Annotation of parser3/src/lib/cord/cordbscs.c, revision 1.2
1.2 ! paf 1: /*
! 2: * Copyright (c) 1993-1994 by Xerox Corporation. All rights reserved.
! 3: *
! 4: * THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY EXPRESSED
! 5: * OR IMPLIED. ANY USE IS AT YOUR OWN RISK.
! 6: *
! 7: * Permission is hereby granted to use or copy this program
! 8: * for any purpose, provided the above notices are retained on all copies.
! 9: * Permission to modify the code and to distribute modified code is granted,
! 10: * provided the above notices are retained, and a notice that the code was
! 11: * modified is included with the above copyright notice.
! 12: *
! 13: * Author: Hans-J. Boehm (boehm@parc.xerox.com)
! 14: */
! 15: /* Boehm, October 3, 1994 5:19 pm PDT */
! 16: # include "gc.h"
! 17: # include "cord.h"
! 18: # include <stdlib.h>
! 19: # include <stdio.h>
! 20: # include <string.h>
! 21:
! 22: /* An implementation of the cord primitives. These are the only */
! 23: /* Functions that understand the representation. We perform only */
! 24: /* minimal checks on arguments to these functions. Out of bounds */
! 25: /* arguments to the iteration functions may result in client functions */
! 26: /* invoked on garbage data. In most cases, client functions should be */
! 27: /* programmed defensively enough that this does not result in memory */
! 28: /* smashes. */
! 29:
! 30: typedef void (* oom_fn)(void);
! 31:
! 32: oom_fn CORD_oom_fn = (oom_fn) 0;
! 33:
! 34: # define OUT_OF_MEMORY { if (CORD_oom_fn != (oom_fn) 0) (*CORD_oom_fn)(); \
! 35: ABORT("Out of memory\n"); }
! 36: # define ABORT(msg) { fprintf(stderr, "%s\n", msg); abort(); }
! 37:
! 38: typedef unsigned long word;
! 39:
! 40: typedef union {
! 41: struct Concatenation {
! 42: char null;
! 43: char header;
! 44: char depth; /* concatenation nesting depth. */
! 45: unsigned char left_len;
! 46: /* Length of left child if it is sufficiently */
! 47: /* short; 0 otherwise. */
! 48: # define MAX_LEFT_LEN 255
! 49: word len;
! 50: CORD left; /* length(left) > 0 */
! 51: CORD right; /* length(right) > 0 */
! 52: } concatenation;
! 53: struct Function {
! 54: char null;
! 55: char header;
! 56: char depth; /* always 0 */
! 57: char left_len; /* always 0 */
! 58: word len;
! 59: CORD_fn fn;
! 60: void * client_data;
! 61: } function;
! 62: struct Generic {
! 63: char null;
! 64: char header;
! 65: char depth;
! 66: char left_len;
! 67: word len;
! 68: } generic;
! 69: char string[1];
! 70: } CordRep;
! 71:
! 72: # define CONCAT_HDR 1
! 73:
! 74: # define FN_HDR 4
! 75: # define SUBSTR_HDR 6
! 76: /* Substring nodes are a special case of function nodes. */
! 77: /* The client_data field is known to point to a substr_args */
! 78: /* structure, and the function is either CORD_apply_access_fn */
! 79: /* or CORD_index_access_fn. */
! 80:
! 81: /* The following may be applied only to function and concatenation nodes: */
! 82: #define IS_CONCATENATION(s) (((CordRep *)s)->generic.header == CONCAT_HDR)
! 83:
! 84: #define IS_FUNCTION(s) ((((CordRep *)s)->generic.header & FN_HDR) != 0)
! 85:
! 86: #define IS_SUBSTR(s) (((CordRep *)s)->generic.header == SUBSTR_HDR)
! 87:
! 88: #define LEN(s) (((CordRep *)s) -> generic.len)
! 89: #define DEPTH(s) (((CordRep *)s) -> generic.depth)
! 90: #define GEN_LEN(s) (CORD_IS_STRING(s) ? strlen(s) : LEN(s))
! 91:
! 92: #define LEFT_LEN(c) ((c) -> left_len != 0? \
! 93: (c) -> left_len \
! 94: : (CORD_IS_STRING((c) -> left) ? \
! 95: (c) -> len - GEN_LEN((c) -> right) \
! 96: : LEN((c) -> left)))
! 97:
! 98: #define SHORT_LIMIT (sizeof(CordRep) - 1)
! 99: /* Cords shorter than this are C strings */
! 100:
! 101:
! 102: /* Dump the internal representation of x to stdout, with initial */
