#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
struct GNUNET_REGEX_Automaton *stack_tail;
};
-/**
- * Type of an automaton.
- */
-enum GNUNET_REGEX_AutomatonType
-{
- NFA,
- DFA
-};
-
-/**
- * Automaton representation.
- */
-struct GNUNET_REGEX_Automaton
-{
- /**
- * Linked list of NFAs used for partial NFA creation.
- */
- struct GNUNET_REGEX_Automaton *prev;
-
- /**
- * Linked list of NFAs used for partial NFA creation.
- */
- struct GNUNET_REGEX_Automaton *next;
-
- /**
- * First state of the automaton. This is mainly used for constructing an NFA,
- * where each NFA itself consists of one or more NFAs linked together.
- */
- struct GNUNET_REGEX_State *start;
-
- /**
- * End state of the partial NFA. This is undefined for DFAs
- */
- struct GNUNET_REGEX_State *end;
-
- /**
- * Number of states in the automaton.
- */
- unsigned int state_count;
-
- /**
- * DLL of states.
- */
- struct GNUNET_REGEX_State *states_head;
-
- /**
- * DLL of states
- */
- struct GNUNET_REGEX_State *states_tail;
-
- /**
- * Type of the automaton.
- */
- enum GNUNET_REGEX_AutomatonType type;
-
- /**
- * Regex
- */
- char *regex;
-
- /**
- * Computed regex (result of RX->NFA->DFA->RX)
- */
- char *computed_regex;
-};
-
-/**
- * A state. Can be used in DFA and NFA automatons.
- */
-struct GNUNET_REGEX_State
-{
- /**
- * This is a linked list.
- */
- struct GNUNET_REGEX_State *prev;
-
- /**
- * This is a linked list.
- */
- struct GNUNET_REGEX_State *next;
-
- /**
- * Unique state id.
- */
- unsigned int id;
-
- /**
- * If this is an accepting state or not.
- */
- int accepting;
-
- /**
- * Marking of the state. This is used for marking all visited states when
- * traversing all states of an automaton and for cases where the state id
- * cannot be used (dfa minimization).
- */
- int marked;
-
- /**
- * Marking the state as contained. This is used for checking, if the state is
- * contained in a set in constant time
- */
- int contained;
-
- /**
- * Marking the state as part of an SCC (Strongly Connected Component). All
- * states with the same scc_id are part of the same SCC. scc_id is 0, if state
- * is not a part of any SCC.
- */
- unsigned int scc_id;
-
- /**
- * Used for SCC detection.
- */
- int index;
-
- /**
- * Used for SCC detection.
- */
- int lowlink;
-
- /**
- * Human readable name of the automaton. Used for debugging and graph
- * creation.
- */
- char *name;
-
- /**
- * Hash of the state.
- */
- struct GNUNET_HashCode hash;
-
- /**
- * State ID for proof creation.
- */
- unsigned int proof_id;
-
- /**
- * Proof for this state.
- */
- char *proof;
-
- /**
- * Number of transitions from this state to other states.
- */
- unsigned int transition_count;
-
- /**
- * DLL of transitions.
- */
- struct Transition *transitions_head;
-
- /**
- * DLL of transitions.
- */
- struct Transition *transitions_tail;
-
- /**
- * Set of states on which this state is based on. Used when creating a DFA out
- * of several NFA states.
- */
- struct GNUNET_REGEX_StateSet *nfa_set;
-};
-
-/**
- * Transition between two states. Each state can have 0-n transitions. If label
- * is 0, this is considered to be an epsilon transition.
- */
-struct Transition
-{
- /**
- * This is a linked list.
- */
- struct Transition *prev;
-
- /**
- * This is a linked list.
- */
- struct Transition *next;
-
- /**
- * Unique id of this transition.
- */
- unsigned int id;
-
- /**
- * Label for this transition. This is basically the edge label for the graph.
- */
- char label;
-
- /**
- * State to which this transition leads.
- */
- struct GNUNET_REGEX_State *to_state;
-
- /**
- * State from which this transition origins.
- */
- struct GNUNET_REGEX_State *from_state;
-
- /**
- * Mark this transition. For example when reversing the automaton.
- */
- int mark;
-};
/**
* Set of states.
unsigned int len;
};
+
/*
* Debug helper functions
*/
+
+/**
+ * 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 *);
+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.
+ */
void
debug_print_state (struct GNUNET_REGEX_State *s)
{
debug_print_transitions (s);
}
+
+/**
+ * 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)
{
debug_print_state (s);
}
+
+/**
+ * Print debug information for given transition 't'.
+ *
+ * @param t transition for which to print debug info.
+ */
void
-debug_print_transition (struct Transition *t)
+debug_print_transition (struct GNUNET_REGEX_Transition *t)
{
char *to_state;
char *from_state;
t->id, from_state, label, to_state);
}
+
void
debug_print_transitions (struct GNUNET_REGEX_State *s)
{
- struct Transition *t;
+ struct GNUNET_REGEX_Transition *t;
for (t = s->transitions_head; NULL != t; t = t->next)
debug_print_transition (t);
}
-/**
- * Recursive function doing DFS with 'v' as a start, detecting all SCCs inside
- * the subgraph reachable from 'v'. Used with scc_tarjan function to detect all
- * SCCs inside an automaton.
