#include "gnunet_container_lib.h"
#include "gnunet_crypto_lib.h"
#include "gnunet_regex_lib.h"
-#include "regex.h"
+#include "regex_internal.h"
+
+
+/**
+ * Constant for how many bits the initial string regex should have.
+ */
+#define INITIAL_BITS 8
-#define initial_bits 10
/**
* Context that contains an id counter for states and transitions as well as a
*/
unsigned int transition_id;
- /**
- * Unique SCC (Strongly Connected Component) id.
- */
- unsigned int scc_id;
-
/**
* DLL of GNUNET_REGEX_Automaton's used as a stack.
*/
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.
- */
- struct GNUNET_REGEX_State *start;
-
- /**
- * End state of the automaton.
- */
- struct GNUNET_REGEX_State *end;
-
- /**
- * Number of states in the automaton.
- */
- unsigned int state_count;
-
- /**
- * DLL of states.
- */
- struct GNUNET_REGEX_State *states_head;
-
- /**
- * DLL of states
+ * Array of states.
*/
- struct GNUNET_REGEX_State *states_tail;
+ struct GNUNET_REGEX_State **states;
/**
- * Type of the automaton.
+ * Length of the 'states' array.
*/
- enum GNUNET_REGEX_automaton_type type;
+ unsigned int len;
};
-/**
- * A state. Can be used in DFA and NFA automatons.
- */
-struct GNUNET_REGEX_State
-{
- /**
- * This is a linked list.
- */
- struct GNUNET_REGEX_State *prev;
-
- /**
- * This is a linked list.
- */
- struct GNUNET_REGEX_State *next;
- /**
- * Unique state id.
- */
- unsigned int id;
-
- /**
- * If this is an accepting state or not.
- */
- int accepting;
-
- /**
- * Marking of the state. This is used for marking all visited states when
- * traversing all states of an automaton and for cases where the state id
- * cannot be used (dfa minimization).
- */
- int marked;
-
- /**
- * Marking the state as contained. This is used for checking, if the state is
- * contained in a set in constant time
- */
- int contained;
-
- /**
- * Marking the state as part of an SCC (Strongly Connected Component). All
- * states with the same scc_id are part of the same SCC. scc_id is 0, if state
- * is not a part of any SCC.
- */
- unsigned int scc_id;
-
- /**
- * Used for SCC detection.
- */
- int index;
-
- /**
- * Used for SCC detection.
- */
- int lowlink;
-
- /**
- * Human readable name of the automaton. Used for debugging and graph
- * creation.
- */
- char *name;
-
- /**
- * Hash of the state.
- */
- GNUNET_HashCode hash;
-
- /**
- * Proof for this state.
- */
- char *proof;
-
- /**
- * Number of transitions from this state to other states.
- */
- unsigned int transition_count;
-
- /**
- * DLL of transitions.
- */
- struct Transition *transitions_head;
-
- /**
- * DLL of transitions.
- */
- struct Transition *transitions_tail;
-
- /**
- * Set of states on which this state is based on. Used when creating a DFA out
- * of several NFA states.
- */
- struct GNUNET_REGEX_StateSet *nfa_set;
-};
+/*
+ * Debug helper functions
+ */
/**
- * Transition between two states. Each state can have 0-n transitions. If label
- * is 0, this is considered to be an epsilon transition.
+ * Print all the transitions of state 's'.
+ *
+ * @param s state for which to print it's transitions.
*/
-struct Transition
-{
- /**
- * This is a linked list.
- */
- struct Transition *prev;
-
- /**
- * This is a linked list.
- */
- struct Transition *next;
-
- /**
- * Unique id of this transition.
- */
- unsigned int id;
-
- /**
- * Label for this transition. This is basically the edge label for the graph.
- */
- char label;
-
- /**
- * State to which this transition leads.
- */
- struct GNUNET_REGEX_State *to_state;
+void
+debug_print_transitions (struct GNUNET_REGEX_State *s);
- /**
- * State from which this transition origins.
- */
- struct GNUNET_REGEX_State *from_state;
-};
/**
- * Set of states.
+ * Print information of the given state 's'.
+ *
+ * @param s state for which debug information should be printed.
*/
-struct GNUNET_REGEX_StateSet
+void
+debug_print_state (struct GNUNET_REGEX_State *s)
{
- /**
- * Array of states.
