static void makedfa(nfagrammar *gr, nfa *nf, dfa *d) { int nbits = nf->nf_nstates; bitset ss; int xx_nstates; ss_state *xx_state, *yy; ss_arc *zz; int istate, jstate, iarc, jarc, ibit; nfastate *st; nfaarc *ar; ss = newbitset(nbits); addclosure(ss, nf, nf->nf_start); xx_state = (ss_state *)PyObject_MALLOC(sizeof(ss_state)); if (xx_state == NULL) Py_FatalError("no mem for xx_state in makedfa"); xx_nstates = 1; yy = &xx_state[0]; yy->ss_ss = ss; yy->ss_narcs = 0; yy->ss_arc = NULL; yy->ss_deleted = 0; yy->ss_finish = testbit(ss, nf->nf_finish); if (yy->ss_finish) printf("Error: nonterminal '%s' may produce empty.\n", nf->nf_name); /* This algorithm is from a book written before the invention of structured programming... */ /* For each unmarked state... */ for (istate = 0; istate < xx_nstates; ++istate) { size_t size; yy = &xx_state[istate]; ss = yy->ss_ss; /* For all its states... */ for (ibit = 0; ibit < nf->nf_nstates; ++ibit) { if (!testbit(ss, ibit)) continue; st = &nf->nf_state[ibit]; /* For all non-empty arcs from this state... */ for (iarc = 0; iarc < st->st_narcs; iarc++) { ar = &st->st_arc[iarc]; if (ar->ar_label == EMPTY) continue; /* Look up in list of arcs from this state */ for (jarc = 0; jarc < yy->ss_narcs; ++jarc) { zz = &yy->ss_arc[jarc]; if (ar->ar_label == zz->sa_label) goto found; } /* Add new arc for this state */ size = sizeof(ss_arc) * (yy->ss_narcs + 1); yy->ss_arc = (ss_arc *)PyObject_REALLOC( yy->ss_arc, size); if (yy->ss_arc == NULL) Py_FatalError("out of mem"); zz = &yy->ss_arc[yy->ss_narcs++]; zz->sa_label = ar->ar_label; zz->sa_bitset = newbitset(nbits); zz->sa_arrow = -1; found: ; /* Add destination */ addclosure(zz->sa_bitset, nf, ar->ar_arrow); } } /* Now look up all the arrow states */ for (jarc = 0; jarc < xx_state[istate].ss_narcs; jarc++) { zz = &xx_state[istate].ss_arc[jarc]; for (jstate = 0; jstate < xx_nstates; jstate++) { if (samebitset(zz->sa_bitset, xx_state[jstate].ss_ss, nbits)) { zz->sa_arrow = jstate; goto done; } } size = sizeof(ss_state) * (xx_nstates + 1); xx_state = (ss_state *)PyObject_REALLOC(xx_state, size); if (xx_state == NULL) Py_FatalError("out of mem"); zz->sa_arrow = xx_nstates; yy = &xx_state[xx_nstates++]; yy->ss_ss = zz->sa_bitset; yy->ss_narcs = 0; yy->ss_arc = NULL; yy->ss_deleted = 0; yy->ss_finish = testbit(yy->ss_ss, nf->nf_finish); done: ; } } if (Py_DebugFlag) printssdfa(xx_nstates, xx_state, nbits, &gr->gr_ll, "before minimizing"); simplify(xx_nstates, xx_state); if (Py_DebugFlag) printssdfa(xx_nstates, xx_state, nbits, &gr->gr_ll, "after minimizing"); convert(d, xx_nstates, xx_state); /* XXX cleanup */ PyObject_FREE(xx_state); }
static void calcfirstset(grammar *g, dfa *d) { int i, j; state *s; arc *a; int nsyms; int *sym; int nbits; static bitset dummy; bitset result; int type; dfa *d1; label *l0; if (Py_DebugFlag) printf("Calculate FIRST set for '%s'\n", d->d_name); if (dummy == NULL) dummy = newbitset(1); if (d->d_first == dummy) { fprintf(stderr, "Left-recursion for '%s'\n", d->d_name); return; } if (d->d_first != NULL) { fprintf(stderr, "Re-calculating FIRST set for '%s' ???\n", d->d_name); } d->d_first = dummy; l0 = g->g_ll.ll_label; nbits = g->g_ll.ll_nlabels; result = newbitset(nbits); sym = (int *)PyObject_MALLOC(sizeof(int)); if (sym == NULL) Py_FatalError("no mem for new sym in calcfirstset"); nsyms = 1; sym[0] = findlabel(&g->g_ll, d->d_type, (char *)NULL); s = &d->d_state[d->d_initial]; for (i = 0; i < s->s_narcs; i++) { a = &s->s_arc[i]; for (j = 0; j < nsyms; j++) { if (sym[j] == a->a_lbl) break; } if (j >= nsyms) { /* New label */ sym = (int *)PyObject_REALLOC(sym, sizeof(int) * (nsyms + 1)); if (sym == NULL) Py_FatalError( "no mem to resize sym in calcfirstset"); sym[nsyms++] = a->a_lbl; type = l0[a->a_lbl].lb_type; if (ISNONTERMINAL(type)) { d1 = PyGrammar_FindDFA(g, type); if (d1->d_first == dummy) { fprintf(stderr, "Left-recursion below '%s'\n", d->d_name); } else { if (d1->d_first == NULL) calcfirstset(g, d1); mergebitset(result, d1->d_first, nbits); } } else if (ISTERMINAL(type)) { addbit(result, a->a_lbl); } } } d->d_first = result; if (Py_DebugFlag) { printf("FIRST set for '%s': {", d->d_name); for (i = 0; i < nbits; i++) { if (testbit(result, i)) printf(" %s", PyGrammar_LabelRepr(&l0[i])); } printf(" }\n"); } PyObject_FREE(sym); }