Guido van Rossum | f70e43a | 1991-02-19 12:39:46 +0000 | [diff] [blame^] | 1 | /*********************************************************** |
| 2 | Copyright 1991 by Stichting Mathematisch Centrum, Amsterdam, The |
| 3 | Netherlands. |
| 4 | |
| 5 | All Rights Reserved |
| 6 | |
| 7 | Permission to use, copy, modify, and distribute this software and its |
| 8 | documentation for any purpose and without fee is hereby granted, |
| 9 | provided that the above copyright notice appear in all copies and that |
| 10 | both that copyright notice and this permission notice appear in |
| 11 | supporting documentation, and that the names of Stichting Mathematisch |
| 12 | Centrum or CWI not be used in advertising or publicity pertaining to |
| 13 | distribution of the software without specific, written prior permission. |
| 14 | |
| 15 | STICHTING MATHEMATISCH CENTRUM DISCLAIMS ALL WARRANTIES WITH REGARD TO |
| 16 | THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND |
| 17 | FITNESS, IN NO EVENT SHALL STICHTING MATHEMATISCH CENTRUM BE LIABLE |
| 18 | FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES |
| 19 | WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN |
| 20 | ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT |
| 21 | OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. |
| 22 | |
| 23 | ******************************************************************/ |
| 24 | |
Guido van Rossum | 85a5fbb | 1990-10-14 12:07:46 +0000 | [diff] [blame] | 25 | /* Parser generator */ |
Guido van Rossum | 3f5da24 | 1990-12-20 15:06:42 +0000 | [diff] [blame] | 26 | /* XXX This file is not yet fully PROTOized */ |
Guido van Rossum | 85a5fbb | 1990-10-14 12:07:46 +0000 | [diff] [blame] | 27 | |
| 28 | /* For a description, see the comments at end of this file */ |
| 29 | |
Guido van Rossum | 3f5da24 | 1990-12-20 15:06:42 +0000 | [diff] [blame] | 30 | #include "pgenheaders.h" |
Guido van Rossum | 85a5fbb | 1990-10-14 12:07:46 +0000 | [diff] [blame] | 31 | #include "assert.h" |
Guido van Rossum | 85a5fbb | 1990-10-14 12:07:46 +0000 | [diff] [blame] | 32 | #include "token.h" |
| 33 | #include "node.h" |
| 34 | #include "grammar.h" |
| 35 | #include "metagrammar.h" |
| 36 | #include "pgen.h" |
| 37 | |
| 38 | extern int debugging; |
| 39 | |
| 40 | |
| 41 | /* PART ONE -- CONSTRUCT NFA -- Cf. Algorithm 3.2 from [Aho&Ullman 77] */ |
| 42 | |
| 43 | typedef struct _nfaarc { |
| 44 | int ar_label; |
| 45 | int ar_arrow; |
| 46 | } nfaarc; |
| 47 | |
| 48 | typedef struct _nfastate { |
| 49 | int st_narcs; |
| 50 | nfaarc *st_arc; |
| 51 | } nfastate; |
| 52 | |
| 53 | typedef struct _nfa { |
| 54 | int nf_type; |
| 55 | char *nf_name; |
| 56 | int nf_nstates; |
| 57 | nfastate *nf_state; |
| 58 | int nf_start, nf_finish; |
