X-Git-Url: https://git.librecmc.org/?a=blobdiff_plain;f=src%2Fregex%2Fregex.c;h=c8b8ad3faedb680268ff3256c91ea8b868a9a449;hb=adbde8be4b6ee0a854177237391936faa8ca61c2;hp=721fe35596b91f906a7258e454ce10552ac973f9;hpb=fc28f65347c3d353c3b60468367f5765b0048c5c;p=oweals%2Fgnunet.git diff --git a/src/regex/regex.c b/src/regex/regex.c index 721fe3559..c8b8ad3fa 100644 --- a/src/regex/regex.c +++ b/src/regex/regex.c @@ -28,10 +28,12 @@ #include "gnunet_regex_lib.h" #include "regex_internal.h" + /** * Constant for how many bits the initial string regex should have. */ -#define INITIAL_BITS 10 +#define INITIAL_BITS 8 + /** * Context that contains an id counter for states and transitions as well as a @@ -60,211 +62,6 @@ struct GNUNET_REGEX_Context struct GNUNET_REGEX_Automaton *stack_tail; }; -/** - * Type of an automaton. - */ -enum GNUNET_REGEX_AutomatonType -{ - NFA, - DFA -}; - -/** - * Automaton representation. - */ -struct GNUNET_REGEX_Automaton -{ - /** - * Linked list of NFAs used for partial NFA creation. - */ - struct GNUNET_REGEX_Automaton *prev; - - /** - * Linked list of NFAs used for partial NFA creation. - */ - struct GNUNET_REGEX_Automaton *next; - - /** - * First state of the automaton. This is mainly used for constructing an NFA, - * where each NFA itself consists of one or more NFAs linked together. - */ - struct GNUNET_REGEX_State *start; - - /** - * End state of the partial NFA. This is undefined for DFAs - */ - struct GNUNET_REGEX_State *end; - - /** - * Number of states in the automaton. - */ - unsigned int state_count; - - /** - * DLL of states. - */ - struct GNUNET_REGEX_State *states_head; - - /** - * DLL of states - */ - struct GNUNET_REGEX_State *states_tail; - - /** - * Type of the automaton. - */ - enum GNUNET_REGEX_AutomatonType type; - - /** - * Regex - */ - char *regex; - - /** - * Canonical regex (result of RX->NFA->DFA->RX) - */ - char *canonical_regex; -}; - -/** - * A state. Can be used in DFA and NFA automatons. - */ -struct GNUNET_REGEX_State -{ - /** - * This is a linked list. - */ - struct GNUNET_REGEX_State *prev; - - /** - * This is a linked list. - */ - struct GNUNET_REGEX_State *next; - - /** - * Unique state id. - */ - unsigned int id; - - /** - * If this is an accepting state or not. - */ - int accepting; - - /** - * Marking of the state. This is used for marking all visited states when - * traversing all states of an automaton and for cases where the state id - * cannot be used (dfa minimization). - */ - int marked; - - /** - * Marking the state as contained. This is used for checking, if the state is - * contained in a set in constant time - */ - int contained; - - /** - * Marking the state as part of an SCC (Strongly Connected Component). All - * states with the same scc_id are part of the same SCC. scc_id is 0, if state - * is not a part of any SCC. - */ - unsigned int scc_id; - - /** - * Used for SCC detection. - */ - int index; - - /** - * Used for SCC detection. - */ - int lowlink; - - /** - * Human readable name of the automaton. Used for debugging and graph - * creation. - */ - char *name; - - /** - * Hash of the state. - */ - struct GNUNET_HashCode hash; - - /** - * State ID for proof creation. - */ - unsigned int proof_id; - - /** - * Proof for this state. - */ - char *proof; - - /** - * Number of transitions from this state to other states. - */ - unsigned int transition_count; - - /** - * DLL of transitions. - */ - struct Transition *transitions_head; - - /** - * DLL of transitions. - */ - struct Transition *transitions_tail; - - /** - * Set of states on which this state is based on. Used when creating a DFA out - * of several NFA states. - */ - struct GNUNET_REGEX_StateSet *nfa_set; -}; - -/** - * Transition between two states. Each state can have 0-n transitions. If label - * is 0, this is considered to be an epsilon transition. - */ -struct Transition -{ - /** - * This is a linked list. - */ - struct Transition *prev; - - /** - * This is a linked list. - */ - struct Transition *next; - - /** - * Unique id of this transition. - */ - unsigned int id; - - /** - * Label for this transition. This is basically the edge label for the graph. - */ - char label; - - /** - * State to which this transition leads. - */ - struct GNUNET_REGEX_State *to_state; - - /** - * State from which this transition origins. - */ - struct GNUNET_REGEX_State *from_state; - - /** - * Mark this transition. For example when reversing the automaton. - */ - int mark; -}; /** * Set of states. @@ -282,6 +79,7 @@ struct GNUNET_REGEX_StateSet unsigned int len; }; + /* * Debug helper functions */ @@ -294,6 +92,7 @@ struct GNUNET_REGEX_StateSet void debug_print_transitions (struct GNUNET_REGEX_State *s); + /** * Print information of the given state 's'. * @@ -318,6 +117,7 @@ debug_print_state (struct GNUNET_REGEX_State *s) debug_print_transitions (s); } + /** * Print debug information for all states contained in the automaton 'a'. * @@ -332,13 +132,14 @@ debug_print_states (struct