X-Git-Url: https://git.librecmc.org/?a=blobdiff_plain;f=src%2Fregex%2Fregex.c;h=c8b8ad3faedb680268ff3256c91ea8b868a9a449;hb=adbde8be4b6ee0a854177237391936faa8ca61c2;hp=e9b262f9594c605960d795b9aa1c8720a31afb4e;hpb=07971e5de7c792b9f591404f3f79e587f48e8d70;p=oweals%2Fgnunet.git diff --git a/src/regex/regex.c b/src/regex/regex.c index e9b262f95..c8b8ad3fa 100644 --- a/src/regex/regex.c +++ b/src/regex/regex.c @@ -26,13 +26,18 @@ #include "gnunet_container_lib.h" #include "gnunet_crypto_lib.h" #include "gnunet_regex_lib.h" -#include "regex.h" +#include "regex_internal.h" -#define initial_bits 10 /** - * Context that contains an id counter for states and transitions - * as well as a DLL of automatons used as a stack for NFA construction. + * Constant for how many bits the initial string regex should have. + */ +#define INITIAL_BITS 8 + + +/** + * Context that contains an id counter for states and transitions as well as a + * DLL of automatons used as a stack for NFA construction. */ struct GNUNET_REGEX_Context { @@ -46,11 +51,6 @@ struct GNUNET_REGEX_Context */ unsigned int transition_id; - /** - * Unique SCC (Strongly Connected Component) id. - */ - unsigned int scc_id; - /** * DLL of GNUNET_REGEX_Automaton's used as a stack. */ @@ -62,337 +62,204 @@ struct GNUNET_REGEX_Context struct GNUNET_REGEX_Automaton *stack_tail; }; -/** - * Type of an automaton. - */ -enum GNUNET_REGEX_automaton_type -{ - NFA, - DFA -}; /** - * Automaton representation. + * Set of states. */ -struct GNUNET_REGEX_Automaton +struct GNUNET_REGEX_StateSet { /** - * This is a linked list. - */ - struct GNUNET_REGEX_Automaton *prev; - - /** - * This is a linked list. - */ - 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. + * Array of states. */ - struct GNUNET_REGEX_State *start; + struct GNUNET_REGEX_State **states; /** - * End state of the automaton. + * Length of the 'states' array. */ - struct GNUNET_REGEX_State *end; + unsigned int len; +}; - /** - * Number of states in the automaton. - */ - unsigned int state_count; - /** - * DLL of states. - */ - struct GNUNET_REGEX_State *states_head; +/* + * Debug helper functions + */ - /** - * DLL of states - */ - struct GNUNET_REGEX_State *states_tail; +/** + * Print all the transitions of state 's'. + * + * @param s state for which to print it's transitions. + */ +void +debug_print_transitions (struct GNUNET_REGEX_State *s); - /** - * Type of the automaton. - */ - enum GNUNET_REGEX_automaton_type type; -}; /** - * A state. Can be used in DFA and NFA automatons. + * Print information of the given state 's'. + * + * @param s state for which debug information should be printed. */ -struct GNUNET_REGEX_State +void +debug_print_state (struct GNUNET_REGEX_State *s) { - /** - * 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. - */ - 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; - - GNUNET_HashCode hash; + char *proof; - /** - * Number of transitions from this state to other states. - */ - unsigned int transition_count; + if (NULL == s->proof) + proof = "NULL"; + else + proof = s->proof; - /** - * DLL of transitions. - */ - struct Transition *transitions_head; + GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, + "State %i: %s marked: %i accepting: %i scc_id: %i transitions: %i proof: %s\n", + s->id, s->name, s->marked, s->accepting, s->scc_id, + s->transition_count, proof); - /** - * DLL of transitions. - */ - struct Transition *transitions_tail; + GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Transitions:\n"); + debug_print_transitions (s); +} - /** - * 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. + * Print debug information for all states contained in the automaton 'a'. + * + * @param a automaton for which debug information of it's states should be printed. */ -struct Transition +void +debug_print_states (struct GNUNET_REGEX_Automaton *a) { - /** - * 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; + struct GNUNET_REGEX_State *s; - /** - * State to which this transition leads. - */ - struct GNUNET_REGEX_State *to_state; + for (s = a->states_head; NULL != s; s = s->next) + debug_print_state (s); +} - /** - * State from which this transition origins. - */ - struct GNUNET_REGEX_State *from_state; -}; /** - * Set of states. + * Print debug information for given transition 't'. + * + * @param t transition for which to print debug info. */ -struct GNUNET_REGEX_StateSet +void +debug_print_transition (struct GNUNET_REGEX_Transition *t) { - /** - * Array of states. - */ - struct GNUNET_REGEX_State **states; + char *to_state; + char *from_state; + char label; - /** - * Length of the 'states' array. - */ - unsigned int len; -}; + if (NULL == t) + return; -static void -debug_print_state (struct GNUNET_REGEX_State *s) -{ - GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, - "State %i: %s marked: %i accepting: %i scc_id: %i\n", s->id, - s->name, s->marked, s->accepting, s->scc_id); -} + if (0 == t->label) + label = '0'; + else + label = t->label; -static void -debug_print_states (struct GNUNET_REGEX_StateSet *sset) -{ - struct GNUNET_REGEX_State *s; - int i; + if (NULL == t->to_state) + to_state = "NULL"; + else + to_state = t->to_state->name; - for (i = 0; i < sset->len; i++) - { - s = sset->states[i]; - debug_print_state (s); - } + if (NULL == t->from_state) + from_state = "NULL"; + else + from_state = t->from_state->name; + + GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Transition %i: From %s on %c to %s\n", + t->id, from_state, label, to_state); } -static void + +void debug_print_transitions (struct GNUNET_REGEX_State *s) { - struct Transition *t; - char *state; - char label; + struct GNUNET_REGEX_Transition *t; for (t = s->transitions_head; NULL != t; t = t->next) - { - if (0 == t->label) - label = '0'; - else - label = t->label; - - if (NULL == t->to_state) - state = "NULL"; - else - state = t->to_state->name; - - GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Transition %i: On %c to %s\n", t->id, - label, state); - } + 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. + * Adds a transition from one state to another on 'label'. Does not add + * duplicate states. * * @param ctx context - * @param v start vertex - * @param index current index - * @param stack stack for saving all SCCs - * @param stack_size current size of the stack + * @param from_state starting state for the transition + * @param label transition label + * @param to_state state to where the transition should point to */ static void -scc_tarjan_strongconnect (struct GNUNET_REGEX_Context *ctx, - struct GNUNET_REGEX_State *v, int *index, - struct GNUNET_REGEX_State **stack, - unsigned int *stack_size) +state_add_transition (struct GNUNET_REGEX_Context *ctx, + struct GNUNET_REGEX_State *from_state, const char label, + struct GNUNET_REGEX_State *to_state) { - struct GNUNET_REGEX_State *w; - struct Transition *t; + int is_dup; + struct GNUNET_REGEX_Transition *t; + struct GNUNET_REGEX_Transition *oth; - v->index = *index; - v->lowlink = *index; - (*index)++; - stack[(*stack_size)++] = v; - v->contained = 1; + if (NULL == from_state) + { + GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not create Transition.