*/
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;
};
+/**
+ * Type of an automaton.
+ */
enum GNUNET_REGEX_automaton_type
{
NFA,
};
/**
- * Automaton representation
+ * Automaton representation.
*/
struct GNUNET_REGEX_Automaton
{
+ /**
+ * 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 State *start;
+
+ /**
+ * End state of the automaton.
+ */
struct State *end;
+ /**
+ * Number of states in the automaton.
+ */
unsigned int state_count;
+
+ /**
+ * DLL of states.
+ */
struct State *states_head;
+
+ /**
+ * DLL of states
+ */
struct State *states_tail;
+ /**
+ * Type of the automaton.
+ */
enum GNUNET_REGEX_automaton_type type;
};
*/
struct State
{
+ /**
+ * This is a linked list.
+ */
struct State *prev;
+
+ /**
+ * This is a linked list.
+ */
struct 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;
+
+ /**
+ * Human readable name of the automaton. Used for debugging
+ * and graph creation.
+ */
char *name;
+ /**
+ * 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 StateSet *nfa_set;
};
*/
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;
+
+ /**
+ * Literal for this transition. This is basically the edge label for
+ * the graph.
+ */
char literal;
+
+ /**
+ * State to which this transition leads.
+ */
struct State *state;
};
/**
- * Set of states
+ * Set of states.
*/
struct StateSet
{
/**
- * Array of states
+ * Array of states.
*/
struct State **states;
+
+ /**
+ * Length of the 'states' array.
+ */
unsigned int len;
};
}
}
+/**
+ * Compare two states. Used for sorting.
+ *
+ * @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)
{
* @param sset1 first state set
* @param sset2 second state set
*
- * @return 0 if they are equal, non 0 otherwise
+ * @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 StateSet *sset1, struct StateSet *sset2)
{
+ int result;
int i;
- if (sset1->len != sset2->len)
+ if (NULL == sset1 || NULL == sset2)
return 1;
+ result = sset1->len - sset2->len;
+
for (i = 0; i < sset1->len; i++)
{
- if (sset1->states[i]->id != sset2->states[i]->id)
- {
- return 1;
- }
+ if (0 != result)
+ break;
+
+ result = state_compare (&sset1->states[i], &sset2->states[i]);
}
- return 0;
+ return result;
}
/**
static void
automaton_fragment_clear (struct GNUNET_REGEX_Automaton *a)
{
+ if (NULL == a)
+ return;
+
a->start = NULL;
a->end = NULL;
a->states_head = NULL;
struct Transition *t;
struct Transition *next_t;
+ if (NULL == s)
+ return;
+
if (NULL != s->name)
GNUNET_free (s->name);
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
automaton_remove_state (struct GNUNET_REGEX_Automaton *a, struct State *s)
{
struct State *ss;
+ struct State *s_check;
+ struct 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->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
+automaton_merge_states (struct GNUNET_REGEX_Context *ctx,
+ struct GNUNET_REGEX_Automaton *a, struct State *s1,
+ struct State *s2)
+{
+ struct State *s_check;
+ struct Transition *t_check;
+ struct Transition *t;
+ char *new_name;
+
+ GNUNET_assert (NULL != ctx && NULL != a && NULL != s1 && NULL != s2);
+
+ // 1. Make all transitions pointing to s2 point to s1
+ 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 (s_check != s1 && s2 == t_check->state)
+ t_check->state = s1;
+ }
+ }
+
+ // 2. Add all transitions from s2 to sX to s1
+ for (t_check = s2->transitions_head; NULL != t_check; t_check = t_check->next)
+ {
+ for (t = s1->transitions_head; NULL != t; t = t->next)
+ {
+ if (t_check->literal != t->literal && NULL != t_check->state &&
+ t_check->state != t->state && t_check->state != s2)
+ {
+ add_transition (ctx, s1, t_check->literal, t_check->state);
+ }
+ }
+ }
+
+ // 3. Rename s1 to {s1,s2}
+ new_name = GNUNET_malloc (strlen (s1->name) + strlen (s2->name) + 1);
+ strncat (new_name, s1->name, strlen (s1->name));
+ strncat (new_name, s2->name, strlen (s2->name));
+ if (NULL != s1->name)
+ GNUNET_free (s1->name);
+ s1->name = new_name;
+
+ // remove state
+ s_check = s2;
+ GNUNET_CONTAINER_DLL_remove (a->states_head, a->states_tail, s_check);
+ a->state_count--;
+ automaton_destroy_state (s_check);
+}
+
+/**
+ * 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 State *s)
{
stack_len++;
while (stack_len > 0)
{
- s = stack[stack_len-1];
+ s = stack[stack_len - 1];
stack_len--;
- s->marked = 1; // mark s as visited
+ s->marked = 1; // mark s as visited
for (t = s->transitions_head; NULL != t; t = t->next)
{
+ // add next states to stack
if (NULL != t->state && 0 == t->state->marked)
- {
- // add next states to stack
- stack[stack_len] = t->state;
- stack_len++;
- }
+ stack[++stack_len] = t->state;
}
}
dfa_remove_dead_states (struct GNUNET_REGEX_Automaton *a)
{
