X-Git-Url: https://git.librecmc.org/?a=blobdiff_plain;f=src%2Fregex%2Fregex.c;h=c8b8ad3faedb680268ff3256c91ea8b868a9a449;hb=adbde8be4b6ee0a854177237391936faa8ca61c2;hp=a524ace297a25ba4141574eac0eb827cf3811a9a;hpb=ac43047941ad44cb3cd408b557a9abdb73d34f06;p=oweals%2Fgnunet.git diff --git a/src/regex/regex.c b/src/regex/regex.c index a524ace29..c8b8ad3fa 100644 --- a/src/regex/regex.c +++ b/src/regex/regex.c @@ -24,336 +24,1290 @@ */ #include "platform.h" #include "gnunet_container_lib.h" +#include "gnunet_crypto_lib.h" #include "gnunet_regex_lib.h" -#include "regex.h" +#include "regex_internal.h" + /** - * 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 { + /** + * Unique state id. + */ unsigned int state_id; + + /** + * Unique transition id. + */ unsigned int transition_id; /** - * DLL of GNUNET_REGEX_Automaton's used as a stack + * DLL of GNUNET_REGEX_Automaton's used as a stack. */ struct GNUNET_REGEX_Automaton *stack_head; + + /** + * DLL of GNUNET_REGEX_Automaton's used as a stack. + */ struct GNUNET_REGEX_Automaton *stack_tail; }; -enum GNUNET_REGEX_automaton_type + +/** + * Set of states. + */ +struct GNUNET_REGEX_StateSet { - NFA, - DFA + /** + * Array of states. + */ + struct GNUNET_REGEX_State **states; + + /** + * Length of the 'states' array. + */ + unsigned int len; }; + +/* + * Debug helper functions + */ + /** - * Automaton representation + * 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); + + +/** + * Print information of the given state 's'. + * + * @param s state for which debug information should be printed. */ -struct GNUNET_REGEX_Automaton +void +debug_print_state (struct GNUNET_REGEX_State *s) { - struct GNUNET_REGEX_Automaton *prev; - struct GNUNET_REGEX_Automaton *next; + char *proof; - struct State *start; - struct State *end; + if (NULL == s->proof) + proof = "NULL"; + else + proof = s->proof; - unsigned int state_count; - struct State *states_head; - struct State *states_tail; + 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); + + GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Transitions:\n"); + debug_print_transitions (s); +} - enum GNUNET_REGEX_automaton_type type; -}; /** - * A state. Can be used in DFA and NFA automatons. + * 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 State +void +debug_print_states (struct GNUNET_REGEX_Automaton *a) { - struct State *prev; - struct State *next; + struct GNUNET_REGEX_State *s; - unsigned int id; - int accepting; - int marked; - char *name; + for (s = a->states_head; NULL != s; s = s->next) + debug_print_state (s); +} - unsigned int transition_count; - struct Transition *transitions_head; - struct Transition *transitions_tail; - struct StateSet *nfa_set; -}; +/** + * Print debug information for given transition 't'. + * + * @param t transition for which to print debug info. + */ +void +debug_print_transition (struct GNUNET_REGEX_Transition *t) +{ + char *to_state; + char *from_state; + char label; + + if (NULL == t) + return; + + if (0 == t->label) + label = '0'; + else + label = t->label; + + if (NULL == t->to_state) + to_state = "NULL"; + else + to_state = t->to_state->name; + + 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); +} + + +void +debug_print_transitions (struct GNUNET_REGEX_State *s) +{ + struct GNUNET_REGEX_Transition *t; + + for (t = s->transitions_head; NULL != t; t = t->next) + debug_print_transition (t); +} + /** - * Transition between two states. Each state can have 0-n transitions. - * If literal is 0, this is considered to be an epsilon transition. + * Adds a transition from one state to another on 'label'. Does not add + * duplicate states. + * + * @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 */ -struct Transition +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 *prev; - struct Transition *next; + int is_dup; + struct GNUNET_REGEX_Transition *t; + struct GNUNET_REGEX_Transition *oth; + + if (NULL == from_state) + { + GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not create Transition.\n"); + return; + } + + // Do not add duplicate state transitions + is_dup = GNUNET_NO; + for (t = from_state->transitions_head; NULL != t; t = t->next) + { + if (t->to_state == to_state && t->label == label && + t->from_state == from_state) + { + is_dup = GNUNET_YES; + break; + } + } + + if (GNUNET_YES == is_dup) + return; + + // 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); +} - unsigned int id; - char literal; - struct State *state; -}; /** - * Set of states + * 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'. */ -struct StateSet +static void +state_remove_transition (struct GNUNET_REGEX_State *state, + struct GNUNET_REGEX_Transition *transition) { - /** - * Array of states - */ - struct State **states; - unsigned int len; -}; + 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); +} + /** - * Initialize a new context + * Compare two states. Used for sorting. * - * @param ctx context + * @param a first state + * @param b second state + * + * @return an integer less than, equal to, or greater than zero + * if the first argument is considered to be respectively + * less than, equal to, or greater than the second. + */ +static int +state_compare (const void *a, const void *b) +{ + struct GNUNET_REGEX_State **s1; + struct GNUNET_REGEX_State **s2; + + s1 = (struct GNUNET_REGEX_State **) a; + s2 = (struct GNUNET_REGEX_State **) 