Annotation of parser3/src/lib/cord/cordbscs.c, revision 1.10
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 */
1.6 paf 16: #include "pa_config_includes.h"
1.2 paf 17: # include "gc.h"
18: # include "cord.h"
19: # include <stdlib.h>
20: # include <stdio.h>
21: # include <string.h>
22:
23: /* An implementation of the cord primitives. These are the only */
24: /* Functions that understand the representation. We perform only */
25: /* minimal checks on arguments to these functions. Out of bounds */
26: /* arguments to the iteration functions may result in client functions */
27: /* invoked on garbage data. In most cases, client functions should be */
28: /* programmed defensively enough that this does not result in memory */
29: /* smashes. */
30:
31: typedef void (* oom_fn)(void);
32:
33: oom_fn CORD_oom_fn = (oom_fn) 0;
34:
35: # define OUT_OF_MEMORY { if (CORD_oom_fn != (oom_fn) 0) (*CORD_oom_fn)(); \
36: ABORT("Out of memory\n"); }
37: # define ABORT(msg) { fprintf(stderr, "%s\n", msg); abort(); }
38:
39: typedef unsigned long word;
40:
41: typedef union {
42: struct Concatenation {
43: char null;
44: char header;
45: char depth; /* concatenation nesting depth. */
46: unsigned char left_len;
47: /* Length of left child if it is sufficiently */
48: /* short; 0 otherwise. */
49: # define MAX_LEFT_LEN 255
50: word len;
51: CORD left; /* length(left) > 0 */
52: CORD right; /* length(right) > 0 */
53: } concatenation;
54: struct Function {
55: char null;
56: char header;
57: char depth; /* always 0 */
58: char left_len; /* always 0 */
59: word len;
60: CORD_fn fn;
61: void * client_data;
62: } function;
63: struct Generic {
64: char null;
65: char header;
66: char depth;
67: char left_len;
68: word len;
69: } generic;
70: char string[1];
71: } CordRep;
72:
73: # define CONCAT_HDR 1
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:
1.8 misha 81: #ifdef CORD_CAT_OPTIMIZATION
82:
83: #define CONCAT_HDR_RO 3
84:
85: #define IS_CONCATENATION(s) ((((CordRep *)s)->generic.header & CONCAT_HDR) != 0)
86: #define IS_RW_CONCATENATION(s) (((CordRep *)s)->generic.header == CONCAT_HDR)
87:
88: #else
89:
1.2 paf 90: /* The following may be applied only to function and concatenation nodes: */
91: #define IS_CONCATENATION(s) (((CordRep *)s)->generic.header == CONCAT_HDR)
92:
1.8 misha 93: #endif
94:
1.2 paf 95: #define IS_FUNCTION(s) ((((CordRep *)s)->generic.header & FN_HDR) != 0)
96:
97: #define IS_SUBSTR(s) (((CordRep *)s)->generic.header == SUBSTR_HDR)
98:
99: #define LEN(s) (((CordRep *)s) -> generic.len)
100: #define DEPTH(s) (((CordRep *)s) -> generic.depth)
101: #define GEN_LEN(s) (CORD_IS_STRING(s) ? strlen(s) : LEN(s))
102:
103: #define LEFT_LEN(c) ((c) -> left_len != 0? \
104: (c) -> left_len \
105: : (CORD_IS_STRING((c) -> left) ? \
106: (c) -> len - GEN_LEN((c) -> right) \
107: : LEN((c) -> left)))
108:
109: #define SHORT_LIMIT (sizeof(CordRep) - 1)
110: /* Cords shorter than this are C strings */
111:
1.3 paf 112: /* paf: using knowledge of interal structure to speedup */
113: char CORD_nul_func(size_t i, void * client_data);
1.2 paf 114:
115: /* Dump the internal representation of x to stdout, with initial */
116: /* indentation level n. */
117: void CORD_dump_inner(CORD x, unsigned n)
118: {
119: register size_t i;
120:
121: for (i = 0; i < (size_t)n; i++) {
122: fputs(" ", stdout);
123: }
124: if (x == 0) {
125: fputs("NIL\n", stdout);
126: } else if (CORD_IS_STRING(x)) {
1.5 paf 127: for (i = 0; i <= SHORT_LIMIT*1000; i++) {
128: if (x[i] == '\0') { putchar('!'); break; }
