+GNUNET_REGEX_automaton_traverse (struct GNUNET_REGEX_Automaton *a,
+ GNUNET_REGEX_traverse_action action,
+ void *action_cls)
+{
+ unsigned int count;
+ struct GNUNET_REGEX_State *s;
+
+ for (s = a->states_head; NULL != s; s = s->next)
+ s->marked = GNUNET_NO;
+ count = 0;
+ automaton_state_traverse (a->start, &count, action, action_cls);
+}
+
+
+/**
+ * Check if the given string 'str' needs parentheses around it when
+ * using it to generate a regex.
+ *
+ * @param str string
+ *
+ * @return GNUNET_YES if parentheses are needed, GNUNET_NO otherwise
+ */
+static int
+needs_parentheses (const char *str)
+{
+ size_t slen;
+ const char *op;
+ const char *cl;
+ const char *pos;
+ unsigned int cnt;
+
+ if ((NULL == str) || ((slen = strlen (str)) < 2))
+ return GNUNET_NO;
+
+ if ('(' != str[0])
+ return GNUNET_YES;
+ cnt = 1;
+ pos = &str[1];
+ while (cnt > 0)
+ {
+ cl = strchr (pos, ')');
+ if (NULL == cl)
+ {
+ GNUNET_break (0);
+ return GNUNET_YES;
+ }
+ op = strchr (pos, '(');
+ if ((NULL != op) && (op < cl))
+ {
+ cnt++;
+ pos = op + 1;
+ continue;
+ }
+ /* got ')' first */
+ cnt--;
+ pos = cl + 1;
+ }
+ return (*pos == '\0') ? GNUNET_NO : GNUNET_YES;
+}
+
+
+/**
+ * Remove parentheses surrounding string 'str'.
+ * Example: "(a)" becomes "a", "(a|b)|(a|c)" stays the same.
+ * You need to GNUNET_free the returned string.
+ *
+ * @param str string, free'd or re-used by this function, can be NULL
+ *
+ * @return string without surrounding parentheses, string 'str' if no preceding
+ * epsilon could be found, NULL if 'str' was NULL
+ */
+static char *
+remove_parentheses (char *str)
+{
+ size_t slen;
+ const char *pos;
+
+ if ((NULL == str) || ('(' != str[0]) ||
+ (str[(slen = strlen (str)) - 1] != ')'))
+ return str;
+
+ pos = strchr (&str[1], ')');
+ if (pos == &str[slen - 1])
+ {
+ memmove (str, &str[1], slen - 2);
+ str[slen - 2] = '\0';
+ }
+ return str;
+}
+
+
+/**
+ * Check if the string 'str' starts with an epsilon (empty string).
+ * Example: "(|a)" is starting with an epsilon.
+ *
+ * @param str string to test
+ *
+ * @return 0 if str has no epsilon, 1 if str starts with '(|' and ends with ')'
+ */
+static int
+has_epsilon (const char *str)
+{
+ return (NULL != str) && ('(' == str[0]) && ('|' == str[1]) &&
+ (')' == str[strlen (str) - 1]);
+}
+
+
+/**
+ * Remove an epsilon from the string str. Where epsilon is an empty string
+ * Example: str = "(|a|b|c)", result: "a|b|c"
+ * The returned string needs to be freed.
+ *
+ * @param str string
+ *
+ * @return string without preceding epsilon, string 'str' if no preceding epsilon
+ * could be found, NULL if 'str' was NULL
+ */
+static char *
+remove_epsilon (const char *str)
+{
+ size_t len;
+
+ if (NULL == str)
+ return NULL;
+ if (('(' == str[0]) && ('|' == str[1]))
+ {
+ len = strlen (str);
+ if (')' == str[len - 1])
+ return GNUNET_strndup (&str[2], len - 3);
+ }
+ return GNUNET_strdup (str);
+}
+
+
+/**
+ * Compare 'str1', starting from position 'k', with whole 'str2'
+ *
+ * @param str1 first string to compare, starting from position 'k'
+ * @param str2 second string for comparison
+ * @param k starting position in 'str1'
+ *
+ * @return -1 if any of the strings is NULL, 0 if equal, non 0 otherwise
+ */
+static int
+strkcmp (const char *str1, const char *str2, size_t k)
+{
+ if ((NULL == str1) || (NULL == str2) || (strlen (str1) < k))
+ return -1;
+ return strcmp (&str1[k], str2);
+}
+
+
+/**
+ * Compare two strings for equality. If either is NULL (or if both are
+ * NULL), they are not equal.