GNUNET_REGEX_Automaton *a) debug_print_state (s); } + /** * Print debug information for given transition 't'. * * @param t transition for which to print debug info. */ void -debug_print_transition (struct Transition *t) +debug_print_transition (struct GNUNET_REGEX_Transition *t) { char *to_state; char *from_state; @@ -366,107 +167,17 @@ debug_print_transition (struct Transition *t) t->id, from_state, label, to_state); } + void debug_print_transitions (struct GNUNET_REGEX_State *s) { - struct Transition *t; + struct GNUNET_REGEX_Transition *t; for (t = s->transitions_head; NULL != t; t = t->next) debug_print_transition (t); } -/** - * Recursive function doing DFS with 'v' as a start, detecting all SCCs inside - * the subgraph reachable from 'v'. Used with scc_tarjan function to detect all - * SCCs inside an automaton. - * - * @param scc_counter counter for numbering the sccs - * @param v start vertex - * @param index current index - * @param stack stack for saving all SCCs - * @param stack_size current size of the stack - */ -static void -scc_tarjan_strongconnect (unsigned int *scc_counter, - struct GNUNET_REGEX_State *v, unsigned int *index, - struct GNUNET_REGEX_State **stack, - unsigned int *stack_size) -{ - struct GNUNET_REGEX_State *w; - struct Transition *t; - - v->index = *index; - v->lowlink = *index; - (*index)++; - stack[(*stack_size)++] = v; - v->contained = 1; - - for (t = v->transitions_head; NULL != t; t = t->next) - { - w = t->to_state; - if (NULL != w && w->index < 0) - { - scc_tarjan_strongconnect (scc_counter, w, index, stack, stack_size); - v->lowlink = (v->lowlink > w->lowlink) ? w->lowlink : v->lowlink; - } - else if (0 != w->contained) - v->lowlink = (v->lowlink > w->index) ? w->index : v->lowlink; - } - - if (v->lowlink == v->index) - { - w = stack[--(*stack_size)]; - w->contained = 0; - - if (v != w) - { - (*scc_counter)++; - while (v != w) - { - w->scc_id = *scc_counter; - w = stack[--(*stack_size)]; - w->contained = 0; - } - w->scc_id = *scc_counter; - } - } -} - - -/** - * Detect all SCCs (Strongly Connected Components) inside the given automaton. - * SCCs will be marked using the scc_id on each state. - * - * @param a the automaton for which SCCs should be computed and assigned. - */ -static void -scc_tarjan (struct GNUNET_REGEX_Automaton *a) -{ - unsigned int index; - unsigned int scc_counter; - struct GNUNET_REGEX_State *v; - struct GNUNET_REGEX_State *stack[a->state_count]; - unsigned int stack_size; - - for (v = a->states_head; NULL != v; v = v->next) - { - v->contained = 0; - v->index = -1; - v->lowlink = -1; - } - - stack_size = 0; - index = 0; - scc_counter = 0; - - for (v = a->states_head; NULL != v; v = v->next) - { - if (v->index < 0) - scc_tarjan_strongconnect (&scc_counter, v, &index, stack, &stack_size); - } -} - /** * Adds a transition from one state to another on 'label'. Does not add * duplicate states. @@ -482,8 +193,8 @@ state_add_transition (struct GNUNET_REGEX_Context *ctx, struct GNUNET_REGEX_State *to_state) { int is_dup; - struct Transition *t; - struct Transition *oth; + struct GNUNET_REGEX_Transition *t; + struct GNUNET_REGEX_Transition *oth; if (NULL == from_state) { @@ -513,7 +224,7 @@ state_add_transition (struct GNUNET_REGEX_Context *ctx, break; } - t = GNUNET_malloc (sizeof (struct Transition)); + t = GNUNET_malloc (sizeof (struct GNUNET_REGEX_Transition)); t->id = ctx->transition_id++; t->label = label; t->to_state = to_state; @@ -525,6 +236,30 @@ state_add_transition (struct GNUNET_REGEX_Context *ctx, from_state->transitions_tail, oth, t); } + +/** + * Remove a 'transition' from 'state'. + * + * @param state state from which the to-be-removed transition originates. + * @param transition transition that should be removed from state 'state'. + */ +static void +state_remove_transition (struct GNUNET_REGEX_State *state, + struct GNUNET_REGEX_Transition *transition) +{ + if (NULL == state || NULL == transition) + return; + + if (transition->from_state != state) + return; + + state->transition_count--; + GNUNET_CONTAINER_DLL_remove (state->transitions_head, state->transitions_tail, + transition); + GNUNET_free (transition); +} + + /** * Compare two states. Used for sorting. * @@ -547,18 +282,20 @@ state_compare (const void *a, const void *b) return (*s1)->id - (*s2)->id; } + /** * Get all edges leaving state 's'. * * @param s state. - * @param edges all edges leaving 's'. + * @param edges all edges leaving 's', expected to be allocated and have enough + * space for s->transitions_count elements. * * @return number of edges. */ static unsigned int state_get_edges (struct GNUNET_REGEX_State *s, struct GNUNET_REGEX_Edge *edges) { - struct Transition *t; + struct GNUNET_REGEX_Transition *t; unsigned int count; if (NULL == s) @@ -578,6 +315,7 @@ state_get_edges (struct GNUNET_REGEX_State *s, struct GNUNET_REGEX_Edge *edges) return count; } + /** * Compare to state sets by comparing the id's of the states that are contained * in each set. Both sets are