- *
- * @param 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
- */
-static void
-scc_tarjan_strongconnect (unsigned int *scc_counter,
- struct GNUNET_REGEX_State *v, unsigned int *index,
- struct GNUNET_REGEX_State **stack,
- unsigned int *stack_size)
-{
- struct GNUNET_REGEX_State *w;
- struct Transition *t;
-
- v->index = *index;
- v->lowlink = *index;
- (*index)++;
- stack[(*stack_size)++] = v;
- v->contained = 1;
-
- for (t = v->transitions_head; NULL != t; t = t->next)
- {
- w = t->to_state;
- if (NULL != w && w->index < 0)
- {
- scc_tarjan_strongconnect (scc_counter, w, index, stack, stack_size);
- v->lowlink = (v->lowlink > w->lowlink) ? w->lowlink : v->lowlink;
- }
- else if (0 != w->contained)
- v->lowlink = (v->lowlink > w->index) ? w->index : v->lowlink;
- }
-
- if (v->lowlink == v->index)
- {
- w = stack[--(*stack_size)];
- w->contained = 0;
-
- if (v != w)
- {
- (*scc_counter)++;
- while (v != w)
- {
- w->scc_id = *scc_counter;
- w = stack[--(*stack_size)];
- w->contained = 0;
- }
- w->scc_id = *scc_counter;
- }
- }
-}
-
-/**
- * Detect all SCCs (Strongly Connected Components) inside the given automaton.
- * SCCs will be marked using the scc_id on each state.
- *
- * @param ctx context
- * @param a automaton
- */
-static void
-scc_tarjan (struct GNUNET_REGEX_Automaton *a)
-{
- unsigned int index;
- unsigned int scc_counter;
- struct GNUNET_REGEX_State *v;
- struct GNUNET_REGEX_State *stack[a->state_count];
- unsigned int stack_size;
-
- for (v = a->states_head; NULL != v; v = v->next)
- {
- v->contained = 0;
- v->index = -1;
- v->lowlink = -1;
- }
-
- stack_size = 0;
- index = 0;
- scc_counter = 0;
-
- for (v = a->states_head; NULL != v; v = v->next)
- {
- if (v->index < 0)
- scc_tarjan_strongconnect (&scc_counter, v, &index, stack, &stack_size);
- }
-}
/**
* Adds a transition from one state to another on 'label'. Does not add
struct GNUNET_REGEX_State *to_state)
{
int is_dup;
- struct Transition *t;
- struct Transition *oth;
+ struct GNUNET_REGEX_Transition *t;
+ struct GNUNET_REGEX_Transition *oth;
if (NULL == from_state)
{
}
}
- if (is_dup)
+ if (GNUNET_YES == is_dup)
return;
// sort transitions by label
break;
}
- t = GNUNET_malloc (sizeof (struct Transition));
+ t = GNUNET_malloc (sizeof (struct GNUNET_REGEX_Transition));
t->id = ctx->transition_id++;
t->label = label;
t->to_state = to_state;
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'
*
}
}
+
/**
* 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;
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
+ // 1. Make all transitions pointing to s2 point to s1, unless this transition
+ // does not already exists, if it already exists remove transition.
for (s_check = a->states_head; NULL != s_check; s_check = s_check->next)
{
- for (t_check = s_check->transitions_head; NULL != t_check;
- t_check = t_check->next)
+ for (t_check = s_check->transitions_head; NULL != t_check; t_check = t_next)
{
+ t_next = t_check->next;
+
if (s2 == t_check->to_state)
- t_check->to_state = s1;
+ {
+ 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);
+ }
}
}
automaton_destroy_state (s2);
}
+
/**
* Add a state to the automaton 'a', always use this function to alter the
* states DLL of the automaton.
a->state_count++;
}
-/**
- * Function that is called with each state, when traversing an automaton.
- *
- * @param cls closure.
- * @param count current count of the state, from 0 to a->state_count -1.
- * @param s state.
- */
-typedef void (*GNUNET_REGEX_traverse_action) (void *cls, unsigned int count,
- struct GNUNET_REGEX_State * s);
/**
* Depth-first traversal of all states that are reachable from state 's'. Expects the states to
* @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)
+automaton_state_traverse (struct GNUNET_REGEX_State *s, unsigned int *count,
+ GNUNET_REGEX_traverse_action action, void *action_cls)
{
- struct Transition *t;
+ struct GNUNET_REGEX_Transition *t;
if (GNUNET_NO != s->marked)
return;
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);
+ automaton_state_traverse (t->to_state, count, action, action_cls);
}
* @param action action to be performed on each state.
* @param action_cls closure for action
*/
-static void
-automaton_traverse (struct GNUNET_REGEX_Automaton *a,
- GNUNET_REGEX_traverse_action action,
- void *action_cls)
+void
+GNUNET_REGEX_automaton_traverse (struct GNUNET_REGEX_Automaton *a,
+ GNUNET_REGEX_traverse_action action,
+ void *action_cls)
{
unsigned int count;
struct GNUNET_REGEX_State *s;
* Check if the given string 'str' needs parentheses around it when
* using it to generate a regex.
*
- * Currently only tests for first and last characters being '()' respectively.
- * FIXME: What about "(ab)|(cd)"?
- *
* @param str string
*
* @return GNUNET_YES if parentheses are needed, GNUNET_NO otherwise
const char *pos;
unsigned int cnt;
- if ( (NULL == str) ||
- ((slen = strlen(str)) < 2) )
+ if ((NULL == str) || ((slen = strlen (str)) < 2))
return GNUNET_NO;
-
+
if ('(' != str[0])
return GNUNET_YES;
cnt = 1;
return GNUNET_YES;
}
op = strchr (pos, '(');
- if ( (NULL != op) && (op < cl))
+ if ((NULL != op) && (op < cl))
{
cnt++;
pos = op + 1;
/**
* Remove parentheses surrounding string 'str'.
- * Example: "(a)" becomes "a".
+ * Example: "(a)" becomes "a", "(a|b)|(a|c)" stays the same.
* You need to GNUNET_free the returned string.
*
- * Currently only tests for first and last characters being '()' respectively.