+ *
+ * @param str1 first string for comparison.
+ * @param str2 second string for comparison.
+ *
+ * @return 0 if the strings are the same, 1 or -1 if not
+ */
+static int
+nullstrcmp (const char *str1, const char *str2)
+{
+ if ((NULL == str1) || (NULL == str2))
+ return -1;
+ return strcmp (str1, str2);
+}
+
+
+/**
+ * Helper function used as 'action' in 'GNUNET_REGEX_automaton_traverse' function to create
+ * the depth-first numbering of the states.
+ *
+ * @param cls states array.
+ * @param count current state counter.
+ * @param s current state.
+ */
+static void
+number_states (void *cls, unsigned int count, struct GNUNET_REGEX_State *s)
+{
+ struct GNUNET_REGEX_State **states = cls;
+
+ s->proof_id = count;
+ if (NULL != states)
+ states[count] = s;
+}
+
+
+/**
+ * Construct the regular expression given the inductive step,
+ * $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^*
+ * R^{(k-1)}_{kj}, and simplify the resulting expression saved in R_cur_ij.
+ *
+ * @param R_last_ij value of $R^{(k-1)_{ij}.
+ * @param R_last_ik value of $R^{(k-1)_{ik}.
+ * @param R_last_kk value of $R^{(k-1)_{kk}.
+ * @param R_last_kj value of $R^{(k-1)_{kj}.
+ * @param R_cur_ij result for this inductive step is saved in R_cur_ij, R_cur_ij
+ * is expected to be NULL when called!
+ */
+static void
+automaton_create_proofs_simplify (char *R_last_ij, char *R_last_ik,
+ char *R_last_kk, char *R_last_kj,
+ char **R_cur_ij)
+{
+ char *R_cur_l;
+ char *R_cur_r;
+ char *temp_a;
+ char *temp_b;
+ char *R_temp_ij;
+ char *R_temp_ik;
+ char *R_temp_kj;
+ char *R_temp_kk;
+
+ int eps_check;
+ int ij_ik_cmp;
+ int ij_kj_cmp;
+
+ int ik_kk_cmp;
+ int kk_kj_cmp;
+ int clean_ik_kk_cmp;
+ int clean_kk_kj_cmp;
+ unsigned int cnt;
+
+ size_t length;
+ size_t length_l;
+ size_t length_r;
+
+ GNUNET_assert (NULL == *R_cur_ij && NULL != R_cur_ij);
+
+ // $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj}
+ // R_last == R^{(k-1)}, R_cur == R^{(k)}
+ // R_cur_ij = R_cur_l | R_cur_r
+ // R_cur_l == R^{(k-1)}_{ij}
+ // R_cur_r == R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj}
+
+ if ((NULL == R_last_ij) && ((NULL == R_last_ik) || (NULL == R_last_kk) || /* technically cannot happen, but looks saner */
+ (NULL == R_last_kj)))
+ {
+ /* R^{(k)}_{ij} = N | N */
+ *R_cur_ij = NULL;
+ return;
+ }
+
+ if ((NULL == R_last_ik) || (NULL == R_last_kk) || /* technically cannot happen, but looks saner */
+ (NULL == R_last_kj))
+ {
+ /* R^{(k)}_{ij} = R^{(k-1)}_{ij} | N */
+ *R_cur_ij = GNUNET_strdup (R_last_ij);
+ return;
+ }
+
+ // $R^{(k)}_{ij} = N | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj} OR
+ // $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj}
+
+ R_cur_r = NULL;
+ R_cur_l = NULL;
+
+ // cache results from strcmp, we might need these many times
+ ij_kj_cmp = nullstrcmp (R_last_ij, R_last_kj);
+ ij_ik_cmp = nullstrcmp (R_last_ij, R_last_ik);
+ ik_kk_cmp = nullstrcmp (R_last_ik, R_last_kk);
+ kk_kj_cmp = nullstrcmp (R_last_kk, R_last_kj);
+
+ // Assign R_temp_(ik|kk|kj) to R_last[][] and remove epsilon as well
+ // as parentheses, so we can better compare the contents
+ R_temp_ik = remove_parentheses (remove_epsilon (R_last_ik));
+ R_temp_kk = remove_parentheses (remove_epsilon (R_last_kk));