| 59 | } nfa; |
| 60 | |
| 61 | static int |
| 62 | addnfastate(nf) |
| 63 | nfa *nf; |
| 64 | { |
| 65 | nfastate *st; |
| 66 | |
| 67 | RESIZE(nf->nf_state, nfastate, nf->nf_nstates + 1); |
| 68 | if (nf->nf_state == NULL) |
| 69 | fatal("out of mem"); |
| 70 | st = &nf->nf_state[nf->nf_nstates++]; |
| 71 | st->st_narcs = 0; |
| 72 | st->st_arc = NULL; |
| 73 | return st - nf->nf_state; |
| 74 | } |
| 75 | |
| 76 | static void |
| 77 | addnfaarc(nf, from, to, lbl) |
| 78 | nfa *nf; |
| 79 | int from, to, lbl; |
| 80 | { |
| 81 | nfastate *st; |
| 82 | nfaarc *ar; |
| 83 | |
| 84 | st = &nf->nf_state[from]; |
| 85 | RESIZE(st->st_arc, nfaarc, st->st_narcs + 1); |
| 86 | if (st->st_arc == NULL) |
| 87 | fatal("out of mem"); |
| 88 | ar = &st->st_arc[st->st_narcs++]; |
| 89 | ar->ar_label = lbl; |
| 90 | ar->ar_arrow = to; |
| 91 | } |
| 92 | |
| 93 | static nfa * |
| 94 | newnfa(name) |
| 95 | char *name; |
| 96 | { |
| 97 | nfa *nf; |
| 98 | static type = NT_OFFSET; /* All types will be disjunct */ |
| 99 | |
| 100 | nf = NEW(nfa, 1); |
| 101 | if (nf == NULL) |
| 102 | fatal("no mem for new nfa"); |
| 103 | nf->nf_type = type++; |
| 104 | nf->nf_name = name; /* XXX strdup(name) ??? */ |
| 105 | nf->nf_nstates = 0; |
| 106 | nf->nf_state = NULL; |
| 107 | nf->nf_start = nf->nf_finish = -1; |
| 108 | return nf; |
| 109 | } |
| 110 | |
| 111 | typedef struct _nfagrammar { |
| 112 | int gr_nnfas; |
| 113 | nfa **gr_nfa; |
| 114 | labellist gr_ll; |
| 115 | } nfagrammar; |
| 116 | |
| 117 | static nfagrammar * |
| 118 | newnfagrammar() |
| 119 | { |
| 120 | nfagrammar *gr; |
| 121 | |
| 122 | gr = NEW(nfagrammar, 1); |
| 123 | if (gr == NULL) |
| 124 | fatal("no mem for new nfa grammar"); |
| 125 | gr->gr_nnfas = 0; |
| 126 | gr->gr_nfa = NULL; |
| 127 | gr->gr_ll.ll_nlabels = 0; |
| 128 | gr->gr_ll.ll_label = NULL; |
| 129 | addlabel(&gr->gr_ll, ENDMARKER, "EMPTY"); |
| 130 | return gr; |
| 131 | } |
| 132 | |
| 133 | static nfa * |
| 134 | addnfa(gr, name) |
| 135 | nfagrammar *gr; |
| 136 | char *name; |
| 137 | { |
| 138 | nfa *nf; |
| 139 | |
| 140 | nf = newnfa(name); |
| 141 | RESIZE(gr->gr_nfa, nfa *, gr->gr_nnfas + 1); |
| 142 | if (gr->gr_nfa == NULL) |
| 143 | fatal("out of mem"); |
| 144 | gr->gr_nfa[gr->gr_nnfas++] = nf; |
| 145 | addlabel(&gr->gr_ll, NAME, nf->nf_name); |
| 146 | return nf; |
| 147 | } |
| 148 | |
| 149 | #ifdef DEBUG |
| 150 | |
| 151 | static char REQNFMT[] = "metacompile: less than %d children\n"; |
| 152 | |
| 153 | #define REQN(i, count) \ |
| 154 | if (i < count) { \ |
| 155 | fprintf(stderr, REQNFMT, count); \ |
| 156 | abort(); \ |
| 157 | } else |
| 158 | |
| 159 | #else |
| 160 | #define REQN(i, count) /* empty */ |
| 161 | #endif |
| 162 | |