\n"); + return; + } - for (t = v->transitions_head; NULL != t; t = t->next) + // Do not add duplicate state transitions + is_dup = GNUNET_NO; + for (t = from_state->transitions_head; NULL != t; t = t->next) { - w = t->to_state; - if (NULL != w && w->index < 0) + if (t->to_state == to_state && t->label == label && + t->from_state == from_state) { - scc_tarjan_strongconnect (ctx, w, index, stack, stack_size); - v->lowlink = (v->lowlink > w->lowlink) ? w->lowlink : v->lowlink; + is_dup = GNUNET_YES; + break; } - 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 (GNUNET_YES == is_dup) + return; - if (v != w) - { - ctx->scc_id++; - while (v != w) - { - w->scc_id = ctx->scc_id; - w = stack[--(*stack_size)]; - w->contained = 0; - } - w->scc_id = ctx->scc_id; - } + // sort transitions by label + for (oth = from_state->transitions_head; NULL != oth; oth = oth->next) + { + if (oth->label > label) + break; } + + t = GNUNET_malloc (sizeof (struct GNUNET_REGEX_Transition)); + t->id = ctx->transition_id++; + t->label = label; + t->to_state = to_state; + t->from_state = from_state; + + // Add outgoing transition to 'from_state' + from_state->transition_count++; + GNUNET_CONTAINER_DLL_insert_before (from_state->transitions_head, + from_state->transitions_tail, oth, t); } + /** - * Detect all SCCs (Strongly Connected Components) inside the given automaton. - * SCCs will be marked using the scc_id on each state. + * Remove a 'transition' from 'state'. * - * @param ctx context - * @param a automaton + * @param state state from which the to-be-removed transition originates. + * @param transition transition that should be removed from state 'state'. */ static void -scc_tarjan (struct GNUNET_REGEX_Context *ctx, struct GNUNET_REGEX_Automaton *a) +state_remove_transition (struct GNUNET_REGEX_State *state, + struct GNUNET_REGEX_Transition *transition) { - unsigned int i; - int index; - 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; - } + if (NULL == state || NULL == transition) + return; - stack_size = 0; - index = 0; + if (transition->from_state != state) + return; - for (i = 0, v = a->states_head; NULL != v; v = v->next) - { - if (v->index < 0) - scc_tarjan_strongconnect (ctx, v, &index, stack, &stack_size); - } + state->transition_count--; + GNUNET_CONTAINER_DLL_remove (state->transitions_head, state->transitions_tail, + transition); + GNUNET_free (transition); } + /** * Compare two states. Used for sorting. * @@ -415,9 +282,43 @@ 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', 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 GNUNET_REGEX_Transition *t; + unsigned int count; + + if (NULL == s) + return 0; + + count = 0; + + for (t = s->transitions_head; NULL != t; t = t->next) + { + if (NULL != t->to_state) + { + edges[count].label = &t->label; + edges[count].destination = t->to_state->hash; + count++; + } + } + 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! + * 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! * * @param sset1 first state set * @param sset2 second state set @@ -431,7 +332,7 @@ state_set_compare (struct GNUNET_REGEX_StateSet *sset1, struct GNUNET_REGEX_StateSet *sset2) { int result; - int i; + unsigned int i; if (NULL == sset1 || NULL == sset2) return 1; @@ -448,6 +349,7 @@ state_set_compare (struct GNUNET_REGEX_StateSet *sset1, return result; } + /** * Clears the given StateSet 'set' * @@ -458,47 +360,15 @@ state_set_clear (struct GNUNET_REGEX_StateSet *set) { if (NULL != set) { - if (NULL != set->states) - GNUNET_free (set->states); + GNUNET_free_non_null (set->states); GNUNET_free (set); } } -/** - * Adds a transition from one state to another on 'label' - * - * @param ctx context - * @param from_state starting state for the transition - * @param label transition label - * @param to_state state to where the transition should point to - */ -static void -state_add_transition (struct GNUNET_REGEX_Context *ctx, - struct GNUNET_REGEX_State *from_state, const char label, - struct GNUNET_REGEX_State *to_state) -{ - struct Transition *t; - - if (NULL == from_state) - { - GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not create Transition.\n"); - return; - } - - t = GNUNET_malloc (sizeof (struct Transition)); - - t->id = ctx->transition_id++; - t->label = label; - t->to_state = to_state; - t->from_state = from_state; - - GNUNET_CONTAINER_DLL_insert (from_state->transitions_head, - from_state->transitions_tail, t); -} /** - * Clears an automaton fragment. Does not destroy the states inside - * the automaton. + * Clears an automaton fragment. Does not destroy the states inside the + * automaton. * * @param a automaton to be cleared */ @@ -516,6 +386,7 @@ automaton_fragment_clear (struct GNUNET_REGEX_Automaton *a) GNUNET_free (a); } + /** * Frees the memory used by State 's' * @@ -524,21 +395,20 @@ 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; - if (NULL != s->name) - GNUNET_free (s->name); + GNUNET_free_non_null (s->name); + GNUNET_free_non_null (s->proof); - for (t = s->transitions_head; NULL != t;) + for (t = s->transitions_head; NULL != t; t = next_t) { next_t = t->next; GNUNET_CONTAINER_DLL_remove (s->transitions_head, s->transitions_tail, t); GNUNET_free (t); - t = next_t; } state_set_clear (s->nfa_set); @@ -546,10 +416,11 @@ 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 to this state, before destroying it. + * 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 + * to this state, before destroying it. * * @param a automaton * @param s state to remove @@ -560,7 +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; @@ -588,8 +459,10 @@ 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'. + * Merge two states into one. Will merge 's1' and 's2' into 's1' and destroy + * 's2'. * * @param ctx context * @param a automaton @@ -603,54 +476,63 @@ 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 GNUNET_REGEX_Transition *t_check; + struct GNUNET_REGEX_Transition *t; + struct GNUNET_REGEX_Transition *t_next; char *new_name; + int is_dup; GNUNET_assert (NULL != ctx && NULL != a && NULL != s1 && NULL != s2); - // 1. Make all transitions pointing to s2 point to s1 + if (s1 == s2) + return; + + // 1. Make all transitions pointing to s2 point to s1, unless this transition + // does not already exists, if it already exists remove transition. for (s_check = a->states_head; NULL != s_check; s_check = s_check->next) { - for (t_check = s_check->transitions_head; NULL != t_check; - t_check = t_check->next) + for (t_check = s_check->transitions_head; NULL != t_check; t_check = t_next) { - if (s_check != s1 && s2 == t_check->to_state) - t_check->to_state = s1; - } + t_next = t_check->next; + + if (s2 == t_check->to_state) + { + is_dup = GNUNET_NO; + for (t = t_check->from_state->transitions_head; NULL != t; t = t->next) + { + if (t->to_state == s1 && t_check->label == t->label) + is_dup = GNUNET_YES; + } + if (GNUNET_NO == is_dup) + t_check->to_state = s1; + else + state_remove_transition (t_check->from_state, t_check); + } + } } // 2. Add all transitions from s2 to sX to s1 for (t_check = s2->transitions_head; NULL != t_check; t_check = t_check->next) { - for (t = s1->transitions_head; NULL != t; t = t->next) - { - if (t_check->label != t->label && NULL != t_check->to_state && - t_check->to_state != t->to_state && t_check->to_state != s2) - { - state_add_transition (ctx, s1, t_check->label, t_check->to_state); - } - } + if (t_check->to_state != s1) + state_add_transition (ctx, s1, t_check->label, t_check->to_state); } // 3. Rename s1 to {s1,s2} - new_name = GNUNET_malloc (strlen (s1->name) + strlen (s2->name) + 1); - strncat (new_name, s1->name, strlen (s1->name)); - strncat (new_name, s2->name, strlen (s2->name)); - if (NULL != s1->name) - GNUNET_free (s1->name); - s1->name = new_name; + new_name = s1->name; + GNUNET_asprintf (&s1->name, "{%s,%s}", new_name, s2->name); + GNUNET_free (new_name); // remove state - s_check = s2; - GNUNET_CONTAINER_DLL_remove (a->states_head, a->states_tail, s_check); + GNUNET_CONTAINER_DLL_remove (a->states_head, a->states_tail, s2); a->state_count--; - automaton_destroy_state (s_check); + automaton_destroy_state (s2); } + /** - * Add a state to the automaton 'a', always use this function to - * alter the states DLL of the automaton. + * Add a state to the automaton 'a', always use this function to alter the + * states DLL of the automaton. * * @param a automaton to add the state to * @param s state that should be added @@ -663,57 +545,742 @@ automaton_add_state (struct GNUNET_REGEX_Automaton *a, a->state_count++; } -typedef void (*GNUNET_REGEX_traverse_action)(void *cls, struct - GNUNET_REGEX_State *s); /** - * Traverses all states that are reachable from state 's'. Expects - * the states to be unmarked (s->marked == GNUNET_NO). Performs - * 'action' on each visited state. + * Depth-first traversal of all states that are reachable from state 's'. Expects the states to + * be unmarked (s->marked == GNUNET_NO). Performs 'action' on each visited + * state. * * @param s start state. + * @param count current count of the state. * @param action action to be performed on each state. + * @param action_cls closure for action + */ +static void +automaton_state_traverse (struct GNUNET_REGEX_State *s, unsigned int *count, + GNUNET_REGEX_traverse_action action, void *action_cls) +{ + struct GNUNET_REGEX_Transition *t; + + if (GNUNET_NO != s->marked) + return; + s->marked = GNUNET_YES; + if (NULL != action) + action (action_cls, *count, s); + (*count)++; + for (t = s->transitions_head; NULL != t; t = t->next) + automaton_state_traverse (t->to_state, count, action, action_cls); +} + + +/** + * Traverses the given automaton from it's start state, visiting all reachable + * states and calling 'action' on each one of them. + * + * @param a automaton. + * @param action action to be performed on each state. + * @param action_cls closure for action + */ +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; + + for (s = a->states_head; NULL != s; s = s->next) + s->marked = GNUNET_NO; + count = 0; + automaton_state_traverse (a->start, &count, action, action_cls); +} + + +/** + * Check if the given string 'str' needs parentheses around it when + * using it to generate a regex. + * + * @param str string + * + * @return GNUNET_YES if parentheses are needed, GNUNET_NO otherwise + */ +static int +needs_parentheses (const char *str) +{ + size_t slen; + const char *op; + const char *cl; + const char *pos; + unsigned int cnt; + + if ((NULL == str) || ((slen = strlen (str)) < 2)) + return GNUNET_NO; + + if ('(' != str[0]) + return GNUNET_YES; + cnt = 1; + pos = &str[1]; + while (cnt > 0) + { + cl = strchr (pos, ')'); + if (NULL == cl) + { + GNUNET_break (0); + return GNUNET_YES; + } + op = strchr (pos, '('); + if ((NULL != op) && (op < cl)) + { + cnt++; + pos = op + 1; + continue; + } + /* got ')' first */ + cnt--; + pos = cl + 1; + } + return (*pos == '\0') ? GNUNET_NO : GNUNET_YES; +} + + +/** + * Remove parentheses surrounding string 'str'. + * Example: "(a)" becomes "a", "(a|b)|(a|c)" stays the same. + * You need to GNUNET_free the returned string. + * + * @param str string, free'd or re-used by this function, can be NULL + * + * @return string without surrounding parentheses, string 'str' if no preceding + * epsilon could be found, NULL if 'str' was NULL + */ +static char * +remove_parentheses (char *str) +{ + size_t slen; + const char *pos; + + if ((NULL == str) || ('(' != str[0]) || + (str[(slen = strlen (str)) - 1] != ')')) + return str; + + pos = strchr (&str[1], ')'); + if (pos == &str[slen - 1]) + { + memmove (str, &str[1], slen - 2); + str[slen - 2] = '\0'; + } + return str; +} + + +/** + * Check if the string 'str' starts with an epsilon (empty string). + * Example: "(|a)" is starting with an epsilon. + * + * @param str string to test + * + * @return 0 if str has no epsilon, 1 if str starts with '(|' and ends with ')' + */ +static int +has_epsilon (const char *str) +{ + return (NULL != str) && ('(' == str[0]) && ('|' == str[1]) && + (')' == str[strlen (str) - 1]); +} + + +/** + * Remove an epsilon from the string str. Where epsilon is an empty string + * Example: str = "(|a|b|c)", result: "a|b|c" + * The returned string needs to be freed. + * + * @param str string + * + * @return string without preceding epsilon, string 'str' if no preceding epsilon + * could be found, NULL if 'str' was NULL + */ +static char * +remove_epsilon (const char *str) +{ + size_t len; + + if (NULL == str) + return NULL; + if (('(' == str[0]) && ('|' == str[1])) + { + len = strlen (str); + if (')' == str[len - 1]) + return GNUNET_strndup (&str[2], len - 3); + } + return GNUNET_strdup (str); +} + + +/** + * Compare 'str1', starting from position 'k', with whole 'str2' + * + * @param str1 first string to compare, starting from position 'k' + * @param str2 second string for comparison + * @param k starting position in 'str1' + * + * @return -1 if any of the strings is NULL, 0 if equal, non 0 otherwise + */ +static int +strkcmp (const char *str1, const char *str2, size_t k) +{ + if ((NULL == str1) || (NULL == str2) || (strlen (str1) < k)) + return -1; + return strcmp (&str1[k], str2); +} + + +/** + * Compare two strings for equality. If either is NULL (or if both are + * NULL), they are not equal. + * + * @param str1 first string for comparison. + * @param str2 second string for comparison. + * + * @return 0 if the strings are the same, 1 or -1 if not + */ +static int +nullstrcmp (const char *str1, const char *str2) +{ + if ((NULL == str1) || (NULL == str2)) + return -1; + return strcmp (str1, str2); +} + + +/** + * Helper function used as 'action' in 'GNUNET_REGEX_automaton_traverse' function to create + * the depth-first numbering of the states. + * + * @param cls states