129: switch(x[i]){
130: case '\n': putchar('|'); break;
131: case '\r': putchar('#'); break;
132: case '\t': putchar('@'); break;
133: default: putchar(x[i]); break;
134: }
1.2 paf 135: }
136: if (x[i] != '\0') fputs("...", stdout);
137: putchar('\n');
138: } else if (IS_CONCATENATION(x)) {
139: register struct Concatenation * conc =
140: &(((CordRep *)x) -> concatenation);
141: printf("Concatenation: %p (len: %d, depth: %d)\n",
142: x, (int)(conc -> len), (int)(conc -> depth));
143: CORD_dump_inner(conc -> left, n+1);
144: CORD_dump_inner(conc -> right, n+1);
145: } else /* function */{
146: register struct Function * func =
147: &(((CordRep *)x) -> function);
148: if (IS_SUBSTR(x)) printf("(Substring) ");
149: printf("Function: %p (len: %d): ", x, (int)(func -> len));
1.5 paf 150: for (i = 0; i < 20*1000 && i < func -> len; i++) {
1.2 paf 151: putchar((*(func -> fn))(i, func -> client_data));
152: }
153: if (i < func -> len) fputs("...", stdout);
154: putchar('\n');
155: }
156: }
157:
158: /* Dump the internal representation of x to stdout */
159: void CORD_dump(CORD x)
160: {
161: CORD_dump_inner(x, 0);
162: fflush(stdout);
163: }
164:
165: CORD CORD_cat_char_star(CORD x, const char* y, size_t leny)
166: {
167: register size_t result_len;
168: register size_t lenx;
169: register int depth;
170:
171: if (x == CORD_EMPTY) return(y);
172: //if (leny == 0) leny=strlen(y); // PAF
173: if (y == 0) ABORT("CORD_cat_char_star(,y,) y==0"); // PAF
174: if (*y == 0) ABORT("CORD_cat_char_star(,y,) y==\"\""); // PAF
175: if (leny == 0) ABORT("CORD_cat_char_star(,y,) leny==0"); // PAF
176:
177: if (CORD_IS_STRING(x)) {
178: lenx = strlen(x);
179: result_len = lenx + leny;
180: if (result_len <= SHORT_LIMIT) {
181: register char * result = GC_MALLOC_ATOMIC(result_len+1);
182:
183: if (result == 0) OUT_OF_MEMORY;
184: memcpy(result, x, lenx);
185: memcpy(result + lenx, y, leny);
186: result[result_len] = '\0';
187: return((CORD) result);
188: } else {
189: depth = 1;
190: }
191: } else {
192: register CORD right;
193: register CORD left;
194: register char * new_right;
195: register size_t right_len;
196:
197: lenx = LEN(x);
198:
199: if (leny <= SHORT_LIMIT/2
200: && IS_CONCATENATION(x)
201: && CORD_IS_STRING(right = ((CordRep *)x) -> concatenation.right)) {
202: /* Merge y into right part of x. */
203: if (!CORD_IS_STRING(left = ((CordRep *)x) -> concatenation.left)) {
204: right_len = lenx - LEN(left);
205: } else if (((CordRep *)x) -> concatenation.left_len != 0) {
206: right_len = lenx - ((CordRep *)x) -> concatenation.left_len;
207: } else {
208: right_len = strlen(right);
209: }
210: result_len = right_len + leny; /* length of new_right */
211: if (result_len <= SHORT_LIMIT) {
212: new_right = GC_MALLOC_ATOMIC(result_len + 1);
1.10 ! misha 213: if (new_right == 0) OUT_OF_MEMORY;
1.2 paf 214: memcpy(new_right, right, right_len);
215: memcpy(new_right + right_len, y, leny);
216: new_right[result_len] = '\0';
217: y = new_right;
218: leny = result_len;
219: x = left;
220: lenx -= right_len;
221: /* Now fall through to concatenate the two pieces: */
222: }
223: if (CORD_IS_STRING(x)) {
224: depth = 1;
225: } else {
226: depth = DEPTH(x) + 1;
227: }
228: } else {
229: depth = DEPTH(x) + 1;
230: }
231: result_len = lenx + leny;
232: }
233: {
234: /* The general case; lenx, result_len is known: */
235: register struct Concatenation * result;
236:
237: result = GC_NEW(struct Concatenation);
238: if (result == 0) OUT_OF_MEMORY;
239: result->header = CONCAT_HDR;
240: result->depth = depth;