expected to be sorted by id! @@ -611,6 +349,7 @@ state_set_compare (struct GNUNET_REGEX_StateSet *sset1, return result; } + /** * Clears the given StateSet 'set' * @@ -626,6 +365,7 @@ state_set_clear (struct GNUNET_REGEX_StateSet *set) } } + /** * Clears an automaton fragment. Does not destroy the states inside the * automaton. @@ -646,6 +386,7 @@ automaton_fragment_clear (struct GNUNET_REGEX_Automaton *a) GNUNET_free (a); } + /** * Frees the memory used by State 's' * @@ -654,8 +395,8 @@ automaton_fragment_clear (struct GNUNET_REGEX_Automaton *a) static void automaton_destroy_state (struct GNUNET_REGEX_State *s) { - struct Transition *t; - struct Transition *next_t; + struct GNUNET_REGEX_Transition *t; + struct GNUNET_REGEX_Transition *next_t; if (NULL == s) return; @@ -675,6 +416,7 @@ automaton_destroy_state (struct GNUNET_REGEX_State *s) GNUNET_free (s); } + /** * Remove a state from the given automaton 'a'. Always use this function when * altering the states of an automaton. Will also remove all transitions leading @@ -689,7 +431,7 @@ automaton_remove_state (struct GNUNET_REGEX_Automaton *a, { struct GNUNET_REGEX_State *ss; struct GNUNET_REGEX_State *s_check; - struct Transition *t_check; + struct GNUNET_REGEX_Transition *t_check; if (NULL == a || NULL == s) return; @@ -717,6 +459,7 @@ automaton_remove_state (struct GNUNET_REGEX_Automaton *a, automaton_destroy_state (ss); } + /** * Merge two states into one. Will merge 's1' and 's2' into 's1' and destroy * 's2'. @@ -733,9 +476,9 @@ automaton_merge_states (struct GNUNET_REGEX_Context *ctx, struct GNUNET_REGEX_State *s2) { struct GNUNET_REGEX_State *s_check; - struct Transition *t_check; - struct Transition *t; - struct Transition *t_next; + struct GNUNET_REGEX_Transition *t_check; + struct GNUNET_REGEX_Transition *t; + struct GNUNET_REGEX_Transition *t_next; char *new_name; int is_dup; @@ -763,11 +506,7 @@ automaton_merge_states (struct GNUNET_REGEX_Context *ctx, if (GNUNET_NO == is_dup) t_check->to_state = s1; else - { - GNUNET_CONTAINER_DLL_remove (t_check->from_state->transitions_head, - t_check->from_state->transitions_tail, - t_check); - } + state_remove_transition (t_check->from_state, t_check); } } } @@ -790,6 +529,7 @@ automaton_merge_states (struct GNUNET_REGEX_Context *ctx, automaton_destroy_state (s2); } + /** * Add a state to the automaton 'a', always use this function to alter the * states DLL of the automaton. @@ -805,15 +545,6 @@ automaton_add_state (struct GNUNET_REGEX_Automaton *a, a->state_count++; } -/** - * Function that is called with each state, when traversing an automaton. - * - * @param cls closure. - * @param count current count of the state, from 0 to a->state_count -1. - * @param s state. - */ -typedef void (*GNUNET_REGEX_traverse_action) (void *cls, unsigned int count, - struct GNUNET_REGEX_State * s); /** * Depth-first traversal of all states that are reachable from state 's'. Expects the states to @@ -829,7 +560,7 @@ static void automaton_state_traverse (struct GNUNET_REGEX_State *s, unsigned int *count, GNUNET_REGEX_traverse_action action, void *action_cls) { - struct Transition *t; + struct GNUNET_REGEX_Transition *t; if (GNUNET_NO != s->marked) return; @@ -850,9 +581,10 @@ automaton_state_traverse (struct GNUNET_REGEX_State *s, unsigned int *count, * @param action action to be performed on each state. * @param action_cls closure for action */ -static void -automaton_traverse (struct GNUNET_REGEX_Automaton *a, - GNUNET_REGEX_traverse_action action, void *action_cls) +void +GNUNET_REGEX_automaton_traverse (struct GNUNET_REGEX_Automaton *a, + GNUNET_REGEX_traverse_action action, + void *action_cls) { unsigned int count; struct GNUNET_REGEX_State *s; @@ -868,9 +600,6 @@ automaton_traverse (struct GNUNET_REGEX_Automaton *a, * Check if the given string 'str' needs parentheses around it when * using it to generate a regex. * - * Currently only tests for first and last characters being '()' respectively. - * FIXME: What about "(ab)|(cd)"? - * * @param str string * * @return GNUNET_YES if parentheses are needed, GNUNET_NO otherwise @@ -916,12 +645,9 @@ needs_parentheses (const char *str) /** * Remove parentheses surrounding string 'str'. - * Example: "(a)" becomes "a". + * Example: "(a)" becomes "a", "(a|b)|(a|c)" stays the same. * You need to GNUNET_free the returned string. * - * Currently only tests for first and last characters being '()' respectively. - * FIXME: What about "(ab)|(cd)"? - * * @param str string, free'd or re-used by this function, can be NULL * * @return string without surrounding parentheses, string 'str' if no preceding @@ -931,12 +657,18 @@ static char * remove_parentheses (char *str) { size_t slen; + const char *pos; if ((NULL == str) || ('(' != str[0]) || (str[(slen = strlen (str)) - 1] != ')')) return str; - memmove (str, &str[1], slen - 2); - str[slen - 2] = '\0'; + + pos = strchr (&str[1], ')'); + if (pos == &str[slen - 1]) + { + memmove (str, &str[1], slen - 2); + str[slen - 2] = '\0'; + } return