_state)
344 from_state = t->from_state->name;
346 GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Transition %i: From %s on %c to %s\n",
347 t->id, from_state, label, to_state);
351 debug_print_transitions (struct GNUNET_REGEX_State *s)
353 struct Transition *t;
355 for (t = s->transitions_head; NULL != t; t = t->next)
356 debug_print_transition (t);
360 * Recursive function doing DFS with 'v' as a start, detecting all SCCs inside
361 * the subgraph reachable from 'v'. Used with scc_tarjan function to detect all
362 * SCCs inside an automaton.
365 * @param v start vertex
366 * @param index current index
367 * @param stack stack for saving all SCCs
368 * @param stack_size current size of the stack
371 scc_tarjan_strongconnect (struct GNUNET_REGEX_Context *ctx,
372 struct GNUNET_REGEX_State *v, int *index,
373 struct GNUNET_REGEX_State **stack,
374 unsigned int *stack_size)
376 struct GNUNET_REGEX_State *w;
377 struct Transition *t;
382 stack[(*stack_size)++] = v;
385 for (t = v->transitions_head; NULL != t; t = t->next)
388 if (NULL != w && w->index < 0)
390 scc_tarjan_strongconnect (ctx, w, index, stack, stack_size);
391 v->lowlink = (v->lowlink > w->lowlink) ? w->lowlink : v->lowlink;
393 else if (0 != w->contained)
394 v->lowlink = (v->lowlink > w->index) ? w->index : v->lowlink;
397 if (v->lowlink == v->index)
399 w = stack[--(*stack_size)];
407 w->scc_id = ctx->scc_id;
408 w = stack[--(*stack_size)];
411 w->scc_id = ctx->scc_id;
417 * Detect all SCCs (Strongly Connected Components) inside the given automaton.
418 * SCCs will be marked using the scc_id on each state.
424 scc_tarjan (struct GNUNET_REGEX_Context *ctx, struct GNUNET_REGEX_Automaton *a)
427 struct GNUNET_REGEX_State *v;
428 struct GNUNET_REGEX_State *stack[a->state_count];
429 unsigned int stack_size;
431 for (v = a->states_head; NULL != v; v = v->next)
441 for (v = a->states_head; NULL != v; v = v->next)
444 scc_tarjan_strongconnect (ctx, v, &index, stack, &stack_size);
449 * Adds a transition from one state to another on 'label'. Does not add
453 * @param from_state starting state for the transition
454 * @param label transition label
455 * @param to_state state to where the transition should point to
458 state_add_transition (struct GNUNET_REGEX_Context *ctx,
459 struct GNUNET_REGEX_State *from_state, const char label,
460 struct GNUNET_REGEX_State *to_state)
463 struct Transition *t;
464 struct Transition *oth;
466 if (NULL == from_state)
468 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not create Transition.\n");
472 // Do not add duplicate state transitions
474 for (t = from_state->transitions_head; NULL != t; t = t->next)
476 if (t->to_state == to_state && t->label == label &&
477 t->from_state == from_state)
487 // sort transitions by label
488 for (oth = from_state->transitions_head; NULL != oth; oth = oth->next)
490 if (oth->label > label)
494 t = GNUNET_malloc (sizeof (struct Transition));
495 t->id = ctx->transition_id++;
497 t->to_state = to_state;
498 t->from_state = from_state;
500 // Add outgoing transition to 'from_state'
501 from_state->transition_count++;
502 GNUNET_CONTAINER_DLL_insert_before (from_state->transitions_head,
503 from_state->transitions_tail, oth, t);
507 * Compare two states. Used for sorting.
509 * @param a first state
510 * @param b second state
512 * @return an integer less than, equal to, or greater than zero
513 * if the first argument is considered to be respectively
514 * less than, equal to, or greater than the second.
517 state_compare (const void *a, const void *b)
519 struct GNUNET_REGEX_State **s1;
520 struct GNUNET_REGEX_State **s2;
522 s1 = (struct GNUNET_REGEX_State **) a;
523 s2 = (struct GNUNET_REGEX_State **) b;
525 return (*s1)->id - (*s2)->id;
529 * Get all edges leaving state 's'.
532 * @param edges all edges leaving 's'.
534 * @return number of edges.
537 state_get_edges (struct GNUNET_REGEX_State *s, struct GNUNET_REGEX_Edge *edges)
539 struct Transition *t;
547 for (t = s->transitions_head; NULL != t; t = t->next)
549 if (NULL != t->to_state)
551 edges[count].label = &t->label;
552 edges[count].destination = t->to_state->hash;
560 * Compare to state sets by comparing the id's of the states that are contained
561 * in each set. Both sets are expected to be sorted by id!
563 * @param sset1 first state set
564 * @param sset2 second state set
566 * @return an integer less than, equal to, or greater than zero
567 * if the first argument is considered to be respectively
568 * less than, equal to, or greater than the second.
571 state_set_compare (struct GNUNET_REGEX_StateSet *sset1,
572 struct GNUNET_REGEX_StateSet *sset2)
577 if (NULL == sset1 || NULL == sset2)
580 result = sset1->len - sset2->len;
582 for (i = 0; i < sset1->len; i++)
587 result = state_compare (&sset1->states[i], &sset2->states[i]);
593 * Clears the given StateSet 'set'
595 * @param set set to be cleared
598 state_set_clear (struct GNUNET_REGEX_StateSet *set)
602 GNUNET_free_non_null (set->states);
608 * Clears an automaton fragment. Does not destroy the states inside the
611 * @param a automaton to be cleared
614 automaton_fragment_clear (struct GNUNET_REGEX_Automaton *a)
621 a->states_head = NULL;
622 a->states_tail = NULL;
628 * Frees the memory used by State 's'
630 * @param s state that should be destroyed
633 automaton_destroy_state (struct GNUNET_REGEX_State *s)
635 struct Transition *t;
636 struct Transition *next_t;
641 GNUNET_free_non_null (s->name);
642 GNUNET_free_non_null (s->proof);
644 for (t = s->transitions_head; NULL != t; t = next_t)
647 GNUNET_CONTAINER_DLL_remove (s->transitions_head, s->transitions_tail, t);
651 state_set_clear (s->nfa_set);
657 * Remove a state from the given automaton 'a'. Always use this function when
658 * altering the states of an automaton. Will also remove all transitions leading
659 * to this state, before destroying it.
662 * @param s state to remove
665 automaton_remove_state (struct GNUNET_REGEX_Automaton *a,
666 struct GNUNET_REGEX_State *s)
668 struct GNUNET_REGEX_State *ss;
669 struct GNUNET_REGEX_State *s_check;
670 struct Transition *t_check;
672 if (NULL == a || NULL == s)
677 GNUNET_CONTAINER_DLL_remove (a->states_head, a->states_tail, s);
680 // remove all transitions leading to this state
681 for (s_check = a->states_head; NULL != s_check; s_check = s_check->next)
683 for (t_check = s_check->transitions_head; NULL != t_check;
684 t_check = t_check->next)
686 if (t_check->to_state == ss)
688 GNUNET_CONTAINER_DLL_remove (s_check->transitions_head,
689 s_check->transitions_tail, t_check);
690 s_check->transition_count--;
695 automaton_destroy_state (ss);
699 * Merge two states into one. Will merge 's1' and 's2' into 's1' and destroy
704 * @param s1 first state
705 * @param s2 second state, will be destroyed
708 automaton_merge_states (struct GNUNET_REGEX_Context *ctx,
709 struct GNUNET_REGEX_Automaton *a,
710 struct GNUNET_REGEX_State *s1,
711 struct GNUNET_REGEX_State *s2)
713 struct GNUNET_REGEX_State *s_check;
714 struct Transition *t_check;
717 GNUNET_assert (NULL != ctx && NULL != a && NULL != s1 && NULL != s2);
722 // 1. Make all transitions pointing to s2 point to s1
723 for (s_check = a->states_head; NULL != s_check; s_check = s_check->next)
725 for (t_check = s_check->transitions_head; NULL != t_check;
726 t_check = t_check->next)
728 if (s2 == t_check->to_state)
729 t_check->to_state = s1;
733 // 2. Add all transitions from s2 to sX to s1
734 for (t_check = s2->transitions_head; NULL != t_check; t_check = t_check->next)
736 if (t_check->to_state != s1)
737 state_add_transition (ctx, s1, t_check->label, t_check->to_state);
740 // 3. Rename s1 to {s1,s2}
742 GNUNET_asprintf (&s1->name, "{%s,%s}", new_name, s2->name);
743 GNUNET_free (new_name);
746 GNUNET_CONTAINER_DLL_remove (a->states_head, a->states_tail, s2);
748 automaton_destroy_state (s2);
752 * Add a state to the automaton 'a', always use this function to alter the
753 * states DLL of the automaton.