| 163 | static nfagrammar * |
| 164 | metacompile(n) |
| 165 | node *n; |
| 166 | { |
| 167 | nfagrammar *gr; |
| 168 | int i; |
| 169 | |
| 170 | printf("Compiling (meta-) parse tree into NFA grammar\n"); |
| 171 | gr = newnfagrammar(); |
| 172 | REQ(n, MSTART); |
| 173 | i = n->n_nchildren - 1; /* Last child is ENDMARKER */ |
| 174 | n = n->n_child; |
| 175 | for (; --i >= 0; n++) { |
| 176 | if (n->n_type != NEWLINE) |
| 177 | compile_rule(gr, n); |
| 178 | } |
| 179 | return gr; |
| 180 | } |
| 181 | |
| 182 | static |
| 183 | compile_rule(gr, n) |
| 184 | nfagrammar *gr; |
| 185 | node *n; |
| 186 | { |
| 187 | nfa *nf; |
| 188 | |
| 189 | REQ(n, RULE); |
| 190 | REQN(n->n_nchildren, 4); |
| 191 | n = n->n_child; |
| 192 | REQ(n, NAME); |
| 193 | nf = addnfa(gr, n->n_str); |
| 194 | n++; |
| 195 | REQ(n, COLON); |
| 196 | n++; |
| 197 | REQ(n, RHS); |
| 198 | compile_rhs(&gr->gr_ll, nf, n, &nf->nf_start, &nf->nf_finish); |
| 199 | n++; |
| 200 | REQ(n, NEWLINE); |
| 201 | } |
| 202 | |
| 203 | static |
| 204 | compile_rhs(ll, nf, n, pa, pb) |
| 205 | labellist *ll; |
| 206 | nfa *nf; |
| 207 | node *n; |
| 208 | int *pa, *pb; |
| 209 | { |
| 210 | int i; |
| 211 | int a, b; |
| 212 | |
| 213 | REQ(n, RHS); |
| 214 | i = n->n_nchildren; |
| 215 | REQN(i, 1); |
| 216 | n = n->n_child; |
| 217 | REQ(n, ALT); |
| 218 | compile_alt(ll, nf, n, pa, pb); |
| 219 | if (--i <= 0) |
| 220 | return; |
| 221 | n++; |
| 222 | a = *pa; |
| 223 | b = *pb; |
| 224 | *pa = addnfastate(nf); |
| 225 | *pb = addnfastate(nf); |
| 226 | addnfaarc(nf, *pa, a, EMPTY); |
| 227 | addnfaarc(nf, b, *pb, EMPTY); |
| 228 | for (; --i >= 0; n++) { |
| 229 | REQ(n, VBAR); |
| 230 | REQN(i, 1); |
| 231 | --i; |
| 232 | n++; |
| 233 | REQ(n, ALT); |
| 234 | compile_alt(ll, nf, n, &a, &b); |
| 235 | addnfaarc(nf, *pa, a, EMPTY); |
| 236 | addnfaarc(nf, b, *pb, EMPTY); |
| 237 | } |
| 238 | } |
| 239 | |
| 240 | static |
| 241 | compile_alt(ll, nf, n, pa, pb) |
| 242 | labellist *ll; |
| 243 | nfa *nf; |
| 244 | node *n; |
| 245 | int *pa, *pb; |
| 246 | { |
| 247 | int i; |
| 248 | int a, b; |
| 249 | |
| 250 | REQ(n, ALT); |
| 251 | i = n->n_nchildren; |
| 252 | REQN(i, 1); |
| 253 | n = n->n_child; |
| 254 | REQ(n, ITEM); |
| 255 | compile_item(ll, nf, n, pa, pb); |
| 256 | --i; |
| 257 | n++; |
| 258 | for (; --i >= 0; n++) { |
| 259 | if (n->n_type == COMMA) { /* XXX Temporary */ |
| 260 | REQN(i, 1); |
| 261 | --i; |
| 262 | n++; |
| 263 | } |
| 264 | REQ(n, ITEM); |
| 265 | compile_item(ll, nf, n, &a, &b); |
| 266 | addnfaarc(nf, *pb, a, EMPTY); |
| 267 | *pb = b; |
| 268 | } |
| 269 | } |
| 270 | |
| 271 | static |
| 272 | compile_item(ll, nf, n, pa, pb) |
| 273 | labellist *ll; |
| 274 | nfa *nf; |
| 275 | node *n; |