! 103: /* indentation level n. */
! 104: void CORD_dump_inner(CORD x, unsigned n)
! 105: {
! 106: register size_t i;
! 107:
! 108: for (i = 0; i < (size_t)n; i++) {
! 109: fputs(" ", stdout);
! 110: }
! 111: if (x == 0) {
! 112: fputs("NIL\n", stdout);
! 113: } else if (CORD_IS_STRING(x)) {
! 114: for (i = 0; i <= SHORT_LIMIT; i++) {
! 115: if (x[i] == '\0') break;
! 116: putchar(x[i]);
! 117: }
! 118: if (x[i] != '\0') fputs("...", stdout);
! 119: putchar('\n');
! 120: } else if (IS_CONCATENATION(x)) {
! 121: register struct Concatenation * conc =
! 122: &(((CordRep *)x) -> concatenation);
! 123: printf("Concatenation: %p (len: %d, depth: %d)\n",
! 124: x, (int)(conc -> len), (int)(conc -> depth));
! 125: CORD_dump_inner(conc -> left, n+1);
! 126: CORD_dump_inner(conc -> right, n+1);
! 127: } else /* function */{
! 128: register struct Function * func =
! 129: &(((CordRep *)x) -> function);
! 130: if (IS_SUBSTR(x)) printf("(Substring) ");
! 131: printf("Function: %p (len: %d): ", x, (int)(func -> len));
! 132: for (i = 0; i < 20 && i < func -> len; i++) {
! 133: putchar((*(func -> fn))(i, func -> client_data));
! 134: }
! 135: if (i < func -> len) fputs("...", stdout);
! 136: putchar('\n');
! 137: }
! 138: }
! 139:
! 140: /* Dump the internal representation of x to stdout */
! 141: void CORD_dump(CORD x)
! 142: {
! 143: CORD_dump_inner(x, 0);
! 144: fflush(stdout);
! 145: }
! 146:
! 147: CORD CORD_cat_char_star(CORD x, const char* y, size_t leny)
! 148: {
! 149: register size_t result_len;
! 150: register size_t lenx;
! 151: register int depth;
! 152:
! 153: if (x == CORD_EMPTY) return(y);
! 154: //if (leny == 0) leny=strlen(y); // PAF
! 155: if (y == 0) ABORT("CORD_cat_char_star(,y,) y==0"); // PAF
! 156: if (*y == 0) ABORT("CORD_cat_char_star(,y,) y==\"\""); // PAF
! 157: if (leny == 0) ABORT("CORD_cat_char_star(,y,) leny==0"); // PAF
! 158:
! 159: if (CORD_IS_STRING(x)) {
! 160: lenx = strlen(x);
! 161: result_len = lenx + leny;
! 162: if (result_len <= SHORT_LIMIT) {
! 163: register char * result = GC_MALLOC_ATOMIC(result_len+1);
! 164:
! 165: if (result == 0) OUT_OF_MEMORY;
! 166: memcpy(result, x, lenx);
! 167: memcpy(result + lenx, y, leny);
! 168: result[result_len] = '\0';
! 169: return((CORD) result);
! 170: } else {
! 171: depth = 1;
! 172: }
! 173: } else {
! 174: register CORD right;
! 175: register CORD left;
! 176: register char * new_right;
! 177: register size_t right_len;
! 178:
! 179: lenx = LEN(x);
! 180:
! 181: if (leny <= SHORT_LIMIT/2
! 182: && IS_CONCATENATION(x)
! 183: && CORD_IS_STRING(right = ((CordRep *)x) -> concatenation.right)) {
! 184: /* Merge y into right part of x. */
! 185: if (!CORD_IS_STRING(left = ((CordRep *)x) -> concatenation.left)) {
! 186: right_len = lenx - LEN(left);
! 187: } else if (((CordRep *)x) -> concatenation.left_len != 0) {
! 188: right_len = lenx - ((CordRep *)x) -> concatenation.left_len;
! 189: } else {
! 190: right_len = strlen(right);
! 191: }
! 192: result_len = right_len + leny; /* length of new_right */
! 193: if (result_len <= SHORT_LIMIT) {
! 194: new_right = GC_MALLOC_ATOMIC(result_len + 1);
! 195: memcpy(new_right, right, right_len);
! 196: memcpy(new_right + right_len, y, leny);
! 197: new_right[result_len] = '\0';
! 198: y = new_right;
! 199: leny = result_len;
! 200: x = left;
! 201: lenx -= right_len;
! 202: /* Now fall through to concatenate the two pieces: */
! 203: }
! 204: if (CORD_IS_STRING(x)) {
! 205: depth = 1;
! 206: } else {
! 207: depth = DEPTH(x) + 1;
! 208: }
! 209: } else {
! 210: depth = DEPTH(x) + 1;
! 211: }
! 212: result_len = lenx + leny;
! 213: }
! 214: {
! 215: /* The general case; lenx, result_len is known: */