755 * @param a automaton to add the state to
756 * @param s state that should be added
759 automaton_add_state (struct GNUNET_REGEX_Automaton *a,
760 struct GNUNET_REGEX_State *s)
762 GNUNET_CONTAINER_DLL_insert (a->states_head, a->states_tail, s);
767 * Function that is called with each state, when traversing an automaton.
769 * @param cls closure.
770 * @param count current count of the state, from 0 to a->state_count -1.
773 typedef void (*GNUNET_REGEX_traverse_action) (void *cls, unsigned int count,
774 struct GNUNET_REGEX_State * s);
777 * Traverses all states that are reachable from state 's'. Expects the states to
778 * be unmarked (s->marked == GNUNET_NO). Performs 'action' on each visited
781 * @param cls closure.
782 * @param s start state.
783 * @param count current count of the state.
784 * @param action action to be performed on each state.
787 automaton_state_traverse (void *cls, struct GNUNET_REGEX_State *s,
789 GNUNET_REGEX_traverse_action action)
791 struct Transition *t;
793 if (GNUNET_NO == s->marked)
795 s->marked = GNUNET_YES;
798 action (cls, *count, s);
802 for (t = s->transitions_head; NULL != t; t = t->next)
803 automaton_state_traverse (cls, t->to_state, count, action);
808 * Traverses the given automaton from it's start state, visiting all reachable
809 * states and calling 'action' on each one of them.
811 * @param cls closure.
812 * @param a automaton.
813 * @param action action to be performed on each state.
816 automaton_traverse (void *cls, struct GNUNET_REGEX_Automaton *a,
817 GNUNET_REGEX_traverse_action action)
820 struct GNUNET_REGEX_State *s;
822 for (s = a->states_head; NULL != s; s = s->next)
823 s->marked = GNUNET_NO;
827 automaton_state_traverse (cls, a->start, &count, action);
832 * Check if the given string 'str' needs parentheses around it when
833 * using it to generate a regex.
835 * Currently only tests for first and last characters being '()' respectively.
836 * FIXME: What about "(ab)|(cd)"?
840 * @return GNUNET_YES if parentheses are needed, GNUNET_NO otherwise
843 needs_parentheses (const char *str)
847 if ( (NULL == str) ||
848 ((slen = strlen(str)) < 2) ||
849 ( ('(' == str[0]) && (')' == str[slen-1]) ) )
856 * Remove parentheses surrounding string 'str'.
857 * Example: "(a)" becomes "a".
858 * You need to GNUNET_free the returned string.
860 * Currently only tests for first and last characters being '()' respectively.
861 * FIXME: What about "(ab)|(cd)"?
863 * @param str string, free'd or re-used by this function, can be NULL
865 * @return string without surrounding parentheses, string 'str' if no preceding
866 * epsilon could be found, NULL if 'str' was NULL
869 remove_parentheses (char *str)
873 if ( (NULL == str) || ('(' != str[0]) || (str[(slen = strlen(str)) - 1] != ')') )
875 memmove (str, &str[1], slen - 2);
876 str[slen - 2] = '\0';
882 * Check if the string 'str' starts with an epsilon (empty string).
883 * Example: "(|a)" is starting with an epsilon.
885 * @param str string to test
887 * @return 0 if str has no epsilon, 1 if str starts with '(|' and ends with ')'
890 has_epsilon (const char *str)
892 return (NULL != str) && ('(' == str[0]) && ('|' == str[1]) && (')' == str[strlen(str) - 1]);
897 * Remove an epsilon from the string str. Where epsilon is an empty string
898 * Example: str = "(|a|b|c)", result: "a|b|c"
899 * The returned string needs to be freed.
903 * @return string without preceding epsilon, string 'str' if no preceding epsilon
904 * could be found, NULL if 'str' was NULL
907 remove_epsilon (const char *str)
913 if ( ('(' == str[0]) && ('|' == str[1]) )
916 if (')' == str[len-1])
917 return GNUNET_strndup (&str[2], len - 3);
919 return GNUNET_strdup (str);
923 * Compare 'str1', starting from position 'k', with whole 'str2'
925 * @param str1 first string to compare, starting from position 'k'
926 * @param str2 second string for comparison
927 * @param k starting position in 'str1'
929 * @return -1 if any of the strings is NULL, 0 if equal, non 0 otherwise
932 strkcmp (const char *str1, const char *str2, size_t k)
934 if (NULL == str1 || NULL == str2)
936 return strcmp (&str1[k], str2);
941 * Compare two strings for equality. If either is NULL (or if both are
942 * NULL), they are not equal.
944 * @return 0 if the strings are the same, 1 or -1 if not
947 nullstrcmp (const char *str1, const char *str2)
949 if ( (NULL == str1) || (NULL == str2) )
951 return strcmp (str1, str2);
956 number_states (void *cls, unsigned int count, struct GNUNET_REGEX_State *s)
958 struct GNUNET_REGEX_State **states;
966 * create proofs for all states in the given automaton. Implementation of the
967 * algorithm descriped in chapter 3.2.1 of "Automata Theory, Languages, and
968 * Computation 3rd Edition" by Hopcroft, Motwani and Ullman.
970 * @param a automaton.