- * FIXME: What about "(ab)|(cd)"?
- *
* @param str string, free'd or re-used by this function, can be NULL
*
* @return string without surrounding parentheses, string 'str' if no preceding
remove_parentheses (char *str)
{
size_t slen;
+ const char *pos;
- if ( (NULL == str) || ('(' != str[0]) || (str[(slen = strlen(str)) - 1] != ')') )
+ if ((NULL == str) || ('(' != str[0]) ||
+ (str[(slen = strlen (str)) - 1] != ')'))
return str;
- memmove (str, &str[1], slen - 2);
- str[slen - 2] = '\0';
+
+ pos = strchr (&str[1], ')');
+ if (pos == &str[slen - 1])
+ {
+ memmove (str, &str[1], slen - 2);
+ str[slen - 2] = '\0';
+ }
return str;
}
static int
has_epsilon (const char *str)
{
- return (NULL != str) && ('(' == str[0]) && ('|' == str[1]) && (')' == str[strlen(str) - 1]);
+ return (NULL != str) && ('(' == str[0]) && ('|' == str[1]) &&
+ (')' == str[strlen (str) - 1]);
}
if (NULL == str)
return NULL;
- if ( ('(' == str[0]) && ('|' == str[1]) )
+ if (('(' == str[0]) && ('|' == str[1]))
{
len = strlen (str);
- if (')' == str[len-1])
+ 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) )
+ 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.
*
- * @return 0 if the strings are the same, 1 or -1 if not
+ * @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) )
+ if ((NULL == str1) || (NULL == str2))
return -1;
return strcmp (str1, str2);
}
-/**
- * Helper function used as 'action' in 'automaton_traverse' function to create
+
+/**
+ * Helper function used as 'action' in 'GNUNET_REGEX_automaton_traverse' function to create
* the depth-first numbering of the states.
- *
+ *
* @param cls states array.
* @param count current state counter.
* @param s current state.
/**
- * 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.
+ * 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 (struct GNUNET_REGEX_Automaton *a)
+automaton_create_proofs_simplify (char *R_last_ij, char *R_last_ik,
+ char *R_last_kk, char *R_last_kj,
+ char **R_cur_ij)
{
- unsigned int n = a->state_count;
- struct GNUNET_REGEX_State *states[n];
- char *R_last[n][n];
- char *R_cur[n][n];
- struct Transition *t;
char *R_cur_l;
char *R_cur_r;
char *temp_a;
char *R_temp_ik;
char *R_temp_kj;
char *R_temp_kk;
+
+ int eps_check;
+ int ij_ik_cmp;
+ int ij_kj_cmp;
+
+ int ik_kk_cmp;
+ int kk_kj_cmp;
+ int clean_ik_kk_cmp;
+ int clean_kk_kj_cmp;
+ unsigned int cnt;
+
+ size_t length;
+ size_t length_l;
+ size_t length_r;
+
+ GNUNET_assert (NULL == *R_cur_ij && NULL != R_cur_ij);
+
+ // $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj}
+ // R_last == R^{(k-1)}, R_cur == R^{(k)}
+ // R_cur_ij = R_cur_l | R_cur_r
+ // R_cur_l == R^{(k-1)}_{ij}
+ // R_cur_r == R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj}
+
+ if ((NULL == R_last_ij) && ((NULL == R_last_ik) || (NULL == R_last_kk) || /* technically cannot happen, but looks saner */
+ (NULL == R_last_kj)))
+ {
+ /* R^{(k)}_{ij} = N | N */
+ *R_cur_ij = NULL;
+ return;
+ }
+
+ if ((NULL == R_last_ik) || (NULL == R_last_kk) || /* technically cannot happen, but looks saner */
+ (NULL == R_last_kj))
+ {
+ /* R^{(k)}_{ij} = R^{(k-1)}_{ij} | N */
+ *R_cur_ij = GNUNET_strdup (R_last_ij);
+ return;
+ }
+
+ // $R^{(k)}_{ij} = N | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj} OR
+ // $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj}
+
+ R_cur_r = NULL;
+ R_cur_l = NULL;
+
+ // cache results from strcmp, we might need these many times
+ ij_kj_cmp = nullstrcmp (R_last_ij, R_last_kj);
+ ij_ik_cmp = nullstrcmp (R_last_ij, R_last_ik);
+ ik_kk_cmp = nullstrcmp (R_last_ik, R_last_kk);
+ kk_kj_cmp = nullstrcmp (R_last_kk, R_last_kj);
+
+ // Assign R_temp_(ik|kk|kj) to R_last[][] and remove epsilon as well
+ // as parentheses, so we can better compare the contents
+ R_temp_ik = remove_parentheses (remove_epsilon (R_last_ik));
+ R_temp_kk = remove_parentheses (remove_epsilon (R_last_kk));
+ R_temp_kj = remove_parentheses (remove_epsilon (R_last_kj));
+
+ clean_ik_kk_cmp = nullstrcmp (R_last_ik, R_temp_kk);
+ clean_kk_kj_cmp = nullstrcmp (R_temp_kk, R_last_kj);
+
+ // construct R_cur_l (and, if necessary R_cur_r)
+ if (NULL != R_last_ij)
+ {
+ // Assign R_temp_ij to R_last_ij and remove epsilon as well
+ // as parentheses, so we can better compare the contents
+ R_temp_ij = remove_parentheses (remove_epsilon (R_last_ij));
+
+ if (0 == strcmp (R_temp_ij, R_temp_ik) && 0 == strcmp (R_temp_ik, R_temp_kk)
+ && 0 == strcmp (R_temp_kk, R_temp_kj))
+ {