| 276 | int *pa, *pb; |
| 277 | { |
| 278 | int i; |
| 279 | int a, b; |
| 280 | |
| 281 | REQ(n, ITEM); |
| 282 | i = n->n_nchildren; |
| 283 | REQN(i, 1); |
| 284 | n = n->n_child; |
| 285 | if (n->n_type == LSQB) { |
| 286 | REQN(i, 3); |
| 287 | n++; |
| 288 | REQ(n, RHS); |
| 289 | *pa = addnfastate(nf); |
| 290 | *pb = addnfastate(nf); |
| 291 | addnfaarc(nf, *pa, *pb, EMPTY); |
| 292 | compile_rhs(ll, nf, n, &a, &b); |
| 293 | addnfaarc(nf, *pa, a, EMPTY); |
| 294 | addnfaarc(nf, b, *pb, EMPTY); |
| 295 | REQN(i, 1); |
| 296 | n++; |
| 297 | REQ(n, RSQB); |
| 298 | } |
| 299 | else { |
| 300 | compile_atom(ll, nf, n, pa, pb); |
| 301 | if (--i <= 0) |
| 302 | return; |
| 303 | n++; |
| 304 | addnfaarc(nf, *pb, *pa, EMPTY); |
| 305 | if (n->n_type == STAR) |
| 306 | *pb = *pa; |
| 307 | else |
| 308 | REQ(n, PLUS); |
| 309 | } |
| 310 | } |
| 311 | |
| 312 | static |
| 313 | compile_atom(ll, nf, n, pa, pb) |
| 314 | labellist *ll; |
| 315 | nfa *nf; |
| 316 | node *n; |
| 317 | int *pa, *pb; |
| 318 | { |
| 319 | int i; |
| 320 | |
| 321 | REQ(n, ATOM); |
| 322 | i = n->n_nchildren; |
| 323 | REQN(i, 1); |
| 324 | n = n->n_child; |
| 325 | if (n->n_type == LPAR) { |
| 326 | REQN(i, 3); |
| 327 | n++; |
| 328 | REQ(n, RHS); |
| 329 | compile_rhs(ll, nf, n, pa, pb); |
| 330 | n++; |
| 331 | REQ(n, RPAR); |
| 332 | } |
| 333 | else if (n->n_type == NAME || n->n_type == STRING) { |
| 334 | *pa = addnfastate(nf); |
| 335 | *pb = addnfastate(nf); |
| 336 | addnfaarc(nf, *pa, *pb, addlabel(ll, n->n_type, n->n_str)); |
| 337 | } |
| 338 | else |
| 339 | REQ(n, NAME); |
| 340 | } |
| 341 | |
| 342 | static void |
| 343 | dumpstate(ll, nf, istate) |
| 344 | labellist *ll; |
| 345 | nfa *nf; |
| 346 | int istate; |
| 347 | { |
| 348 | nfastate *st; |
| 349 | int i; |
| 350 | nfaarc *ar; |
| 351 | |
| 352 | printf("%c%2d%c", |
| 353 | istate == nf->nf_start ? '*' : ' ', |
| 354 | istate, |
| 355 | istate == nf->nf_finish ? '.' : ' '); |
| 356 | st = &nf->nf_state[istate]; |
| 357 | ar = st->st_arc; |
| 358 | for (i = 0; i < st->st_narcs; i++) { |
| 359 | if (i > 0) |
| 360 | printf("\n "); |
| 361 | printf("-> %2d %s", ar->ar_arrow, |
| 362 | labelrepr(&ll->ll_label[ar->ar_label])); |
| 363 | ar++; |
| 364 | } |
| 365 | printf("\n"); |
| 366 | } |
| 367 | |
| 368 | static void |
| 369 | dumpnfa(ll, nf) |
| 370 | labellist *ll; |
| 371 | nfa *nf; |
| 372 | { |
| 373 | int i; |
| 374 | |
| 375 | printf("NFA '%s' has %d states; start %d, finish %d\n", |
| 376 | nf->nf_name, nf->nf_nstates, nf->nf_start, nf->nf_finish); |
| 377 | for (i = 0; i < nf->nf_nstates; i++) |
| 378 | dumpstate(ll, nf, i); |
| 379 | } |
| 380 | |
| 381 | |
| 382 | /* PART TWO -- CONSTRUCT DFA -- Algorithm 3.1 from [Aho&Ullman 77] */ |
| 383 | |