241: if (lenx <= MAX_LEFT_LEN) result->left_len = lenx;
242: result->len = result_len;
243: result->left = x;
244: result->right = y;
245: if (depth >= MAX_DEPTH) {
246: return(CORD_balance((CORD)result));
247: } else {
248: return((CORD) result);
249: }
250: }
251: }
252:
253: CORD CORD_cat(CORD x, CORD y)
254: {
255: register size_t result_len;
256: register int depth;
257: register size_t lenx;
258:
1.8 misha 259: if (x == CORD_EMPTY) return(y);
1.2 paf 260: if (y == CORD_EMPTY) return(x);
261: if (CORD_IS_STRING(y)) {
262: return(CORD_cat_char_star(x, y, strlen(y)));
263: } else if (CORD_IS_STRING(x)) {
264: lenx = strlen(x);
265: depth = DEPTH(y) + 1;
266: } else {
267: register int depthy = DEPTH(y);
268:
269: lenx = LEN(x);
270: depth = DEPTH(x) + 1;
271: if (depthy >= depth) depth = depthy + 1;
272: }
273: result_len = lenx + LEN(y);
274: {
275: register struct Concatenation * result;
276:
277: result = GC_NEW(struct Concatenation);
278: if (result == 0) OUT_OF_MEMORY;
279: result->header = CONCAT_HDR;
280: result->depth = depth;
281: // printf("depth=%d\n", depth);
282: if (lenx <= MAX_LEFT_LEN) result->left_len = lenx;
283: result->len = result_len;
284: result->left = x;
285: result->right = y;
286: // PAF@design.ru bug fix:
287: if (depth >= MAX_DEPTH) {
288: return(CORD_balance((CORD)result));
289: } else {
290: return((CORD) result);
291: }
292: }
293: }
294:
1.8 misha 295: #ifdef CORD_CAT_OPTIMIZATION
296: void CORD_concatenation_protect(CORD x){
297: if(IS_RW_CONCATENATION(x))
298: ((struct Concatenation*)x)->header = CONCAT_HDR_RO;
299: }
300:
301: /* Optimized version to be called from parser code */
302: CORD CORD_cat_char_star_optimized(CORD x, const char* y, size_t leny)
303: {
304: register size_t result_len;
305: register size_t lenx;
306: register int depth;
307:
308: if (x == CORD_EMPTY) return(y);
309: //if (leny == 0) leny=strlen(y); // PAF
310: if (y == 0) ABORT("CORD_cat_char_star(,y,) y==0"); // PAF
311: if (*y == 0) ABORT("CORD_cat_char_star(,y,) y==\"\""); // PAF
312: if (leny == 0) ABORT("CORD_cat_char_star(,y,) leny==0"); // PAF
313:
314: if (CORD_IS_STRING(x)) {
315: lenx = strlen(x);
316: result_len = lenx + leny;
317: if (result_len <= SHORT_LIMIT) {
318: register char * result = GC_MALLOC_ATOMIC(result_len+1);
319:
320: if (result == 0) OUT_OF_MEMORY;
321: memcpy(result, x, lenx);
322: memcpy(result + lenx, y, leny);
323: result[result_len] = '\0';
324: return((CORD) result);
325: } else {
326: depth = 1;
327: }
328: } else {
329: register CORD right;
330: register CORD left;
331: register char * new_right;
332: register size_t right_len;
1.10 ! misha 333:
1.8 misha 334: lenx = LEN(x);
1.10 ! misha 335:
1.8 misha 336: if (
337: leny <= SHORT_LIMIT/2
338: && IS_CONCATENATION(x)
339: && CORD_IS_STRING(right = ((CordRep *)x) -> concatenation.right)
340: ){
341: /* Merge y into right part of x. */
342: if (!CORD_IS_STRING(left = ((CordRep *)x) -> concatenation.left)) {
343: right_len = lenx - LEN(left);
344: } else if (((CordRep *)x) -> concatenation.left_len != 0) {
345: right_len = lenx - ((CordRep *)x) -> concatenation.left_len;
346: } else {
347: right_len = strlen(right);
348: }
349: result_len = right_len + leny; /* length of new_right */
350: if (result_len <= SHORT_LIMIT) {
351: new_right = GC_MALLOC_ATOMIC(result_len + 1);
1.10 ! misha 352: if (new_right == 0) OUT_OF_MEMORY;
1.8 misha 353: memcpy(new_right, right, right_len);
354: memcpy(new_right + right_len, y, leny);
355: new_right[result_len] = '\0';
356:
357: if (IS_RW_CONCATENATION(x)) {