+ if (0 == strlen (R_temp_ij))
+ {
+ R_cur_r = GNUNET_strdup ("");
+ }
+ else if ((0 == strncmp (R_last_ij, "(|", 2)) ||
+ (0 == strncmp (R_last_ik, "(|", 2) &&
+ 0 == strncmp (R_last_kj, "(|", 2)))
+ {
+ // a|(e|a)a*(e|a) = a*
+ // a|(e|a)(e|a)*(e|a) = a*
+ // (e|a)|aa*a = a*
+ // (e|a)|aa*(e|a) = a*
+ // (e|a)|(e|a)a*a = a*
+ // (e|a)|(e|a)a*(e|a) = a*
+ // (e|a)|(e|a)(e|a)*(e|a) = a*
+ if (GNUNET_YES == needs_parentheses (R_temp_ij))
+ GNUNET_asprintf (&R_cur_r, "(%s)*", R_temp_ij);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s*", R_temp_ij);
+ }
+ else
+ {
+ // a|aa*a = a+
+ // a|(e|a)a*a = a+
+ // a|aa*(e|a) = a+
+ // a|(e|a)(e|a)*a = a+
+ // a|a(e|a)*(e|a) = a+
+ if (GNUNET_YES == needs_parentheses (R_temp_ij))
+ GNUNET_asprintf (&R_cur_r, "(%s)+", R_temp_ij);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s+", R_temp_ij);
+ }
+ }
+ else if (0 == ij_ik_cmp && 0 == clean_kk_kj_cmp && 0 != clean_ik_kk_cmp)
+ {
+ // a|ab*b = ab*
+ if (strlen (R_last_kk) < 1)
+ R_cur_r = GNUNET_strdup (R_last_ij);
+ else if (GNUNET_YES == needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last_ij, R_temp_kk);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s%s*", R_last_ij, R_last_kk);
+
+ R_cur_l = NULL;
+ }
+ else if (0 == ij_kj_cmp && 0 == clean_ik_kk_cmp && 0 != clean_kk_kj_cmp)
+ {
+ // a|bb*a = b*a
+ if (strlen (R_last_kk) < 1)
+ R_cur_r = GNUNET_strdup (R_last_kj);
+ else if (GNUNET_YES == needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk, R_last_kj);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last_kj);
+
+ R_cur_l = NULL;
+ }
+ else if (0 == ij_ik_cmp && 0 == kk_kj_cmp && !has_epsilon (R_last_ij) &&
+ has_epsilon (R_last_kk))
+ {
+ // a|a(e|b)*(e|b) = a|ab* = a|a|ab|abb|abbb|... = ab*
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last_ij, R_temp_kk);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s%s*", R_last_ij, R_temp_kk);
+
+ R_cur_l = NULL;
+ }
+ else if (0 == ij_kj_cmp && 0 == ik_kk_cmp && !has_epsilon (R_last_ij) &&
+ has_epsilon (R_last_kk))
+ {
+ // a|(e|b)(e|b)*a = a|b*a = a|a|ba|bba|bbba|... = b*a
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk, R_last_ij);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last_ij);
+
+ R_cur_l = NULL;
+ }
+ else
+ {
+ temp_a = (NULL == R_last_ij) ? NULL : GNUNET_strdup (R_last_ij);
+ temp_a = remove_parentheses (temp_a);
+ R_cur_l = temp_a;
+ }
+
+ GNUNET_free_non_null (R_temp_ij);
+ }
+ else
+ {
+ // we have no left side
+ R_cur_l = NULL;
+ }
+
+ // construct R_cur_r, if not already constructed
+ if (NULL == R_cur_r)
+ {
+ length = strlen (R_temp_kk) - strlen (R_last_ik);
+
+ // a(ba)*bx = (ab)+x
+ if (length > 0 && NULL != R_last_kk && 0 < strlen (R_last_kk) &&
+ NULL != R_last_kj && 0 < strlen (R_last_kj) && NULL != R_last_ik &&
+ 0 < strlen (R_last_ik) && 0 == strkcmp (R_temp_kk, R_last_ik, length) &&
+ 0 == strncmp (R_temp_kk, R_last_kj, length))
+ {
+ temp_a = GNUNET_malloc (length + 1);
+ temp_b = GNUNET_malloc ((strlen (R_last_kj) - length) + 1);
+
+ length_l = 0;
+ length_r = 0;
+
+ for (cnt = 0; cnt < strlen (R_last_kj); cnt++)
+ {
+ if (cnt < length)
+ {
+ temp_a[length_l] = R_last_kj[cnt];
+ length_l++;
+ }
+ else
+ {
+ temp_b[length_r] = R_last_kj[cnt];
+ length_r++;
+ }
+ }
+ temp_a[length_l] = '\0';
+ temp_b[length_r] = '\0';
+
+ // e|(ab)+ = (ab)*
+ if (NULL != R_cur_l && 0 == strlen (R_cur_l) && 0 == strlen (temp_b))
+ {
+ GNUNET_asprintf (&R_cur_r, "(%s%s)*", R_last_ik, temp_a);
+ GNUNET_free (R_cur_l);
+ R_cur_l = NULL;
+ }
+ else
+ {
+ GNUNET_asprintf (&R_cur_r, "(%s%s)+%s", R_last_ik, temp_a, temp_b);
+ }
+ GNUNET_free (temp_a);
+ GNUNET_free (temp_b);
+ }
+ else if (0 == strcmp (R_temp_ik, R_temp_kk) &&
+ 0 == strcmp (R_temp_kk, R_temp_kj))
+ {
+ // (e|a)a*(e|a) = a*
+ // (e|a)(e|a)*(e|a) = a*
+ if (has_epsilon (R_last_ik) && has_epsilon (R_last_kj))
+ {
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "(%s)*", R_temp_kk);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s*", R_temp_kk);
+ }
+ // aa*a = a+a
+ else if (0 == clean_ik_kk_cmp && 0 == clean_kk_kj_cmp &&
+ !has_epsilon (R_last_ik))
+ {
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_temp_kk, R_temp_kk);
+ else
+ GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_temp_kk, R_temp_kk);
+ }
+ // (e|a)a*a = a+
+ // aa*(e|a) = a+
+ // a(e|a)*(e|a) = a+
+ // (e|a)a*a = a+
+ else
+ {
+ eps_check =
+ (has_epsilon (R_last_ik) + has_epsilon (R_last_kk) +