| 384 | static int |
| 385 | addclosure(ss, nf, istate) |
| 386 | bitset ss; |
| 387 | nfa *nf; |
| 388 | int istate; |
| 389 | { |
| 390 | if (addbit(ss, istate)) { |
| 391 | nfastate *st = &nf->nf_state[istate]; |
| 392 | nfaarc *ar = st->st_arc; |
| 393 | int i; |
| 394 | |
| 395 | for (i = st->st_narcs; --i >= 0; ) { |
| 396 | if (ar->ar_label == EMPTY) |
| 397 | addclosure(ss, nf, ar->ar_arrow); |
| 398 | ar++; |
| 399 | } |
| 400 | } |
| 401 | } |
| 402 | |
| 403 | typedef struct _ss_arc { |
| 404 | bitset sa_bitset; |
| 405 | int sa_arrow; |
| 406 | int sa_label; |
| 407 | } ss_arc; |
| 408 | |
| 409 | typedef struct _ss_state { |
| 410 | bitset ss_ss; |
| 411 | int ss_narcs; |
| 412 | ss_arc *ss_arc; |
| 413 | int ss_deleted; |
| 414 | int ss_finish; |
| 415 | int ss_rename; |
| 416 | } ss_state; |
| 417 | |
| 418 | typedef struct _ss_dfa { |
| 419 | int sd_nstates; |
| 420 | ss_state *sd_state; |
| 421 | } ss_dfa; |
| 422 | |
| 423 | static |
| 424 | makedfa(gr, nf, d) |
| 425 | nfagrammar *gr; |
| 426 | nfa *nf; |
| 427 | dfa *d; |
| 428 | { |
| 429 | int nbits = nf->nf_nstates; |
| 430 | bitset ss; |
| 431 | int xx_nstates; |
| 432 | ss_state *xx_state, *yy; |
| 433 | ss_arc *zz; |
| 434 | int istate, jstate, iarc, jarc, ibit; |
| 435 | nfastate *st; |
| 436 | nfaarc *ar; |
| 437 | |
| 438 | ss = newbitset(nbits); |
| 439 | addclosure(ss, nf, nf->nf_start); |
| 440 | xx_state = NEW(ss_state, 1); |
| 441 | if (xx_state == NULL) |
| 442 | fatal("no mem for xx_state in makedfa"); |
| 443 | xx_nstates = 1; |
| 444 | yy = &xx_state[0]; |
| 445 | yy->ss_ss = ss; |
| 446 | yy->ss_narcs = 0; |
| 447 | yy->ss_arc = NULL; |
| 448 | yy->ss_deleted = 0; |
| 449 | yy->ss_finish = testbit(ss, nf->nf_finish); |
| 450 | if (yy->ss_finish) |
| 451 | printf("Error: nonterminal '%s' may produce empty.\n", |
| 452 | nf->nf_name); |
| 453 | |
| 454 | /* This algorithm is from a book written before |
| 455 | the invention of structured programming... */ |
| 456 | |
| 457 | /* For each unmarked state... */ |
| 458 | for (istate = 0; istate < xx_nstates; ++istate) { |
| 459 | yy = &xx_state[istate]; |
| 460 | ss = yy->ss_ss; |
| 461 | /* For all its states... */ |
| 462 | for (ibit = 0; ibit < nf->nf_nstates; ++ibit) { |
| 463 | if (!testbit(ss, ibit)) |
| 464 | continue; |
| 465 | st = &nf->nf_state[ibit]; |
| 466 | /* For all non-empty arcs from this state... */ |
| 467 | for (iarc = 0; iarc < st->st_narcs; iarc++) { |
| 468 | ar = &st->st_arc[iarc]; |
| 469 | if (ar->ar_label == EMPTY) |
| 470 | continue; |
| 471 | /* Look up in list of arcs from this state */ |
| 472 | for (jarc = 0; jarc < yy->ss_narcs; ++jarc) { |
| 473 | zz = &yy->ss_arc[jarc]; |
| 474 | if (ar->ar_label == zz->sa_label) |
| 475 | goto found; |