358: // Optimization: instead of new Concatenation current is modified
359: ((CordRep *)x) -> concatenation.right=new_right;
360: ((CordRep *)x) -> concatenation.len += leny;
361: return x;
362: }
363:
364: y = new_right;
365: leny = result_len;
366: x = left;
367: lenx -= right_len;
368: /* Now fall through to concatenate the two pieces: */
369: }
370: if (CORD_IS_STRING(x)) {
371: depth = 1;
372: } else {
373: depth = DEPTH(x) + 1;
374: }
375: } else {
376: depth = DEPTH(x) + 1;
377: }
378: result_len = lenx + leny;
379: }
380: {
381: /* The general case; lenx, result_len is known: */
382: register struct Concatenation * result;
1.10 ! misha 383:
1.8 misha 384: result = GC_NEW(struct Concatenation);
385: if (result == 0) OUT_OF_MEMORY;
386: result->header = CONCAT_HDR;
387: result->depth = depth;
388: if (lenx <= MAX_LEFT_LEN) result->left_len = lenx;
389: result->len = result_len;
390: result->left = x;
391: result->right = y;
392: if (depth >= MAX_DEPTH) {
393: return(CORD_balance((CORD)result));
394: } else {
395: return((CORD) result);
396: }
397: }
398: }
399:
400: /* Optimized version to be called from parser code */
401: CORD CORD_cat_optimized(CORD x, CORD y)
402: {
403: register size_t result_len;
404: register int depth;
405: register size_t lenx;
406:
407: if (x == CORD_EMPTY){
408: CORD_concatenation_protect(y); // to guarantee y won't be modified
409: return(y);
410: }
411: if (y == CORD_EMPTY) return(x);
412: if (CORD_IS_STRING(y)) {
413: return(CORD_cat_char_star_optimized(x, y, strlen(y))); // optimized version is called
414: } else if (CORD_IS_STRING(x)) {
415: lenx = strlen(x);
416: depth = DEPTH(y) + 1;
417: } else {
418: register int depthy = DEPTH(y);
419:
420: lenx = LEN(x);
421: depth = DEPTH(x) + 1;
422: if (depthy >= depth) depth = depthy + 1;
423: }
424: result_len = lenx + LEN(y);
425: {
426: register struct Concatenation * result;
427:
428: result = GC_NEW(struct Concatenation);
429: if (result == 0) OUT_OF_MEMORY;
430: result->header = CONCAT_HDR;
431: result->depth = depth;
432: // printf("depth=%d\n", depth);
433: if (lenx <= MAX_LEFT_LEN) result->left_len = lenx;
434: result->len = result_len;
435: result->left = x;
436: result->right = y;
437: // PAF@design.ru bug fix:
438: if (depth >= MAX_DEPTH) {
439: return(CORD_balance((CORD)result));
440: } else {
441: return((CORD) result);
442: }
443: }
444: }
445: #endif
446:
1.2 paf 447: CORD CORD_from_fn(CORD_fn fn, void * client_data, size_t len)
448: {
449: if (len <= 0) return(0);
450: if (len <= SHORT_LIMIT) {
451: register char * result;
452: register size_t i;
453: char buf[SHORT_LIMIT+1];
454: register char c;
455:
456: for (i = 0; i < len; i++) {
457: c = (*fn)(i, client_data);
458: if (c == '\0') goto gen_case;
459: buf[i] = c;
460: }
461: buf[i] = '\0';
462: result = GC_MALLOC_ATOMIC(len+1);
463: if (result == 0) OUT_OF_MEMORY;
464: strcpy(result, buf);
465: result[len] = '\0';
466: return((CORD) result);
467: }
468: gen_case:
469: {
470: register struct Function * result;
471:
472: result = GC_NEW(struct Function);
473: if (result == 0) OUT_OF_MEMORY;
474: result->header = FN_HDR;
475: /* depth is already 0 */
476: result->len = len;
477: result->fn = fn;
478: result->client_data = client_data;
479: return((CORD) result);
480: }
481: }
482:
483: size_t CORD_len(CORD x)
484: {
485: if (x == 0) {
486: return(0);
487: } else {
488: return(GEN_LEN(x));
489: }
490: }
491:
492: struct substr_args {
493: CordRep * sa_cord;
494: size_t sa_index;
495: };
496:
497: char CORD_index_access_fn(size_t i, void * client_data)