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! + * + * @param sset1 first state set + * @param sset2 second state set + * + * @return an integer less than, equal to, or greater than zero + * if the first argument is considered to be respectively + * less than, equal to, or greater than the second. + */ +static int +state_set_compare (struct GNUNET_REGEX_StateSet *sset1, + struct GNUNET_REGEX_StateSet *sset2) +{ + int result; + unsigned int i; + + if (NULL == sset1 || NULL == sset2) + return 1; + + result = sset1->len - sset2->len; + + for (i = 0; i < sset1->len; i++) + { + if (0 != result) + break; + + result = state_compare (&sset1->states[i], &sset2->states[i]); + } + return result; +} + + +/** + * Clears the given StateSet 'set' + * + * @param set set to be cleared */ static void -GNUNET_REGEX_context_init (struct GNUNET_REGEX_Context *ctx) +state_set_clear (struct GNUNET_REGEX_StateSet *set) { - if (NULL == ctx) + if (NULL != set) { - GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Context was NULL!"); + GNUNET_free_non_null (set->states); + GNUNET_free (set); + } +} + + +/** + * Clears an automaton fragment. Does not destroy the states inside the + * automaton. + * + * @param a automaton to be cleared + */ +static void +automaton_fragment_clear (struct GNUNET_REGEX_Automaton *a) +{ + if (NULL == a) return; + + a->start = NULL; + a->end = NULL; + a->states_head = NULL; + a->states_tail = NULL; + a->state_count = 0; + GNUNET_free (a); +} + + +/** + * Frees the memory used by State 's' + * + * @param s state that should be destroyed + */ +static void +automaton_destroy_state (struct GNUNET_REGEX_State *s) +{ + struct GNUNET_REGEX_Transition *t; + struct GNUNET_REGEX_Transition *next_t; + + if (NULL == s) + return; + + GNUNET_free_non_null (s->name); + GNUNET_free_non_null (s->proof); + + 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); } - ctx->state_id = 0; - ctx->transition_id = 0; - ctx->stack_head = NULL; - ctx->stack_tail = NULL; + + state_set_clear (s->nfa_set); + + 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. + * + * @param a automaton + * @param s state to remove + */ static void -debug_print_state (struct State *s) +automaton_remove_state (struct GNUNET_REGEX_Automaton *a, + struct GNUNET_REGEX_State *s) { - GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, - "State %i: %s marked: %i accepting: %i\n", s->id, s->name, - s->marked, s->accepting); + struct GNUNET_REGEX_State *ss; + struct GNUNET_REGEX_State *s_check; + struct GNUNET_REGEX_Transition *t_check; + + if (NULL == a || NULL == s) + return; + + // remove state + ss = s; + GNUNET_CONTAINER_DLL_remove (a->states_head, a->states_tail, s); + a->state_count--; + + // remove all transitions leading to this state + 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) + { + if (t_check->to_state == ss) + { + GNUNET_CONTAINER_DLL_remove (s_check->transitions_head, + s_check->transitions_tail, t_check); + s_check->transition_count--; + } + } + } + + automaton_destroy_state (ss); } + +/** + * Merge two states into one. Will merge 's1' and 's2' into 's1' and destroy + * 's2'. + * + * @param ctx context + * @param a automaton + * @param s1 first state + * @param s2 second state, will be destroyed + */ static void -debug_print_states (struct StateSet *sset) +automaton_merge_states (struct GNUNET_REGEX_Context *ctx, + struct GNUNET_REGEX_Automaton *a, + struct GNUNET_REGEX_State *s1, + struct GNUNET_REGEX_State *s2) { - struct State *s; - int i; + struct GNUNET_REGEX_State *s_check; + struct GNUNET_REGEX_Transition *t_check; + struct GNUNET_REGEX_Transition *t; + struct GNUNET_REGEX_Transition *t_next; + char *new_name; + int is_dup; - for (i = 0; i < sset->len; i++) + GNUNET_assert (NULL != ctx && NULL != a && NULL != s1 && NULL != s2); + + 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) { - s = sset->states[i]; - debug_print_state (s); + for (t_check = s_check->transitions_head; NULL != t_check; t_check = t_next) + { + 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) + { + 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 = s1->name; + GNUNET_asprintf (&s1->name, "{%s,%s}", new_name, s2->name); + GNUNET_free (new_name); + + // remove state + GNUNET_CONTAINER_DLL_remove (a->states_head, a->states_tail, s2); + a->state_count--; + automaton_destroy_state (s2); +} + + +/** + * 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 + */ +static void +automaton_add_state (struct GNUNET_REGEX_Automaton *a, + struct GNUNET_REGEX_State *s) +{ + GNUNET_CONTAINER_DLL_insert (a->states_head, a->states_tail, s); + a->state_count++; } + +/** + * 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 -debug_print_transitions (struct State *s) +automaton_state_traverse (struct GNUNET_REGEX_State *s, unsigned int *count, + GNUNET_REGEX_traverse_action action, void *action_cls) { - struct Transition *t; - char *state; - char literal; + 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) { - if (0 == t->literal) - literal = '0'; - else - literal = t->literal; + 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 == t->state) - state = "NULL"; + 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 - state = t->state->name; + { + temp_a = (NULL == R_last_ij) ? NULL : GNUNET_strdup (R_last_ij); + temp_a = remove_parentheses (temp_a); + R_cur_l = temp_a; + } - GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Transition %i: On %c to %s\n", t->id, - literal, state); + GNUNET_free_non_null (R_temp_ij); + } + else + { + // we have no left side + R_cur_l = NULL; } -} -static int -state_compare (const void *a, const void *b) -{ - struct State **s1; - struct State **s2; + // construct R_cur_r, if not already constructed + if (NULL == R_cur_r) + { + length = strlen (R_temp_kk) - strlen (R_last_ik); - s1 = (struct State **) a; - s2 = (struct State **) b; + // 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); - return (*s1)->id - (*s2)->id; -} + length_l = 0; + length_r = 0; -/** - * 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 - * - * @return 0 if they are equal, non 0 otherwise - */ -static int -state_set_compare (struct StateSet *sset1, struct StateSet *sset2) -{ - int i; + 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'; - if (sset1->len != sset2->len) - return 1; + // 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)); - for (i = 0; i < sset1->len; i++) - { - if (sset1->states[i]->id != sset2->states[i]->id) + 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 { - return 1; + 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); + } } } - return 0; -} -/** - * Checks if 'elem' is contained in 'set' - * - * @param set set of states - * @param elem state - * - * @return GNUNET_YES if 'set' contains 'elem, GNUNET_NO otherwise - */ -static int -state_set_contains (struct StateSet *set, struct State *elem) -{ - struct State *s; - int i; + GNUNET_free_non_null (R_temp_ik); + GNUNET_free_non_null (R_temp_kk); + GNUNET_free_non_null (R_temp_kj); - for (i = 0; i < set->len; i++) + if (NULL == R_cur_l && NULL == R_cur_r) { - s = set->states[i]; - if (0 == memcmp (s, elem, sizeof (struct State))) - return GNUNET_YES; + *R_cur_ij = NULL; + return; } - return GNUNET_NO; -} -/** - * Clears the given StateSet 'set' - * - * @param set set to be cleared - */ -static void -state_set_clear (struct StateSet *set) -{ - if (NULL != set) + if (NULL != R_cur_l && NULL == R_cur_r) { - if (NULL != set->states) - GNUNET_free (set->states); - GNUNET_free (set); + *R_cur_ij = R_cur_l; + return; } -} - -/** - * Adds a transition from one state to another on 'literal' - * - * @param ctx context - * @param from_state starting state for the transition - * @param literal transition label - * @param to_state state to where the transition should point to - */ -static void -add_transition (struct GNUNET_REGEX_Context *ctx, struct State *from_state, - const char literal, struct State *to_state) -{ - struct Transition *t; - if (NULL == from_state) + if (NULL == R_cur_l && NULL != R_cur_r) { - GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not create Transition.\n"); + *R_cur_ij = R_cur_r; return; } - t = GNUNET_malloc (sizeof (struct Transition)); + if (0 == nullstrcmp (R_cur_l, R_cur_r)) + { + *R_cur_ij = R_cur_l; + GNUNET_free (R_cur_r); + return; + } - t->id = ctx->transition_id++; - t->literal = literal; - t->state = to_state; + GNUNET_asprintf (R_cur_ij, "(%s|%s)", R_cur_l, R_cur_r); - GNUNET_CONTAINER_DLL_insert (from_state->transitions_head, - from_state->transitions_tail, t); + GNUNET_free (R_cur_l); + GNUNET_free (R_cur_r); } -/** - * Clears an automaton fragment. Does not destroy the states inside - * the automaton. - * - * @param a automaton to be cleared - */ -static void -automaton_fragment_clear (struct GNUNET_REGEX_Automaton *a) -{ - a->start = NULL; - a->end = NULL; - a->states_head = NULL; - a->states_tail = NULL; - GNUNET_free (a); -} /** - * Frees the memory used by State 's' + * 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 s state that should be destroyed + * @param a automaton. */ static void -automaton_destroy_state (struct State *s) +automaton_create_proofs (struct GNUNET_REGEX_Automaton *a) { - struct Transition *t; - struct Transition *next_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]); + } + } - if (NULL != s->name) - GNUNET_free (s->name); + // 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; + } + } + } - for (t = s->transitions_head; NULL != t;) + // assign proofs and hashes + for (i = 0; i < n; i++) { - next_t = t->next; - GNUNET_CONTAINER_DLL_remove (s->transitions_head, s->transitions_tail, t); - GNUNET_free (t); - t = next_t; + 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); + } } - state_set_clear (s->nfa_set); + // complete regex for whole DFA: union of all pairs (start state/accepting state(s)). + complete_regex = NULL; + for (i = 0; i < n; i++) + { + 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; - GNUNET_free (s); + // cleanup + for (i = 0; i < n; i++) + { + for (j = 0; j < n; j++) + GNUNET_free_non_null (R_last[i][j]); + } } + /** - * 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 * * @return new DFA state */ -static struct State * -dfa_state_create (struct GNUNET_REGEX_Context *ctx, struct StateSet *nfa_states) +static struct GNUNET_REGEX_State * +dfa_state_create (struct GNUNET_REGEX_Context *ctx, + struct GNUNET_REGEX_StateSet *nfa_states) { - struct State *s; + struct GNUNET_REGEX_State *s; char *name; int len = 0; - struct State *cstate; - struct Transition *ctran; - int insert = 1; - struct Transition *t; - int i; + struct GNUNET_REGEX_State *cstate; + struct GNUNET_REGEX_Transition *ctran; + unsigned int i; - s = GNUNET_malloc (sizeof (struct State)); + s = GNUNET_malloc (sizeof (struct GNUNET_REGEX_State)); s->id = ctx->state_id++; s->accepting = 0; s->marked = 0; s->name = NULL; + s->scc_id = 0; + s->index = -1; + s->lowlink = -1; + s->contained = 0; + s->proof = NULL; if (NULL == nfa_states) { @@ -385,28 +1339,15 @@ dfa_state_create (struct GNUNET_REGEX_Context *ctx, struct StateSet *nfa_states) name = NULL; } - // Add a transition for each distinct literal to NULL state + // Add a transition for each distinct label to NULL state for (ctran = cstate->transitions_head; NULL != ctran; ctran = ctran->next) { - if (0 != ctran->literal) - { - insert = 1; - - for (t = s->transitions_head; NULL != t; t = t->next) - { - if (t->literal == ctran->literal) - { - insert = 0; - break; - } - } - - if (insert) - add_transition (ctx, s, ctran->literal, NULL); - } + if (0 != ctran->label) + 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; } @@ -416,11 +1357,20 @@ dfa_state_create (struct GNUNET_REGEX_Context *ctx, struct StateSet *nfa_states) return s; } -static struct State * -dfa_move (struct State *s, const char literal) + +/** + * Move from the given state 's' to the next state on transition 'label' + * + * @param s starting state + * @param label edge label to follow + * + * @return new state or NULL, if transition on label not possible + */ +static struct GNUNET_REGEX_State * +dfa_move (struct GNUNET_REGEX_State *s, const char label) { - struct Transition *t; - struct State *new_s; + struct GNUNET_REGEX_Transition *t; + struct GNUNET_REGEX_State *new_s; if (NULL == s) return NULL; @@ -429,9 +1379,9 @@ dfa_move (struct State *s, const char literal) for (t = s->transitions_head; NULL != t; t = t->next) { - if (literal == t->literal) + if (label == t->label) { - new_s = t->state; + new_s = t->to_state; break; } } @@ -439,46 +1389,47 @@ dfa_move (struct State *s, const char literal) return new_s; } + +/** + * Remove all unreachable states from DFA 'a'. Unreachable states are those + * states that are not reachable from the starting state. + * + * @param a DFA automaton + */ static void dfa_remove_unreachable_states (struct GNUNET_REGEX_Automaton *a) { - /* - * struct StateSet *unreachable; - * struct State *stack[a->size]; - * struct State *s; - * - * unreachable = GNUNET_malloc (sizeof (struct StateSet)); - * - * // 1. add all states to unreachable set - * for (s = a->states_head; NULL != s; s = s->next) - * { - * GNUNET_array_append (unreachable->states, unreachable->len; s); - * } - * - * // 2. traverse dfa from start state and remove every visited state from unreachable set - * s = a->start; - * // push a->start - * while (stack->len > 0) - * { - * s = stack->states[stack->len - 1]; - * GNUNET_array_grow (stack->states; stack->len; stack->len-1); - * GNUNET_array_ - * for (t = s->transitions_head; NULL != t; t = t->next) - * { - * - * } - * } - * // 3. delete all states that are still in the unreachable set - */ + struct GNUNET_REGEX_State *s; + struct GNUNET_REGEX_State *s_next; + + // 1. unmark all states + for (s = a->states_head; NULL != s; s = s->next) + s->marked = GNUNET_NO; + + // 2. traverse dfa from start state and mark all visited states + 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) + 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 themselves. + * + * @param a DFA automaton + */ static void dfa_remove_dead_states (struct GNUNET_REGEX_Automaton *a) { - struct State *s; - struct State *s_check; - struct Transition *t; - struct Transition *t_check; + struct GNUNET_REGEX_State *s; + struct GNUNET_REGEX_Transition *t; int dead; GNUNET_assert (DFA == a->type); @@ -491,7 +1442,7 @@ dfa_remove_dead_states (struct GNUNET_REGEX_Automaton *a) dead = 1; for (t = s->transitions_head; NULL != t; t = t->next) { - if (NULL != t->state && t->state != s) + if (NULL != t->to_state && t->to_state != s) { dead = 0; break; @@ -502,33 +1453,120 @@ dfa_remove_dead_states (struct GNUNET_REGEX_Automaton *a) continue; // state s is dead, remove it - // 1. remove all transitions to this state - for (s_check = a->states_head; NULL != s_check; s_check = s_check->next) + automaton_remove_state (a, s); + } +} + + +/** + * Merge all non distinguishable states in the DFA 'a' + * + * @param ctx context + * @param a DFA automaton + */ +static void +dfa_merge_nondistinguishable_states (struct GNUNET_REGEX_Context *ctx, + struct GNUNET_REGEX_Automaton *a) +{ + unsigned int i; + int table[a->state_count][a->state_count]; + struct GNUNET_REGEX_State *s1; + struct GNUNET_REGEX_State *s2; + 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; + + 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; s2 = s2->next) + { + table[s1->marked][s2->marked] = 0; + + if ((s1->accepting && !s2->accepting) || + (!s1->accepting && s2->accepting)) + { + table[s1->marked][s2->marked] = 1; + } + } + } + + // Find all equal states + change = 1; + while (0 != change) + { + change = 0; + for (s1 = a->states_head; NULL != s1; s1 = s1->next) { - for (t_check = s_check->transitions_head; NULL != t_check; - t_check = t_check->next) + for (s2 = a->states_head; NULL != s2 && s1 != s2; s2 = s2->next) { - if (t_check->state == s) + 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) + { + 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) { - GNUNET_CONTAINER_DLL_remove (s_check->transitions_head, - s_check->transitions_tail, t_check); + // Make sure ALL edges of