| 476 | } |
| 477 | /* Add new arc for this state */ |
| 478 | RESIZE(yy->ss_arc, ss_arc, yy->ss_narcs + 1); |
| 479 | if (yy->ss_arc == NULL) |
| 480 | fatal("out of mem"); |
| 481 | zz = &yy->ss_arc[yy->ss_narcs++]; |
| 482 | zz->sa_label = ar->ar_label; |
| 483 | zz->sa_bitset = newbitset(nbits); |
| 484 | zz->sa_arrow = -1; |
| 485 | found: ; |
| 486 | /* Add destination */ |
| 487 | addclosure(zz->sa_bitset, nf, ar->ar_arrow); |
| 488 | } |
| 489 | } |
| 490 | /* Now look up all the arrow states */ |
| 491 | for (jarc = 0; jarc < xx_state[istate].ss_narcs; jarc++) { |
| 492 | zz = &xx_state[istate].ss_arc[jarc]; |
| 493 | for (jstate = 0; jstate < xx_nstates; jstate++) { |
| 494 | if (samebitset(zz->sa_bitset, |
| 495 | xx_state[jstate].ss_ss, nbits)) { |
| 496 | zz->sa_arrow = jstate; |
| 497 | goto done; |
| 498 | } |
| 499 | } |
| 500 | RESIZE(xx_state, ss_state, xx_nstates + 1); |
| 501 | if (xx_state == NULL) |
| 502 | fatal("out of mem"); |
| 503 | zz->sa_arrow = xx_nstates; |
| 504 | yy = &xx_state[xx_nstates++]; |
| 505 | yy->ss_ss = zz->sa_bitset; |
| 506 | yy->ss_narcs = 0; |
| 507 | yy->ss_arc = NULL; |
| 508 | yy->ss_deleted = 0; |
| 509 | yy->ss_finish = testbit(yy->ss_ss, nf->nf_finish); |
| 510 | done: ; |
| 511 | } |
| 512 | } |
| 513 | |
| 514 | if (debugging) |
| 515 | printssdfa(xx_nstates, xx_state, nbits, &gr->gr_ll, |
| 516 | "before minimizing"); |
| 517 | |
| 518 | simplify(xx_nstates, xx_state); |
| 519 | |
| 520 | if (debugging) |
| 521 | printssdfa(xx_nstates, xx_state, nbits, &gr->gr_ll, |
| 522 | "after minimizing"); |
| 523 | |
| 524 | convert(d, xx_nstates, xx_state); |
| 525 | |
| 526 | /* XXX cleanup */ |
| 527 | } |
| 528 | |
| 529 | static |
| 530 | printssdfa(xx_nstates, xx_state, nbits, ll, msg) |
| 531 | int xx_nstates; |
| 532 | ss_state *xx_state; |
| 533 | int nbits; |
| 534 | labellist *ll; |
| 535 | char *msg; |
| 536 | { |
| 537 | int i, ibit, iarc; |
| 538 | ss_state *yy; |
| 539 | ss_arc *zz; |
| 540 | |
| 541 | printf("Subset DFA %s\n", msg); |
| 542 | for (i = 0; i < xx_nstates; i++) { |
| 543 | yy = &xx_state[i]; |
| 544 | if (yy->ss_deleted) |
| 545 | continue; |
| 546 | printf(" Subset %d", i); |
| 547 | if (yy->ss_finish) |
| 548 | printf(" (finish)"); |
| 549 | printf(" { "); |
| 550 | for (ibit = 0; ibit < nbits; ibit++) { |
| 551 | if (testbit(yy->ss_ss, ibit)) |
| 552 | printf("%d ", ibit); |
| 553 | } |
| 554 | printf("}\n"); |
| 555 | for (iarc = 0; iarc < yy->ss_narcs; iarc++) { |
| 556 | zz = &yy->ss_arc[iarc]; |
| 557 | printf(" Arc to state %d, label %s\n", |
| 558 | zz->sa_arrow, |
| 559 | labelrepr(&ll->ll_label[zz->sa_label])); |
| 560 | } |
| 561 | } |
| 562 | } |
| 563 | |
| 564 | |
| 565 | /* PART THREE -- SIMPLIFY DFA */ |