+ has_epsilon (R_last_kj));
+
+ if (eps_check == 1)
+ {
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "(%s)+", R_temp_kk);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s+", R_temp_kk);
+ }
+ }
+ }
+ // aa*b = a+b
+ // (e|a)(e|a)*b = a*b
+ else if (0 == strcmp (R_temp_ik, R_temp_kk))
+ {
+ if (has_epsilon (R_last_ik))
+ {
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk, R_last_kj);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last_kj);
+ }
+ else
+ {
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_temp_kk, R_last_kj);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s+%s", R_temp_kk, R_last_kj);
+ }
+ }
+ // ba*a = ba+
+ // b(e|a)*(e|a) = ba*
+ else if (0 == strcmp (R_temp_kk, R_temp_kj))
+ {
+ if (has_epsilon (R_last_kj))
+ {
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last_ik, R_temp_kk);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s%s*", R_last_ik, R_temp_kk);
+ }
+ else
+ {
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_last_ik, R_temp_kk);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s+%s", R_last_ik, R_temp_kk);
+ }
+ }
+ else
+ {
+ if (strlen (R_temp_kk) > 0)
+ {
+ if (needs_parentheses (R_temp_kk))
+ {
+ GNUNET_asprintf (&R_cur_r, "%s(%s)*%s", R_last_ik, R_temp_kk,
+ R_last_kj);
+ }
+ else
+ {
+ GNUNET_asprintf (&R_cur_r, "%s%s*%s", R_last_ik, R_temp_kk,
+ R_last_kj);
+ }
+ }
+ else
+ {
+ GNUNET_asprintf (&R_cur_r, "%s%s", R_last_ik, R_last_kj);
+ }
+ }
+ }
+
+ GNUNET_free_non_null (R_temp_ik);
+ GNUNET_free_non_null (R_temp_kk);
+ GNUNET_free_non_null (R_temp_kj);
+
+ if (NULL == R_cur_l && NULL == R_cur_r)
+ {
+ *R_cur_ij = NULL;
+ return;
+ }
+
+ if (NULL != R_cur_l && NULL == R_cur_r)
+ {
+ *R_cur_ij = R_cur_l;
+ return;
+ }
+
+ if (NULL == R_cur_l && NULL != R_cur_r)
+ {
+ *R_cur_ij = R_cur_r;
+ return;
+ }
+
+ if (0 == nullstrcmp (R_cur_l, R_cur_r))
+ {
+ *R_cur_ij = R_cur_l;
+ GNUNET_free (R_cur_r);
+ return;
+ }
+
+ GNUNET_asprintf (R_cur_ij, "(%s|%s)", R_cur_l, R_cur_r);
+
+ GNUNET_free (R_cur_l);
+ GNUNET_free (R_cur_r);
+}
+
+
+/**
+ * create proofs for all states in the given automaton. Implementation of the
+ * algorithm descriped in chapter 3.2.1 of "Automata Theory, Languages, and
+ * Computation 3rd Edition" by Hopcroft, Motwani and Ullman.
+ *
+ * @param a automaton.
+ */
+static void
+automaton_create_proofs (struct GNUNET_REGEX_Automaton *a)
+{
+ 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;
- int cnt;
- int eps_check;
- int ij_ik_cmp;
- int ij_kj_cmp;
- int ik_kj_cmp;
- int ik_kk_cmp;
- int kk_kj_cmp;
- int clean_ik_kk_cmp;
- int clean_kk_kj_cmp;
- int length;
- int length_l;
- int length_r;
+
/* create depth-first numbering of the states, initializes 'state' */
- automaton_traverse (a, &number_states, states);
+ GNUNET_REGEX_automaton_traverse (a, &number_states, states);
/* Compute regular expressions of length "1" between each pair of states */
for (i = 0; i < n; i++)
{
- for (j=0;j<n;j++)
+ 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;
+ j = t->to_state->proof_id;
if (NULL == R_last[i][j])
- GNUNET_asprintf (&R_last[i][j], "%c", t->label);
+ GNUNET_asprintf (&R_last[i][j], "%c", t->label);
else
- {
- temp_a = R_last[i][j];
- GNUNET_asprintf (&R_last[i][j], "%s|%c", R_last[i][j], t->label);
- GNUNET_free (temp_a);
- }
- if (GNUNET_YES == needs_parentheses (R_last[i][j]))
- {
- temp_a = R_last[i][j];
- GNUNET_asprintf (&R_last[i][j], "(%s)", R_last[i][j]);
- GNUNET_free (temp_a);
- }
+ {
+ 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_a = R_last[i][i];
- GNUNET_asprintf (&R_last[i][i], "(|%s)", R_last[i][i]);
- GNUNET_free (temp_a);
- }
+ {
+ 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);
+ }
-
- // INDUCTION
+ /* 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++)
{
- /* GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, */
- /* ">>> R_last[i][j] = %s R_last[i][k] = %s " */
- /* "R_last[k][k] = %s R_last[k][j] = %s\n", R_last[i][j], */
- /* R_last[i][k], R_last[k][k], R_last[k][j]); */
-
- R_cur[i][j] = NULL;
- R_cur_r = NULL;
- R_cur_l = NULL;
-
- // cache results from strcmp, we might need these many times
- ij_kj_cmp = nullstrcmp (R_last[i][j], R_last[k][j]);
- ij_ik_cmp = nullstrcmp (R_last[i][j], R_last[i][k]);
- ik_kk_cmp = nullstrcmp (R_last[i][k], R_last[k][k]);
- ik_kj_cmp = nullstrcmp (R_last[i][k], R_last[k][j]);
- kk_kj_cmp = nullstrcmp (R_last[k][k], R_last[k][j]);
-
- // $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk})^* R^{(k-1)}_{kj}