+ R_temp_kj = remove_parentheses (remove_epsilon (R_last_kj));
+
+ clean_ik_kk_cmp = nullstrcmp (R_last_ik, R_temp_kk);
+ clean_kk_kj_cmp = nullstrcmp (R_temp_kk, R_last_kj);
+
+ // construct R_cur_l (and, if necessary R_cur_r)
+ if (NULL != R_last_ij)
+ {
+ // Assign R_temp_ij to R_last_ij and remove epsilon as well
+ // as parentheses, so we can better compare the contents
+ R_temp_ij = remove_parentheses (remove_epsilon (R_last_ij));
+
+ if (0 == strcmp (R_temp_ij, R_temp_ik) && 0 == strcmp (R_temp_ik, R_temp_kk)
+ && 0 == strcmp (R_temp_kk, R_temp_kj))
+ {
+ if (0 == strlen (R_temp_ij))
+ {
+ R_cur_r = GNUNET_strdup ("");
+ }
+ else if ((0 == strncmp (R_last_ij, "(|", 2)) ||
+ (0 == strncmp (R_last_ik, "(|", 2) &&
+ 0 == strncmp (R_last_kj, "(|", 2)))
+ {
+ // a|(e|a)a*(e|a) = a*
+ // a|(e|a)(e|a)*(e|a) = a*
+ // (e|a)|aa*a = a*
+ // (e|a)|aa*(e|a) = a*
+ // (e|a)|(e|a)a*a = a*
+ // (e|a)|(e|a)a*(e|a) = a*
+ // (e|a)|(e|a)(e|a)*(e|a) = a*
+ if (GNUNET_YES == needs_parentheses (R_temp_ij))
+ GNUNET_asprintf (&R_cur_r, "(%s)*", R_temp_ij);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s*", R_temp_ij);
+ }
+ else
+ {
+ // a|aa*a = a+
+ // a|(e|a)a*a = a+
+ // a|aa*(e|a) = a+
+ // a|(e|a)(e|a)*a = a+
+ // a|a(e|a)*(e|a) = a+
+ if (GNUNET_YES == needs_parentheses (R_temp_ij))
+ GNUNET_asprintf (&R_cur_r, "(%s)+", R_temp_ij);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s+", R_temp_ij);
+ }
+ }
+ else if (0 == ij_ik_cmp && 0 == clean_kk_kj_cmp && 0 != clean_ik_kk_cmp)
+ {
+ // a|ab*b = ab*
+ if (strlen (R_last_kk) < 1)
+ R_cur_r = GNUNET_strdup (R_last_ij);
+ else if (GNUNET_YES == needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last_ij, R_temp_kk);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s%s*", R_last_ij, R_last_kk);
+
+ R_cur_l = NULL;
+ }
+ else if (0 == ij_kj_cmp && 0 == clean_ik_kk_cmp && 0 != clean_kk_kj_cmp)
+ {
+ // a|bb*a = b*a
+ if (strlen (R_last_kk) < 1)
+ R_cur_r = GNUNET_strdup (R_last_kj);
+ else if (GNUNET_YES == needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk, R_last_kj);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last_kj);
+
+ R_cur_l = NULL;
+ }
+ else if (0 == ij_ik_cmp && 0 == kk_kj_cmp && !has_epsilon (R_last_ij) &&
+ has_epsilon (R_last_kk))
+ {
+ // a|a(e|b)*(e|b) = a|ab* = a|a|ab|abb|abbb|... = ab*
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last_ij, R_temp_kk);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s%s*", R_last_ij, R_temp_kk);
+
+ R_cur_l = NULL;
+ }
+ else if (0 == ij_kj_cmp && 0 == ik_kk_cmp && !has_epsilon (R_last_ij) &&
+ has_epsilon (R_last_kk))
+ {
+ // a|(e|b)(e|b)*a = a|b*a = a|a|ba|bba|bbba|... = b*a
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk, R_last_ij);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last_ij);
+
+ R_cur_l = NULL;
+ }
+ else
+ {
+ temp_a = (NULL == R_last_ij) ? NULL : GNUNET_strdup (R_last_ij);
+ temp_a = remove_parentheses (temp_a);
+ R_cur_l = temp_a;
+ }
+
+ GNUNET_free_non_null (R_temp_ij);
+ }
+ else
+ {
+ // we have no left side
+ R_cur_l = NULL;
+ }
+
+ // construct R_cur_r, if not already constructed
+ if (NULL == R_cur_r)
+ {
+ length = strlen (R_temp_kk) - strlen (R_last_ik);
+
+ // a(ba)*bx = (ab)+x
+ if (length > 0 && NULL != R_last_kk && 0 < strlen (R_last_kk) &&