! 216: register struct Concatenation * result;
! 217:
! 218: result = GC_NEW(struct Concatenation);
! 219: if (result == 0) OUT_OF_MEMORY;
! 220: result->header = CONCAT_HDR;
! 221: result->depth = depth;
! 222: if (lenx <= MAX_LEFT_LEN) result->left_len = lenx;
! 223: result->len = result_len;
! 224: result->left = x;
! 225: result->right = y;
! 226: if (depth >= MAX_DEPTH) {
! 227: return(CORD_balance((CORD)result));
! 228: } else {
! 229: return((CORD) result);
! 230: }
! 231: }
! 232: }
! 233:
! 234:
! 235: CORD CORD_cat(CORD x, CORD y)
! 236: {
! 237: register size_t result_len;
! 238: register int depth;
! 239: register size_t lenx;
! 240:
! 241: if (x == CORD_EMPTY) return(y);
! 242: if (y == CORD_EMPTY) return(x);
! 243: if (CORD_IS_STRING(y)) {
! 244: return(CORD_cat_char_star(x, y, strlen(y)));
! 245: } else if (CORD_IS_STRING(x)) {
! 246: lenx = strlen(x);
! 247: depth = DEPTH(y) + 1;
! 248: } else {
! 249: register int depthy = DEPTH(y);
! 250:
! 251: lenx = LEN(x);
! 252: depth = DEPTH(x) + 1;
! 253: if (depthy >= depth) depth = depthy + 1;
! 254: }
! 255: result_len = lenx + LEN(y);
! 256: {
! 257: register struct Concatenation * result;
! 258:
! 259: result = GC_NEW(struct Concatenation);
! 260: if (result == 0) OUT_OF_MEMORY;
! 261: result->header = CONCAT_HDR;
! 262: result->depth = depth;
! 263: // printf("depth=%d\n", depth);
! 264: if (lenx <= MAX_LEFT_LEN) result->left_len = lenx;
! 265: result->len = result_len;
! 266: result->left = x;
! 267: result->right = y;
! 268: // PAF@design.ru bug fix:
! 269: if (depth >= MAX_DEPTH) {
! 270: return(CORD_balance((CORD)result));
! 271: } else {
! 272: return((CORD) result);
! 273: }
! 274: }
! 275: }
! 276:
! 277:
! 278:
! 279: CORD CORD_from_fn(CORD_fn fn, void * client_data, size_t len)
! 280: {
! 281: if (len <= 0) return(0);
! 282: if (len <= SHORT_LIMIT) {
! 283: register char * result;
! 284: register size_t i;
! 285: char buf[SHORT_LIMIT+1];
! 286: register char c;
! 287:
! 288: for (i = 0; i < len; i++) {
! 289: c = (*fn)(i, client_data);
! 290: if (c == '\0') goto gen_case;
! 291: buf[i] = c;
! 292: }
! 293: buf[i] = '\0';
! 294: result = GC_MALLOC_ATOMIC(len+1);
! 295: if (result == 0) OUT_OF_MEMORY;
! 296: strcpy(result, buf);
! 297: result[len] = '\0';
! 298: return((CORD) result);
! 299: }
! 300: gen_case:
! 301: {
! 302: register struct Function * result;
! 303:
! 304: result = GC_NEW(struct Function);
! 305: if (result == 0) OUT_OF_MEMORY;
! 306: result->header = FN_HDR;
! 307: /* depth is already 0 */
! 308: result->len = len;
! 309: result->fn = fn;
! 310: result->client_data = client_data;
! 311: return((CORD) result);
! 312: }
! 313: }
! 314:
! 315: size_t CORD_len(CORD x)
! 316: {
! 317: if (x == 0) {
! 318: return(0);
! 319: } else {
! 320: return(GEN_LEN(x));
! 321: }
! 322: }
! 323:
! 324: struct substr_args {
! 325: CordRep * sa_cord;
! 326: size_t sa_index;
! 327: };
! 328:
! 329: char CORD_index_access_fn(size_t i, void * client_data)
! 330: {
! 331: register struct substr_args *descr = (struct substr_args *)client_data;
! 332:
! 333: return(((char *)(descr->sa_cord))[i + descr->sa_index]);
! 334: }
! 335:
! 336: char CORD_apply_access_fn(size_t i, void * client_data)
! 337: {
! 338: register struct substr_args *descr = (struct substr_args *)client_data;
! 339: register struct Function * fn_cord = &(descr->sa_cord->function);
! 340:
! 341: return((*(fn_cord->fn))(i + descr->sa_index, fn_cord->client_data));
! 342: }
! 343:
! 344: /* A version of CORD_substr that simply returns a function node, thus */
! 345: /* postponing its work. The fourth argument is a function that may */
! 346: /* be used for efficient access to the ith character. */