973 automaton_create_proofs (struct GNUNET_REGEX_Automaton *a)
975 struct GNUNET_REGEX_State *states[a->state_count];
976 struct Transition *t;
977 char *R_last[a->state_count][a->state_count];
978 char *R_cur[a->state_count][a->state_count];
987 char *complete_regex;
1000 int clean_kk_kj_cmp;
1008 // number the states
1009 automaton_traverse (states, a, number_states);
1012 for (i = 0; i < n; i++)
1014 for (j = 0; j < n; j++)
1017 R_last[i][j] = NULL;
1018 for (t = states[i]->transitions_head; NULL != t; t = t->next)
1020 if (t->to_state == states[j])
1022 if (NULL == R_last[i][j])
1023 GNUNET_asprintf (&R_last[i][j], "%c", t->label);
1026 temp_a = R_last[i][j];
1027 GNUNET_asprintf (&R_last[i][j], "%s|%c", R_last[i][j], t->label);
1028 GNUNET_free (temp_a);
1035 if (NULL == R_last[i][j])
1036 GNUNET_asprintf (&R_last[i][j], "");
1039 temp_a = R_last[i][j];
1040 GNUNET_asprintf (&R_last[i][j], "(|%s)", R_last[i][j]);
1041 GNUNET_free (temp_a);
1044 else if (GNUNET_YES == needs_parentheses (R_last[i][j]))
1046 temp_a = R_last[i][j];
1047 GNUNET_asprintf (&R_last[i][j], "(%s)", R_last[i][j]);
1048 GNUNET_free (temp_a);
1055 for (k = 0; k < n; k++)
1057 for (i = 0; i < n; i++)
1059 for (j = 0; j < n; j++)
1061 /* GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, */
1062 /* ">>> R_last[i][j] = %s R_last[i][k] = %s " */
1063 /* "R_last[k][k] = %s R_last[k][j] = %s\n", R_last[i][j], */
1064 /* R_last[i][k], R_last[k][k], R_last[k][j]); */
1070 // cache results from strcmp, we might need these many times
1071 ij_kj_cmp = nullstrcmp (R_last[i][j], R_last[k][j]);
1072 ij_ik_cmp = nullstrcmp (R_last[i][j], R_last[i][k]);
1073 ik_kk_cmp = nullstrcmp (R_last[i][k], R_last[k][k]);
1074 ik_kj_cmp = nullstrcmp (R_last[i][k], R_last[k][j]);
1075 kk_kj_cmp = nullstrcmp (R_last[k][k], R_last[k][j]);
1077 // $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk})^* R^{(k-1)}_{kj}
1078 // With: R_cur[i][j] = R_cur_l | R_cur_r
1079 // Rij(k) = Rij(k-1), because right side (R_cur_r) is empty set (NULL)
1080 if ((NULL == R_last[i][k] || NULL == R_last[k][j] ||
1081 NULL == R_last[k][k]) && NULL != R_last[i][j])
1083 R_cur[i][j] = GNUNET_strdup (R_last[i][j]);
1085 // Everything is NULL, so Rij(k) = NULL
1086 else if ((NULL == R_last[i][k] || NULL == R_last[k][j] ||
1087 NULL == R_last[k][k]) && NULL == R_last[i][j])
1091 // Right side (R_cur_r) not NULL
1094 /* GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, */
1095 /* "R_temp_ij = %s R_temp_ik = %s R_temp_kk = %s R_temp_kj = %s\n", */
1096 /* R_temp_ij, R_temp_ik, R_temp_kk, R_temp_kj); */
1098 // Assign R_temp_(ik|kk|kj) to R_last[][] and remove epsilon as well
1099 // as parentheses, so we can better compare the contents
1100 R_temp_ik = remove_parentheses (remove_epsilon (R_last[i][k]));
1101 R_temp_kk = remove_parentheses (remove_epsilon (R_last[k][k]));
1102 R_temp_kj = remove_parentheses (remove_epsilon (R_last[k][j]));
1104 clean_ik_kk_cmp = nullstrcmp (R_last[i][k], R_temp_kk);
1105 clean_kk_kj_cmp = nullstrcmp (R_temp_kk, R_last[k][j]);
1107 // construct R_cur_l (and, if necessary R_cur_r)
1108 if (NULL != R_last[i][j])
1110 // Assign R_temp_ij to R_last[i][j] and remove epsilon as well
1111 // as parentheses, so we can better compare the contents
1112 R_temp_ij = remove_parentheses (remove_epsilon (R_last[i][j]));
1114 if (0 == strcmp (R_temp_ij, R_temp_ik) &&
1115 0 == strcmp (R_temp_ik, R_temp_kk) &&
1116 0 == strcmp (R_temp_kk, R_temp_kj))
1118 if (0 == strlen (R_temp_ij))
1120 R_cur_r = GNUNET_strdup ("");
1122 // a|(e|a)a*(e|a) = a*
1123 // a|(e|a)(e|a)*(e|a) = a*
1125 // (e|a)|aa*(e|a) = a*
1126 // (e|a)|(e|a)a*a = a*
1127 // (e|a)|(e|a)a*(e|a) = a*
1128 // (e|a)|(e|a)(e|a)*(e|a) = a*
1129 else if ((0 == strncmp (R_last[i][j], "(|", 2)) ||
1130 (0 == strncmp (R_last[i][k], "(|", 2) &&
1131 0 == strncmp (R_last[k][j], "(|", 2)))
1133 if (GNUNET_YES == needs_parentheses (R_temp_ij))
1134 GNUNET_asprintf (&R_cur_r, "(%s)*", R_temp_ij);
1136 GNUNET_asprintf (&R_cur_r, "%s*", R_temp_ij);
1141 // a|(e|a)(e|a)*a = a+
1142 // a|a(e|a)*(e|a) = a+
1145 if (GNUNET_YES == needs_parentheses (R_temp_ij))
1146 GNUNET_asprintf (&R_cur_r, "(%s)+", R_temp_ij);
1148 GNUNET_asprintf (&R_cur_r, "%s+", R_temp_ij);
1152 else if (0 == ij_ik_cmp && 0 == clean_kk_kj_cmp &&
1153 0 != clean_ik_kk_cmp)
1155 if (strlen (R_last[k][k]) < 1)
1156 R_cur_r = GNUNET_strdup (R_last[i][j]);
1157 else if (GNUNET_YES == needs_parentheses (R_temp_kk))
1158 GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last[i][j], R_temp_kk);
1160 GNUNET_asprintf (&R_cur_r, "%s%s*", R_last[i][j], R_last[k][k]);
1165 else if (0 == ij_kj_cmp && 0 == clean_ik_kk_cmp &&
1166 0 != clean_kk_kj_cmp)
1168 if (strlen (R_last[k][k]) < 1)
1169 R_cur_r = GNUNET_strdup (R_last[k][j]);
1170 else if (GNUNET_YES == needs_parentheses (R_temp_kk))
1171 GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk, R_last[k][j]);
1173 GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last[k][j]);
1177 // a|a(e|b)*(e|b) = a|ab* = a|a|ab|abb|abbb|... = ab*
1178 else if (0 == ij_ik_cmp && 0 == kk_kj_cmp &&
1179 !has_epsilon (R_last[i][j]) && has_epsilon (R_last[k][k]))
1181 if (needs_parentheses (R_temp_kk))
1182 GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last[i][j], R_temp_kk);
1184 GNUNET_asprintf (&R_cur_r, "%s%s*", R_last[i][j], R_temp_kk);
1188 // a|(e|b)(e|b)*a = a|b*a = a|a|ba|bba|bbba|... = b*a
1189 else if (0 == ij_kj_cmp && 0 == ik_kk_cmp &&
1190 !has_epsilon (R_last[i][j]) && has_epsilon (R_last[k][k]))
1192 if (needs_parentheses (R_temp_kk))
1193 GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk, R_last[i][j]);
1195 GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last[i][j]);
1201 /* GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "NO SIMPLIFICATION\n"); */
1202 temp_a = (NULL == R_last[i][j]) ? NULL : GNUNET_strdup (R_last[i][j]);
1203 temp_a = remove_parentheses (temp_a);
1207 GNUNET_free_non_null (R_temp_ij);
1209 // we have no left side
1215 // construct R_cur_r, if not already constructed
1216 if (NULL == R_cur_r)
1218 length = strlen (R_temp_kk) - strlen (R_last[i][k]);
1220 // a(ba)*bx = (ab)+x
1221 if (length > 0 && NULL != R_last[k][k] && 0 < strlen (R_last[k][k])
1222 && NULL != R_last[k][j] && 0 < strlen (R_last[k][j]) &&
1223 NULL != R_last[i][k] && 0 < strlen (R_last[i][k]) &&
1224 0 == strkcmp (R_temp_kk, R_last[i][k], length) &&
1225 0 == strncmp (R_temp_kk, R_last[k][j], length))
1227 temp_a = GNUNET_malloc (length + 1);
1228 temp_b = GNUNET_malloc ((strlen (R_last[k][j]) - length) + 1);
1233 for (cnt = 0; cnt < strlen (R_last[k][j]); cnt++)
1237 temp_a[length_l] = R_last[k][j][cnt];
1242 temp_b[length_r] = R_last[k][j][cnt];
1246 temp_a[length_l] = '\0';
1247 temp_b[length_r] = '\0';
1250 if (NULL != R_cur_l && 0 == strlen (R_cur_l) &&
1251 0 == strlen (temp_b))
1253 GNUNET_asprintf (&R_cur_r, "(%s%s)*", R_last[i][k], temp_a);