- */
- struct GNUNET_REGEX_State **states;
+ char *proof;
- /**
- * Length of the 'states' array.
- */
- unsigned int len;
-};
+ if (NULL == s->proof)
+ proof = "NULL";
+ else
+ proof = s->proof;
-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 transitions: %i\n",
+ "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);
+ s->transition_count, proof);
+
+ GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Transitions:\n");
+ debug_print_transitions (s);
}
-static void
-debug_print_states (struct GNUNET_REGEX_StateSet *sset)
+
+/**
+ * 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.
+ */
+void
+debug_print_states (struct GNUNET_REGEX_Automaton *a)
{
struct GNUNET_REGEX_State *s;
- int i;
- for (i = 0; i < sset->len; i++)
- {
- s = sset->states[i];
+ for (s = a->states_head; NULL != s; s = s->next)
debug_print_state (s);
- }
}
-static void
-debug_print_transitions (struct GNUNET_REGEX_State *s)
-{
- struct Transition *t;
- char *state;
- char label;
-
- 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);
- }
-}
/**
- * 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.
+ * Print debug information for given transition 't'.
*
- * @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 t transition for which to print debug info.
*/
-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)
+void
+debug_print_transition (struct GNUNET_REGEX_Transition *t)
{
- struct GNUNET_REGEX_State *w;
- struct Transition *t;
+ char *to_state;
+ char *from_state;
+ char label;
- v->index = *index;
- v->lowlink = *index;
- (*index)++;
- stack[(*stack_size)++] = v;
- v->contained = 1;
+ if (NULL == t)
+ return;
- for (t = v->transitions_head; NULL != t; t = t->next)
- {
- w = t->to_state;
- if (NULL != w && w->index < 0)
- {
- scc_tarjan_strongconnect (ctx, w, index, stack, stack_size);
- v->lowlink = (v->lowlink > w->lowlink) ? w->lowlink : v->lowlink;
- }
- else if (0 != w->contained)
- v->lowlink = (v->lowlink > w->index) ? w->index : v->lowlink;
- }
+ if (0 == t->label)
+ label = '0';
+ else
+ label = t->label;
- if (v->lowlink == v->index)
- {
- w = stack[--(*stack_size)];
- w->contained = 0;
+ if (NULL == t->to_state)
+ to_state = "NULL";
+ else
+ to_state = t->to_state->name;
- 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;
- }
- }
-}
+ if (NULL == t->from_state)
+ from_state = "NULL";
+ else
+ from_state = t->from_state->name;
-/**
- * Detect all SCCs (Strongly Connected Components) inside the given automaton.
- * SCCs will be marked using the scc_id on each state.
- *
- * @param ctx context
- * @param a automaton
- */
-static void
-scc_tarjan (struct GNUNET_REGEX_Context *ctx, struct GNUNET_REGEX_Automaton *a)
-{
- unsigned int i;
- int index;
- struct GNUNET_REGEX_State *v;
- struct GNUNET_REGEX_State *stack[a->state_count];
- unsigned int stack_size;
+ GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Transition %i: From %s on %c to %s\n",
+ t->id, from_state, label, to_state);
+}
- for (v = a->states_head; NULL != v; v = v->next)
- {
- v->contained = 0;
- v->index = -1;
- v->lowlink = -1;
- }
- stack_size = 0;
- index = 0;
+void
+debug_print_transitions (struct GNUNET_REGEX_State *s)
+{
+ struct GNUNET_REGEX_Transition *t;
- 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);
- }
+ for (t = s->transitions_head; NULL != t; t = t->next)
+ debug_print_transition (t);
}
+
/**
* Adds a transition from one state to another on 'label'. Does not add
* duplicate states.
struct GNUNET_REGEX_State *to_state)
{
int is_dup;
- struct Transition *t;
+ struct GNUNET_REGEX_Transition *t;
+ struct GNUNET_REGEX_Transition *oth;
if (NULL == from_state)
{
}
}
- if (is_dup)
+ if (GNUNET_YES == is_dup)
return;
- t = GNUNET_malloc (sizeof (struct Transition));
+ // 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;
// Add outgoing transition to 'from_state'
from_state->transition_count++;
- GNUNET_CONTAINER_DLL_insert (from_state->transitions_head,
- from_state->transitions_tail, t);
+ GNUNET_CONTAINER_DLL_insert_before (from_state->transitions_head,
+ from_state->transitions_tail, oth, t);
}
+
+/**
+ * Remove a 'transition' from 'state'.