- // With: R_cur[i][j] = R_cur_l | R_cur_r
- // Rij(k) = Rij(k-1), because right side (R_cur_r) is empty set (NULL)
- if ((NULL == R_last[i][k] || NULL == R_last[k][j] ||
- NULL == R_last[k][k]) && NULL != R_last[i][j])
- {
- R_cur[i][j] = GNUNET_strdup (R_last[i][j]);
- }
- // Everything is NULL, so Rij(k) = NULL
- else if ((NULL == R_last[i][k] || NULL == R_last[k][j] ||
- NULL == R_last[k][k]) && NULL == R_last[i][j])
- {
- R_cur[i][j] = NULL;
- }
- // Right side (R_cur_r) not NULL
- else
- {
- /* GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, */
- /* "R_temp_ij = %s R_temp_ik = %s R_temp_kk = %s R_temp_kj = %s\n", */
- /* R_temp_ij, R_temp_ik, R_temp_kk, R_temp_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[i][k]));
- R_temp_kk = remove_parentheses (remove_epsilon (R_last[k][k]));
- R_temp_kj = remove_parentheses (remove_epsilon (R_last[k][j]));
-
- clean_ik_kk_cmp = nullstrcmp (R_last[i][k], R_temp_kk);
- clean_kk_kj_cmp = nullstrcmp (R_temp_kk, R_last[k][j]);
-
- // construct R_cur_l (and, if necessary R_cur_r)
- if (NULL != R_last[i][j])
- {
- // Assign R_temp_ij to R_last[i][j] and remove epsilon as well
- // as parentheses, so we can better compare the contents
- R_temp_ij = remove_parentheses (remove_epsilon (R_last[i][j]));
-
- 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 ("");
- }
- // 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*
- else if ((0 == strncmp (R_last[i][j], "(|", 2)) ||
- (0 == strncmp (R_last[i][k], "(|", 2) &&
- 0 == strncmp (R_last[k][j], "(|", 2)))
- {
- 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);
- }
- // 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+
- else
- {
- 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);
- }
- }
- // a|ab*b = ab*
- else if (0 == ij_ik_cmp && 0 == clean_kk_kj_cmp &&
- 0 != clean_ik_kk_cmp)
- {
- if (strlen (R_last[k][k]) < 1)
- R_cur_r = GNUNET_strdup (R_last[i][j]);
- else if (GNUNET_YES == needs_parentheses (R_temp_kk))
- GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last[i][j], R_temp_kk);
- else
- GNUNET_asprintf (&R_cur_r, "%s%s*", R_last[i][j], R_last[k][k]);
-
- R_cur_l = NULL;
- }
- // a|bb*a = b*a
- else if (0 == ij_kj_cmp && 0 == clean_ik_kk_cmp &&
- 0 != clean_kk_kj_cmp)
- {
- if (strlen (R_last[k][k]) < 1)
- R_cur_r = GNUNET_strdup (R_last[k][j]);
- else if (GNUNET_YES == needs_parentheses (R_temp_kk))
- GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk, R_last[k][j]);
- else
- GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last[k][j]);
-
- R_cur_l = NULL;
- }
- // a|a(e|b)*(e|b) = a|ab* = a|a|ab|abb|abbb|... = ab*
- else if (0 == ij_ik_cmp && 0 == kk_kj_cmp &&
- !has_epsilon (R_last[i][j]) && has_epsilon (R_last[k][k]))
- {
- if (needs_parentheses (R_temp_kk))
- GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last[i][j], R_temp_kk);
- else
- GNUNET_asprintf (&R_cur_r, "%s%s*", R_last[i][j], R_temp_kk);
-
- R_cur_l = NULL;
- }
- // a|(e|b)(e|b)*a = a|b*a = a|a|ba|bba|bbba|... = b*a
- else if (0 == ij_kj_cmp && 0 == ik_kk_cmp &&
- !has_epsilon (R_last[i][j]) && has_epsilon (R_last[k][k]))
- {
- if (needs_parentheses (R_temp_kk))
- GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk, R_last[i][j]);
- else
- GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last[i][j]);
-
- R_cur_l = NULL;
- }
- else
- {
- /* GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "NO SIMPLIFICATION\n"); */
- temp_a = (NULL == R_last[i][j]) ? NULL : GNUNET_strdup (R_last[i][j]);
- temp_a = remove_parentheses (temp_a);
- R_cur_l = temp_a;
- }
-
- GNUNET_free_non_null (R_temp_ij);
- }
- // we have no left side
- else
- {
- 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[i][k]);
-
- // a(ba)*bx = (ab)+x
- if (length > 0 && NULL != R_last[k][k] && 0 < strlen (R_last[k][k])
- && NULL != R_last[k][j] && 0 < strlen (R_last[k][j]) &&
- NULL != R_last[i][k] && 0 < strlen (R_last[i][k]) &&
- 0 == strkcmp (R_temp_kk, R_last[i][k], length) &&
- 0 == strncmp (R_temp_kk, R_last[k][j], length))
- {
- temp_a = GNUNET_malloc (length + 1);
- temp_b = GNUNET_malloc ((strlen (R_last[k][j]) - length) + 1);
-
- length_l = 0;
- length_r = 0;
-
- for (cnt = 0; cnt < strlen (R_last[k][j]); cnt++)
- {
- if (cnt < length)
- {
- temp_a[length_l] = R_last[k][j][cnt];
- length_l++;
- }
- else
- {
- temp_b[length_r] = R_last[k][j][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[i][k], temp_a);