1254 GNUNET_free (R_cur_l);
1259 GNUNET_asprintf (&R_cur_r, "(%s%s)+%s", R_last[i][k], temp_a,
1262 GNUNET_free (temp_a);
1263 GNUNET_free (temp_b);
1265 else if (0 == strcmp (R_temp_ik, R_temp_kk) &&
1266 0 == strcmp (R_temp_kk, R_temp_kj))
1268 // (e|a)a*(e|a) = a*
1269 // (e|a)(e|a)*(e|a) = a*
1270 if (has_epsilon (R_last[i][k]) && has_epsilon (R_last[k][j]))
1272 if (needs_parentheses (R_temp_kk))
1273 GNUNET_asprintf (&R_cur_r, "(%s)*", R_temp_kk);
1275 GNUNET_asprintf (&R_cur_r, "%s*", R_temp_kk);
1278 else if (0 == clean_ik_kk_cmp && 0 == clean_kk_kj_cmp &&
1279 !has_epsilon (R_last[i][k]))
1281 if (needs_parentheses (R_temp_kk))
1282 GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_temp_kk, R_temp_kk);
1284 GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_temp_kk, R_temp_kk);
1288 // a(e|a)*(e|a) = a+
1293 (has_epsilon (R_last[i][k]) + has_epsilon (R_last[k][k]) +
1294 has_epsilon (R_last[k][j]));
1298 if (needs_parentheses (R_temp_kk))
1299 GNUNET_asprintf (&R_cur_r, "(%s)+", R_temp_kk);
1301 GNUNET_asprintf (&R_cur_r, "%s+", R_temp_kk);
1306 // (e|a)(e|a)*b = a*b
1307 else if (0 == strcmp (R_temp_ik, R_temp_kk))
1309 if (has_epsilon (R_last[i][k]))
1311 if (needs_parentheses (R_temp_kk))
1312 GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk,
1315 GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last[k][j]);
1319 if (needs_parentheses (R_temp_kk))
1320 GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_temp_kk,
1323 GNUNET_asprintf (&R_cur_r, "%s+%s", R_temp_kk, R_last[k][j]);
1327 // b(e|a)*(e|a) = ba*
1328 else if (0 == strcmp (R_temp_kk, R_temp_kj))
1330 if (has_epsilon (R_last[k][j]))
1332 if (needs_parentheses (R_temp_kk))
1333 GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last[i][k],
1336 GNUNET_asprintf (&R_cur_r, "%s%s*", R_last[i][k], R_temp_kk);
1340 if (needs_parentheses (R_temp_kk))
1341 GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_last[i][k],
1344 GNUNET_asprintf (&R_cur_r, "%s+%s", R_last[i][k], R_temp_kk);
1349 if (strlen (R_temp_kk) > 0)
1351 if (needs_parentheses (R_temp_kk))
1353 GNUNET_asprintf (&R_cur_r, "%s(%s)*%s", R_last[i][k],
1354 R_temp_kk, R_last[k][j]);
1358 GNUNET_asprintf (&R_cur_r, "%s%s*%s", R_last[i][k], R_temp_kk,
1364 GNUNET_asprintf (&R_cur_r, "%s%s", R_last[i][k], R_last[k][j]);
1369 /* GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "R_cur_l: %s\n", R_cur_l); */
1370 /* GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "R_cur_r: %s\n", R_cur_r); */
1372 // putting it all together
1373 if (NULL != R_cur_l && NULL != R_cur_r)
1376 if (0 == strcmp (R_cur_l, R_cur_r))
1378 R_cur[i][j] = GNUNET_strdup (R_cur_l);
1380 // R_cur_l | R_cur_r
1383 GNUNET_asprintf (&R_cur[i][j], "(%s|%s)", R_cur_l, R_cur_r);
1386 else if (NULL != R_cur_l)
1388 R_cur[i][j] = GNUNET_strdup (R_cur_l);
1390 else if (NULL != R_cur_r)
1392 R_cur[i][j] = GNUNET_strdup (R_cur_r);
1399 GNUNET_free_non_null (R_cur_l);
1400 GNUNET_free_non_null (R_cur_r);
1402 GNUNET_free_non_null (R_temp_ik);
1403 GNUNET_free_non_null (R_temp_kk);
1404 GNUNET_free_non_null (R_temp_kj);
1409 // set R_last = R_cur
1410 for (i = 0; i < n; i++)
1412 for (j = 0; j < n; j++)
1414 GNUNET_free_non_null (R_last[i][j]);
1415 R_last[i][j] = R_cur[i][j];
1421 // assign proofs and hashes
1422 for (i = 0; i < n; i++)
1424 if (NULL != R_last[a->start->proof_id][i])
1426 states[i]->proof = GNUNET_strdup (R_last[a->start->proof_id][i]);
1427 GNUNET_CRYPTO_hash (states[i]->proof, strlen (states[i]->proof),
1432 // complete regex for whole DFA: union of all pairs (start state/accepting state(s)).
1433 complete_regex = NULL;
1434 for (i = 0; i < n; i++)
1436 if (states[i]->accepting)
1438 if (NULL == complete_regex && 0 < strlen (R_last[a->start->proof_id][i]))
1439 GNUNET_asprintf (&complete_regex, "%s", R_last[a->start->proof_id][i]);
1440 else if (NULL != R_last[a->start->proof_id][i] &&
1441 0 < strlen (R_last[a->start->proof_id][i]))
1443 temp_a = complete_regex;
1444 GNUNET_asprintf (&complete_regex, "%s|%s", complete_regex,
1445 R_last[a->start->proof_id][i]);
1446 GNUNET_free (temp_a);
1450 a->computed_regex = complete_regex;
1452 GNUNET_log (GNUNET_ERROR_TYPE_DEBUG,
1453 "---------------------------------------------\n");
1454 GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Regex: %s\n", a->regex);
1455 GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Complete Regex: %s\n", complete_regex);
1456 GNUNET_log (GNUNET_ERROR_TYPE_DEBUG,
1457 "---------------------------------------------\n");
1460 for (i = 0; i < n; i++)
1462 for (j = 0; j < n; j++)
1463 GNUNET_free_non_null (R_last[i][j]);
1468 * Creates a new DFA state based on a set of NFA states. Needs to be freed using
1469 * automaton_destroy_state.
1471 * @param ctx context
1472 * @param nfa_states set of NFA states on which the DFA should be based on
1474 * @return new DFA state
1476 static struct GNUNET_REGEX_State *
1477 dfa_state_create (struct GNUNET_REGEX_Context *ctx,
1478 struct GNUNET_REGEX_StateSet *nfa_states)
1480 struct GNUNET_REGEX_State *s;
1483 struct GNUNET_REGEX_State *cstate;
1484 struct Transition *ctran;
1486 struct Transition *t;
1489 s = GNUNET_malloc (sizeof (struct GNUNET_REGEX_State));
1490 s->id = ctx->state_id++;
1500 if (NULL == nfa_states)
1502 GNUNET_asprintf (&s->name, "s%i", s->id);
1506 s->nfa_set = nfa_states;
1508 if (nfa_states->len < 1)
1511 // Create a name based on 'sset'
1512 s->name = GNUNET_malloc (sizeof (char) * 2);
1513 strcat (s->name, "{");
1516 for (i = 0; i < nfa_states->len; i++)
1518 cstate = nfa_states->states[i];
1519 GNUNET_asprintf (&name, "%i,", cstate->id);
1523 len = strlen (s->name) + strlen (name) + 1;
1524 s->name = GNUNET_realloc (s->name, len);
1525 strcat (s->name, name);
1530 // Add a transition for each distinct label to NULL state
1531 for (ctran = cstate->transitions_head; NULL != ctran; ctran = ctran->next)
1533 if (0 != ctran->label)
1537 for (t = s->transitions_head; NULL != t; t = t->next)
1539 if (t->label == ctran->label)
1547 state_add_transition (ctx, s, ctran->label, NULL);
1551 // If the nfa_states contain an accepting state, the new dfa state is also
1553 if (cstate->accepting)
1557 s->name[strlen (s->name) - 1] = '}';
1563 * Move from the given state 's' to the next state on transition 'label'
1565 * @param s starting state
1566 * @param label edge label to follow
1568 * @return new state or NULL, if transition on label not possible
1570 static struct GNUNET_REGEX_State *
1571 dfa_move (struct GNUNET_REGEX_State *s, const char label)
1573 struct Transition *t;
1574 struct GNUNET_REGEX_State *new_s;
1581 for (t = s->transitions_head; NULL != t; t = t->next)
1583 if (label == t->label)
1585 new_s = t->to_state;
1594 * Remove all unreachable states from DFA 'a'. Unreachable states are those
1595 * states that are not reachable from the starting state.