str; } @@ -983,6 +715,7 @@ remove_epsilon (const char *str) return GNUNET_strdup (str); } + /** * Compare 'str1', starting from position 'k', with whole 'str2' * @@ -1018,8 +751,9 @@ nullstrcmp (const char *str1, const char *str2) return strcmp (str1, str2); } + /** - * Helper function used as 'action' in 'automaton_traverse' function to create + * Helper function used as 'action' in 'GNUNET_REGEX_automaton_traverse' function to create * the depth-first numbering of the states. * * @param cls states array. @@ -1032,14 +766,16 @@ number_states (void *cls, unsigned int count, struct GNUNET_REGEX_State *s) struct GNUNET_REGEX_State **states = cls; s->proof_id = count; - states[count] = s; + if (NULL != states) + states[count] = s; } -/** + +/** * Construct the regular expression given the inductive step, * $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* * R^{(k-1)}_{kj}, and simplify the resulting expression saved in R_cur_ij. - * + * * @param R_last_ij value of $R^{(k-1)_{ij}. * @param R_last_ik value of $R^{(k-1)_{ik}. * @param R_last_kk value of $R^{(k-1)_{kk}. @@ -1065,7 +801,6 @@ automaton_create_proofs_simplify (char *R_last_ij, char *R_last_ik, int ij_ik_cmp; int ij_kj_cmp; - //int ik_kj_cmp; int ik_kk_cmp; int kk_kj_cmp; int clean_ik_kk_cmp; @@ -1110,7 +845,6 @@ automaton_create_proofs_simplify (char *R_last_ij, char *R_last_ik, ij_kj_cmp = nullstrcmp (R_last_ij, R_last_kj); ij_ik_cmp = nullstrcmp (R_last_ij, R_last_ik); ik_kk_cmp = nullstrcmp (R_last_ik, R_last_kk); - //ik_kj_cmp = nullstrcmp (R_last_ik, R_last_kj); kk_kj_cmp = nullstrcmp (R_last_kk, R_last_kj); // Assign R_temp_(ik|kk|kj) to R_last[][] and remove epsilon as well @@ -1373,6 +1107,10 @@ automaton_create_proofs_simplify (char *R_last_ij, char *R_last_ik, } } + GNUNET_free_non_null (R_temp_ik); + GNUNET_free_non_null (R_temp_kk); + GNUNET_free_non_null (R_temp_kj); + if (NULL == R_cur_l && NULL == R_cur_r) { *R_cur_ij = NULL; @@ -1381,33 +1119,30 @@ automaton_create_proofs_simplify (char *R_last_ij, char *R_last_ik, if (NULL != R_cur_l && NULL == R_cur_r) { - *R_cur_ij = GNUNET_strdup (R_cur_l); + *R_cur_ij = R_cur_l; return; } if (NULL == R_cur_l && NULL != R_cur_r) { - *R_cur_ij = GNUNET_strdup (R_cur_r); + *R_cur_ij = R_cur_r; return; } if (0 == nullstrcmp (R_cur_l, R_cur_r)) { - *R_cur_ij = GNUNET_strdup (R_cur_l); + *R_cur_ij = R_cur_l; + GNUNET_free (R_cur_r); return; } GNUNET_asprintf (R_cur_ij, "(%s|%s)", R_cur_l, R_cur_r); - GNUNET_free_non_null (R_cur_l); - GNUNET_free_non_null (R_cur_r); - - GNUNET_free_non_null (R_temp_ik); - GNUNET_free_non_null (R_temp_kk); - GNUNET_free_non_null (R_temp_kj); - + GNUNET_free (R_cur_l); + GNUNET_free (R_cur_r); } + /** * create proofs for all states in the given automaton. Implementation of the * algorithm descriped in chapter 3.2.1 of "Automata Theory, Languages, and @@ -1423,7 +1158,7 @@ automaton_create_proofs (struct GNUNET_REGEX_Automaton *a) char *R_last[n][n]; char *R_cur[n][n]; char *temp; - struct Transition *t; + struct GNUNET_REGEX_Transition *t; char *complete_regex; unsigned int i; unsigned int j; @@ -1431,7 +1166,7 @@ automaton_create_proofs (struct GNUNET_REGEX_Automaton *a) /* create depth-first numbering of the states, initializes 'state' */ - automaton_traverse (a, &number_states, states); + GNUNET_REGEX_automaton_traverse (a, &number_states, states); /* Compute regular expressions of length "1" between each pair of states */ for (i = 0; i < n; i++) @@ -1534,13 +1269,6 @@ automaton_create_proofs (struct GNUNET_REGEX_Automaton *a) } a->canonical_regex = complete_regex; - GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, - "---------------------------------------------\n"); - GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Regex: %s\n", a->regex); - GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Complete Regex: %s\n", complete_regex); - GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, - "---------------------------------------------\n"); - // cleanup for (i = 0; i < n; i++) { @@ -1549,6 +1277,7 @@ automaton_create_proofs (struct GNUNET_REGEX_Automaton *a) } } + /** * Creates a new DFA state based on a set of NFA states. Needs to be freed using * automaton_destroy_state. @@ -1566,7 +1295,7 @@ dfa_state_create (struct GNUNET_REGEX_Context *ctx, char *name; int len = 0; struct GNUNET_REGEX_State *cstate; - struct Transition *ctran; + struct GNUNET_REGEX_Transition *ctran; unsigned int i; s = GNUNET_malloc (sizeof (struct GNUNET_REGEX_State)); @@ -1628,6 +1357,7 @@ dfa_state_create (struct GNUNET_REGEX_Context *ctx, return s; } + /** * Move from the given state 's' to the next state on transition 'label' * @@ -1639,7 +1369,7 @@ dfa_state_create (struct GNUNET_REGEX_Context *ctx, static struct GNUNET_REGEX_State * dfa_move (struct GNUNET_REGEX_State *s, const char label) { - struct Transition *t; + struct GNUNET_REGEX_Transition *t; struct GNUNET_REGEX_State *new_s; if (NULL == s) @@ -1659,6 +1389,7 @@ dfa_move (struct