! 347: /* Assumes i >= 0 and i + n < length(x). */
! 348: CORD CORD_substr_closure(CORD x, size_t i, size_t n, CORD_fn f)
! 349: {
! 350: register struct substr_args * sa = GC_NEW(struct substr_args);
! 351: CORD result;
! 352:
! 353: if (sa == 0) OUT_OF_MEMORY;
! 354: sa->sa_cord = (CordRep *)x;
! 355: sa->sa_index = i;
! 356: result = CORD_from_fn(f, (void *)sa, n);
! 357: ((CordRep *)result) -> function.header = SUBSTR_HDR;
! 358: return (result);
! 359: }
! 360:
! 361: # define SUBSTR_LIMIT (10 * SHORT_LIMIT)
! 362: /* Substrings of function nodes and flat strings shorter than */
! 363: /* this are flat strings. Othewise we use a functional */
! 364: /* representation, which is significantly slower to access. */
! 365:
! 366: /* A version of CORD_substr that assumes i >= 0, n > 0, and i + n < length(x).*/
! 367: CORD CORD_substr_checked(CORD x, size_t i, size_t n)
! 368: {
! 369: if (CORD_IS_STRING(x)) {
! 370: if (n > SUBSTR_LIMIT) {
! 371: return(CORD_substr_closure(x, i, n, CORD_index_access_fn));
! 372: } else {
! 373: register char * result = GC_MALLOC_ATOMIC(n+1);
! 374:
! 375: if (result == 0) OUT_OF_MEMORY;
! 376: strncpy(result, x+i, n);
! 377: result[n] = '\0';
! 378: return(result);
! 379: }
! 380: } else if (IS_CONCATENATION(x)) {
! 381: register struct Concatenation * conc
! 382: = &(((CordRep *)x) -> concatenation);
! 383: register size_t left_len;
! 384: register size_t right_len;
! 385:
! 386: left_len = LEFT_LEN(conc);
! 387: right_len = conc -> len - left_len;
! 388: if (i >= left_len) {
! 389: if (n == right_len) return(conc -> right);
! 390: return(CORD_substr_checked(conc -> right, i - left_len, n));
! 391: } else if (i+n <= left_len) {
! 392: if (n == left_len) return(conc -> left);
! 393: return(CORD_substr_checked(conc -> left, i, n));
! 394: } else {
! 395: /* Need at least one character from each side. */
! 396: register CORD left_part;
! 397: register CORD right_part;
! 398: register size_t left_part_len = left_len - i;
! 399:
! 400: if (i == 0) {
! 401: left_part = conc -> left;
! 402: } else {
! 403: left_part = CORD_substr_checked(conc -> left, i, left_part_len);
! 404: }
! 405: if (i + n == right_len + left_len) {
! 406: right_part = conc -> right;
! 407: } else {
! 408: right_part = CORD_substr_checked(conc -> right, 0,
! 409: n - left_part_len);
! 410: }
! 411: return(CORD_cat(left_part, right_part));
! 412: }
! 413: } else /* function */ {
! 414: if (n > SUBSTR_LIMIT) {
! 415: if (IS_SUBSTR(x)) {
! 416: /* Avoid nesting substring nodes. */
! 417: register struct Function * f = &(((CordRep *)x) -> function);
! 418: register struct substr_args *descr =
! 419: (struct substr_args *)(f -> client_data);
! 420:
! 421: return(CORD_substr_closure((CORD)descr->sa_cord,
! 422: i + descr->sa_index,
! 423: n, f -> fn));
! 424: } else {
! 425: return(CORD_substr_closure(x, i, n, CORD_apply_access_fn));
! 426: }
! 427: } else {
! 428: char * result;
! 429: register struct Function * f = &(((CordRep *)x) -> function);
! 430: char buf[SUBSTR_LIMIT+1];
! 431: register char * p = buf;
! 432: register char c;
! 433: register int j;
! 434: register int lim = i + n;
! 435:
! 436: for (j = i; j < lim; j++) {
! 437: c = (*(f -> fn))(j, f -> client_data);
! 438: if (c == '\0') {
! 439: return(CORD_substr_closure(x, i, n, CORD_apply_access_fn));
! 440: }
! 441: *p++ = c;
! 442: }
! 443: *p = '\0';
! 444: result = GC_MALLOC_ATOMIC(n+1);
! 445: if (result == 0) OUT_OF_MEMORY;
! 446: strcpy(result, buf);
! 447: return(result);
! 448: }
! 449: }
! 450: }
! 451:
! 452: CORD CORD_substr(CORD x, size_t i, size_t n)
! 453: {
! 454: register size_t len = CORD_len(x);
! 455:
! 456: if (i >= len || n <= 0) return(0);