1597 * @param a DFA automaton
1600 dfa_remove_unreachable_states (struct GNUNET_REGEX_Automaton *a)
1602 struct GNUNET_REGEX_State *s;
1603 struct GNUNET_REGEX_State *s_next;
1605 // 1. unmark all states
1606 for (s = a->states_head; NULL != s; s = s->next)
1607 s->marked = GNUNET_NO;
1609 // 2. traverse dfa from start state and mark all visited states
1610 automaton_traverse (NULL, a, NULL);
1612 // 3. delete all states that were not visited
1613 for (s = a->states_head; NULL != s; s = s_next)
1616 if (GNUNET_NO == s->marked)
1617 automaton_remove_state (a, s);
1622 * Remove all dead states from the DFA 'a'. Dead states are those states that do
1623 * not transition to any other state but themselfes.
1625 * @param a DFA automaton
1628 dfa_remove_dead_states (struct GNUNET_REGEX_Automaton *a)
1630 struct GNUNET_REGEX_State *s;
1631 struct Transition *t;
1634 GNUNET_assert (DFA == a->type);
1636 for (s = a->states_head; NULL != s; s = s->next)
1642 for (t = s->transitions_head; NULL != t; t = t->next)
1644 if (NULL != t->to_state && t->to_state != s)
1654 // state s is dead, remove it
1655 automaton_remove_state (a, s);
1660 * Merge all non distinguishable states in the DFA 'a'
1662 * @param ctx context
1663 * @param a DFA automaton
1666 dfa_merge_nondistinguishable_states (struct GNUNET_REGEX_Context *ctx,
1667 struct GNUNET_REGEX_Automaton *a)
1670 int table[a->state_count][a->state_count];
1671 struct GNUNET_REGEX_State *s1;
1672 struct GNUNET_REGEX_State *s2;
1673 struct Transition *t1;
1674 struct Transition *t2;
1675 struct GNUNET_REGEX_State *s1_next;
1676 struct GNUNET_REGEX_State *s2_next;
1678 int num_equal_edges;
1680 for (i = 0, s1 = a->states_head; i < a->state_count && NULL != s1;
1686 // Mark all pairs of accepting/!accepting states
1687 for (s1 = a->states_head; NULL != s1; s1 = s1->next)
1689 for (s2 = a->states_head; NULL != s2; s2 = s2->next)
1691 table[s1->marked][s2->marked] = 0;
1693 if ((s1->accepting && !s2->accepting) ||
1694 (!s1->accepting && s2->accepting))
1696 table[s1->marked][s2->marked] = 1;
1701 // Find all equal states
1706 for (s1 = a->states_head; NULL != s1; s1 = s1->next)
1708 for (s2 = a->states_head; NULL != s2 && s1 != s2; s2 = s2->next)
1710 if (0 != table[s1->marked][s2->marked])
1713 num_equal_edges = 0;
1714 for (t1 = s1->transitions_head; NULL != t1; t1 = t1->next)
1716 for (t2 = s2->transitions_head; NULL != t2; t2 = t2->next)
1718 if (t1->label == t2->label)
1721 if (0 != table[t1->to_state->marked][t2->to_state->marked] ||
1722 0 != table[t2->to_state->marked][t1->to_state->marked])
1724 table[s1->marked][s2->marked] = t1->label != 0 ? t1->label : 1;
1730 if (num_equal_edges != s1->transition_count ||
1731 num_equal_edges != s2->transition_count)
1733 // Make sure ALL edges of possible equal states are the same
1734 table[s1->marked][s2->marked] = -2;
1740 // Merge states that are equal
1741 for (s1 = a->states_head; NULL != s1; s1 = s1_next)
1744 for (s2 = a->states_head; NULL != s2 && s1 != s2; s2 = s2_next)
1747 if (table[s1->marked][s2->marked] == 0)
1748 automaton_merge_states (ctx, a, s1, s2);
1754 * Minimize the given DFA 'a' by removing all unreachable states, removing all
1755 * dead states and merging all non distinguishable states
1757 * @param ctx context
1758 * @param a DFA automaton
1761 dfa_minimize (struct GNUNET_REGEX_Context *ctx,
1762 struct GNUNET_REGEX_Automaton *a)
1767 GNUNET_assert (DFA == a->type);
1769 // 1. remove unreachable states
1770 dfa_remove_unreachable_states (a);
1772 // 2. remove dead states
1773 dfa_remove_dead_states (a);
1775 // 3. Merge nondistinguishable states
1776 dfa_merge_nondistinguishable_states (ctx, a);
1780 * Creates a new NFA fragment. Needs to be cleared using
1781 * automaton_fragment_clear.
1783 * @param start starting state
1784 * @param end end state
1786 * @return new NFA fragment
1788 static struct GNUNET_REGEX_Automaton *
1789 nfa_fragment_create (struct GNUNET_REGEX_State *start,
1790 struct GNUNET_REGEX_State *end)
1792 struct GNUNET_REGEX_Automaton *n;
1794 n = GNUNET_malloc (sizeof (struct GNUNET_REGEX_Automaton));
1800 if (NULL == start && NULL == end)
1803 automaton_add_state (n, end);
1804 automaton_add_state (n, start);
1813 * Adds a list of states to the given automaton 'n'.
1815 * @param n automaton to which the states should be added
1816 * @param states_head head of the DLL of states
1817 * @param states_tail tail of the DLL of states
1820 nfa_add_states (struct GNUNET_REGEX_Automaton *n,
1821 struct GNUNET_REGEX_State *states_head,
1822 struct GNUNET_REGEX_State *states_tail)
1824 struct GNUNET_REGEX_State *s;
1826 if (NULL == n || NULL == states_head)
1828 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not add states\n");
1832 if (NULL == n->states_head)
1834 n->states_head = states_head;
1835 n->states_tail = states_tail;
1839 if (NULL != states_head)
1841 n->states_tail->next = states_head;
1842 n->states_tail = states_tail;
1845 for (s = states_head; NULL != s; s = s->next)
1850 * Creates a new NFA state. Needs to be freed using automaton_destroy_state.
1852 * @param ctx context
1853 * @param accepting is it an accepting state or not
1855 * @return new NFA state
1857 static struct GNUNET_REGEX_State *
1858 nfa_state_create (struct GNUNET_REGEX_Context *ctx, int accepting)
1860 struct GNUNET_REGEX_State *s;
1862 s = GNUNET_malloc (sizeof (struct GNUNET_REGEX_State));
1863 s->id = ctx->state_id++;
1864 s->accepting = accepting;
1871 GNUNET_asprintf (&s->name, "s%i", s->id);
1877 * Calculates the NFA closure set for the given state.