array. + * @param count current state counter. + * @param s current state. + */ +static void +number_states (void *cls, unsigned int count, struct GNUNET_REGEX_State *s) +{ + struct GNUNET_REGEX_State **states = cls; + + s->proof_id = count; + 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}. + * @param R_last_kj value of $R^{(k-1)_{kj}. + * @param R_cur_ij result for this inductive step is saved in R_cur_ij, R_cur_ij + * is expected to be NULL when called! + */ +static void +automaton_create_proofs_simplify (char *R_last_ij, char *R_last_ik, + char *R_last_kk, char *R_last_kj, + char **R_cur_ij) +{ + char *R_cur_l; + char *R_cur_r; + char *temp_a; + char *temp_b; + char *R_temp_ij; + char *R_temp_ik; + char *R_temp_kj; + char *R_temp_kk; + + int eps_check; + int ij_ik_cmp; + int ij_kj_cmp; + + int ik_kk_cmp; + int kk_kj_cmp; + int clean_ik_kk_cmp; + int clean_kk_kj_cmp; + unsigned int cnt; + + size_t length; + size_t length_l; + size_t length_r; + + GNUNET_assert (NULL == *R_cur_ij && NULL != R_cur_ij); + + // $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj} + // R_last == R^{(k-1)}, R_cur == R^{(k)} + // R_cur_ij = R_cur_l | R_cur_r + // R_cur_l == R^{(k-1)}_{ij} + // R_cur_r == R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj} + + if ((NULL == R_last_ij) && ((NULL == R_last_ik) || (NULL == R_last_kk) || /* technically cannot happen, but looks saner */ + (NULL == R_last_kj))) + { + /* R^{(k)}_{ij} = N | N */ + *R_cur_ij = NULL; + return; + } + + if ((NULL == R_last_ik) || (NULL == R_last_kk) || /* technically cannot happen, but looks saner */ + (NULL == R_last_kj)) + { + /* R^{(k)}_{ij} = R^{(k-1)}_{ij} | N */ + *R_cur_ij = GNUNET_strdup (R_last_ij); + return; + } + + // $R^{(k)}_{ij} = N | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj} OR + // $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj} + + R_cur_r = NULL; + R_cur_l = NULL; + + // cache results from strcmp, we might need these many times + 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); + kk_kj_cmp = nullstrcmp (R_last_kk, R_last_kj); + + // Assign R_temp_(ik|kk|kj) to R_last[][] and remove epsilon as well + // as parentheses, so we can better compare the contents + R_temp_ik = remove_parentheses (remove_epsilon (R_last_ik)); + R_temp_kk = remove_parentheses (remove_epsilon (R_last_kk)); + R_temp_kj = remove_parentheses (remove_epsilon (R_last_kj)); + + clean_ik_kk_cmp = nullstrcmp (R_last_ik, R_temp_kk); + clean_kk_kj_cmp = nullstrcmp (R_temp_kk, R_last_kj); + + // construct R_cur_l (and, if necessary R_cur_r) + if (NULL != R_last_ij) + { + // Assign R_temp_ij to R_last_ij and remove epsilon as well + // as parentheses, so we can better compare the contents + R_temp_ij = remove_parentheses (remove_epsilon (R_last_ij)); + + if (0 == strcmp (R_temp_ij, R_temp_ik) && 0 == strcmp (R_temp_ik, R_temp_kk) + && 0 == strcmp (R_temp_kk, R_temp_kj)) + { + if (0 == strlen (R_temp_ij)) + { + R_cur_r = GNUNET_strdup (""); + } + else if ((0 == strncmp (R_last_ij, "(|", 2)) || + (0 == strncmp (R_last_ik, "(|", 2) && + 0 == strncmp (R_last_kj, "(|", 2))) + { + // a|(e|a)a*(e|a) = a* + // a|(e|a)(e|a)*(e|a) = a* + // (e|a)|aa*a = a* + // (e|a)|aa*(e|a) = a* + // (e|a)|(e|a)a*a = a* + // (e|a)|(e|a)a*(e|a) = a* + // (e|a)|(e|a)(e|a)*(e|a) = a* + if (GNUNET_YES == needs_parentheses (R_temp_ij)) + GNUNET_asprintf (&R_cur_r, "(%s)*", R_temp_ij); + else + GNUNET_asprintf (&R_cur_r, "%s*", R_temp_ij); + } + else + { + // a|aa*a = a+ + // a|(e|a)a*a = a+ + // a|aa*(e|a) = a+ + // a|(e|a)(e|a)*a = a+ + // a|a(e|a)*(e|a) = a+ + if (GNUNET_YES == needs_parentheses (R_temp_ij)) + GNUNET_asprintf (&R_cur_r, "(%s)+", R_temp_ij); + else + GNUNET_asprintf (&R_cur_r, "%s+", R_temp_ij); + } + } + else if (0 == ij_ik_cmp && 0 == clean_kk_kj_cmp && 0 != clean_ik_kk_cmp) + { + // a|ab*b = ab* + if (strlen (R_last_kk) < 1) + R_cur_r = GNUNET_strdup (R_last_ij); + else if (GNUNET_YES == needs_parentheses (R_temp_kk)) + GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last_ij, R_temp_kk); + else + GNUNET_asprintf (&R_cur_r, "%s%s*", R_last_ij, R_last_kk); + + R_cur_l = NULL; + } + else if (0 == ij_kj_cmp && 0 == clean_ik_kk_cmp && 0 != clean_kk_kj_cmp) + { + // a|bb*a = b*a + if (strlen (R_last_kk) < 1) + R_cur_r = GNUNET_strdup (R_last_kj); + else if (GNUNET_YES == needs_parentheses (R_temp_kk)) + GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk, R_last_kj); + else + GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last_kj); + + R_cur_l = NULL; + } + else if (0 == ij_ik_cmp && 0 == kk_kj_cmp && !has_epsilon (R_last_ij) && + has_epsilon (R_last_kk)) + { + // a|a(e|b)*(e|b) = a|ab* = a|a|ab|abb|abbb|... = ab* + if (needs_parentheses (R_temp_kk)) + GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last_ij, R_temp_kk); + else + GNUNET_asprintf (&R_cur_r, "%s%s*", R_last_ij, R_temp_kk); + + R_cur_l = NULL; + } + else if (0 == ij_kj_cmp && 0 == ik_kk_cmp && !has_epsilon (R_last_ij) && + has_epsilon (R_last_kk)) + { + // a|(e|b)(e|b)*a = a|b*a = a|a|ba|bba|bbba|... = b*a + if (needs_parentheses (R_temp_kk)) + GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk, R_last_ij); + else + GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last_ij); + + R_cur_l = NULL; + } + else + { + temp_a = (NULL == R_last_ij) ? NULL : GNUNET_strdup (R_last_ij); + temp_a = remove_parentheses (temp_a); + R_cur_l = temp_a; + } + + GNUNET_free_non_null (R_temp_ij); + } + else + { + // we have no left side + R_cur_l = NULL; + } + + // construct R_cur_r, if not already constructed + if (NULL == R_cur_r) + { + length = strlen (R_temp_kk) - strlen (R_last_ik); + + // a(ba)*bx = (ab)+x + if (length > 0 && NULL != R_last_kk && 0 < strlen (R_last_kk) && + NULL != R_last_kj && 0 < strlen (R_last_kj) && NULL != R_last_ik && + 0 < strlen (R_last_ik) && 0 == strkcmp (R_temp_kk, R_last_ik, length) && + 0 == strncmp (R_temp_kk, R_last_kj, length)) + { + temp_a = GNUNET_malloc (length + 1); + temp_b = GNUNET_malloc ((strlen (R_last_kj) - length) + 1); + + length_l = 0; + length_r = 0; + + for (cnt = 0; cnt < strlen (R_last_kj); cnt++) + { + if (cnt < length) + { + temp_a[length_l] = R_last_kj[cnt]; + length_l++; + } + else + { + temp_b[length_r] = R_last_kj[cnt]; + length_r++; + } + } + temp_a[length_l] = '\0'; + temp_b[length_r] = '\0'; + + // e|(ab)+ = (ab)* + if (NULL != R_cur_l && 0 == strlen (R_cur_l) && 0 == strlen (temp_b)) + { + GNUNET_asprintf (&R_cur_r, "(%s%s)*", R_last_ik, temp_a); + GNUNET_free (R_cur_l); + R_cur_l = NULL; + } + else + { + GNUNET_asprintf (&R_cur_r, "(%s%s)+%s", R_last_ik, temp_a, temp_b); + } + GNUNET_free (temp_a); + GNUNET_free (temp_b); + } + else if (0 == strcmp (R_temp_ik, R_temp_kk) && + 0 == strcmp (R_temp_kk, R_temp_kj)) + { + // (e|a)a*(e|a) = a* + // (e|a)(e|a)*(e|a) = a* + if (has_epsilon (R_last_ik) && has_epsilon (R_last_kj)) + { + if (needs_parentheses (R_temp_kk)) + GNUNET_asprintf (&R_cur_r, "(%s)*", R_temp_kk); + else + GNUNET_asprintf (&R_cur_r, "%s*", R_temp_kk); + } + // aa*a = a+a + else if (0 == clean_ik_kk_cmp && 0 == clean_kk_kj_cmp && + !has_epsilon (R_last_ik)) + { + if (needs_parentheses (R_temp_kk)) + GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_temp_kk, R_temp_kk); + else + GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_temp_kk, R_temp_kk); + } + // (e|a)a*a = a+ + // aa*(e|a) = a+ + // a(e|a)*(e|a) = a+ + // (e|a)a*a = a+ + else + { + eps_check = + (has_epsilon (R_last_ik) + has_epsilon (R_last_kk) + + has_epsilon (R_last_kj)); + + if (eps_check == 1) + { + if (needs_parentheses (R_temp_kk)) + GNUNET_asprintf (&R_cur_r, "(%s)+", R_temp_kk); + else + GNUNET_asprintf (&R_cur_r, "%s+", R_temp_kk); + } + } + } + // aa*b = a+b + // (e|a)(e|a)*b = a*b + else if (0 == strcmp (R_temp_ik, R_temp_kk)) + { + if (has_epsilon (R_last_ik)) + { + if (needs_parentheses (R_temp_kk)) + GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk, R_last_kj); + else + GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last_kj); + } + else + { + if (needs_parentheses (R_temp_kk)) + GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_temp_kk, R_last_kj); + else + GNUNET_asprintf (&R_cur_r, "%s+%s", R_temp_kk, R_last_kj); + } + } + // ba*a = ba+ + // b(e|a)*(e|a) = ba* + else if (0 == strcmp (R_temp_kk, R_temp_kj)) + { + if (has_epsilon (R_last_kj)) + { + if (needs_parentheses (R_temp_kk)) + GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last_ik, R_temp_kk); + else + GNUNET_asprintf (&R_cur_r, "%s%s*", R_last_ik, R_temp_kk); + } + else + { + if (needs_parentheses (R_temp_kk)) + GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_last_ik, R_temp_kk); + else + GNUNET_asprintf (&R_cur_r, "%s+%s", R_last_ik, R_temp_kk); + } + } + else + { + if (strlen (R_temp_kk) > 0) + { + if (needs_parentheses (R_temp_kk)) + { + GNUNET_asprintf (&R_cur_r, "%s(%s)*%s", R_last_ik, R_temp_kk, + R_last_kj); + } + else + { + GNUNET_asprintf (&R_cur_r, "%s%s*%s", R_last_ik, R_temp_kk, + R_last_kj); + } + } + else + { + GNUNET_asprintf (&R_cur_r, "%s%s", R_last_ik, R_last_kj); + } + } + } + + 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; + return; + } + + if (NULL != R_cur_l && NULL == R_cur_r) + { + *R_cur_ij = R_cur_l; + return; + } + + if (NULL == R_cur_l && NULL != R_cur_r) + { + *R_cur_ij = R_cur_r; + return; + } + + if (0 == nullstrcmp (R_cur_l, R_cur_r)) + { + *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 (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 + * Computation 3rd Edition" by Hopcroft, Motwani and Ullman. + * + * @param a automaton. */ static void -automaton_state_traverse (void *cls, struct GNUNET_REGEX_State *s, - GNUNET_REGEX_traverse_action action) +automaton_create_proofs (struct GNUNET_REGEX_Automaton *a) { - struct Transition *t; + unsigned int n = a->state_count; + struct GNUNET_REGEX_State *states[n]; + char *R_last[n][n]; + char *R_cur[n][n]; + char *temp; + struct GNUNET_REGEX_Transition *t; + char *complete_regex; + unsigned int i; + unsigned int j; + unsigned int k; + + + /* create depth-first numbering of the states, initializes 'state' */ + 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++) + { + for (j = 0; j < n; j++) + { + R_cur[i][j] = NULL; + R_last[i][j] = NULL; + } + for (t = states[i]->transitions_head; NULL != t; t = t->next) + { + j = t->to_state->proof_id; + if (NULL == R_last[i][j]) + GNUNET_asprintf (&R_last[i][j], "%c", t->label); + else + { + temp = R_last[i][j]; + GNUNET_asprintf (&R_last[i][j], "%s|%c", R_last[i][j], t->label); + GNUNET_free (temp); + } + } + if (NULL == R_last[i][i]) + GNUNET_asprintf (&R_last[i][i], ""); + else + { + temp = R_last[i][i]; + GNUNET_asprintf (&R_last[i][i], "(|%s)", R_last[i][i]); + GNUNET_free (temp); + } + } + for (i = 0; i < n; i++) + for (j = 0; j < n; j++) + if (needs_parentheses (R_last[i][j])) + { + temp = R_last[i][j]; + GNUNET_asprintf (&R_last[i][j], "(%s)", R_last[i][j]); + GNUNET_free (temp); + } + + /* Compute regular expressions of length "k" between each pair of states per induction */ + for (k = 0; k < n; k++) + { + for (i = 0; i < n; i++) + { + for (j = 0; j < n; j++) + { + // Basis for the recursion: + // $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj} + // R_last == R^{(k-1)}, R_cur == R^{(k)} + + // Create R_cur[i][j] and simplify the expression + automaton_create_proofs_simplify (R_last[i][j], R_last[i][k], + R_last[k][k], R_last[k][j], + &R_cur[i][j]); + } + } + + // set R_last = R_cur + for (i = 0; i < n; i++) + { + for (j = 0; j < n; j++) + { + GNUNET_free_non_null (R_last[i][j]); + R_last[i][j] = R_cur[i][j]; + R_cur[i][j] = NULL; + } + } + } + + // assign proofs and hashes + for (i = 0; i < n; i++) + { + if (NULL != R_last[a->start->proof_id][i]) + { + states[i]->proof = GNUNET_strdup (R_last[a->start->proof_id][i]); + GNUNET_CRYPTO_hash (states[i]->proof, strlen (states[i]->proof), + &states[i]->hash); + } + } - if (GNUNET_NO == s->marked) + // complete regex for whole DFA: union of all pairs (start state/accepting state(s)). + complete_regex = NULL; + for (i = 0; i < n; i++) { - s->marked = GNUNET_YES; - - if (NULL != action) - action (cls, s); + if (states[i]->accepting) + { + if (NULL == complete_regex && 0 < strlen (R_last[a->start->proof_id][i])) + { + GNUNET_asprintf (&complete_regex, "%s", R_last[a->start->proof_id][i]); + } + else if (NULL != R_last[a->start->proof_id][i] && + 0 < strlen (R_last[a->start->proof_id][i])) + { + temp = complete_regex; + GNUNET_asprintf (&complete_regex, "%s|%s", complete_regex, + R_last[a->start->proof_id][i]); + GNUNET_free (temp); + } + } + } + a->canonical_regex = complete_regex; - for (t = s->transitions_head; NULL != t; t = t->next) - automaton_state_traverse (cls, t->to_state, action); + // cleanup + for (i = 0; i < n; i++) + { + for (j = 0; j < n; j++) + GNUNET_free_non_null (R_last[i][j]); } } -/** - * Traverses the given automaton from it's start state, visiting all - * reachable states and calling 'action' on each one of them. - * - * @param a automaton. - * @param action action to be performed on each state. - */ -static void -automaton_traverse (void *cls, struct GNUNET_REGEX_Automaton *a, - GNUNET_REGEX_traverse_action action) -{ - struct GNUNET_REGEX_State *s; - - for (s = a->start; NULL != s; s = s->next) - s->marked = GNUNET_NO; - - automaton_state_traverse (cls, a->start, action); -} /** - * Creates a new DFA state based on a set of NFA states. Needs to be freed - * using automaton_destroy_state. + * Creates a new DFA state based on a set of NFA states. Needs to be freed using + * automaton_destroy_state. * * @param ctx context * @param nfa_states set of NFA states on which the DFA should be based on @@ -728,10 +1295,8 @@ dfa_state_create (struct GNUNET_REGEX_Context *ctx, char *name; int len = 0; struct GNUNET_REGEX_State *cstate; - struct Transition *ctran; - int insert = 1; - struct Transition *t; - int i; + struct GNUNET_REGEX_Transition *ctran; + unsigned int i; s = GNUNET_malloc (sizeof (struct