! 457: /* n < 0 is impossible in a correct C implementation, but */
! 458: /* quite possible under SunOS 4.X. */
! 459: if (i + n > len) n = len - i;
! 460: # ifndef __STDC__
! 461: if (i < 0) ABORT("CORD_substr: second arg. negative");
! 462: /* Possible only if both client and C implementation are buggy. */
! 463: /* But empirically this happens frequently. */
! 464: # endif
! 465: return(CORD_substr_checked(x, i, n));
! 466: }
! 467:
! 468: /* See cord.h for definition. We assume i is in range. */
! 469: int CORD_iter5(CORD x, size_t i, CORD_iter_fn f1,
! 470: CORD_batched_iter_fn f2, void * client_data)
! 471: {
! 472: if (x == 0) return(0);
! 473: if (CORD_IS_STRING(x)) {
! 474: register const char* p = x+i;
! 475:
! 476: if (*p == '\0') ABORT("2nd arg to CORD_iter5 too big");
! 477: if (f2 != CORD_NO_FN) {
! 478: return((*f2)(p, client_data));
! 479: } else {
! 480: while (*p) {
! 481: if ((*f1)(*p, client_data)) return(1);
! 482: p++;
! 483: }
! 484: return(0);
! 485: }
! 486: } else if (IS_CONCATENATION(x)) {
! 487: register struct Concatenation * conc
! 488: = &(((CordRep *)x) -> concatenation);
! 489:
! 490:
! 491: if (i > 0) {
! 492: register size_t left_len = LEFT_LEN(conc);
! 493:
! 494: if (i >= left_len) {
! 495: return(CORD_iter5(conc -> right, i - left_len, f1, f2,
! 496: client_data));
! 497: }
! 498: }
! 499: if (CORD_iter5(conc -> left, i, f1, f2, client_data)) {
! 500: return(1);
! 501: }
! 502: return(CORD_iter5(conc -> right, 0, f1, f2, client_data));
! 503: } else /* function */ {
! 504: register struct Function * f = &(((CordRep *)x) -> function);
! 505: register size_t j;
! 506: register size_t lim = f -> len;
! 507:
! 508: for (j = i; j < lim; j++) {
! 509: if ((*f1)((*(f -> fn))(j, f -> client_data), client_data)) {
! 510: return(1);
! 511: }
! 512: }
! 513: return(0);
! 514: }
! 515: }
! 516:
! 517: #undef CORD_iter
! 518: int CORD_iter(CORD x, CORD_iter_fn f1, void * client_data)
! 519: {
! 520: return(CORD_iter5(x, 0, f1, CORD_NO_FN, client_data));
! 521: }
! 522:
! 523: int CORD_riter4(CORD x, size_t i, CORD_iter_fn f1, void * client_data)
! 524: {
! 525: if (x == 0) return(0);
! 526: if (CORD_IS_STRING(x)) {
! 527: register const char* p = x + i;
! 528: register char c;
! 529:
! 530: for(;;) {
! 531: c = *p;
! 532: if (c == '\0') ABORT("2nd arg to CORD_riter4 too big");
! 533: if ((*f1)(c, client_data)) return(1);
! 534: if (p == x) break;
! 535: p--;
! 536: }
! 537: return(0);
! 538: } else if (IS_CONCATENATION(x)) {
! 539: register struct Concatenation * conc
! 540: = &(((CordRep *)x) -> concatenation);
! 541: register CORD left_part = conc -> left;
! 542: register size_t left_len;
! 543:
! 544: left_len = LEFT_LEN(conc);
! 545: if (i >= left_len) {
! 546: if (CORD_riter4(conc -> right, i - left_len, f1, client_data)) {
! 547: return(1);
! 548: }
! 549: return(CORD_riter4(left_part, left_len - 1, f1, client_data));
! 550: } else {
! 551: return(CORD_riter4(left_part, i, f1, client_data));
! 552: }
! 553: } else /* function */ {
! 554: register struct Function * f = &(((CordRep *)x) -> function);
! 555: register size_t j;
! 556:
! 557: for (j = i; ; j--) {
! 558: if ((*f1)((*(f -> fn))(j, f -> client_data), client_data)) {
! 559: return(1);
! 560: }
! 561: if (j == 0) return(0);
! 562: }
! 563: }
! 564: }
! 565:
! 566: int CORD_riter(CORD x, CORD_iter_fn f1, void * client_data)
! 567: {
! 568: return(CORD_riter4(x, CORD_len(x) - 1, f1, client_data));
! 569: }
! 570:
! 571: /*
! 572: * The following functions are concerned with balancing cords.
! 573: * Strategy:
! 574: * Scan the cord from left to right, keeping the cord scanned so far
! 575: * as a forest of balanced trees of exponentialy decreasing length.