| 566 | |
| 567 | /* Simplify the DFA by repeatedly eliminating states that are |
| 568 | equivalent to another oner. This is NOT Algorithm 3.3 from |
| 569 | [Aho&Ullman 77]. It does not always finds the minimal DFA, |
| 570 | but it does usually make a much smaller one... (For an example |
| 571 | of sub-optimal behaviour, try S: x a b+ | y a b+.) |
| 572 | */ |
| 573 | |
| 574 | static int |
| 575 | samestate(s1, s2) |
| 576 | ss_state *s1, *s2; |
| 577 | { |
| 578 | int i; |
| 579 | |
| 580 | if (s1->ss_narcs != s2->ss_narcs || s1->ss_finish != s2->ss_finish) |
| 581 | return 0; |
| 582 | for (i = 0; i < s1->ss_narcs; i++) { |
| 583 | if (s1->ss_arc[i].sa_arrow != s2->ss_arc[i].sa_arrow || |
| 584 | s1->ss_arc[i].sa_label != s2->ss_arc[i].sa_label) |
| 585 | return 0; |
| 586 | } |
| 587 | return 1; |
| 588 | } |
| 589 | |
| 590 | static void |
| 591 | renamestates(xx_nstates, xx_state, from, to) |
| 592 | int xx_nstates; |
| 593 | ss_state *xx_state; |
| 594 | int from, to; |
| 595 | { |
| 596 | int i, j; |
| 597 | |
| 598 | if (debugging) |
| 599 | printf("Rename state %d to %d.\n", from, to); |
| 600 | for (i = 0; i < xx_nstates; i++) { |
| 601 | if (xx_state[i].ss_deleted) |
| 602 | continue; |
| 603 | for (j = 0; j < xx_state[i].ss_narcs; j++) { |
| 604 | if (xx_state[i].ss_arc[j].sa_arrow == from) |
| 605 | xx_state[i].ss_arc[j].sa_arrow = to; |
| 606 | } |
| 607 | } |
| 608 | } |
| 609 | |
| 610 | static |
| 611 | simplify(xx_nstates, xx_state) |
| 612 | int xx_nstates; |
| 613 | ss_state *xx_state; |
| 614 | { |
| 615 | int changes; |
| 616 | int i, j, k; |
| 617 | |
| 618 | do { |
| 619 | changes = 0; |
| 620 | for (i = 1; i < xx_nstates; i++) { |
| 621 | if (xx_state[i].ss_deleted) |
| 622 | continue; |
| 623 | for (j = 0; j < i; j++) { |
| 624 | if (xx_state[j].ss_deleted) |
| 625 | continue; |
| 626 | if (samestate(&xx_state[i], &xx_state[j])) { |
| 627 | xx_state[i].ss_deleted++; |
| 628 | renamestates(xx_nstates, xx_state, i, j); |
| 629 | changes++; |
| 630 | break; |
| 631 | } |
| 632 | } |
| 633 | } |
| 634 | } while (changes); |
| 635 | } |
| 636 | |
| 637 | |
| 638 | /* PART FOUR -- GENERATE PARSING TABLES */ |
| 639 | |
| 640 | /* Convert the DFA into a grammar that can be used by our parser */ |
| 641 | |
| 642 | static |
| 643 | convert(d, xx_nstates, xx_state) |
| 644 | dfa *d; |
| 645 | int xx_nstates; |
| 646 | ss_state *xx_state; |
| 647 | { |
| 648 | int i, j; |
| 649 | ss_state *yy; |
| 650 | ss_arc *zz; |
| 651 | |
| 652 | for (i = 0; i < xx_nstates; i++) { |
| 653 | yy = &xx_state[i]; |
| 654 | if (yy->ss_deleted) |
| 655 | continue; |
| 656 | yy->ss_rename = addstate(d); |
| 657 | } |
| 658 | |
| 659 | for (i = 0; i < xx_nstates; i++) { |
| 660 | yy = &xx_state[i]; |