GNUNET_REGEX_State *s, const char label) return new_s; } + /** * Remove all unreachable states from DFA 'a'. Unreachable states are those * states that are not reachable from the starting state. @@ -1676,7 +1407,7 @@ dfa_remove_unreachable_states (struct GNUNET_REGEX_Automaton *a) s->marked = GNUNET_NO; // 2. traverse dfa from start state and mark all visited states - automaton_traverse (a, NULL, NULL); + GNUNET_REGEX_automaton_traverse (a, NULL, NULL); // 3. delete all states that were not visited for (s = a->states_head; NULL != s; s = s_next) @@ -1687,9 +1418,10 @@ dfa_remove_unreachable_states (struct GNUNET_REGEX_Automaton *a) } } + /** * Remove all dead states from the DFA 'a'. Dead states are those states that do - * not transition to any other state but themselfes. + * not transition to any other state but themselves. * * @param a DFA automaton */ @@ -1697,7 +1429,7 @@ static void dfa_remove_dead_states (struct GNUNET_REGEX_Automaton *a) { struct GNUNET_REGEX_State *s; - struct Transition *t; + struct GNUNET_REGEX_Transition *t; int dead; GNUNET_assert (DFA == a->type); @@ -1725,6 +1457,7 @@ dfa_remove_dead_states (struct GNUNET_REGEX_Automaton *a) } } + /** * Merge all non distinguishable states in the DFA 'a' * @@ -1739,8 +1472,8 @@ dfa_merge_nondistinguishable_states (struct GNUNET_REGEX_Context *ctx, int table[a->state_count][a->state_count]; struct GNUNET_REGEX_State *s1; struct GNUNET_REGEX_State *s2; - struct Transition *t1; - struct Transition *t2; + struct GNUNET_REGEX_Transition *t1; + struct GNUNET_REGEX_Transition *t2; struct GNUNET_REGEX_State *s1_next; struct GNUNET_REGEX_State *s2_next; int change; @@ -1819,6 +1552,7 @@ dfa_merge_nondistinguishable_states (struct GNUNET_REGEX_Context *ctx, } } + /** * Minimize the given DFA 'a' by removing all unreachable states, removing all * dead states and merging all non distinguishable states @@ -1845,6 +1579,7 @@ dfa_minimize (struct GNUNET_REGEX_Context *ctx, dfa_merge_nondistinguishable_states (ctx, a); } + /** * Creates a new NFA fragment. Needs to be cleared using * automaton_fragment_clear. @@ -1866,7 +1601,7 @@ nfa_fragment_create (struct GNUNET_REGEX_State *start, n->start = NULL; n->end = NULL; - if (NULL == start && NULL == end) + if (NULL == start || NULL == end) return n; automaton_add_state (n, end); @@ -1878,6 +1613,7 @@ nfa_fragment_create (struct GNUNET_REGEX_State *start, return n; } + /** * Adds a list of states to the given automaton 'n'. * @@ -1915,6 +1651,7 @@ nfa_add_states (struct GNUNET_REGEX_Automaton *n, n->state_count++; } + /** * Creates a new NFA state. Needs to be freed using automaton_destroy_state. * @@ -1942,6 +1679,7 @@ nfa_state_create (struct GNUNET_REGEX_Context *ctx, int accepting) return s; } + /** * Calculates the NFA closure set for the given state. * @@ -1960,7 +1698,7 @@ nfa_closure_create (struct GNUNET_REGEX_Automaton *nfa, struct GNUNET_REGEX_StateSet *cls_check; struct GNUNET_REGEX_State *clsstate; struct GNUNET_REGEX_State *currentstate; - struct Transition *ctran; + struct GNUNET_REGEX_Transition *ctran; if (NULL == s) return NULL; @@ -2008,6 +1746,7 @@ nfa_closure_create (struct GNUNET_REGEX_Automaton *nfa, return cls; } + /** * Calculates the closure set for the given set of states. * @@ -2064,6 +1803,7 @@ nfa_closure_set_create (struct GNUNET_REGEX_Automaton *nfa, return cls; } + /** * Pops two NFA fragments (a, b) from the stack and concatenates them (ab) * @@ -2074,28 +1814,31 @@ nfa_add_concatenation (struct GNUNET_REGEX_Context *ctx) { struct GNUNET_REGEX_Automaton *a; struct GNUNET_REGEX_Automaton *b; - struct GNUNET_REGEX_Automaton *new; + struct GNUNET_REGEX_Automaton *new_nfa; b = ctx->stack_tail; + GNUNET_assert (NULL != b); GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, b); a = ctx->stack_tail; + GNUNET_assert (NULL != a); GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a); state_add_transition (ctx, a->end, 0, b->start); a->end->accepting = 0; b->end->accepting = 1; - new = nfa_fragment_create (NULL, NULL); - nfa_add_states (new, a->states_head, a->states_tail); - nfa_add_states (new, b->states_head, b->states_tail); - new->start = a->start; - new->end = b->end; + new_nfa = nfa_fragment_create (NULL, NULL); + nfa_add_states (new_nfa, a->states_head, a->states_tail); + nfa_add_states (new_nfa, b->states_head, b->states_tail); + new_nfa->start = a->start; + new_nfa->end = b->end; automaton_fragment_clear (a); automaton_fragment_clear (b); - GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new); + GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new_nfa); } + /** * Pops a NFA fragment from the stack (a) and adds a new fragment (a*) * @@ -2105,12 +1848,11 @@ static void nfa_add_star_op (struct GNUNET_REGEX_Context *ctx) { struct GNUNET_REGEX_Automaton *a; - struct GNUNET_REGEX_Automaton *new; + struct GNUNET_REGEX_Automaton *new_nfa; struct GNUNET_REGEX_State *start; struct GNUNET_REGEX_State *end; a = ctx->stack_tail; - GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a); if (NULL == a) { @@ -2119,6 +1861,8 @@ nfa_add_star_op (struct GNUNET_REGEX_Context *ctx) return; } + GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a); + start = nfa_state_create (ctx, 0); end = nfa_state_create (ctx, 1); @@ -2130,13 +1874,14 @@ nfa_add_star_op (struct GNUNET_REGEX_Context *ctx) a->end->accepting = 0; end->accepting = 1; - new = nfa_fragment_create (start, end); - nfa_add_states (new, a->states_head, a->states_tail); + new_nfa = nfa_fragment_create (start, end); + nfa_add_states (new_nfa, a->states_head, a->states_tail); automaton_fragment_clear (a); - GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new); + GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new_nfa); } + /** * Pops an NFA fragment (a) from the stack and adds a new fragment (a+) * @@ -2155,6 +1900,7 @@ nfa_add_plus_op (struct GNUNET_REGEX_Context *ctx) GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, a); } + /** * Pops an NFA fragment (a) from the stack and adds a new fragment (a?) * @@ -2164,12 +1910,11 @@ static void nfa_add_question_op (struct GNUNET_REGEX_Context *ctx) { struct GNUNET_REGEX_Automaton *a; - struct GNUNET_REGEX_Automaton *new; + struct GNUNET_REGEX_Automaton *new_nfa; struct GNUNET_REGEX_State *start; struct GNUNET_REGEX_State *end; a = ctx->stack_tail; - GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a); if (NULL == a) { @@ -2178,6 +1923,8 @@ nfa_add_question_op (struct GNUNET_REGEX_Context *ctx) return; } + GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a); + start = nfa_state_create (ctx, 0); end = nfa_state_create (ctx, 1); @@ -2187,13 +1934,14 @@ nfa_add_question_op (struct GNUNET_REGEX_Context *ctx) a->end->accepting = 0; - new = nfa_fragment_create (start, end); - nfa_add_states (new, a->states_head, a->states_tail); + new_nfa = nfa_fragment_create (start, end); + nfa_add_states (new_nfa, a->states_head, a->states_tail); automaton_fragment_clear (a); - GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new); + GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new_nfa); } + /** * Pops two NFA fragments (a, b) from the stack and adds a new NFA fragment that * alternates between a and b (a|b) @@ -2205,13 +1953,15 @@ nfa_add_alternation (struct GNUNET_REGEX_Context *ctx) { struct GNUNET_REGEX_Automaton *a; struct GNUNET_REGEX_Automaton *b; - struct GNUNET_REGEX_Automaton *new; + struct GNUNET_REGEX_Automaton *new_nfa; struct GNUNET_REGEX_State *start; struct GNUNET_REGEX_State *end; b = ctx->stack_tail; + GNUNET_assert (NULL != b); GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, b); a = ctx->stack_tail; + GNUNET_assert (NULL != a); GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a); start = nfa_state_create (ctx, 0); @@ -2226,15 +1976,16 @@ nfa_add_alternation (struct GNUNET_REGEX_Context *ctx) b->end->accepting = 0; end->accepting = 1; - new = nfa_fragment_create (start, end); - nfa_add_states (new, a->states_head, a->states_tail); - nfa_add_states (new, b->states_head, b->states_tail); + new_nfa = nfa_fragment_create (start, end); + nfa_add_states (new_nfa, a->states_head, a->states_tail); + nfa_add_states (new_nfa, b->states_head, b->states_tail); automaton_fragment_clear (a); automaton_fragment_clear (b); - GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new); + GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new_nfa); } + /** * Adds a new nfa fragment to the stack * @@ -2258,6 +2009,7 @@ nfa_add_label (struct GNUNET_REGEX_Context *ctx, const char lit) GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, n); } + /** * Initialize a new context * @@ -2277,6 +2029,7 @@ GNUNET_REGEX_context_init (struct GNUNET_REGEX_Context *ctx) ctx->stack_tail = NULL; } + /** * Construct an NFA by parsing the regex string of length 'len'. * @@ -2387,6 +2140,7 @@ GNUNET_REGEX_construct_nfa (const char *regex, const size_t len) case 92: /* escape: \ */ regexp++; count++; + /* fall through! */ default: if (atomcount > 1) { @@ -2421,6 +2175,9 @@ GNUNET_REGEX_construct_nfa (const char *regex, const size_t len) nfa->regex = GNUNET_strdup (regex); + /* create depth-first numbering of the states for pretty printing */ + GNUNET_REGEX_automaton_traverse (nfa, &number_states, NULL); + return nfa; error: @@ -2439,6 +2196,7 @@ error: return NULL; } + /** * Create DFA states based on given 'nfa' and starting with 'dfa_state'. * @@ -2454,7 +2212,7 @@ construct_dfa_states (struct GNUNET_REGEX_Context *ctx, struct GNUNET_REGEX_Automaton *dfa, struct GNUNET_REGEX_State *dfa_state) { - struct Transition *ctran; + struct GNUNET_REGEX_Transition *ctran; struct GNUNET_REGEX_State *state_iter; struct GNUNET_REGEX_State *new_dfa_state; struct GNUNET_REGEX_State *state_contains; @@ -2492,6 +2250,7 @@ construct_dfa_states (struct GNUNET_REGEX_Context *ctx, } } + /** * Construct DFA for the given 'regex' of length 'len' * @@ -2542,6 +2301,7 @@ GNUNET_REGEX_construct_dfa (const char *regex, const size_t len) return dfa; } + /** * Free the