- GNUNET_free (R_cur_l);
- R_cur_l = NULL;
- }
- else
- {
- GNUNET_asprintf (&R_cur_r, "(%s%s)+%s", R_last[i][k], 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[i][k]) && has_epsilon (R_last[k][j]))
- {
- 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[i][k]))
- {
- 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[i][k]) + has_epsilon (R_last[k][k]) +
- has_epsilon (R_last[k][j]));
-
- 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[i][k]))
- {
- if (needs_parentheses (R_temp_kk))
- GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk,
- R_last[k][j]);
- else
- GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last[k][j]);
- }
- else
- {
- if (needs_parentheses (R_temp_kk))
- GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_temp_kk,
- R_last[k][j]);
- else
- GNUNET_asprintf (&R_cur_r, "%s+%s", R_temp_kk, R_last[k][j]);
- }
- }
- // ba*a = ba+
- // b(e|a)*(e|a) = ba*
- else if (0 == strcmp (R_temp_kk, R_temp_kj))
- {
- if (has_epsilon (R_last[k][j]))
- {
- if (needs_parentheses (R_temp_kk))
- GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last[i][k],
- R_temp_kk);
- else
- GNUNET_asprintf (&R_cur_r, "%s%s*", R_last[i][k], R_temp_kk);
- }
- else
- {
- if (needs_parentheses (R_temp_kk))
- GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_last[i][k],
- R_temp_kk);
- else
- GNUNET_asprintf (&R_cur_r, "%s+%s", R_last[i][k], 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[i][k],
- R_temp_kk, R_last[k][j]);
- }
- else
- {
- GNUNET_asprintf (&R_cur_r, "%s%s*%s", R_last[i][k], R_temp_kk,
- R_last[k][j]);
- }
- }
- else
- {
- GNUNET_asprintf (&R_cur_r, "%s%s", R_last[i][k], R_last[k][j]);
- }
- }
- }
-
- /* GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "R_cur_l: %s\n", R_cur_l); */
- /* GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "R_cur_r: %s\n", R_cur_r); */
-
- // putting it all together
- if (NULL != R_cur_l && NULL != R_cur_r)
- {
- // a|a = a
- if (0 == strcmp (R_cur_l, R_cur_r))
- {
- R_cur[i][j] = GNUNET_strdup (R_cur_l);
- }
- // R_cur_l | R_cur_r
- else
- {
- GNUNET_asprintf (&R_cur[i][j], "(%s|%s)", R_cur_l, R_cur_r);
- }
- }
- else if (NULL != R_cur_l)
- {
- R_cur[i][j] = GNUNET_strdup (R_cur_l);
- }
- else if (NULL != R_cur_r)
- {
- R_cur[i][j] = GNUNET_strdup (R_cur_r);
- }
- else
- {
- R_cur[i][j] = NULL;
- }
-
- GNUNET_free_non_null (R_cur_l);
- GNUNET_free_non_null (R_cur_r);
-
- GNUNET_free_non_null (R_temp_ik);
- GNUNET_free_non_null (R_temp_kk);
- GNUNET_free_non_null (R_temp_kj);
- }
+ // 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]);
}
}
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;
+ R_last[i][j] = R_cur[i][j];
+ R_cur[i][j] = NULL;
}
}
}
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_a = complete_regex;
+ temp = complete_regex;
GNUNET_asprintf (&complete_regex, "%s|%s", complete_regex,
R_last[a->start->proof_id][i]);
- GNUNET_free (temp_a);
+ GNUNET_free (temp);
}
}
}
- a->computed_regex = complete_regex;
-
- GNUNET_log (GNUNET_ERROR_TYPE_DEBUG,
- "---------------------------------------------\n");
- GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Regex: %s\n", a->regex);
- GNUNET_log (GNUNET_ERROR_TYPE_DEBUG, "Complete Regex: %s\n", complete_regex);
- GNUNET_log (GNUNET_ERROR_TYPE_DEBUG,
- "---------------------------------------------\n");
+ a->canonical_regex = complete_regex;
// cleanup
for (i = 0; i < n; i++)
}
}
+
/**
* Creates a new DFA state based on a set of NFA states. Needs to be freed using
* automaton_destroy_state.
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++;
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)
return new_s;
}
+
/**
* Remove all unreachable states from DFA 'a'. Unreachable states are those
* states that are not reachable from the starting state.
s->marked = GNUNET_NO;
// 2. traverse dfa from start state and mark all visited states
- automaton_traverse (a, NULL, NULL);
+ GNUNET_REGEX_automaton_traverse (a, NULL, NULL);
// 3. delete all states that were not visited
for (s = a->states_head; NULL != s; s = s_next)
}
}
+
/**
* 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 num_equal_edges;
+ unsigned int num_equal_edges;
for (i = 0, s1 = a->states_head; i < a->state_count && NULL != s1;
i++, s1 = s1->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;
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->stack_tail = NULL;
}
+
/**
* Construct an NFA by parsing the regex string of length 'len'.