GNUNET_REGEX_State)); s->id = ctx->state_id++; @@ -742,6 +1307,7 @@ dfa_state_create (struct GNUNET_REGEX_Context *ctx, s->index = -1; s->lowlink = -1; s->contained = 0; + s->proof = NULL; if (NULL == nfa_states) { @@ -777,24 +1343,11 @@ dfa_state_create (struct GNUNET_REGEX_Context *ctx, for (ctran = cstate->transitions_head; NULL != ctran; ctran = ctran->next) { if (0 != ctran->label) - { - insert = 1; - - for (t = s->transitions_head; NULL != t; t = t->next) - { - if (t->label == ctran->label) - { - insert = 0; - break; - } - } - - if (insert) - state_add_transition (ctx, s, ctran->label, NULL); - } + state_add_transition (ctx, s, ctran->label, NULL); } - // If the nfa_states contain an accepting state, the new dfa state is also accepting + // If the nfa_states contain an accepting state, the new dfa state is also + // accepting if (cstate->accepting) s->accepting = 1; } @@ -804,9 +1357,9 @@ dfa_state_create (struct GNUNET_REGEX_Context *ctx, return s; } + /** - * Move from the given state 's' to the next state on - * transition 'label' + * Move from the given state 's' to the next state on transition 'label' * * @param s starting state * @param label edge label to follow @@ -816,7 +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) @@ -838,8 +1391,8 @@ dfa_move (struct GNUNET_REGEX_State *s, const char label) /** - * Remove all unreachable states from DFA 'a'. Unreachable states - * are those states that are not reachable from the starting state. + * Remove all unreachable states from DFA 'a'. Unreachable states are those + * states that are not reachable from the starting state. * * @param a DFA automaton */ @@ -854,23 +1407,21 @@ 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 (NULL, a, 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) { s_next = s->next; if (GNUNET_NO == s->marked) - { - GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Removed state %s\n", s->name); automaton_remove_state (a, s); - } } } + /** - * Remove all dead states from the DFA 'a'. Dead states are those - * states that do not transition to any other state but themselfes. + * Remove all dead states from the DFA 'a'. Dead states are those states that do + * not transition to any other state but themselves. * * @param a DFA automaton */ @@ -878,7 +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); @@ -906,6 +1457,7 @@ dfa_remove_dead_states (struct GNUNET_REGEX_Automaton *a) } } + /** * Merge all non distinguishable states in the DFA 'a' * @@ -916,34 +1468,40 @@ static void dfa_merge_nondistinguishable_states (struct GNUNET_REGEX_Context *ctx, struct GNUNET_REGEX_Automaton *a) { - int i; + unsigned int i; 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; + unsigned int num_equal_edges; - change = 1; for (i = 0, s1 = a->states_head; i < a->state_count && NULL != s1; i++, s1 = s1->next) + { s1->marked = i; + } // Mark all pairs of accepting/!accepting states for (s1 = a->states_head; NULL != s1; s1 = s1->next) { - for (s2 = a->states_head; NULL != s2 && s1 != s2; s2 = s2->next) + for (s2 = a->states_head; NULL != s2; s2 = s2->next) { + table[s1->marked][s2->marked] = 0; + if ((s1->accepting && !s2->accepting) || (!s1->accepting && s2->accepting)) { table[s1->marked][s2->marked] = 1; } - else - table[s1->marked][s2->marked] = 0; } } + // Find all equal states + change = 1; while (0 != change) { change = 0; @@ -954,44 +1512,50 @@ dfa_merge_nondistinguishable_states (struct GNUNET_REGEX_Context *ctx, if (0 != table[s1->marked][s2->marked]) continue; + num_equal_edges = 0; for (t1 = s1->transitions_head; NULL != t1; t1 = t1->next) { for (t2 = s2->transitions_head; NULL != t2; t2 = t2->next) { - if (t1->label == t2->label && t1->to_state == t2->to_state && - (0 != table[t1->to_state->marked][t2->to_state->marked] || - 0 != table[t2->to_state->marked][t1->to_state->marked])) - { - table[s1->marked][s2->marked] = t1->label; - change = 1; - } - else if (t1->label != t2->label && t1->to_state != t2->to_state) + if (t1->label == t2->label) { - table[s1->marked][s2->marked] = -1; - change = 1; + num_equal_edges++; + if (0 != table[t1->to_state->marked][t2->to_state->marked] || + 0 != table[t2->to_state->marked][t1->to_state->marked]) + { + table[s1->marked][s2->marked] = t1->label != 0 ? t1->label : 1; + change = 1; + } } } } + if (num_equal_edges != s1->transition_count || + num_equal_edges != s2->transition_count) + { + // Make sure ALL edges of possible equal states are the same + table[s1->marked][s2->marked] = -2; + } } } } - struct GNUNET_REGEX_State *s2_next; - - for (i = 0, s1 = a->states_head; NULL != s1; s1 = s1->next) + // Merge states that are equal + for (s1 = a->states_head; NULL != s1; s1 = s1_next) { + s1_next = s1->next; for (s2 = a->states_head; NULL != s2 && s1 != s2; s2 = s2_next) { s2_next = s2->next; - if (s1 != s2 && table[s1->marked][s2->marked] == 0) + if (table[s1->marked][s2->marked] == 0) automaton_merge_states (ctx, a, s1, s2); } } } + /** - * Minimize the given DFA 'a' by removing all unreachable states, - * removing all dead states and merging all non distinguishable states + * Minimize the given DFA 'a' by removing all unreachable states, removing all + * dead states and merging all non distinguishable states * * @param ctx context * @param a DFA automaton @@ -1015,8 +1579,10 @@ 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. + * Creates a new NFA fragment. Needs to be cleared using + * automaton_fragment_clear. * * @param start starting state * @param end end state @@ -1035,7 +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); @@ -1047,6 +1613,7 @@ nfa_fragment_create (struct GNUNET_REGEX_State *start, return n; } + /** * Adds a list of states to the given automaton 'n'. * @@ -1084,6 +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. * @@ -1111,6 +1679,7 @@ nfa_state_create (struct GNUNET_REGEX_Context *ctx, int accepting) return s; } + /** * Calculates the NFA closure set for the given state. * @@ -1129,7 +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; @@ -1169,6 +1738,7 @@ nfa_closure_create (struct GNUNET_REGEX_Automaton *nfa, GNUNET_assert (0 == cls_check->len); GNUNET_free (cls_check); + // sort the states if (cls->len > 1) qsort (cls->states, cls->len, sizeof (struct GNUNET_REGEX_State *), state_compare); @@ -1176,6 +1746,7 @@ nfa_closure_create (struct GNUNET_REGEX_Automaton *nfa, return cls; } + /** * Calculates the closure set for the given set of states. * @@ -1188,16 +1759,15 @@ nfa_closure_create (struct GNUNET_REGEX_Automaton *nfa, */ static struct GNUNET_REGEX_StateSet * nfa_closure_set_create (struct GNUNET_REGEX_Automaton *nfa, - struct GNUNET_REGEX_StateSet *states, - const char label) + struct GNUNET_REGEX_StateSet *states, const char label) { struct GNUNET_REGEX_State *s; struct GNUNET_REGEX_StateSet *sset; struct GNUNET_REGEX_StateSet *cls; - int i; - int j; - int k; - int contains; + unsigned int i; + unsigned int j; + unsigned int k; + unsigned int contains; if (NULL == states) return