1879 * @param nfa the NFA containing 's'
1880 * @param s starting point state
1881 * @param label transitioning label on which to base the closure on,
1882 * pass 0 for epsilon transition
1884 * @return sorted nfa closure on 'label' (epsilon closure if 'label' is 0)
1886 static struct GNUNET_REGEX_StateSet *
1887 nfa_closure_create (struct GNUNET_REGEX_Automaton *nfa,
1888 struct GNUNET_REGEX_State *s, const char label)
1890 struct GNUNET_REGEX_StateSet *cls;
1891 struct GNUNET_REGEX_StateSet *cls_check;
1892 struct GNUNET_REGEX_State *clsstate;
1893 struct GNUNET_REGEX_State *currentstate;
1894 struct Transition *ctran;
1899 cls = GNUNET_malloc (sizeof (struct GNUNET_REGEX_StateSet));
1900 cls_check = GNUNET_malloc (sizeof (struct GNUNET_REGEX_StateSet));
1902 for (clsstate = nfa->states_head; NULL != clsstate; clsstate = clsstate->next)
1903 clsstate->contained = 0;
1905 // Add start state to closure only for epsilon closure
1907 GNUNET_array_append (cls->states, cls->len, s);
1909 GNUNET_array_append (cls_check->states, cls_check->len, s);
1910 while (cls_check->len > 0)
1912 currentstate = cls_check->states[cls_check->len - 1];
1913 GNUNET_array_grow (cls_check->states, cls_check->len, cls_check->len - 1);
1915 for (ctran = currentstate->transitions_head; NULL != ctran;
1916 ctran = ctran->next)
1918 if (NULL != ctran->to_state && label == ctran->label)
1920 clsstate = ctran->to_state;
1922 if (NULL != clsstate && 0 == clsstate->contained)
1924 GNUNET_array_append (cls->states, cls->len, clsstate);
1925 GNUNET_array_append (cls_check->states, cls_check->len, clsstate);
1926 clsstate->contained = 1;
1931 GNUNET_assert (0 == cls_check->len);
1932 GNUNET_free (cls_check);
1936 qsort (cls->states, cls->len, sizeof (struct GNUNET_REGEX_State *),
1943 * Calculates the closure set for the given set of states.
1945 * @param nfa the NFA containing 's'
1946 * @param states list of states on which to base the closure on
1947 * @param label transitioning label for which to base the closure on,
1948 * pass 0 for epsilon transition
1950 * @return sorted nfa closure on 'label' (epsilon closure if 'label' is 0)
1952 static struct GNUNET_REGEX_StateSet *
1953 nfa_closure_set_create (struct GNUNET_REGEX_Automaton *nfa,
1954 struct GNUNET_REGEX_StateSet *states, const char label)
1956 struct GNUNET_REGEX_State *s;
1957 struct GNUNET_REGEX_StateSet *sset;
1958 struct GNUNET_REGEX_StateSet *cls;
1967 cls = GNUNET_malloc (sizeof (struct GNUNET_REGEX_StateSet));
1969 for (i = 0; i < states->len; i++)
1971 s = states->states[i];
1972 sset = nfa_closure_create (nfa, s, label);
1974 for (j = 0; j < sset->len; j++)
1977 for (k = 0; k < cls->len; k++)
1979 if (sset->states[j]->id == cls->states[k]->id)
1986 GNUNET_array_append (cls->states, cls->len, sset->states[j]);
1988 state_set_clear (sset);
1992 qsort (cls->states, cls->len, sizeof (struct GNUNET_REGEX_State *),
1999 * Pops two NFA fragments (a, b) from the stack and concatenates them (ab)
2001 * @param ctx context
2004 nfa_add_concatenation (struct GNUNET_REGEX_Context *ctx)
2006 struct GNUNET_REGEX_Automaton *a;
2007 struct GNUNET_REGEX_Automaton *b;
2008 struct GNUNET_REGEX_Automaton *new;
2010 b = ctx->stack_tail;
2011 GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, b);
2012 a = ctx->stack_tail;
2013 GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a);
2015 state_add_transition (ctx, a->end, 0, b->start);
2016 a->end->accepting = 0;
2017 b->end->accepting = 1;
2019 new = nfa_fragment_create (NULL, NULL);
2020 nfa_add_states (new, a->states_head, a->states_tail);
2021 nfa_add_states (new, b->states_head, b->states_tail);
2022 new->start = a->start;
2024 automaton_fragment_clear (a);
2025 automaton_fragment_clear (b);
2027 GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new);
2031 * Pops a NFA fragment from the stack (a) and adds a new fragment (a*)
2033 * @param ctx context
2036 nfa_add_star_op (struct GNUNET_REGEX_Context *ctx)
2038 struct GNUNET_REGEX_Automaton *a;
2039 struct GNUNET_REGEX_Automaton *new;
2040 struct GNUNET_REGEX_State *start;
2041 struct GNUNET_REGEX_State *end;
2043 a = ctx->stack_tail;
2044 GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a);
2048 GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
2049 "nfa_add_star_op failed, because there was no element on the stack");
2053 start = nfa_state_create (ctx, 0);
2054 end = nfa_state_create (ctx, 1);
2056 state_add_transition (ctx, start, 0, a->start);
2057 state_add_transition (ctx, start, 0, end);
2058 state_add_transition (ctx, a->end, 0, a->start);
2059 state_add_transition (ctx, a->end, 0, end);
2061 a->end->accepting = 0;
2064 new = nfa_fragment_create (start, end);
2065 nfa_add_states (new, a->states_head, a->states_tail);
2066 automaton_fragment_clear (a);
2068 GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new);
2072 * Pops an NFA fragment (a) from the stack and adds a new fragment (a+)
2074 * @param ctx context
2077 nfa_add_plus_op (struct GNUNET_REGEX_Context *ctx)
2079 struct GNUNET_REGEX_Automaton *a;
2081 a = ctx->stack_tail;
2082 GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a);
2084 state_add_transition (ctx, a->end, 0, a->start);
2086 GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, a);
2090 * Pops an NFA fragment (a) from the stack and adds a new fragment (a?)
2092 * @param ctx context
2095 nfa_add_question_op (struct GNUNET_REGEX_Context *ctx)
2097 struct GNUNET_REGEX_Automaton *a;
2098 struct GNUNET_REGEX_Automaton *new;
2099 struct GNUNET_REGEX_State *start;
2100 struct GNUNET_REGEX_State *end;
2102 a = ctx->stack_tail;
2103 GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a);
2107 GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
2108 "nfa_add_question_op failed, because there was no element on the stack");
2112 start = nfa_state_create (ctx, 0);
2113 end = nfa_state_create (ctx, 1);
2115 state_add_transition (ctx, start, 0, a->start);
2116 state_add_transition (ctx, start, 0, end);
2117 state_add_transition (ctx, a->end, 0, end);
2119 a->end->accepting = 0;
2121 new = nfa_fragment_create (start, end);
2122 nfa_add_states (new, a->states_head, a->states_tail);
2123 automaton_fragment_clear (a);
2125 GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new);
2129 * Pops two NFA fragments (a, b) from the stack and adds a new NFA fragment that
2130 * alternates between a and b (a|b)
2132 * @param ctx context
2135 nfa_add_alternation (struct GNUNET_REGEX_Context *ctx)
2137 struct GNUNET_REGEX_Automaton *a;
2138 struct GNUNET_REGEX_Automaton *b;
2139 struct GNUNET_REGEX_Automaton *new;
2140 struct GNUNET_REGEX_State *start;
2141 struct GNUNET_REGEX_State *end;
2143 b = ctx->stack_tail;
2144 GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, b);
2145 a = ctx->stack_tail;
2146 GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a);
2148 start = nfa_state_create (ctx, 0);
2149 end = nfa_state_create (ctx, 1);
2150 state_add_transition (ctx, start, 0, a->start);
2151 state_add_transition (ctx, start, 0, b->start);
2153 state_add_transition (ctx, a->end, 0, end);
2154 state_add_transition (ctx, b->end, 0, end);
2156 a->end->accepting = 0;
2157 b->end->accepting = 0;
2160 new = nfa_fragment_create (start, end);
2161 nfa_add_states (new, a->states_head, a->states_tail);
2162 nfa_add_states (new, b->states_head, b->states_tail);
2163 automaton_fragment_clear (a);
2164 automaton_fragment_clear (b);
2166 GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new);
2170 * Adds a new nfa fragment to the stack
2172 * @param ctx context
2173 * @param lit label for nfa transition
2176 nfa_add_label (struct GNUNET_REGEX_Context *ctx, const char lit)
2178 struct GNUNET_REGEX_Automaton *n;
2179 struct GNUNET_REGEX_State *start;
2180 struct GNUNET_REGEX_State *end;
2182 GNUNET_assert (NULL != ctx);
2184 start = nfa_state_create (ctx, 0);
2185 end = nfa_state_create (ctx, 1);
2186 state_add_transition (ctx, start, lit, end);
2187 n = nfa_fragment_create (start, end);
2188 GNUNET_assert (NULL != n);
2189 GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, n);
2193 * Initialize a new context
2195 * @param ctx context
2198 GNUNET_REGEX_context_init (struct GNUNET_REGEX_Context *ctx)
2202 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Context was NULL!");
2206 ctx->transition_id = 0;
2208 ctx->stack_head = NULL;
2209 ctx->stack_tail = NULL;
2213 * Construct an NFA by parsing the regex string of length 'len'.