possible equal states are the same + table[s1->marked][s2->marked] = -2; } } } - // 2. remove state - GNUNET_CONTAINER_DLL_remove (a->states_head, a->states_tail, s); } -} - -static void -dfa_merge_nondistinguishable_states (struct GNUNET_REGEX_Automaton *a) -{ + // 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 (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 + * + * @param ctx context + * @param a DFA automaton + */ static void -dfa_minimize (struct GNUNET_REGEX_Automaton *a) +dfa_minimize (struct GNUNET_REGEX_Context *ctx, + struct GNUNET_REGEX_Automaton *a) { + if (NULL == a) + return; + GNUNET_assert (DFA == a->type); // 1. remove unreachable states @@ -538,11 +1576,13 @@ dfa_minimize (struct GNUNET_REGEX_Automaton *a) dfa_remove_dead_states (a); // 3. Merge nondistinguishable states - dfa_merge_nondistinguishable_states (a); + 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 @@ -550,7 +1590,8 @@ dfa_minimize (struct GNUNET_REGEX_Automaton *a) * @return new NFA fragment */ static struct GNUNET_REGEX_Automaton * -nfa_fragment_create (struct State *start, struct State *end) +nfa_fragment_create (struct GNUNET_REGEX_State *start, + struct GNUNET_REGEX_State *end) { struct GNUNET_REGEX_Automaton *n; @@ -560,11 +1601,11 @@ nfa_fragment_create (struct State *start, struct State *end) n->start = NULL; n->end = NULL; - if (NULL == start && NULL == end) + if (NULL == start || NULL == end) return n; - GNUNET_CONTAINER_DLL_insert (n->states_head, n->states_tail, end); - GNUNET_CONTAINER_DLL_insert (n->states_head, n->states_tail, start); + automaton_add_state (n, end); + automaton_add_state (n, start); n->start = start; n->end = end; @@ -572,6 +1613,7 @@ nfa_fragment_create (struct State *start, struct State *end) return n; } + /** * Adds a list of states to the given automaton 'n'. * @@ -580,9 +1622,12 @@ nfa_fragment_create (struct State *start, struct State *end) * @param states_tail tail of the DLL of states */ static void -nfa_add_states (struct GNUNET_REGEX_Automaton *n, struct State *states_head, - struct State *states_tail) +nfa_add_states (struct GNUNET_REGEX_Automaton *n, + struct GNUNET_REGEX_State *states_head, + struct GNUNET_REGEX_State *states_tail) { + struct GNUNET_REGEX_State *s; + if (NULL == n || NULL == states_head) { GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not add states\n"); @@ -601,8 +1646,12 @@ nfa_add_states (struct GNUNET_REGEX_Automaton *n, struct State *states_head, n->states_tail->next = states_head; n->states_tail = states_tail; } + + for (s = states_head; NULL != s; s = s->next) + n->state_count++; } + /** * Creates a new NFA state. Needs to be freed using automaton_destroy_state. * @@ -611,47 +1660,57 @@ nfa_add_states (struct GNUNET_REGEX_Automaton *n, struct State *states_head, * * @return new NFA state */ -static struct State * +static struct GNUNET_REGEX_State * nfa_state_create (struct GNUNET_REGEX_Context *ctx, int accepting) { - struct State *s; + struct GNUNET_REGEX_State *s; - s = GNUNET_malloc (sizeof (struct State)); + s = GNUNET_malloc (sizeof (struct GNUNET_REGEX_State)); s->id = ctx->state_id++; s->accepting = accepting; s->marked = 0; + s->contained = 0; + s->index = -1; + s->lowlink = -1; + s->scc_id = 0; s->name = NULL; GNUNET_asprintf (&s->name, "s%i", s->id); return s; } + /** - * Calculates the NFA closure set for the given state + * Calculates the NFA closure set for the given state. * + * @param nfa the NFA containing 's' * @param s starting point state - * @param literal transitioning literal on which to base the closure on, + * @param label transitioning label on which to base the closure on, * pass 0 for epsilon transition * - * @return nfa closure on 'literal' (epsilon closure if 'literal' is 0) + * @return sorted nfa closure on 'label' (epsilon closure if 'label' is 0) */ -static struct StateSet * -nfa_closure_create (struct State *s, const char literal) +static struct GNUNET_REGEX_StateSet * +nfa_closure_create (struct GNUNET_REGEX_Automaton *nfa, + struct GNUNET_REGEX_State *s, const char label) { - struct StateSet *cls; - struct StateSet *cls_check; - struct State *clsstate; - struct State *currentstate; - struct Transition *ctran; + struct GNUNET_REGEX_StateSet *cls; + struct GNUNET_REGEX_StateSet *cls_check; + struct GNUNET_REGEX_State *clsstate; + struct GNUNET_REGEX_State *currentstate; + struct GNUNET_REGEX_Transition *ctran; if (NULL == s) return NULL; - cls = GNUNET_malloc (sizeof (struct StateSet)); - cls_check = GNUNET_malloc (sizeof (struct StateSet)); + cls = GNUNET_malloc (sizeof (struct GNUNET_REGEX_StateSet)); + cls_check = GNUNET_malloc (sizeof (struct GNUNET_REGEX_StateSet)); + + for (clsstate = nfa->states_head; NULL != clsstate; clsstate = clsstate->next) + clsstate->contained = 0; // Add start state to closure only for epsilon closure - if (0 == literal) + if (0 == label) GNUNET_array_append (cls->states, cls->len, s); GNUNET_array_append (cls_check->states, cls_check->len, s); @@ -663,15 +1722,15 @@ nfa_closure_create (struct State *s, const char literal) for (ctran = currentstate->transitions_head; NULL != ctran; ctran = ctran->next) { - if (NULL != ctran->state && literal == ctran->literal) + if (NULL != ctran->to_state && label == ctran->label) { - clsstate = ctran->state; + clsstate = ctran->to_state; - if (NULL != clsstate && - GNUNET_YES != state_set_contains (cls, clsstate)) + if (NULL != clsstate && 0 == clsstate->contained) { GNUNET_array_append (cls->states, cls->len, clsstate); GNUNET_array_append (cls_check->states, cls_check->len, clsstate); + clsstate->contained = 1; } } } @@ -679,41 +1738,46 @@ nfa_closure_create (struct State *s, const char literal) GNUNET_assert (0 == cls_check->len); GNUNET_free (cls_check); + // sort the states if (cls->len > 1) - qsort (cls->states, cls->len, sizeof (struct State *), state_compare); + qsort (cls->states, cls->len, sizeof (struct GNUNET_REGEX_State *), + state_compare); return cls; } + /** * Calculates the closure set for the given set of states. * + * @param nfa the NFA containing 's' * @param states list of states on which to base the closure on - * @param literal transitioning literal for which to base the closure on, + * @param label transitioning label for which to base the closure on, * pass 0 for epsilon transition * - * @return nfa closure on 'literal' (epsilon closure if 'literal' is 0) + * @return sorted nfa closure on 'label' (epsilon closure if 'label' is 0) */ -static struct StateSet * -nfa_closure_set_create (struct StateSet *states, const char literal) +static struct GNUNET_REGEX_StateSet * +nfa_closure_set_create (struct GNUNET_REGEX_Automaton *nfa, + struct GNUNET_REGEX_StateSet *states, const char label) { - struct State *s; - struct StateSet *sset; - struct StateSet *cls; - int i; - int j; - int k; - int contains; + struct GNUNET_REGEX_State *s; + struct GNUNET_REGEX_StateSet *sset; + struct GNUNET_REGEX_StateSet *cls; + unsigned int i; + unsigned int j; + unsigned int k; + unsigned int contains; if (NULL == states) return NULL; - cls = GNUNET_malloc (sizeof (struct StateSet)); + cls = GNUNET_malloc (sizeof (struct GNUNET_REGEX_StateSet)); for (i = 0; i < states->len; i++) { s = states->states[i]; - sset = nfa_closure_create (s, literal); + sset = nfa_closure_create (nfa, s, label); for (j = 0; j < sset->len; j++) { @@ -721,7 +1785,10 @@ nfa_closure_set_create (struct StateSet *states, const char literal) for (k = 0; k < cls->len; k++) { if (sset->states[j]->id == cls->states[k]->id) + { contains = 1; + break; + } } if (!contains) GNUNET_array_append (cls->states, cls->len, sset->states[j]); @@ -730,11 +1797,13 @@ nfa_closure_set_create (struct StateSet *states, const char literal) } if (cls->len > 1) - qsort (cls->states, cls->len, sizeof (struct State *), state_compare); + qsort (cls->states, cls->len, sizeof (struct GNUNET_REGEX_State *), + state_compare); return cls; } + /** * Pops two NFA fragments (a, b) from the stack and concatenates them (ab) * @@ -745,28 +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); - add_transition (ctx, a->end, 0, b->start); + 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*) * @@ -776,12 +1848,11 @@ static void nfa_add_star_op (struct GNUNET_REGEX_Context *ctx) { struct GNUNET_REGEX_Automaton *a; - struct GNUNET_REGEX_Automaton *new; - struct State *start; - struct State *end; + 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) { @@ -790,24 +1861,27 @@ 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); - add_transition (ctx, start, 0, a->start); - add_transition (ctx, start, 0, end); - add_transition (ctx, a->end, 0, a->start); - add_transition (ctx, a->end, 0, end); + state_add_transition (ctx, start, 0, a->start); + state_add_transition (ctx, start, 0, end); + state_add_transition (ctx, a->end, 0, a->start); + state_add_transition (ctx, a->end, 0, end); 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+) * @@ -821,14 +1895,56 @@ nfa_add_plus_op (struct GNUNET_REGEX_Context *ctx) a = ctx->stack_tail; GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a); - add_transition (ctx, a->end, 0, a->start); + state_add_transition (ctx, a->end, 0, a->start); 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?) + * + * @param ctx context + */ +static void +nfa_add_question_op (struct GNUNET_REGEX_Context *ctx) +{ + struct GNUNET_REGEX_Automaton *a; + struct GNUNET_REGEX_Automaton *new_nfa; + struct GNUNET_REGEX_State *start; + struct GNUNET_REGEX_State *end; + + a = ctx->stack_tail; + + if (NULL == a) + { + GNUNET_log (GNUNET_ERROR_TYPE_ERROR, + "nfa_add_question_op failed, because there was no element on the stack"); + return; + } + + GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a); + + start = nfa_state_create (ctx, 0); + end = nfa_state_create (ctx, 1); + + state_add_transition (ctx, start, 0, a->start); + state_add_transition (ctx, start, 0, end); + state_add_transition (ctx, a->end, 0, end); + + a->end->accepting = 0; + + 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_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 */ @@ -837,59 +1953,83 @@ nfa_add_alternation (struct GNUNET_REGEX_Context *ctx) { struct GNUNET_REGEX_Automaton *a; struct GNUNET_REGEX_Automaton *b; - struct GNUNET_REGEX_Automaton *new; - struct State *start; - struct State *end; + 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); end = nfa_state_create (ctx, 1); - add_transition (ctx, start, 0, a->start); - add_transition (ctx, start, 0, b->start); + state_add_transition (ctx, start, 0, a->start); + state_add_transition (ctx, start, 0, b->start); - add_transition (ctx, a->end, 0, end); - add_transition (ctx, b->end, 0, end); + state_add_transition (ctx, a->end, 0, end); + state_add_transition (ctx, b->end, 0, end); a->end->accepting = 0; 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 * * @param ctx context - * @param lit literal for nfa transition + * @param lit label for nfa transition */ static void -nfa_add_literal (struct GNUNET_REGEX_Context *ctx, const char lit) +nfa_add_label (struct GNUNET_REGEX_Context *ctx, const char lit) { struct GNUNET_REGEX_Automaton *n; - struct State *start; - struct State *end; + struct GNUNET_REGEX_State *start; + struct GNUNET_REGEX_State *end; GNUNET_assert (NULL != ctx); start = nfa_state_create (ctx, 0); end = nfa_state_create (ctx, 1); - add_transition (ctx, start, lit, end); + state_add_transition (ctx, start, lit, end); n = nfa_fragment_create (start, end); GNUNET_assert (NULL != n); GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, n); } + +/** + * Initialize a new context + * + * @param ctx context + */ +static void +GNUNET_REGEX_context_init (struct GNUNET_REGEX_Context *ctx) +{ + if (NULL == ctx) + { + GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Context was NULL!"); + return; + } + ctx->state_id = 0; + ctx->transition_id = 0; + ctx->stack_head = NULL; + ctx->stack_tail = NULL; +} + + /** * Construct an NFA by parsing the regex string of length 'len'. * @@ -976,7 +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); @@ -989,16 +2129,25 @@ GNUNET_REGEX_construct_nfa (const char *regex, const size_t len) } nfa_add_plus_op (&ctx); break; + case '?': + if (atomcount == 0) + { + error_msg = "Cannot append '?' to nothing"; + goto error; + } + nfa_add_question_op (&ctx); + break; case 92: /* escape: \ */ regexp++; count++; + /* fall through! */ default: if (atomcount > 1) { --atomcount; nfa_add_concatenation (&ctx); } - nfa_add_literal (&ctx, *regexp); + nfa_add_label (&ctx, *regexp); atomcount++; break; } @@ -1013,64 +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; } - GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, - "Created NFA with %i States and a total of %i Transitions\n", - ctx.state_id, ctx.transition_id); + 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); - 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; } + /** - * Free the memory allocated by constructing the GNUNET_REGEX_Automaton - * data structure. + * Create DFA states based on given 'nfa' and starting with 'dfa_state'. * - * @param a automaton to be destroyed + * @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. */ -void -GNUNET_REGEX_automaton_destroy (struct GNUNET_REGEX_Automaton *a) +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 State *s; - struct State *next_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; - if (NULL == a) - return; + 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; + } - for (s = a->states_head; NULL != s;) - { - next_state = s->next; - automaton_destroy_state (s); - s = next_state; + 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); + } } - - GNUNET_free (a); } + /** * Construct DFA for the given 'regex' of length 'len' * @@ -1085,187 +2265,99 @@ 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 StateSet *tmp; - struct StateSet *nfa_set; - struct StateSet *dfa_stack; - struct Transition *ctran; - struct State *dfa_state; - struct State *new_dfa_state; - struct State *state_contains; - struct State *state_iter; + struct GNUNET_REGEX_StateSet *nfa_set; GNUNET_REGEX_context_init (&ctx); // Create NFA nfa = GNUNET_REGEX_construct_nfa (regex, len); + if (NULL == nfa) + { + GNUNET_log (GNUNET_ERROR_TYPE_ERROR, + "Could not create DFA, because NFA creation failed\n"); + return NULL; + } + 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 StateSet)); - nfa_set = nfa_closure_create (nfa->start, 0); + nfa_set = nfa_closure_create (nfa, nfa->start, 0); dfa->start = dfa_state_create (&ctx, nfa_set); - GNUNET_CONTAINER_DLL_insert (dfa->states_head, dfa->states_tail, dfa->start); - GNUNET_array_append (dfa_stack->states, dfa_stack->len, dfa->start); - 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->literal && NULL == ctran->state) - { - tmp = nfa_closure_set_create (dfa_state->nfa_set, ctran->literal); - nfa_set = nfa_closure_set_create (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; - } + automaton_add_state (dfa, dfa->start); - if (NULL == state_contains) - { - GNUNET_CONTAINER_DLL_insert_tail (dfa->states_head, dfa->states_tail, - new_dfa_state); - GNUNET_array_append (dfa_stack->states, dfa_stack->len, - new_dfa_state); - ctran->state = new_dfa_state; - } - else - { - ctran->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); - dfa_minimize (dfa); + // Minimize DFA + dfa_minimize (&ctx, dfa); - GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Created DFA with %i States\n", - ctx.state_id); + // Create proofs for all states + automaton_create_proofs (dfa); return dfa; } + /** - * Save the given automaton as a GraphViz dot file + * Free the memory allocated by constructing the GNUNET_REGEX_Automaton data + * structure. * - * @param a the automaton to be saved - * @param filename where to save the file + * @param a automaton to be destroyed */ void -GNUNET_REGEX_automaton_save_graph (struct GNUNET_REGEX_Automaton *a, - const char *filename) +GNUNET_REGEX_automaton_destroy (struct GNUNET_REGEX_Automaton *a) { - struct State *s; - struct Transition *ctran; - char *s_acc = NULL; - char *s_tran = NULL; - char *start; - char *end; - FILE *p; + struct GNUNET_REGEX_State *s; + struct GNUNET_REGEX_State *next_state; 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); + GNUNET_free_non_null (a->regex); + GNUNET_free_non_null (a->canonical_regex); - for (s = a->states_head; NULL != s; s = s->next) + for (s = a->states_head; NULL != s;) { - if (s->accepting) - { - GNUNET_asprintf (&s_acc, "\"%s\" [shape=doublecircle];\n", s->name); - 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->state) - { - GNUNET_log (GNUNET_ERROR_TYPE_ERROR, - "Transition from State %i has has no state for transitioning\n", - s->id); - continue; - } - - if (ctran->literal == 0) - { - GNUNET_asprintf (&s_tran, "\"%s\" -> \"%s\" [label = \"epsilon\"];\n", - s->name, ctran->state->name); - } - else - { - GNUNET_asprintf (&s_tran, "\"%s\" -> \"%s\" [label = \"%c\"];\n", - s->name, ctran->state->name, ctran->literal); - } - - fwrite (s_tran, strlen (s_tran), 1, p); - GNUNET_free (s_tran); - } + next_state = s->next; + automaton_destroy_state (s); + s = next_state; } - end = "\n}\n"; - fwrite (end, strlen (end), 1, p); - fclose (p); + GNUNET_free (a); } + /** * Evaluates the given string using the given DFA automaton * * @param a automaton, type must be DFA * @param string string that should be evaluated * - * @return GNUNET_YES if string matches, GNUNET_NO if not, GNUNET_SYSERR otherwise + * @return 0 if string matches, non 0 otherwise */ static int evaluate_dfa (struct GNUNET_REGEX_Automaton *a, const char *string) { const char *strp; - struct State *s; + struct GNUNET_REGEX_State *s; if (DFA != a->type) { GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Tried to evaluate DFA, but NFA automaton given"); - return GNUNET_SYSERR; + return -1; } 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); @@ -1274,46 +2366,49 @@ evaluate_dfa (struct GNUNET_REGEX_Automaton *a, const char *string) } if (NULL != s && s->accepting) - return GNUNET_YES; + return 0; - return GNUNET_NO; + return 1; } + /** * Evaluates the given string using the given NFA automaton * * @param a automaton, type must be NFA * @param string string that should be evaluated * - * @return GNUNET_YES if string matches, GNUNET_NO if not, GNUNET_SYSERR otherwise + * @return 0 if string matches, non 0 otherwise */ static int evaluate_nfa (struct GNUNET_REGEX_Automaton *a, const char *string) { const char *strp; - struct State *s; - struct StateSet *sset; - struct StateSet *new_sset; - int i; + struct GNUNET_REGEX_State *s; + struct GNUNET_REGEX_StateSet *sset; + struct GNUNET_REGEX_StateSet *new_sset; + unsigned int i; int result; if (NFA != a->type) { GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Tried to evaluate NFA, but DFA automaton given"); - return GNUNET_SYSERR; + return -1; } - result = GNUNET_NO; - strp = string; - sset = GNUNET_malloc (sizeof (struct StateSet)); - GNUNET_array_append (sset->states, sset->len, a->start); + // 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; + sset = nfa_closure_create (a, a->start, 0); for (strp = string; NULL != strp && *strp; strp++) { - new_sset = nfa_closure_set_create (sset, *strp); + new_sset = nfa_closure_set_create (a, sset, *strp); state_set_clear (sset); - sset = nfa_closure_set_create (new_sset, 0); + sset = nfa_closure_set_create (a, new_sset, 0); state_set_clear (new_sset); } @@ -1322,7 +2417,7 @@ evaluate_nfa (struct GNUNET_REGEX_Automaton *a, const char *string) s = sset->states[i]; if (NULL != s && s->accepting) { - result = GNUNET_YES; + result = 0; break; } } @@ -1338,7 +2433,7 @@ evaluate_nfa (struct GNUNET_REGEX_Automaton *a, const char *string) * @param a automaton * @param string string to check * - * @return GNUNET_YES if 'a' matches 'string', GNUNET_NO otherwise + * @return 0 if string matches, non 0 otherwise */ int GNUNET_REGEX_eval (struct GNUNET_REGEX_Automaton *a, const char *string) @@ -1362,3 +2457,245 @@ 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_string'. + * + * @param input_string string. + * @param string_len length of the 'input_string'. + * @param key pointer to where to write the hash code. + * + * @return number of bits of 'input_string' that have been consumed + * to construct the 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; + + if (NULL == input_string) + { + GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Given input string was NULL!\n"); + return 0; + } + + GNUNET_CRYPTO_hash (input_string, size, key); + + return size; +} + + +/** + * Check if the given 'proof' matches the given 'key'. + * + * @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. + */ +int +GNUNET_REGEX_check_proof (const char *proof, const struct GNUNET_HashCode *key) +{ + 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. + * + * @param s state. + * @param iterator iterator function called for each edge. + * @param iterator_cls closure. + */ +static void +iterate_edge (struct GNUNET_REGEX_State *s, GNUNET_REGEX_KeyIterator iterator, + void *iterator_cls) +{ + struct GNUNET_REGEX_Transition *t; + struct GNUNET_REGEX_Edge edges[s->transition_count]; + unsigned int num_edges; + + if (GNUNET_YES != s->marked) + { + s->marked = GNUNET_YES; + + num_edges = state_get_edges (s, 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. + * + * @param a automaton. + * @param iterator iterator called for each edge. + * @param iterator_cls closure. + */ +void +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); +}