+ NULL != R_last_kj && 0 < strlen (R_last_kj) && NULL != R_last_ik &&
+ 0 < strlen (R_last_ik) && 0 == strkcmp (R_temp_kk, R_last_ik, length) &&
+ 0 == strncmp (R_temp_kk, R_last_kj, length))
+ {
+ temp_a = GNUNET_malloc (length + 1);
+ temp_b = GNUNET_malloc ((strlen (R_last_kj) - length) + 1);
+
+ length_l = 0;
+ length_r = 0;
+
+ for (cnt = 0; cnt < strlen (R_last_kj); cnt++)
+ {
+ if (cnt < length)
+ {
+ temp_a[length_l] = R_last_kj[cnt];
+ length_l++;
+ }
+ else
+ {
+ temp_b[length_r] = R_last_kj[cnt];
+ length_r++;
+ }
+ }
+ temp_a[length_l] = '\0';
+ temp_b[length_r] = '\0';
+
+ // e|(ab)+ = (ab)*
+ if (NULL != R_cur_l && 0 == strlen (R_cur_l) && 0 == strlen (temp_b))
+ {
+ GNUNET_asprintf (&R_cur_r, "(%s%s)*", R_last_ik, temp_a);
+ GNUNET_free (R_cur_l);
+ R_cur_l = NULL;
+ }
+ else
+ {
+ GNUNET_asprintf (&R_cur_r, "(%s%s)+%s", R_last_ik, temp_a, temp_b);
+ }
+ GNUNET_free (temp_a);
+ GNUNET_free (temp_b);
+ }
+ else if (0 == strcmp (R_temp_ik, R_temp_kk) &&
+ 0 == strcmp (R_temp_kk, R_temp_kj))
+ {
+ // (e|a)a*(e|a) = a*
+ // (e|a)(e|a)*(e|a) = a*
+ if (has_epsilon (R_last_ik) && has_epsilon (R_last_kj))
+ {
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "(%s)*", R_temp_kk);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s*", R_temp_kk);
+ }
+ // aa*a = a+a
+ else if (0 == clean_ik_kk_cmp && 0 == clean_kk_kj_cmp &&
+ !has_epsilon (R_last_ik))
+ {
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_temp_kk, R_temp_kk);
+ else
+ GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_temp_kk, R_temp_kk);
+ }
+ // (e|a)a*a = a+
+ // aa*(e|a) = a+
+ // a(e|a)*(e|a) = a+
+ // (e|a)a*a = a+
+ else
+ {
+ eps_check =
+ (has_epsilon (R_last_ik) + has_epsilon (R_last_kk) +
+ has_epsilon (R_last_kj));
+
+ if (eps_check == 1)
+ {
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "(%s)+", R_temp_kk);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s+", R_temp_kk);
+ }
+ }
+ }
+ // aa*b = a+b
+ // (e|a)(e|a)*b = a*b
+ else if (0 == strcmp (R_temp_ik, R_temp_kk))
+ {
+ if (has_epsilon (R_last_ik))
+ {
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "(%s)*%s", R_temp_kk, R_last_kj);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s*%s", R_temp_kk, R_last_kj);
+ }
+ else
+ {
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_temp_kk, R_last_kj);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s+%s", R_temp_kk, R_last_kj);
+ }
+ }
+ // ba*a = ba+
+ // b(e|a)*(e|a) = ba*
+ else if (0 == strcmp (R_temp_kk, R_temp_kj))
+ {
+ if (has_epsilon (R_last_kj))
+ {
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "%s(%s)*", R_last_ik, R_temp_kk);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s%s*", R_last_ik, R_temp_kk);
+ }
+ else
+ {
+ if (needs_parentheses (R_temp_kk))
+ GNUNET_asprintf (&R_cur_r, "(%s)+%s", R_last_ik, R_temp_kk);
+ else
+ GNUNET_asprintf (&R_cur_r, "%s+%s", R_last_ik, R_temp_kk);
+ }
+ }
+ else
+ {
+ if (strlen (R_temp_kk) > 0)
+ {
+ if (needs_parentheses (R_temp_kk))
+ {
+ GNUNET_asprintf (&R_cur_r, "%s(%s)*%s", R_last_ik, R_temp_kk,
+ R_last_kj);
+ }
+ else
+ {
+ GNUNET_asprintf (&R_cur_r, "%s%s*%s", R_last_ik, R_temp_kk,
+ R_last_kj);
+ }
+ }
+ else
+ {
+ GNUNET_asprintf (&R_cur_r, "%s%s", R_last_ik, R_last_kj);