+ *
+ * @param state state from which the to-be-removed transition originates.
+ * @param transition transition that should be removed from state 'state'.
+ */
+static void
+state_remove_transition (struct GNUNET_REGEX_State *state,
+ struct GNUNET_REGEX_Transition *transition)
+{
+ if (NULL == state || NULL == transition)
+ return;
+
+ if (transition->from_state != state)
+ return;
+
+ state->transition_count--;
+ GNUNET_CONTAINER_DLL_remove (state->transitions_head, state->transitions_tail,
+ transition);
+ GNUNET_free (transition);
+}
+
+
/**
* Compare two states. Used for sorting.
*
return (*s1)->id - (*s2)->id;
}
+
/**
* Get all edges leaving state 's'.
*
* @param s state.
- * @param edges all edges leaving 's'.
+ * @param edges all edges leaving 's', expected to be allocated and have enough
+ * space for s->transitions_count elements.
*
* @return number of edges.
*/
static unsigned int
state_get_edges (struct GNUNET_REGEX_State *s, struct GNUNET_REGEX_Edge *edges)
{
- struct Transition *t;
+ struct GNUNET_REGEX_Transition *t;
unsigned int count;
if (NULL == s)
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!
struct GNUNET_REGEX_StateSet *sset2)
{
int result;
- int i;
+ unsigned int i;
if (NULL == sset1 || NULL == sset2)
return 1;
return result;
}
+
/**
* Clears the given StateSet 'set'
*
{
if (NULL != set)
{
- if (NULL != set->states)
- GNUNET_free (set->states);
+ GNUNET_free_non_null (set->states);
GNUNET_free (set);
}
}
+
/**
* Clears an automaton fragment. Does not destroy the states inside the
* automaton.
GNUNET_free (a);
}
+
/**
* Frees the memory used by State 's'
*
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; t = next_t)
{
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
{
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;
automaton_destroy_state (ss);
}
+
/**
* Merge two states into one. Will merge 's1' and 's2' into 's1' and destroy
* 's2'.
struct GNUNET_REGEX_State *s2)
{
struct GNUNET_REGEX_State *s_check;
- struct Transition *t_check;
+ 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);
+
+ 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_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
+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;
+ 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_assert (NULL != ctx && NULL != a && NULL != s1 && NULL != s2);
+ GNUNET_free_non_null (R_temp_ik);
+ GNUNET_free_non_null (R_temp_kk);
+ GNUNET_free_non_null (R_temp_kj);
- // 1. Make all transitions pointing to s2 point to s1
- for (s_check = a->states_head; NULL != s_check; s_check = s_check->next)
+ if (NULL == R_cur_l && NULL == R_cur_r)
{
- for (t_check = s_check->transitions_head; NULL != t_check;
- t_check = t_check->next)
- {
- if (s2 == t_check->to_state)
- t_check->to_state = s1;
- }
+ *R_cur_ij = NULL;
+ return;
}
- // 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 (NULL != R_cur_l && NULL == R_cur_r)
{
- if (t_check->to_state != s1)
- state_add_transition (ctx, s1, t_check->label, t_check->to_state);
+ *R_cur_ij = R_cur_l;
+ return;
}
- // 3. Rename s1 to {s1,s2}
- new_name = GNUNET_strdup (s1->name);
- if (NULL != s1->name)
+ if (NULL == R_cur_l && NULL != R_cur_r)
{
- GNUNET_free (s1->name);
- s1->name = NULL;
+ *R_cur_ij = R_cur_r;
+ return;
}
- 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);
+ 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);
}
+
/**
- * Add a state to the automaton 'a', always use this function to alter the
- * states DLL of the automaton.
+ * 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 to add the state to
- * @param s state that should be added
+ * @param a automaton.