! 576: * When a new subtree needs to be added to the forest, we concatenate all
! 577: * shorter ones to the new tree in the appropriate order, and then insert
! 578: * the result into the forest.
! 579: * Crucial invariants:
! 580: * 1. The concatenation of the forest (in decreasing order) with the
! 581: * unscanned part of the rope is equal to the rope being balanced.
! 582: * 2. All trees in the forest are balanced.
! 583: * 3. forest[i] has depth at most i.
! 584: */
! 585:
! 586: typedef struct {
! 587: CORD c;
! 588: size_t len; /* Actual length of c */
! 589: } ForestElement;
! 590:
! 591: static size_t min_len [ MAX_DEPTH ];
! 592:
! 593: static int min_len_init = 0;
! 594:
! 595: int CORD_max_len;
! 596:
! 597: typedef ForestElement Forest [ MAX_DEPTH ];
! 598: /* forest[i].len >= fib(i+1) */
! 599: /* The string is the concatenation */
! 600: /* of the forest in order of DECREASING */
! 601: /* indices. */
! 602:
! 603: void CORD_init_min_len()
! 604: {
! 605: register int i;
! 606: register size_t last, previous, current;
! 607:
! 608: min_len[0] = previous = 1;
! 609: min_len[1] = last = 2;
! 610: for (i = 2; i < MAX_DEPTH; i++) {
! 611: current = last + previous;
! 612: if (current < last) /* overflow */ current = last;
! 613: min_len[i] = current;
! 614: previous = last;
! 615: last = current;
! 616: }
! 617: CORD_max_len = last - 1;
! 618: min_len_init = 1;
! 619: }
! 620:
! 621:
! 622: void CORD_init_forest(ForestElement * forest, size_t max_len)
! 623: {
! 624: register int i;
! 625:
! 626: for (i = 0; i < MAX_DEPTH; i++) {
! 627: forest[i].c = 0;
! 628: if (min_len[i] > max_len) return;
! 629: }
! 630: ABORT("Cord too long");
! 631: }
! 632:
! 633: /* Add a leaf to the appropriate level in the forest, cleaning */
! 634: /* out lower levels as necessary. */
! 635: /* Also works if x is a balanced tree of concatenations; however */
! 636: /* in this case an extra concatenation node may be inserted above x; */
! 637: /* This node should not be counted in the statement of the invariants. */
! 638: void CORD_add_forest(ForestElement * forest, CORD x, size_t len)
! 639: {
! 640: register int i = 0;
! 641: register CORD sum = CORD_EMPTY;
! 642: register size_t sum_len = 0;
! 643:
! 644: while (len > min_len[i + 1]) {
! 645: if (forest[i].c != 0) {
! 646: sum = CORD_cat(forest[i].c, sum);
! 647: sum_len += forest[i].len;
! 648: forest[i].c = 0;
! 649: }
! 650: i++;
! 651: }
! 652: /* Sum has depth at most 1 greter than what would be required */
! 653: /* for balance. */
! 654: sum = CORD_cat(sum, x);
! 655: sum_len += len;
! 656: /* If x was a leaf, then sum is now balanced. To see this */
! 657: /* consider the two cases in which forest[i-1] either is or is */
! 658: /* not empty. */
! 659: while (sum_len >= min_len[i]) {
! 660: if (forest[i].c != 0) {
! 661: sum = CORD_cat(forest[i].c, sum);
! 662: sum_len += forest[i].len;
! 663: /* This is again balanced, since sum was balanced, and has */
! 664: /* allowable depth that differs from i by at most 1. */
! 665: forest[i].c = 0;
! 666: }
! 667: i++;
! 668: }
! 669: i--;
! 670: forest[i].c = sum;
! 671: forest[i].len = sum_len;
! 672: }
! 673:
! 674: CORD CORD_concat_forest(ForestElement * forest, size_t expected_len)
! 675: {
! 676: register int i = 0;
! 677: CORD sum = 0;
! 678: size_t sum_len = 0;
! 679:
! 680: while (sum_len != expected_len) {
! 681: if (forest[i].c != 0) {
! 682: sum = CORD_cat(forest[i].c, sum);
! 683: sum_len += forest[i].len;
! 684: }
! 685: i++;
! 686: }
! 687: return(sum);
! 688: }
! 689:
! 690: /* Insert the frontier of x into forest. Balanced subtrees are */
! 691: /* treated as leaves. This potentially adds one to the depth */