struct State *s;
- struct State *s_check;
struct Transition *t;
- struct Transition *t_check;
int dead;
GNUNET_assert (DFA == a->type);
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)
- {
- for (t_check = s_check->transitions_head; NULL != t_check;
- t_check = t_check->next)
- {
- if (t_check->state == s)
- {
- GNUNET_CONTAINER_DLL_remove (s_check->transitions_head,
- s_check->transitions_tail, t_check);
- }
- }
- }
- // 2. remove state
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_Automaton *a)
+dfa_merge_nondistinguishable_states (struct GNUNET_REGEX_Context *ctx,
+ struct GNUNET_REGEX_Automaton *a)
{
+ int i;
+ int table[a->state_count][a->state_count];
+ struct State *s1;
+ struct State *s2;
+ struct Transition *t1;
+ struct Transition *t2;
+ int change;
+
+ change = 1;
+ for (i = 0, s1 = a->states_head; i < a->state_count && NULL != s1;
+ i++, s1 = s1->next)
+ s1->marked = i;
+
+ // Mark all pairs of accepting/!accepting states
+ for (s1 = a->states_head; NULL != s1; s1 = s1->next)
+ {
+ for (s2 = a->states_head; NULL != s2 && s1 != s2; s2 = s2->next)
+ {
+ if ((s1->accepting && !s2->accepting) ||
+ (!s1->accepting && s2->accepting))
+ {
+ table[s1->marked][s2->marked] = 1;
+ }
+ else
+ table[s1->marked][s2->marked] = 0;
+ }
+ }
+
+ while (0 != change)
+ {
+ change = 0;
+ for (s1 = a->states_head; NULL != s1; s1 = s1->next)
+ {
+ for (s2 = a->states_head; NULL != s2 && s1 != s2; s2 = s2->next)
+ {
+ if (0 != table[s1->marked][s2->marked])
+ continue;
+
+ for (t1 = s1->transitions_head; NULL != t1; t1 = t1->next)
+ {
+ for (t2 = s2->transitions_head; NULL != t2; t2 = t2->next)
+ {
+ if (t1->literal == t2->literal && t1->state == t2->state &&
+ (0 != table[t1->state->marked][t2->state->marked] ||
+ 0 != table[t2->state->marked][t1->state->marked]))
+ {
+ table[s1->marked][s2->marked] = t1->literal;
+ change = 1;
+ }
+ else if (t1->literal != t2->literal && t1->state != t2->state)
+ {
+ table[s1->marked][s2->marked] = -1;
+ change = 1;
+ }
+ }
+ }
+ }
+ }
+ }
+ struct State *s2_next;
+
+ for (i = 0, s1 = a->states_head; NULL != s1; s1 = s1->next)
+ {
+ for (s2 = a->states_head; NULL != s2 && s1 != s2; s2 = s2_next)
+ {
+ s2_next = s2->next;
+ if (s1 != s2 && table[s1->marked][s2->marked] == 0)
+ 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
dfa_remove_dead_states (a);
// 3. Merge nondistinguishable states
- dfa_merge_nondistinguishable_states (a);
+ dfa_merge_nondistinguishable_states (ctx, a);
}
/**
}
/**
- * Calculates the NFA closure set for the given state
+ * Calculates the NFA closure set for the given state.
*
* @param s starting point state
* @param literal transitioning literal 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 'literal' (epsilon closure if 'literal' is 0)
*/
static struct StateSet *
nfa_closure_create (struct State *s, const char literal)
* @param literal transitioning literal 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 'literal' (epsilon closure if 'literal' is 0)
*/
static struct StateSet *
nfa_closure_set_create (struct StateSet *states, const char literal)
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;
+ struct State *start;
+ struct State *end;
+
+ a = ctx->stack_tail;
+ GNUNET_CONTAINER_DLL_remove (ctx->stack_head, ctx->stack_tail, a);
+
+ if (NULL == a)
+ {
+ GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
+ "nfa_add_question_op failed, because there was no element on the stack");
+ return;
+ }
+
+ 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, end);
+
+ a->end->accepting = 0;
+
+ new = nfa_fragment_create (start, end);
+ nfa_add_states (new, a->states_head, a->states_tail);
+ automaton_fragment_clear (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)
}
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++;
GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not parse regex\n");
if (NULL != error_msg)
GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "%s\n", error_msg);
- GNUNET_free (p);
+ if (NULL != p)
+ GNUNET_free (p);
while (NULL != ctx.stack_tail)
{
GNUNET_REGEX_automaton_destroy (ctx.stack_tail);
// 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->start = dfa_state_create (&ctx, nfa_set);
automaton_add_state (dfa, dfa->start);
GNUNET_array_append (dfa_stack->states, dfa_stack->len, dfa->start);
+
+ // Create dfa states by combining nfa states
while (dfa_stack->len > 0)
{
dfa_state = dfa_stack->states[dfa_stack->len - 1];
GNUNET_free (dfa_stack);
GNUNET_REGEX_automaton_destroy (nfa);
- dfa_minimize (dfa);
+ dfa_minimize (&ctx, dfa);
return dfa;
}
return result;
}
-
/**
* Evaluates the given 'string' against the given compiled regex
*
projects / oweals / gnunet.git / blobdiff