*/
static void
-automaton_add_state (struct GNUNET_REGEX_Automaton *a,
- struct GNUNET_REGEX_State *s)
+automaton_create_proofs (struct GNUNET_REGEX_Automaton *a)
{
- GNUNET_CONTAINER_DLL_insert (a->states_head, a->states_tail, s);
- a->state_count++;
-}
+ 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;
-typedef void (*GNUNET_REGEX_traverse_action) (void *cls,
- struct GNUNET_REGEX_State * s);
-static void
-automaton_state_traverse_backward (void *cls, struct GNUNET_REGEX_State *s,
- GNUNET_REGEX_traverse_action action)
-{
+ /* create depth-first numbering of the states, initializes 'state' */
+ GNUNET_REGEX_automaton_traverse (a, &number_states, states);
- GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Traversing backwards...\n");
+ /* 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]);
+ }
+ }
-/**
- * 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.
- *
- * @param cls closure.
- * @param s start state.
- * @param action action to be performed on each state.
- */
-static void
-automaton_state_traverse (void *cls, struct GNUNET_REGEX_State *s,
- GNUNET_REGEX_traverse_action action)
-{
- struct Transition *t;
+ // 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;
+ }
+ }
+ }
- if (GNUNET_NO == s->marked)
+ // assign proofs and hashes
+ for (i = 0; i < n; i++)
{
- s->marked = GNUNET_YES;
+ 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 (NULL != action)
- action (cls, s);
+ // 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;
- 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 cls closure.
- * @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
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++;
s->index = -1;
s->lowlink = -1;
s->contained = 0;
+ s->proof = NULL;
if (NULL == nfa_states)
{
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
return s;
}
+
/**
* Move from the given state 's' to the next state on transition 'label'
*
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)
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.
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);
}
}
+
/**
* Merge all non distinguishable states in the DFA 'a'
*
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;
- int common_labels;
+ unsigned int num_equal_edges;
for (i = 0, s1 = a->states_head; i < a->state_count && NULL != s1;
i++, s1 = s1->next)
// 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)
{
if (0 != table[s1->marked][s2->marked])
continue;
- common_labels = GNUNET_NO;
+ 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)
{
- common_labels = GNUNET_YES;
-
+ num_equal_edges++;
if (0 != table[t1->to_state->marked][t2->to_state->marked] ||
0 != table[t2->to_state->marked][t1->to_state->marked])
{
}
}
}
- if (GNUNET_NO == common_labels)
+ 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 (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;
}
}
+
/**
* Minimize the given DFA 'a' by removing all unreachable states, removing all
* dead states and merging all non distinguishable states
dfa_merge_nondistinguishable_states (ctx, a);
}
+
/**
* Creates a new NFA fragment. Needs to be cleared using
* automaton_fragment_clear.
return n;
}
+
/**
* Adds a list of states to the given automaton 'n'.
*
n->state_count++;
}
+
/**
* Creates a new NFA state. Needs to be freed using automaton_destroy_state.
*
return s;
}
+
/**
* Calculates the NFA closure set for the given state.
*
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;
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);
return cls;
}
+
/**
* Calculates the closure set for the given set of states.
*
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;
return cls;
}
+
/**
* Pops two NFA fragments (a, b) from the stack and concatenates them (ab)
*
struct GNUNET_REGEX_Automaton *new;
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);
GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new);
}
+
/**
* Pops a NFA fragment from the stack (a) and adds a new fragment (a*)
*
GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new);
}
+
/**
* Pops an NFA fragment (a) from the stack and adds a new fragment (a+)
*
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?)
*
GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new);
}
+
/**
* Pops two NFA fragments (a, b) from the stack and adds a new NFA fragment that
* alternates between a and b (a|b)
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);
GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, new);
}
+
/**
* Adds a new nfa fragment to the stack
*
GNUNET_CONTAINER_DLL_insert_tail (ctx->stack_head, ctx->stack_tail, n);
}
+
/**
* Initialize a new context
*
}
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'.
*
case '*':
if (atomcount == 0)
{
- error_msg = "Cannot append '+' to nothing";
+ error_msg = "Cannot append '*' to nothing";
goto error;
}
nfa_add_star_op (&ctx);
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);
+
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'.