! 692: /* of the final tree. */
! 693: void CORD_balance_insert(CORD x, size_t len, ForestElement * forest)
! 694: {
! 695: register int depth;
! 696:
! 697: if (CORD_IS_STRING(x)) {
! 698: CORD_add_forest(forest, x, len);
! 699: } else if (IS_CONCATENATION(x)
! 700: && ((depth = DEPTH(x)) >= MAX_DEPTH
! 701: || len < min_len[depth])) {
! 702: register struct Concatenation * conc
! 703: = &(((CordRep *)x) -> concatenation);
! 704: size_t left_len = LEFT_LEN(conc);
! 705:
! 706: CORD_balance_insert(conc -> left, left_len, forest);
! 707: CORD_balance_insert(conc -> right, len - left_len, forest);
! 708: } else /* function or balanced */ {
! 709: CORD_add_forest(forest, x, len);
! 710: }
! 711: }
! 712:
! 713:
! 714: CORD CORD_balance(CORD x)
! 715: {
! 716: Forest forest;
! 717: register size_t len;
! 718:
! 719: if (x == 0) return(0);
! 720: if (CORD_IS_STRING(x)) return(x);
! 721: if (!min_len_init) CORD_init_min_len();
! 722: len = LEN(x);
! 723: CORD_init_forest(forest, len);
! 724: CORD_balance_insert(x, len, forest);
! 725: return(CORD_concat_forest(forest, len));
! 726: }
! 727:
! 728:
! 729: /* Position primitives */
! 730:
! 731: /* Private routines to deal with the hard cases only: */
! 732:
! 733: /* P contains a prefix of the path to cur_pos. Extend it to a full */
! 734: /* path and set up leaf info. */
! 735: /* Return 0 if past the end of cord, 1 o.w. */
! 736: void CORD__extend_path(register CORD_pos p)
! 737: {
! 738: register struct CORD_pe * current_pe = &(p[0].path[p[0].path_len]);
! 739: register CORD top = current_pe -> pe_cord;
! 740: register size_t pos = p[0].cur_pos;
! 741: register size_t top_pos = current_pe -> pe_start_pos;
! 742: register size_t top_len = GEN_LEN(top);
! 743:
! 744: /* Fill in the rest of the path. */
! 745: while(!CORD_IS_STRING(top) && IS_CONCATENATION(top)) {
! 746: register struct Concatenation * conc =
! 747: &(((CordRep *)top) -> concatenation);
! 748: register size_t left_len;
! 749:
! 750: left_len = LEFT_LEN(conc);
! 751: current_pe++;
! 752: if (pos >= top_pos + left_len) {
! 753: current_pe -> pe_cord = top = conc -> right;
! 754: current_pe -> pe_start_pos = top_pos = top_pos + left_len;
! 755: top_len -= left_len;
! 756: } else {
! 757: current_pe -> pe_cord = top = conc -> left;
! 758: current_pe -> pe_start_pos = top_pos;
! 759: top_len = left_len;
! 760: }
! 761: p[0].path_len++;
! 762: }
! 763: /* Fill in leaf description for fast access. */
! 764: if (CORD_IS_STRING(top)) {
! 765: p[0].cur_leaf = top;
! 766: p[0].cur_start = top_pos;
! 767: p[0].cur_end = top_pos + top_len;
! 768: } else {
! 769: p[0].cur_end = 0;
! 770: }
! 771: if (pos >= top_pos + top_len) p[0].path_len = CORD_POS_INVALID;
! 772: }
! 773:
! 774: char CORD__pos_fetch(register CORD_pos p)
! 775: {
! 776: /* Leaf is a function node */
! 777: struct CORD_pe * pe = &((p)[0].path[(p)[0].path_len]);
! 778: CORD leaf = pe -> pe_cord;
! 779: register struct Function * f = &(((CordRep *)leaf) -> function);
! 780:
! 781: if (!IS_FUNCTION(leaf)) ABORT("CORD_pos_fetch: bad leaf");
! 782: return ((*(f -> fn))(p[0].cur_pos - pe -> pe_start_pos, f -> client_data));
! 783: }
! 784:
! 785: void CORD__next(register CORD_pos p)
! 786: {
! 787: register size_t cur_pos = p[0].cur_pos + 1;
! 788: register struct CORD_pe * current_pe = &((p)[0].path[(p)[0].path_len]);
! 789: register CORD leaf = current_pe -> pe_cord;
! 790:
! 791: /* Leaf is not a string or we're at end of leaf */
! 792: p[0].cur_pos = cur_pos;
! 793: if (!CORD_IS_STRING(leaf)) {
! 794: /* Function leaf */
! 795: register struct Function * f = &(((CordRep *)leaf) -> function);