memory allocated by constructing the GNUNET_REGEX_Automaton data * structure. @@ -2570,132 +2330,6 @@ GNUNET_REGEX_automaton_destroy (struct GNUNET_REGEX_Automaton *a) GNUNET_free (a); } -/** - * Save a state to an open file pointer. cls is expected to be a file pointer to - * an open file. Used only in conjunction with - * GNUNET_REGEX_automaton_save_graph. - * - * @param cls file pointer. - * @param count current count of the state, not used. - * @param s state. - */ -void -GNUNET_REGEX_automaton_save_graph_step (void *cls, unsigned int count, - struct GNUNET_REGEX_State *s) -{ - FILE *p; - struct Transition *ctran; - char *s_acc = NULL; - char *s_tran = NULL; - - p = cls; - - if (s->accepting) - { - GNUNET_asprintf (&s_acc, - "\"%s(%i)\" [shape=doublecircle, color=\"0.%i 0.8 0.95\"];\n", - s->name, s->proof_id, s->scc_id); - } - else - { - GNUNET_asprintf (&s_acc, "\"%s(%i)\" [color=\"0.%i 0.8 0.95\"];\n", s->name, - s->proof_id, s->scc_id); - } - - if (NULL == s_acc) - { - GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not print state %s\n", s->name); - return; - } - fwrite (s_acc, strlen (s_acc), 1, p); - GNUNET_free (s_acc); - s_acc = NULL; - - for (ctran = s->transitions_head; NULL != ctran; ctran = ctran->next) - { - if (NULL == ctran->to_state) - { - GNUNET_log (GNUNET_ERROR_TYPE_ERROR, - "Transition from State %i has no state for transitioning\n", - s->id); - continue; - } - - if (ctran->label == 0) - { - GNUNET_asprintf (&s_tran, - "\"%s(%i)\" -> \"%s(%i)\" [label = \"epsilon\", color=\"0.%i 0.8 0.95\"];\n", - s->name, s->proof_id, ctran->to_state->name, - ctran->to_state->proof_id, s->scc_id); - } - else - { - GNUNET_asprintf (&s_tran, - "\"%s(%i)\" -> \"%s(%i)\" [label = \"%c\", color=\"0.%i 0.8 0.95\"];\n", - s->name, s->proof_id, ctran->to_state->name, - ctran->to_state->proof_id, ctran->label, s->scc_id); - } - - if (NULL == s_tran) - { - GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not print state %s\n", - s->name); - return; - } - - fwrite (s_tran, strlen (s_tran), 1, p); - GNUNET_free (s_tran); - s_tran = NULL; - } -} - -/** - * Save the given automaton as a GraphViz dot file - * - * @param a the automaton to be saved - * @param filename where to save the file - */ -void -GNUNET_REGEX_automaton_save_graph (struct GNUNET_REGEX_Automaton *a, - const char *filename) -{ - char *start; - char *end; - FILE *p; - - if (NULL == a) - { - GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not print NFA, was NULL!"); - return; - } - - if (NULL == filename || strlen (filename) < 1) - { - GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "No Filename given!"); - return; - } - - p = fopen (filename, "w"); - - if (NULL == p) - { - GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not open file for writing: %s", - filename); - return; - } - - /* First add the SCCs to the automaton, so we can color them nicely */ - scc_tarjan (a); - - start = "digraph G {\nrankdir=LR\n"; - fwrite (start, strlen (start), 1, p); - - automaton_traverse (a, &GNUNET_REGEX_automaton_save_graph_step, p); - - end = "\n}\n"; - fwrite (end, strlen (end), 1, p); - fclose (p); -} /** * Evaluates the given string using the given DFA automaton @@ -2737,6 +2371,7 @@ evaluate_dfa (struct GNUNET_REGEX_Automaton *a, const char *string) return 1; } + /** * Evaluates the given string using the given NFA automaton * @@ -2767,7 +2402,6 @@ evaluate_nfa (struct GNUNET_REGEX_Automaton *a, const char *string) return 0; result = 1; - strp = string; sset = nfa_closure_create (a, a->start, 0); for (strp = string; NULL != strp && *strp; strp++) @@ -2792,6 +2426,7 @@ evaluate_nfa (struct GNUNET_REGEX_Automaton *a, const char *string) return result; } + /** * Evaluates the given 'string' against the given compiled regex * @@ -2844,9 +2479,10 @@ GNUNET_REGEX_get_canonical_regex (struct GNUNET_REGEX_Automaton *a) return a->canonical_regex; } + /** * Get the first key for the given 'input_string'. This hashes the first x bits - * of the 'input_strings'. + * of the 'input_string'. * * @param input_string string. * @param string_len length of the 'input_string'. @@ -2855,9 +2491,9 @@ GNUNET_REGEX_get_canonical_regex (struct GNUNET_REGEX_Automaton *a) * @return number of bits of 'input_string' that have been consumed * to construct the key */ -unsigned int -GNUNET_REGEX_get_first_key (const char *input_string, unsigned int string_len, - struct GNUNET_HashCode *key) +size_t +GNUNET_REGEX_get_first_key (const char *input_string, size_t string_len, + struct GNUNET_HashCode * key) { unsigned int size; @@ -2874,20 +2510,142 @@ GNUNET_REGEX_get_first_key (const char *input_string, unsigned int string_len, return size; } + /** * Check if the given 'proof' matches the given 'key'. * - * @param proof partial regex - * @param key hash + * @param proof partial regex of a state. + * @param key hash of a state. * - * @return GNUNET_OK if the proof is valid for the given key + * @return GNUNET_OK if the proof is valid for the given