NULL; @@ -1233,6 +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) * @@ -1243,28 +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*) * @@ -1274,12 +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) { @@ -1288,6 +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); @@ -1299,13 +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+) * @@ -1324,6 +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?) * @@ -1333,12 +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) { @@ -1347,6 +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); @@ -1356,16 +1934,17 @@ 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) + * Pops two NFA fragments (a, b) from the stack and adds a new NFA fragment that + * alternates between a and b (a|b) * * @param ctx context */ @@ -1374,13 +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); @@ -1395,15 +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 * @@ -1427,6 +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 * @@ -1442,11 +2025,11 @@ GNUNET_REGEX_context_init (struct GNUNET_REGEX_Context *ctx) } ctx->state_id = 0; ctx->transition_id = 0; - ctx->scc_id = 0; ctx->stack_head = NULL; ctx->stack_tail = NULL; } + /** * Construct an NFA by parsing the regex string of length 'len'. * @@ -1533,7 +2116,7 @@ GNUNET_REGEX_construct_nfa (const char *regex, const size_t len) case '*': if (atomcount == 0) { - error_msg = "Cannot append '+' to nothing"; + error_msg = "Cannot append '*' to nothing"; goto error; } nfa_add_star_op (&ctx); @@ -1557,6 +2140,7 @@ GNUNET_REGEX_construct_nfa (const char *regex, const size_t len) case 92: /* escape: \ */ regexp++; count++; + /* fall through! */ default: if (atomcount > 1) { @@ -1578,36 +2162,95 @@ GNUNET_REGEX_construct_nfa (const char *regex, const size_t len) for (; altcount > 0; altcount--) nfa_add_alternation (&ctx); - if (NULL != p) - GNUNET_free (p); + GNUNET_free_non_null (p); nfa = ctx.stack_tail; GNUNET_CONTAINER_DLL_remove (ctx.stack_head, ctx.stack_tail, nfa); - if (NULL != ctx.stack_head) { error_msg = "Creating the NFA failed. NFA stack was not empty!"; goto error; } + 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: - GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not parse regex\n"); + GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not parse regex: %s\n", regex); if (NULL != error_msg) GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "%s\n", error_msg); - if (NULL != p) - GNUNET_free (p); - while (NULL != ctx.stack_tail) + + GNUNET_free_non_null (p); + + while (NULL != (nfa = ctx.stack_head)) { - GNUNET_REGEX_automaton_destroy (ctx.stack_tail); - GNUNET_CONTAINER_DLL_remove (ctx.stack_head, ctx.stack_tail, - ctx.stack_tail); + GNUNET_CONTAINER_DLL_remove (ctx.stack_head, ctx.stack_tail, nfa); + GNUNET_REGEX_automaton_destroy (nfa); } + return NULL; } + +/** + * Create DFA states based on given 'nfa' and starting with 'dfa_state'. + * + * @param ctx context. + * @param nfa NFA automaton. + * @param dfa DFA automaton. + * @param dfa_state current dfa state, pass epsilon closure of first nfa state + * for starting. + */ +static void +construct_dfa_states (struct GNUNET_REGEX_Context *ctx, + struct GNUNET_REGEX_Automaton *nfa, + struct GNUNET_REGEX_Automaton *dfa, + struct GNUNET_REGEX_State *dfa_state) +{ + struct GNUNET_REGEX_Transition *ctran; + struct GNUNET_REGEX_State *state_iter; + struct GNUNET_REGEX_State *new_dfa_state; + struct GNUNET_REGEX_State *state_contains; + struct GNUNET_REGEX_StateSet *tmp; + struct GNUNET_REGEX_StateSet *nfa_set; + + for (ctran = dfa_state->transitions_head; NULL != ctran; ctran = ctran->next) + { + if (0 == ctran->label || NULL != ctran->to_state) + continue; + + tmp = nfa_closure_set_create (nfa, dfa_state->nfa_set, ctran->label); + nfa_set = nfa_closure_set_create (nfa, tmp, 0); + state_set_clear (tmp); + new_dfa_state = dfa_state_create (ctx, nfa_set); + state_contains = NULL; + for (state_iter = dfa->states_head; NULL != state_iter; + state_iter = state_iter->next) + { + if (0 == state_set_compare (state_iter->nfa_set, new_dfa_state->nfa_set)) + state_contains = state_iter; + } + + if (NULL == state_contains) + { + automaton_add_state (dfa, new_dfa_state); + ctran->to_state = new_dfa_state; + construct_dfa_states (ctx, nfa, dfa, new_dfa_state); + } + else + { + ctran->to_state = state_contains; + automaton_destroy_state (new_dfa_state); + } + } +} + + /** * Construct DFA for the given 'regex' of length 'len' * @@ -1622,14 +2265,7 @@ GNUNET_REGEX_construct_dfa (const char *regex, const size_t len) struct GNUNET_REGEX_Context ctx; struct GNUNET_REGEX_Automaton *dfa; struct GNUNET_REGEX_Automaton *nfa; - struct GNUNET_REGEX_StateSet *tmp; struct GNUNET_REGEX_StateSet *nfa_set; - struct GNUNET_REGEX_StateSet *dfa_stack; - struct Transition *ctran; - struct GNUNET_REGEX_State *dfa_state; - struct GNUNET_REGEX_State *new_dfa_state; - struct GNUNET_REGEX_State *state_contains; - struct GNUNET_REGEX_State *state_iter; GNUNET_REGEX_context_init (&ctx); @@ -1645,66 +2281,30 @@ GNUNET_REGEX_construct_dfa (const char *regex, const size_t len) dfa = GNUNET_malloc (sizeof (struct GNUNET_REGEX_Automaton)); dfa->type = DFA; + dfa->regex = GNUNET_strdup (regex); // Create DFA start state from epsilon closure - dfa_stack = GNUNET_malloc (sizeof (struct GNUNET_REGEX_StateSet)); nfa_set = nfa_closure_create (nfa, nfa->start, 0); dfa->start = dfa_state_create (&ctx, nfa_set); automaton_add_state (dfa, dfa->start); - GNUNET_array_append (dfa_stack->states, dfa_stack->len, dfa->start); - - // Create dfa states by combining nfa states - while (dfa_stack->len > 0) - { - dfa_state = dfa_stack->states[dfa_stack->len - 1]; - GNUNET_array_grow (dfa_stack->states, dfa_stack->len, dfa_stack->len - 1); - - for (ctran = dfa_state->transitions_head; NULL != ctran; - ctran = ctran->next) - { - if (0 != ctran->label && NULL == ctran->to_state) - { - tmp = nfa_closure_set_create (nfa, dfa_state->nfa_set, ctran->label); - nfa_set = nfa_closure_set_create (nfa, tmp, 0); - state_set_clear (tmp); - new_dfa_state = dfa_state_create (&ctx, nfa_set); - state_contains = NULL; - for (state_iter = dfa->states_head; NULL != state_iter; - state_iter = state_iter->next) - { - if (0 == - state_set_compare (state_iter->nfa_set, new_dfa_state->nfa_set)) - state_contains = state_iter; - } - if (NULL == state_contains) - { - automaton_add_state (dfa, new_dfa_state); - GNUNET_array_append (dfa_stack->states, dfa_stack->len, - new_dfa_state); - ctran->to_state = new_dfa_state; - } - else - { - ctran->to_state = state_contains; - automaton_destroy_state (new_dfa_state); - } - } - } - } + construct_dfa_states (&ctx, nfa, dfa, dfa->start); - GNUNET_free (dfa_stack); GNUNET_REGEX_automaton_destroy (nfa); + // Minimize DFA dfa_minimize (&ctx, dfa); - scc_tarjan (&ctx, dfa); + + // Create proofs for all states + automaton_create_proofs (dfa); return dfa; } + /** - * Free the memory allocated by constructing the GNUNET_REGEX_Automaton - * data