2215 * @param regex regular expression string
2216 * @param len length of the string
2218 * @return NFA, needs to be freed using GNUNET_REGEX_destroy_automaton
2220 struct GNUNET_REGEX_Automaton *
2221 GNUNET_REGEX_construct_nfa (const char *regex, const size_t len)
2223 struct GNUNET_REGEX_Context ctx;
2224 struct GNUNET_REGEX_Automaton *nfa;
2228 unsigned int altcount;
2229 unsigned int atomcount;
2230 unsigned int pcount;
2237 GNUNET_REGEX_context_init (&ctx);
2246 for (count = 0; count < len && *regexp; count++, regexp++)
2254 nfa_add_concatenation (&ctx);
2256 GNUNET_array_grow (p, pcount, pcount + 1);
2257 p[pcount - 1].altcount = altcount;
2258 p[pcount - 1].atomcount = atomcount;
2265 error_msg = "Cannot append '|' to nothing";
2268 while (--atomcount > 0)
2269 nfa_add_concatenation (&ctx);
2275 error_msg = "Missing opening '('";
2280 // Ignore this: "()"
2282 altcount = p[pcount].altcount;
2283 atomcount = p[pcount].atomcount;
2286 while (--atomcount > 0)
2287 nfa_add_concatenation (&ctx);
2288 for (; altcount > 0; altcount--)
2289 nfa_add_alternation (&ctx);
2291 altcount = p[pcount].altcount;
2292 atomcount = p[pcount].atomcount;
2298 error_msg = "Cannot append '*' to nothing";
2301 nfa_add_star_op (&ctx);
2306 error_msg = "Cannot append '+' to nothing";
2309 nfa_add_plus_op (&ctx);
2314 error_msg = "Cannot append '?' to nothing";
2317 nfa_add_question_op (&ctx);
2319 case 92: /* escape: \ */
2326 nfa_add_concatenation (&ctx);
2328 nfa_add_label (&ctx, *regexp);
2335 error_msg = "Unbalanced parenthesis";
2338 while (--atomcount > 0)
2339 nfa_add_concatenation (&ctx);
2340 for (; altcount > 0; altcount--)
2341 nfa_add_alternation (&ctx);
2343 GNUNET_free_non_null (p);
2345 nfa = ctx.stack_tail;
2346 GNUNET_CONTAINER_DLL_remove (ctx.stack_head, ctx.stack_tail, nfa);
2348 if (NULL != ctx.stack_head)
2350 error_msg = "Creating the NFA failed. NFA stack was not empty!";
2354 nfa->regex = GNUNET_strdup (regex);
2359 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not parse regex: %s\n", regex);
2360 if (NULL != error_msg)
2361 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "%s\n", error_msg);
2363 GNUNET_free_non_null (p);
2365 while (NULL != (nfa = ctx.stack_head))
2367 GNUNET_CONTAINER_DLL_remove (ctx.stack_head, ctx.stack_tail, nfa);
2368 GNUNET_REGEX_automaton_destroy (nfa);
2375 * Create DFA states based on given 'nfa' and starting with 'dfa_state'.
2377 * @param ctx context.
2378 * @param nfa NFA automaton.
2379 * @param dfa DFA automaton.
2380 * @param dfa_state current dfa state, pass epsilon closure of first nfa state
2384 construct_dfa_states (struct GNUNET_REGEX_Context *ctx,
2385 struct GNUNET_REGEX_Automaton *nfa,
2386 struct GNUNET_REGEX_Automaton *dfa,
2387 struct GNUNET_REGEX_State *dfa_state)
2389 struct Transition *ctran;
2390 struct GNUNET_REGEX_State *state_iter;
2391 struct GNUNET_REGEX_State *new_dfa_state;
2392 struct GNUNET_REGEX_State *state_contains;
2393 struct GNUNET_REGEX_StateSet *tmp;
2394 struct GNUNET_REGEX_StateSet *nfa_set;
2396 for (ctran = dfa_state->transitions_head; NULL != ctran; ctran = ctran->next)
2398 if (0 == ctran->label || NULL != ctran->to_state)
2401 tmp = nfa_closure_set_create (nfa, dfa_state->nfa_set, ctran->label);
2402 nfa_set = nfa_closure_set_create (nfa, tmp, 0);
2403 state_set_clear (tmp);
2404 new_dfa_state = dfa_state_create (ctx, nfa_set);
2405 state_contains = NULL;
2406 for (state_iter = dfa->states_head; NULL != state_iter;
2407 state_iter = state_iter->next)
2409 if (0 == state_set_compare (state_iter->nfa_set, new_dfa_state->nfa_set))
2410 state_contains = state_iter;
2413 if (NULL == state_contains)
2415 automaton_add_state (dfa, new_dfa_state);
2416 ctran->to_state = new_dfa_state;
2417 construct_dfa_states (ctx, nfa, dfa, new_dfa_state);
2421 ctran->to_state = state_contains;
2422 automaton_destroy_state (new_dfa_state);
2428 * Construct DFA for the given 'regex' of length 'len'
2430 * @param regex regular expression string
2431 * @param len length of the regular expression
2433 * @return DFA, needs to be freed using GNUNET_REGEX_destroy_automaton
2435 struct GNUNET_REGEX_Automaton *
2436 GNUNET_REGEX_construct_dfa (const char *regex, const size_t len)
2438 struct GNUNET_REGEX_Context ctx;
2439 struct GNUNET_REGEX_Automaton *dfa;
2440 struct GNUNET_REGEX_Automaton *nfa;
2441 struct GNUNET_REGEX_StateSet *nfa_set;
2443 GNUNET_REGEX_context_init (&ctx);
2446 nfa = GNUNET_REGEX_construct_nfa (regex, len);
2450 GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
2451 "Could not create DFA, because NFA creation failed\n");
2455 dfa = GNUNET_malloc (sizeof (struct GNUNET_REGEX_Automaton));
2457 dfa->regex = GNUNET_strdup (regex);
2459 // Create DFA start state from epsilon closure
2460 nfa_set = nfa_closure_create (nfa, nfa->start, 0);
2461 dfa->start = dfa_state_create (&ctx, nfa_set);
2462 automaton_add_state (dfa, dfa->start);
2464 construct_dfa_states (&ctx, nfa, dfa, dfa->start);
2466 GNUNET_REGEX_automaton_destroy (nfa);
2469 dfa_minimize (&ctx, dfa);
2472 scc_tarjan (&ctx, dfa);
2474 // Create proofs for all states
2475 automaton_create_proofs (dfa);
2481 * Free the memory allocated by constructing the GNUNET_REGEX_Automaton data
2484 * @param a automaton to be destroyed
2487 GNUNET_REGEX_automaton_destroy (struct GNUNET_REGEX_Automaton *a)
2489 struct GNUNET_REGEX_State *s;
2490 struct GNUNET_REGEX_State *next_state;
2495 GNUNET_free_non_null (a->regex);
2496 GNUNET_free_non_null (a->computed_regex);
2498 for (s = a->states_head; NULL != s;)
2500 next_state = s->next;
2501 automaton_destroy_state (s);
2509 * Save a state to an open file pointer. cls is expected to be a file pointer to
2510 * an open file. Used only in conjunction with
2511 * GNUNET_REGEX_automaton_save_graph.