*
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;
}
}
+
/**
* Construct DFA for the given 'regex' of length 'len'
*
return dfa;
}
+
/**
* Free the memory allocated by constructing the GNUNET_REGEX_Automaton data
* structure.
return;
GNUNET_free_non_null (a->regex);
- GNUNET_free_non_null (a->computed_regex);
+ GNUNET_free_non_null (a->canonical_regex);
for (s = a->states_head; NULL != s;)
{
GNUNET_free (a);
}
-/**
- * Save a state to an open file pointer. cls is expected to be a file pointer to
- * an open file. Used only in conjunction with
- * GNUNET_REGEX_automaton_save_graph.
- *
- * @param cls file pointer.
- * @param count current count of the state, not used.
- * @param s state.
- */
-void
-GNUNET_REGEX_automaton_save_graph_step (void *cls, unsigned int count,
- struct GNUNET_REGEX_State *s)
-{
- FILE *p;
- struct Transition *ctran;
- char *s_acc = NULL;
- char *s_tran = NULL;
-
- p = cls;
-
- if (s->accepting)
- {
- GNUNET_asprintf (&s_acc,
- "\"%s(%i)\" [shape=doublecircle, color=\"0.%i 0.8 0.95\"];\n",
- s->name, s->proof_id, s->scc_id);
- }
- else
- {
- GNUNET_asprintf (&s_acc, "\"%s(%i)\" [color=\"0.%i 0.8 0.95\"];\n", s->name,
- s->proof_id, s->scc_id);
- }
-
- if (NULL == s_acc)
- {
- GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not print state %s\n", s->name);
- return;
- }
- fwrite (s_acc, strlen (s_acc), 1, p);
- GNUNET_free (s_acc);
- s_acc = NULL;
-
- for (ctran = s->transitions_head; NULL != ctran; ctran = ctran->next)
- {
- if (NULL == ctran->to_state)
- {
- GNUNET_log (GNUNET_ERROR_TYPE_ERROR,
- "Transition from State %i has no state for transitioning\n",
- s->id);
- continue;
- }
-
- if (ctran->label == 0)
- {
- GNUNET_asprintf (&s_tran,
- "\"%s(%i)\" -> \"%s(%i)\" [label = \"epsilon\", color=\"0.%i 0.8 0.95\"];\n",
- s->name, s->proof_id, ctran->to_state->name,
- ctran->to_state->proof_id, s->scc_id);
- }
- else
- {
- GNUNET_asprintf (&s_tran,
- "\"%s(%i)\" -> \"%s(%i)\" [label = \"%c\", color=\"0.%i 0.8 0.95\"];\n",
- s->name, s->proof_id, ctran->to_state->name,
- ctran->to_state->proof_id, ctran->label, s->scc_id);
- }
-
- if (NULL == s_tran)
- {
- GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not print state %s\n",
- s->name);
- return;
- }
-
- fwrite (s_tran, strlen (s_tran), 1, p);
- GNUNET_free (s_tran);
- s_tran = NULL;
- }
-}
-
-/**
- * Save the given automaton as a GraphViz dot file
- *
- * @param a the automaton to be saved
- * @param filename where to save the file
- */
-void
-GNUNET_REGEX_automaton_save_graph (struct GNUNET_REGEX_Automaton *a,
- const char *filename)
-{
- char *start;
- char *end;
- FILE *p;
-
- if (NULL == a)
- {
- GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not print NFA, was NULL!");
- return;
- }
-
- if (NULL == filename || strlen (filename) < 1)
- {
- GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "No Filename given!");
- return;
- }
-
- p = fopen (filename, "w");
-
- if (NULL == p)
- {
- GNUNET_log (GNUNET_ERROR_TYPE_ERROR, "Could not open file for writing: %s",
- filename);
- return;
- }
-
- /* First add the SCCs to the automaton, so we can color them nicely */
- scc_tarjan (a);
-
- start = "digraph G {\nrankdir=LR\n";
- fwrite (start, strlen (start), 1, p);
-
- automaton_traverse (a, &GNUNET_REGEX_automaton_save_graph_step, p);
-
- end = "\n}\n";
- fwrite (end, strlen (end), 1, p);
- fclose (p);
-}
/**
* Evaluates the given string using the given DFA automaton
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 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 computed regex of the given automaton.
+ * 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 computed regex is only useful for testing.
+ * 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_computed_regex (struct GNUNET_REGEX_Automaton *a)
+GNUNET_REGEX_get_canonical_regex (struct GNUNET_REGEX_Automaton *a)
{
if (NULL == a)
return NULL;
- return a->computed_regex;
+ 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,
- struct GNUNET_HashCode *key)
+size_t
+GNUNET_REGEX_get_first_key (const char *input_string, size_t string_len,
+ struct GNUNET_HashCode * key)
{
unsigned int size;
return size;
}
+
/**
* Check if the given 'proof' matches the given 'key'.
*
- * @param proof partial regex
- * @param key hash
+ * @param proof partial regex of a state.
+ * @param key hash of a state.
*
- * @return GNUNET_OK if the proof is valid for the given key
+ * @return GNUNET_OK if the proof is valid for the given key.
*/
int
GNUNET_REGEX_check_proof (const char *proof, const struct GNUNET_HashCode *key)
{
- return GNUNET_OK;
+ struct GNUNET_HashCode key_check;
+
+ GNUNET_CRYPTO_hash (proof, strlen (proof), &key_check);
+ return (0 ==
+ GNUNET_CRYPTO_hash_cmp (key, &key_check)) ? GNUNET_OK : GNUNET_NO;
+}
+
+/**
+ * Recursive helper function for iterate_initial_edges. Will call iterator
+ * function for each initial state.
+ *
+ * @param 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)
+{
+ iterate_initial_edge (initial_len + 1, 0, NULL, a->start, 0, iterator,
+ iterator_cls);
}
+
/**
* Iterate over all edges helper function starting from state 's', calling
* iterator on for each edge.
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;
}
}
+
/**
* 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);
}