key. */ int GNUNET_REGEX_check_proof (const char *proof, const struct GNUNET_HashCode *key) { - return GNUNET_OK; + struct GNUNET_HashCode key_check; + + GNUNET_CRYPTO_hash (proof, strlen (proof), &key_check); + return (0 == + GNUNET_CRYPTO_hash_cmp (key, &key_check)) ? GNUNET_OK : GNUNET_NO; +} + + +/** + * Recursive helper function for iterate_initial_edges. Will call iterator + * function for each initial state. + * + * @param min_len minimum length of the path in the graph. + * @param max_len maximum length of the path in the graph. + * @param cur_len current length of the path already traversed. + * @param consumed_string string consumed by traversing the graph till this state. + * @param state current state of the automaton. + * @param iterator iterator function called for each edge. + * @param iterator_cls closure for the iterator function. + */ +static void +iterate_initial_edge (const unsigned int min_len, const unsigned int max_len, + unsigned int cur_len, char *consumed_string, + struct GNUNET_REGEX_State *state, + GNUNET_REGEX_KeyIterator iterator, void *iterator_cls) +{ + unsigned int i; + char label[state->transition_count][2]; + char *temp; + struct GNUNET_REGEX_Transition *t; + unsigned int num_edges = state->transition_count; + struct GNUNET_REGEX_Edge edges[num_edges]; + struct GNUNET_HashCode hash; + + if (cur_len > min_len && NULL != consumed_string && cur_len <= max_len) + { + for (i = 0, t = state->transitions_head; NULL != t; t = t->next, i++) + { + label[i][0] = t->label; + label[i][1] = '\0'; + edges[i].label = label[i]; + edges[i].destination = t->to_state->hash; + } + + GNUNET_CRYPTO_hash (consumed_string, strlen (consumed_string), &hash); + iterator (iterator_cls, &hash, consumed_string, state->accepting, num_edges, + edges); + } + + if (cur_len < max_len) + { + cur_len++; + for (t = state->transitions_head; NULL != t; t = t->next) + { + if (NULL != consumed_string) + GNUNET_asprintf (&temp, "%s%c", consumed_string, t->label); + else + GNUNET_asprintf (&temp, "%c", t->label); + + iterate_initial_edge (min_len, max_len, cur_len, temp, t->to_state, + iterator, iterator_cls); + GNUNET_free (temp); + } + } +} + + +/** + * Iterate over all initial edges that aren't actually part of the automaton. + * This is needed to find the initial states returned by + * GNUNET_REGEX_get_first_key. Iteration will start at the first branch state (a + * state that has more than one outgoing edge, can be the first state), because + * all previous states will have the same proof and be iterated over in + * iterate_all_edges. + * + * @param a the automaton for which the initial states should be computed. + * @param initial_len length of the initial state string. + * @param iterator iterator function called for each edge. + * @param iterator_cls closure for the iterator function. + */ +void +iterate_initial_edges (struct GNUNET_REGEX_Automaton *a, + const unsigned int initial_len, + GNUNET_REGEX_KeyIterator iterator, void *iterator_cls) +{ + char *consumed_string; + char *temp; + struct GNUNET_REGEX_State *s; + unsigned int cur_len; + + if (1 > initial_len) + return; + + consumed_string = NULL; + s = a->start; + cur_len = 0; + + if (1 == s->transition_count) + { + do + { + if (NULL != consumed_string) + { + temp = consumed_string; + GNUNET_asprintf (&consumed_string, "%s%c", consumed_string, + s->transitions_head->label); + GNUNET_free (temp); + } + else + GNUNET_asprintf (&consumed_string, "%c", s->transitions_head->label); + + s = s->transitions_head->to_state; + cur_len++; + } + while (cur_len < initial_len && 1 == s->transition_count); + } + + iterate_initial_edge (cur_len, initial_len, cur_len, consumed_string, s, + iterator, iterator_cls); + + GNUNET_free_non_null (consumed_string); } + /** * Iterate over all edges helper function starting from state 's', calling * iterator on for each edge. @@ -2900,7 +2658,7 @@ static void iterate_edge (struct GNUNET_REGEX_State *s, GNUNET_REGEX_KeyIterator iterator, void *iterator_cls) { - struct Transition *t; + struct GNUNET_REGEX_Transition *t; struct GNUNET_REGEX_Edge edges[s->transition_count]; unsigned int num_edges; @@ -2910,13 +2668,16 @@ iterate_edge (struct GNUNET_REGEX_State *s, GNUNET_REGEX_KeyIterator iterator, num_edges = state_get_edges (s, edges); - iterator (iterator_cls, &s->hash, s->proof, s->accepting, num_edges, edges); + if (0 < strlen (s->proof) || s->accepting) + iterator (iterator_cls, &s->hash, s->proof, s->accepting, num_edges, + edges); for (t = s->transitions_head; NULL != t; t = t->next) iterate_edge (t->to_state, iterator, iterator_cls); } } + /** * Iterate over all edges starting from start state of automaton 'a'. Calling * iterator for each edge. @@ -2935,5 +2696,6 @@ GNUNET_REGEX_iterate_all_edges (struct GNUNET_REGEX_Automaton *a, for (s = a->states_head; NULL != s; s = s->next) s->marked = GNUNET_NO; + iterate_initial_edges (a, INITIAL_BITS, iterator, iterator_cls); iterate_edge (a->start, iterator, iterator_cls); }