structure. + * Free the memory allocated by constructing the GNUNET_REGEX_Automaton data + * structure. * * @param a automaton to be destroyed */ @@ -1717,6 +2317,9 @@ GNUNET_REGEX_automaton_destroy (struct GNUNET_REGEX_Automaton *a) if (NULL == a) return; + GNUNET_free_non_null (a->regex); + GNUNET_free_non_null (a->canonical_regex); + for (s = a->states_head; NULL != s;) { next_state = s->next; @@ -1727,101 +2330,6 @@ GNUNET_REGEX_automaton_destroy (struct GNUNET_REGEX_Automaton *a) GNUNET_free (a); } -/** - * 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) -{ - struct GNUNET_REGEX_State *s; - struct Transition *ctran; - char *s_acc = NULL; - char *s_tran = NULL; - 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 (p == NULL) - { - GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not open file for writing: %s", - filename); - return; - } - - start = "digraph G {\nrankdir=LR\n"; - fwrite (start, strlen (start), 1, p); - - for (s = a->states_head; NULL != s; s = s->next) - { - if (s->accepting) - { - GNUNET_asprintf (&s_acc, - "\"%s\" [shape=doublecircle, color=\"0.%i 0.8 0.95\"];\n", - s->name, s->scc_id); - fwrite (s_acc, strlen (s_acc), 1, p); - GNUNET_free (s_acc); - } - else - { - GNUNET_asprintf (&s_acc, "\"%s\" [color=\"0.%i 0.8 0.95\"];\n", s->name, - s->scc_id); - fwrite (s_acc, strlen (s_acc), 1, p); - GNUNET_free (s_acc); - } - - s->marked = 1; - - 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 has no state for transitioning\n", - s->id); - continue; - } - - if (ctran->label == 0) - { - GNUNET_asprintf (&s_tran, - "\"%s\" -> \"%s\" [label = \"epsilon\", color=\"0.%i 0.8 0.95\"];\n", - s->name, ctran->to_state->name, s->scc_id); - } - else - { - GNUNET_asprintf (&s_tran, - "\"%s\" -> \"%s\" [label = \"%c\", color=\"0.%i 0.8 0.95\"];\n", - s->name, ctran->to_state->name, ctran->label, - s->scc_id); - } - - fwrite (s_tran, strlen (s_tran), 1, p); - GNUNET_free (s_tran); - } - } - - end = "\n}\n"; - fwrite (end, strlen (end), 1, p); - fclose (p); -} /** * Evaluates the given string using the given DFA automaton @@ -1846,6 +2354,10 @@ evaluate_dfa (struct GNUNET_REGEX_Automaton *a, const char *string) s = a->start; + // If the string is empty but the starting state is accepting, we accept. + if ((NULL == string || 0 == strlen (string)) && s->accepting) + return 0; + for (strp = string; NULL != strp && *strp; strp++) { s = dfa_move (s, *strp); @@ -1859,6 +2371,7 @@ evaluate_dfa (struct GNUNET_REGEX_Automaton *a, const char *string) return 1; } + /** * Evaluates the given string using the given NFA automaton * @@ -1874,7 +2387,7 @@ evaluate_nfa (struct GNUNET_REGEX_Automaton *a, const char *string) struct GNUNET_REGEX_State *s; struct GNUNET_REGEX_StateSet *sset; struct GNUNET_REGEX_StateSet *new_sset; - int i; + unsigned int i; int result; if (NFA != a->type) @@ -1884,8 +2397,11 @@ evaluate_nfa (struct GNUNET_REGEX_Automaton *a, const char *string) return -1; } + // If the string is empty but the starting state is accepting, we accept. + if ((NULL == string || 0 == strlen (string)) && a->start->accepting) + return 0; + result = 1; - strp = string; sset = nfa_closure_create (a, a->start, 0); for (strp = string; NULL != strp && *strp; strp++) @@ -1910,6 +2426,7 @@ evaluate_nfa (struct GNUNET_REGEX_Automaton *a, const char *string) return result; } + /** * Evaluates the given 'string' against the given compiled regex * @@ -1941,9 +2458,31 @@ GNUNET_REGEX_eval (struct GNUNET_REGEX_Automaton *a, const char *string) return result; } + +/** + * Get the canonical regex of the given automaton. + * When constructing the automaton a proof is computed for each state, + * consisting of the regular expression leading to this state. A complete + * regex for the automaton can be computed by combining these proofs. + * As of now this function is only useful for testing. + * + * @param a automaton for which the canonical regex should be returned. + * + * @return + */ +const char * +GNUNET_REGEX_get_canonical_regex (struct GNUNET_REGEX_Automaton *a) +{ + if (NULL == a) + return NULL; + + return a->canonical_regex; +} + + /** - * Get the first key for the given 'input_string'. This hashes - * the first x bits of the 'input_strings'. + * Get the first key for the given 'input_string'. This hashes the first x bits + * of the 'input_string'. * * @param input_string string. * @param string_len length of the 'input_string'. @@ -1952,14 +2491,13 @@ GNUNET_REGEX_eval (struct GNUNET_REGEX_Automaton *a, const char *string) * @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, - 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; - size = string_len < initial_bits ? string_len : initial_bits; + size = string_len < INITIAL_BITS ? string_len : INITIAL_BITS; if (NULL == input_string) { @@ -1972,52 +2510,142 @@ GNUNET_REGEX_get_first_key (const char *input_string, 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 GNUNET_HashCode *key) +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; } + /** - * Get all edges leaving state 's'. + * Recursive helper function for iterate_initial_edges. Will call iterator + * function for each initial state. * - * @param s state. - * @param edges all edges leaving 's'. + * @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. * - * @return number of 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. */ -static unsigned int -state_get_edges (struct GNUNET_REGEX_State *s, struct GNUNET_REGEX_Edge *edges) +void +iterate_initial_edges (struct GNUNET_REGEX_Automaton *a, + const unsigned int initial_len, + GNUNET_REGEX_KeyIterator iterator, void *iterator_cls) { - struct Transition *t; - unsigned int count; + char *consumed_string; + char *temp; + struct GNUNET_REGEX_State *s; + unsigned int cur_len; - if (NULL == s) - return 0; + if (1 > initial_len) + return; - count = 0; + consumed_string = NULL; + s = a->start; + cur_len = 0; - for (t = s->transitions_head; NULL != t; t = t->next) + if (1 == s->transition_count) { - if (NULL != t->to_state) + do { - edges[count].label = &t->label; - edges[count].destination = t->to_state->hash; - count++; + 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); } - return 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. @@ -2027,33 +2655,29 @@ state_get_edges (struct GNUNET_REGEX_State *s, struct GNUNET_REGEX_Edge *edges) * @param iterator_cls closure. */ static void -iterate_edge (struct GNUNET_REGEX_State *s, - GNUNET_REGEX_KeyIterator iterator, +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; - if (GNUNET_NO == s->marked) + if (GNUNET_YES != s->marked) { s->marked = GNUNET_YES; num_edges = state_get_edges (s, edges); - iterator (iterator_cls, - &s->hash, - NULL, - - 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. @@ -2067,5 +2691,11 @@ GNUNET_REGEX_iterate_all_edges (struct GNUNET_REGEX_Automaton *a, GNUNET_REGEX_KeyIterator iterator, void *iterator_cls) { + struct GNUNET_REGEX_State *s; + + 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); }