| 661 | if (yy->ss_deleted) |
| 662 | continue; |
| 663 | for (j = 0; j < yy->ss_narcs; j++) { |
| 664 | zz = &yy->ss_arc[j]; |
| 665 | addarc(d, yy->ss_rename, |
| 666 | xx_state[zz->sa_arrow].ss_rename, |
| 667 | zz->sa_label); |
| 668 | } |
| 669 | if (yy->ss_finish) |
| 670 | addarc(d, yy->ss_rename, yy->ss_rename, 0); |
| 671 | } |
| 672 | |
| 673 | d->d_initial = 0; |
| 674 | } |
| 675 | |
| 676 | |
| 677 | /* PART FIVE -- GLUE IT ALL TOGETHER */ |
| 678 | |
| 679 | static grammar * |
| 680 | maketables(gr) |
| 681 | nfagrammar *gr; |
| 682 | { |
| 683 | int i; |
| 684 | nfa *nf; |
| 685 | dfa *d; |
| 686 | grammar *g; |
| 687 | |
| 688 | if (gr->gr_nnfas == 0) |
| 689 | return NULL; |
| 690 | g = newgrammar(gr->gr_nfa[0]->nf_type); |
| 691 | /* XXX first rule must be start rule */ |
| 692 | g->g_ll = gr->gr_ll; |
| 693 | |
| 694 | for (i = 0; i < gr->gr_nnfas; i++) { |
| 695 | nf = gr->gr_nfa[i]; |
| 696 | if (debugging) { |
| 697 | printf("Dump of NFA for '%s' ...\n", nf->nf_name); |
| 698 | dumpnfa(&gr->gr_ll, nf); |
| 699 | } |
| 700 | printf("Making DFA for '%s' ...\n", nf->nf_name); |
| 701 | d = adddfa(g, nf->nf_type, nf->nf_name); |
| 702 | makedfa(gr, gr->gr_nfa[i], d); |
| 703 | } |
| 704 | |
| 705 | return g; |
| 706 | } |
| 707 | |
| 708 | grammar * |
| 709 | pgen(n) |
| 710 | node *n; |
| 711 | { |
| 712 | nfagrammar *gr; |
| 713 | grammar *g; |
| 714 | |
| 715 | gr = metacompile(n); |
| 716 | g = maketables(gr); |
| 717 | translatelabels(g); |
| 718 | addfirstsets(g); |
| 719 | return g; |
| 720 | } |
| 721 | |
| 722 | |
| 723 | /* |
| 724 | |
| 725 | Description |
| 726 | ----------- |
| 727 | |
| 728 | Input is a grammar in extended BNF (using * for repetition, + for |
| 729 | at-least-once repetition, [] for optional parts, | for alternatives and |
| 730 | () for grouping). This has already been parsed and turned into a parse |
| 731 | tree. |
| 732 | |
| 733 | Each rule is considered as a regular expression in its own right. |
| 734 | It is turned into a Non-deterministic Finite Automaton (NFA), which |
| 735 | is then turned into a Deterministic Finite Automaton (DFA), which is then |
| 736 | optimized to reduce the number of states. See [Aho&Ullman 77] chapter 3, |
| 737 | or similar compiler books (this technique is more often used for lexical |
| 738 | analyzers). |
| 739 | |
| 740 | The DFA's are used by the parser as parsing tables in a special way |
| 741 | that's probably unique. Before they are usable, the FIRST sets of all |
| 742 | non-terminals are computed. |
| 743 | |
| 744 | Reference |
| 745 | --------- |
| 746 | |
| 747 | [Aho&Ullman 77] |
| 748 | Aho&Ullman, Principles of Compiler Design, Addison-Wesley 1977 |
| 749 | (first edition) |
| 750 | |
| 751 | */ |