+ }
+ }
+ }
+
+ GNUNET_free_non_null (R_temp_ik);
+ GNUNET_free_non_null (R_temp_kk);
+ GNUNET_free_non_null (R_temp_kj);
+
+ if (NULL == R_cur_l && NULL == R_cur_r)
+ {
+ *R_cur_ij = NULL;
+ return;
+ }
+
+ if (NULL != R_cur_l && NULL == R_cur_r)
+ {
+ *R_cur_ij = R_cur_l;
+ return;
+ }
+
+ if (NULL == R_cur_l && NULL != R_cur_r)
+ {
+ *R_cur_ij = R_cur_r;
+ return;
+ }
+
+ if (0 == nullstrcmp (R_cur_l, R_cur_r))
+ {
+ *R_cur_ij = R_cur_l;
+ GNUNET_free (R_cur_r);
+ return;
+ }
+
+ GNUNET_asprintf (R_cur_ij, "(%s|%s)", R_cur_l, R_cur_r);
+
+ GNUNET_free (R_cur_l);
+ GNUNET_free (R_cur_r);
+}
+
+
+/**
+ * create proofs for all states in the given automaton. Implementation of the
+ * algorithm descriped in chapter 3.2.1 of "Automata Theory, Languages, and
+ * Computation 3rd Edition" by Hopcroft, Motwani and Ullman.
+ *
+ * @param a automaton.
+ */
+static void
+automaton_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;
+
+
+ /* create depth-first numbering of the states, initializes 'state' */
+ GNUNET_REGEX_automaton_traverse (a, &number_states, states);
+
+ /* Compute regular expressions of length "1" between each pair of states */
+ for (i = 0; i < n; i++)
+ {
+ for (j = 0; j < n; j++)
+ {
+ R_cur[i][j] = NULL;
+ R_last[i][j] = NULL;
+ }
+ for (t = states[i]->transitions_head; NULL != t; t = t->next)
+ {
+ j = t->to_state->proof_id;
+ if (NULL == R_last[i][j])
+ GNUNET_asprintf (&R_last[i][j], "%c", t->label);
+ else
+ {
+ temp = R_last[i][j];
+ GNUNET_asprintf (&R_last[i][j], "%s|%c", R_last[i][j], t->label);
+ GNUNET_free (temp);
+ }
+ }
+ if (NULL == R_last[i][i])
+ GNUNET_asprintf (&R_last[i][i], "");
+ else
+ {
+ temp = R_last[i][i];
+ GNUNET_asprintf (&R_last[i][i], "(|%s)", R_last[i][i]);
+ GNUNET_free (temp);
+ }
+ }
+ for (i = 0; i < n; i++)
+ for (j = 0; j < n; j++)
+ if (needs_parentheses (R_last[i][j]))
+ {
+ temp = R_last[i][j];
+ GNUNET_asprintf (&R_last[i][j], "(%s)", R_last[i][j]);
+ GNUNET_free (temp);
+ }
+
+ /* Compute regular expressions of length "k" between each pair of states per induction */
+ for (k = 0; k < n; k++)
+ {
+ for (i = 0; i < n; i++)
+ {
+ for (j = 0; j < n; j++)
+ {
+ // Basis for the recursion:
+ // $R^{(k)}_{ij} = R^{(k-1)}_{ij} | R^{(k-1)}_{ik} ( R^{(k-1)}_{kk} )^* R^{(k-1)}_{kj}
+ // R_last == R^{(k-1)}, R_cur == R^{(k)}
+
+ // Create R_cur[i][j] and simplify the expression
+ automaton_create_proofs_simplify (R_last[i][j], R_last[i][k],
+ R_last[k][k], R_last[k][j],
+ &R_cur[i][j]);
+ }
+ }
+
+ // set R_last = R_cur
+ for (i = 0; i < n; i++)
+ {
+ for (j = 0; j < n; j++)
+ {
+ GNUNET_free_non_null (R_last[i][j]);
+ R_last[i][j] = R_cur[i][j];
+ R_cur[i][j] = NULL;
+ }
+ }
+ }
+
+ // assign proofs and hashes
+ for (i = 0; i < n; i++)
+ {
+ if (NULL != R_last[a->start->proof_id][i])
+ {
+ states[i]->proof = GNUNET_strdup (R_last[a->start->proof_id][i]);
+ GNUNET_CRYPTO_hash (states[i]->proof, strlen (states[i]->proof),
+ &states[i]->hash);
+ }
+ }
+
+ // 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;
+
+ // cleanup
+ for (i = 0; i < n; i++)
+ {
+ for (j = 0; j < n; j++)
+ GNUNET_free_non_null (R_last[i][j]);
+ }
+}
+
+
+/**
+ * Creates a new DFA state based on a set of NFA states. Needs to be freed using
+ * automaton_destroy_state.