*
struct GNUNET_REGEX_Automaton *dfa,
struct GNUNET_REGEX_State *dfa_state)
{
- struct Transition *ctran;
+ struct GNUNET_REGEX_Transition *ctran;
struct GNUNET_REGEX_State *state_iter;
struct GNUNET_REGEX_State *new_dfa_state;
struct GNUNET_REGEX_State *state_contains;
if (NULL == state_contains)
{
automaton_add_state (dfa, new_dfa_state);
- construct_dfa_states (ctx, nfa, dfa, new_dfa_state);
ctran->to_state = new_dfa_state;
+ construct_dfa_states (ctx, nfa, dfa, new_dfa_state);
}
else
{
}
}
+
/**
* Construct DFA for the given 'regex' of length 'len'
*
dfa = GNUNET_malloc (sizeof (struct GNUNET_REGEX_Automaton));
dfa->type = DFA;
+ dfa->regex = GNUNET_strdup (regex);
// Create DFA start state from epsilon closure
nfa_set = nfa_closure_create (nfa, nfa->start, 0);
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.
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;
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 (NULL == p)
- {
- 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);
- }
- else
- {
- GNUNET_asprintf (&s_acc, "\"%s\" [color=\"0.%i 0.8 0.95\"];\n", s->name,
- s->scc_id);
- }
-
- if (NULL == s_acc)
- {
- GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not print state %s\n",
- s->name);
- return;
- }
- fwrite (s_acc, strlen (s_acc), 1, p);
- GNUNET_free (s_acc);
- s_acc = NULL;
-
- for (ctran = s->transitions_head; NULL != ctran; ctran = ctran->next)
- {
- if (NULL == ctran->to_state)
- {
- GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
- "Transition from State %i has 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);
- }
-
- if (NULL == s_tran)
- {
- GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not print state %s\n",
- s->name);
- return;
- }
-
- fwrite (s_tran, strlen (s_tran), 1, p);
- GNUNET_free (s_tran);
- s_tran = NULL;
- }
- }
-
- end = "\n}\n";
- fwrite (end, strlen (end), 1, p);
- fclose (p);
-}
/**
* Evaluates the given string using the given DFA automaton
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);
return 1;
}
+
/**
* Evaluates the given string using the given NFA automaton
*
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)
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++)
return result;
}
+
/**
* Evaluates the given 'string' against the given compiled regex
*
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'.
+ * of the 'input_string'.
*
* @param input_string string.
* @param string_len length of the 'input_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)
{
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)
+{
+ 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 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 next_state next state in the graph that is reachable with 'label' transition.
+ * @param label label of the transition to the next state.
+ * @param iterator iterator function called for each edge.
+ * @param iterator_cls closure for the iterator function.
+ */
+static void
+iterate_initial_edge (const unsigned int max_len, unsigned int cur_len,
+ char *consumed_string,
+ struct GNUNET_REGEX_State *next_state, const char label,
+ GNUNET_REGEX_KeyIterator iterator, void *iterator_cls)
+{
+ char *temp;
+ struct GNUNET_REGEX_Transition *t;
+ struct GNUNET_REGEX_Edge edge;
+ struct GNUNET_HashCode hash;
+ size_t constr_len;
+
+ GNUNET_log (GNUNET_ERROR_TYPE_DEBUG,
+ "max_len: %u, cur_len: %u, consumed_string: %s\n", max_len,
+ cur_len, consumed_string);
+
+ if (max_len == cur_len)
+ {
+ constr_len = strlen (consumed_string);
+ GNUNET_CRYPTO_hash (consumed_string, constr_len - 1, &hash);
+ GNUNET_asprintf (&temp, "%c", label);
+ edge.label = temp;
+ edge.destination = next_state->hash;
+ consumed_string[constr_len - 1] = '\0';
+ iterator (iterator_cls, &hash, consumed_string, 0, 1, &edge);
+ GNUNET_free (temp);
+
+ return;
+ }
+ else if (cur_len <= max_len)
+ {
+ cur_len++;
+ for (t = next_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 (max_len, cur_len, temp, t->to_state, t->label,
+ 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.
+ *
+ * @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)
{
- return GNUNET_OK;
+ iterate_initial_edge (initial_len + 1, 0, NULL, a->start, 0, iterator,
+ iterator_cls);
}
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;
num_edges = state_get_edges (s, edges);
- iterator (iterator_cls, &s->hash, NULL, s->accepting, num_edges, edges);
+ 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.
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);
}