2513 * @param cls file pointer.
2514 * @param count current count of the state, not used.
2518 GNUNET_REGEX_automaton_save_graph_step (void *cls, unsigned int count,
2519 struct GNUNET_REGEX_State *s)
2522 struct Transition *ctran;
2524 char *s_tran = NULL;
2530 GNUNET_asprintf (&s_acc,
2531 "\"%s(%i)\" [shape=doublecircle, color=\"0.%i 0.8 0.95\"];\n",
2532 s->name, s->proof_id, s->scc_id);
2536 GNUNET_asprintf (&s_acc, "\"%s(%i)\" [color=\"0.%i 0.8 0.95\"];\n", s->name,
2537 s->proof_id, s->scc_id);
2542 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not print state %s\n", s->name);
2545 fwrite (s_acc, strlen (s_acc), 1, p);
2546 GNUNET_free (s_acc);
2549 for (ctran = s->transitions_head; NULL != ctran; ctran = ctran->next)
2551 if (NULL == ctran->to_state)
2553 GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
2554 "Transition from State %i has no state for transitioning\n",
2559 if (ctran->label == 0)
2561 GNUNET_asprintf (&s_tran,
2562 "\"%s(%i)\" -> \"%s(%i)\" [label = \"epsilon\", color=\"0.%i 0.8 0.95\"];\n",
2563 s->name, s->proof_id, ctran->to_state->name,
2564 ctran->to_state->proof_id, s->scc_id);
2568 GNUNET_asprintf (&s_tran,
2569 "\"%s(%i)\" -> \"%s(%i)\" [label = \"%c\", color=\"0.%i 0.8 0.95\"];\n",
2570 s->name, s->proof_id, ctran->to_state->name,
2571 ctran->to_state->proof_id, ctran->label, s->scc_id);
2576 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not print state %s\n",
2581 fwrite (s_tran, strlen (s_tran), 1, p);
2582 GNUNET_free (s_tran);
2588 * Save the given automaton as a GraphViz dot file
2590 * @param a the automaton to be saved
2591 * @param filename where to save the file
2594 GNUNET_REGEX_automaton_save_graph (struct GNUNET_REGEX_Automaton *a,
2595 const char *filename)
2603 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not print NFA, was NULL!");
2607 if (NULL == filename || strlen (filename) < 1)
2609 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "No Filename given!");
2613 p = fopen (filename, "w");
2617 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not open file for writing: %s",
2622 start = "digraph G {\nrankdir=LR\n";
2623 fwrite (start, strlen (start), 1, p);
2625 automaton_traverse (p, a, GNUNET_REGEX_automaton_save_graph_step);
2628 fwrite (end, strlen (end), 1, p);
2633 * Evaluates the given string using the given DFA automaton
2635 * @param a automaton, type must be DFA
2636 * @param string string that should be evaluated
2638 * @return 0 if string matches, non 0 otherwise
2641 evaluate_dfa (struct GNUNET_REGEX_Automaton *a, const char *string)
2644 struct GNUNET_REGEX_State *s;
2648 GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
2649 "Tried to evaluate DFA, but NFA automaton given");
2655 // If the string is empty but the starting state is accepting, we accept.
2656 if ((NULL == string || 0 == strlen (string)) && s->accepting)
2659 for (strp = string; NULL != strp && *strp; strp++)
2661 s = dfa_move (s, *strp);
2666 if (NULL != s && s->accepting)
2673 * Evaluates the given string using the given NFA automaton
2675 * @param a automaton, type must be NFA
2676 * @param string string that should be evaluated
2678 * @return 0 if string matches, non 0 otherwise
2681 evaluate_nfa (struct GNUNET_REGEX_Automaton *a, const char *string)
2684 struct GNUNET_REGEX_State *s;
2685 struct GNUNET_REGEX_StateSet *sset;
2686 struct GNUNET_REGEX_StateSet *new_sset;
2692 GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
2693 "Tried to evaluate NFA, but DFA automaton given");
2697 // If the string is empty but the starting state is accepting, we accept.
2698 if ((NULL == string || 0 == strlen (string)) && a->start->accepting)
2703 sset = nfa_closure_create (a, a->start, 0);
2705 for (strp = string; NULL != strp && *strp; strp++)
2707 new_sset = nfa_closure_set_create (a, sset, *strp);
2708 state_set_clear (sset);
2709 sset = nfa_closure_set_create (a, new_sset, 0);
2710 state_set_clear (new_sset);
2713 for (i = 0; i < sset->len; i++)
2715 s = sset->states[i];
2716 if (NULL != s && s->accepting)
2723 state_set_clear (sset);
2728 * Evaluates the given 'string' against the given compiled regex
2730 * @param a automaton
2731 * @param string string to check
2733 * @return 0 if string matches, non 0 otherwise
2736 GNUNET_REGEX_eval (struct GNUNET_REGEX_Automaton *a, const char *string)
2743 result = evaluate_dfa (a, string);
2746 result = evaluate_nfa (a, string);
2749 GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
2750 "Evaluating regex failed, automaton has no type!\n");
2751 result = GNUNET_SYSERR;
2759 * Get the computed regex of the given automaton.
2760 * When constructing the automaton a proof is computed for each state,
2761 * consisting of the regular expression leading to this state. A complete
2762 * regex for the automaton can be computed by combining these proofs.
2763 * As of now this computed regex is only useful for testing.
2766 GNUNET_REGEX_get_computed_regex (struct GNUNET_REGEX_Automaton *a)
2771 return a->computed_regex;
2775 * Get the first key for the given 'input_string'. This hashes the first x bits
2776 * of the 'input_strings'.
2778 * @param input_string string.
2779 * @param string_len length of the 'input_string'.
2780 * @param key pointer to where to write the hash code.
2782 * @return number of bits of 'input_string' that have been consumed
2783 * to construct the key
2786 GNUNET_REGEX_get_first_key (const char *input_string, unsigned int string_len,
2787 struct GNUNET_HashCode *key)
2791 size = string_len < INITIAL_BITS ? string_len : INITIAL_BITS;
2793 if (NULL == input_string)
2795 GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Given input string was NULL!\n");
2799 GNUNET_CRYPTO_hash (input_string, size, key);
2805 * Check if the given 'proof' matches the given 'key'.
2807 * @param proof partial regex
2810 * @return GNUNET_OK if the proof is valid for the given key
2813 GNUNET_REGEX_check_proof (const char *proof, const struct GNUNET_HashCode *key)
2819 * Iterate over all edges helper function starting from state 's', calling
2820 * iterator on for each edge.
2823 * @param iterator iterator function called for each edge.
2824 * @param iterator_cls closure.
2827 iterate_edge (struct GNUNET_REGEX_State *s, GNUNET_REGEX_KeyIterator iterator,
2830 struct Transition *t;
2831 struct GNUNET_REGEX_Edge edges[s->transition_count];
2832 unsigned int num_edges;
2834 if (GNUNET_YES != s->marked)
2836 s->marked = GNUNET_YES;
2838 num_edges = state_get_edges (s, edges);
2840 iterator (iterator_cls, &s->hash, s->proof, s->accepting, num_edges, edges);
2842 for (t = s->transitions_head; NULL != t; t = t->next)
2843 iterate_edge (t->to_state, iterator, iterator_cls);
2848 * Iterate over all edges starting from start state of automaton 'a'. Calling
2849 * iterator for each edge.
2851 * @param a automaton.
2852 * @param iterator iterator called for each edge.
2853 * @param iterator_cls closure.
2856 GNUNET_REGEX_iterate_all_edges (struct GNUNET_REGEX_Automaton *a,
2857 GNUNET_REGEX_KeyIterator iterator,
2860 struct GNUNET_REGEX_State *s;
2862 for (s = a->states_head; NULL != s; s = s->next)
2863 s->marked = GNUNET_NO;
2865 iterate_edge (a->start, iterator, iterator_cls);