! 796: register size_t start_pos = current_pe -> pe_start_pos;
! 797: register size_t end_pos = start_pos + f -> len;
! 798:
! 799: if (cur_pos < end_pos) {
! 800: /* Fill cache and return. */
! 801: register size_t i;
! 802: register size_t limit = cur_pos + FUNCTION_BUF_SZ;
! 803: register CORD_fn fn = f -> fn;
! 804: register void * client_data = f -> client_data;
! 805:
! 806: if (limit > end_pos) {
! 807: limit = end_pos;
! 808: }
! 809: for (i = cur_pos; i < limit; i++) {
! 810: p[0].function_buf[i - cur_pos] =
! 811: (*fn)(i - start_pos, client_data);
! 812: }
! 813: p[0].cur_start = cur_pos;
! 814: p[0].cur_leaf = p[0].function_buf;
! 815: p[0].cur_end = limit;
! 816: return;
! 817: }
! 818: }
! 819: /* End of leaf */
! 820: /* Pop the stack until we find two concatenation nodes with the */
! 821: /* same start position: this implies we were in left part. */
! 822: {
! 823: while (p[0].path_len > 0
! 824: && current_pe[0].pe_start_pos != current_pe[-1].pe_start_pos) {
! 825: p[0].path_len--;
! 826: current_pe--;
! 827: }
! 828: if (p[0].path_len == 0) {
! 829: p[0].path_len = CORD_POS_INVALID;
! 830: return;
! 831: }
! 832: }
! 833: p[0].path_len--;
! 834: CORD__extend_path(p);
! 835: }
! 836:
! 837: void CORD__prev(register CORD_pos p)
! 838: {
! 839: register struct CORD_pe * pe = &(p[0].path[p[0].path_len]);
! 840:
! 841: if (p[0].cur_pos == 0) {
! 842: p[0].path_len = CORD_POS_INVALID;
! 843: return;
! 844: }
! 845: p[0].cur_pos--;
! 846: if (p[0].cur_pos >= pe -> pe_start_pos) return;
! 847:
! 848: /* Beginning of leaf */
! 849:
! 850: /* Pop the stack until we find two concatenation nodes with the */
! 851: /* different start position: this implies we were in right part. */
! 852: {
! 853: register struct CORD_pe * current_pe = &((p)[0].path[(p)[0].path_len]);
! 854:
! 855: while (p[0].path_len > 0
! 856: && current_pe[0].pe_start_pos == current_pe[-1].pe_start_pos) {
! 857: p[0].path_len--;
! 858: current_pe--;
! 859: }
! 860: }
! 861: p[0].path_len--;
! 862: CORD__extend_path(p);
! 863: }
! 864:
! 865: #undef CORD_pos_fetch
! 866: #undef CORD_next
! 867: #undef CORD_prev
! 868: #undef CORD_pos_to_index
! 869: #undef CORD_pos_to_cord
! 870: #undef CORD_pos_valid
! 871:
! 872: char CORD_pos_fetch(register CORD_pos p)
! 873: {
! 874: if (p[0].cur_start <= p[0].cur_pos && p[0].cur_pos < p[0].cur_end) {
! 875: return(p[0].cur_leaf[p[0].cur_pos - p[0].cur_start]);
! 876: } else {
! 877: return(CORD__pos_fetch(p));
! 878: }
! 879: }
! 880:
! 881: void CORD_next(CORD_pos p)
! 882: {
! 883: if (p[0].cur_pos < p[0].cur_end - 1) {
! 884: p[0].cur_pos++;
! 885: } else {
! 886: CORD__next(p);
! 887: }
! 888: }
! 889:
! 890: void CORD_prev(CORD_pos p)
! 891: {
! 892: if (p[0].cur_end != 0 && p[0].cur_pos > p[0].cur_start) {
! 893: p[0].cur_pos--;
! 894: } else {
! 895: CORD__prev(p);
! 896: }
! 897: }
! 898:
! 899: size_t CORD_pos_to_index(CORD_pos p)
! 900: {
! 901: return(p[0].cur_pos);
! 902: }
! 903:
! 904: CORD CORD_pos_to_cord(CORD_pos p)
! 905: {
! 906: return(p[0].path[0].pe_cord);
! 907: }
! 908:
! 909: int CORD_pos_valid(CORD_pos p)
! 910: {
! 911: return(p[0].path_len != CORD_POS_INVALID);
! 912: }
! 913:
! 914: void CORD_set_pos(CORD_pos p, CORD x, size_t i)
! 915: {
! 916: if (x == CORD_EMPTY) {
! 917: p[0].path_len = CORD_POS_INVALID;
! 918: return;
! 919: }
! 920: p[0].path[0].pe_cord = x;
! 921: p[0].path[0].pe_start_pos = 0;
! 922: p[0].path_len = 0;
! 923: p[0].cur_pos = i;
! 924: CORD__extend_path(p);
! 925: }
E-mail:
![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to cordbscs.c CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [parser3project] / parser3 / src / lib / cord