+ *
+ * @param ctx context
+ * @param nfa_states set of NFA states on which the DFA should be based on
+ *
+ * @return new DFA state
+ */
+static struct GNUNET_REGEX_State *
+dfa_state_create (struct GNUNET_REGEX_Context *ctx,
+ struct GNUNET_REGEX_StateSet *nfa_states)
+{
+ struct GNUNET_REGEX_State *s;
+ char *name;
+ int len = 0;
+ struct GNUNET_REGEX_State *cstate;
+ struct GNUNET_REGEX_Transition *ctran;
+ unsigned int i;
+
+ s = GNUNET_malloc (sizeof (struct GNUNET_REGEX_State));
+ s->id = ctx->state_id++;
+ s->accepting = 0;
+ s->marked = 0;
+ s->name = NULL;
+ s->scc_id = 0;
+ s->index = -1;
+ s->lowlink = -1;
+ s->contained = 0;
+ s->proof = NULL;
+
+ if (NULL == nfa_states)
+ {
+ GNUNET_asprintf (&s->name, "s%i", s->id);
+ return s;
+ }
+
+ s->nfa_set = nfa_states;
+
+ if (nfa_states->len < 1)
+ return s;
+
+ // Create a name based on 'sset'
+ s->name = GNUNET_malloc (sizeof (char) * 2);
+ strcat (s->name, "{");
+ name = NULL;
+
+ for (i = 0; i < nfa_states->len; i++)
+ {
+ cstate = nfa_states->states[i];
+ GNUNET_asprintf (&name, "%i,", cstate->id);
+
+ if (NULL != name)
+ {
+ len = strlen (s->name) + strlen (name) + 1;
+ s->name = GNUNET_realloc (s->name, len);
+ strcat (s->name, name);
+ GNUNET_free (name);
+ name = NULL;
+ }
+
+ // Add a transition for each distinct label to NULL state
+ for (ctran = cstate->transitions_head; NULL != ctran; ctran = ctran->next)
+ {
+ if (0 != ctran->label)
+ state_add_transition (ctx, s, ctran->label, NULL);
+ }
+
+ // If the nfa_states contain an accepting state, the new dfa state is also
+ // accepting
+ if (cstate->accepting)
+ s->accepting = 1;
+ }
+
+ s->name[strlen (s->name) - 1] = '}';
+
+ return s;
+}
+
+
+/**
+ * Move from the given state 's' to the next state on transition 'label'
+ *
+ * @param s starting state
+ * @param label edge label to follow
+ *
+ * @return new state or NULL, if transition on label not possible
+ */
+static struct GNUNET_REGEX_State *
+dfa_move (struct GNUNET_REGEX_State *s, const char label)
+{
+ struct GNUNET_REGEX_Transition *t;
+ struct GNUNET_REGEX_State *new_s;
+
+ if (NULL == s)
+ return NULL;
+
+ new_s = NULL;
+
+ for (t = s->transitions_head; NULL != t; t = t->next)
+ {
+ if (label == t->label)
+ {
+ new_s = t->to_state;
+ break;
+ }
+ }
+
+ return new_s;
+}
+
+
+/**
+ * Remove all unreachable states from DFA 'a'. Unreachable states are those
+ * states that are not reachable from the starting state.
+ *
+ * @param a DFA automaton
+ */
+static void
+dfa_remove_unreachable_states (struct GNUNET_REGEX_Automaton *a)