498: {
499: register struct substr_args *descr = (struct substr_args *)client_data;
500:
501: return(((char *)(descr->sa_cord))[i + descr->sa_index]);
502: }
503:
504: char CORD_apply_access_fn(size_t i, void * client_data)
505: {
506: register struct substr_args *descr = (struct substr_args *)client_data;
507: register struct Function * fn_cord = &(descr->sa_cord->function);
508:
509: return((*(fn_cord->fn))(i + descr->sa_index, fn_cord->client_data));
510: }
511:
512: /* A version of CORD_substr that simply returns a function node, thus */
513: /* postponing its work. The fourth argument is a function that may */
514: /* be used for efficient access to the ith character. */
515: /* Assumes i >= 0 and i + n < length(x). */
516: CORD CORD_substr_closure(CORD x, size_t i, size_t n, CORD_fn f)
517: {
518: register struct substr_args * sa = GC_NEW(struct substr_args);
519: CORD result;
520:
521: if (sa == 0) OUT_OF_MEMORY;
522: sa->sa_cord = (CordRep *)x;
523: sa->sa_index = i;
524: result = CORD_from_fn(f, (void *)sa, n);
525: ((CordRep *)result) -> function.header = SUBSTR_HDR;
526: return (result);
527: }
528:
529: # define SUBSTR_LIMIT (10 * SHORT_LIMIT)
530: /* Substrings of function nodes and flat strings shorter than */
531: /* this are flat strings. Othewise we use a functional */
532: /* representation, which is significantly slower to access. */
533:
534: /* A version of CORD_substr that assumes i >= 0, n > 0, and i + n < length(x).*/
535: CORD CORD_substr_checked(CORD x, size_t i, size_t n)
536: {
537: if (CORD_IS_STRING(x)) {
538: if (n > SUBSTR_LIMIT) {
539: return(CORD_substr_closure(x, i, n, CORD_index_access_fn));
540: } else {
541: register char * result = GC_MALLOC_ATOMIC(n+1);
542:
543: if (result == 0) OUT_OF_MEMORY;
544: strncpy(result, x+i, n);
545: result[n] = '\0';
546: return(result);
547: }
548: } else if (IS_CONCATENATION(x)) {
549: register struct Concatenation * conc
550: = &(((CordRep *)x) -> concatenation);
551: register size_t left_len;
552: register size_t right_len;
553:
554: left_len = LEFT_LEN(conc);
555: right_len = conc -> len - left_len;
556: if (i >= left_len) {
557: if (n == right_len) return(conc -> right);
558: return(CORD_substr_checked(conc -> right, i - left_len, n));
559: } else if (i+n <= left_len) {
560: if (n == left_len) return(conc -> left);
561: return(CORD_substr_checked(conc -> left, i, n));
562: } else {
563: /* Need at least one character from each side. */
564: register CORD left_part;
565: register CORD right_part;
566: register size_t left_part_len = left_len - i;
567:
568: if (i == 0) {
569: left_part = conc -> left;
570: } else {
571: left_part = CORD_substr_checked(conc -> left, i, left_part_len);
572: }
573: if (i + n == right_len + left_len) {
574: right_part = conc -> right;
575: } else {
576: right_part = CORD_substr_checked(conc -> right, 0,
577: n - left_part_len);
578: }
579: return(CORD_cat(left_part, right_part));
580: }
581: } else /* function */ {
582: if (n > SUBSTR_LIMIT) {
583: if (IS_SUBSTR(x)) {
584: /* Avoid nesting substring nodes. */
585: register struct Function * f = &(((CordRep *)x) -> function);
586: register struct substr_args *descr =
587: (struct substr_args *)(f -> client_data);
588:
589: return(CORD_substr_closure((CORD)descr->sa_cord,
590: i + descr->sa_index,
591: n, f -> fn));
592: } else {
593: return(CORD_substr_closure(x, i, n, CORD_apply_access_fn));
594: }
595: } else {
596: char * result;
597: register struct Function * f = &(((CordRep *)x) -> function);
598: char buf[SUBSTR_LIMIT+1];
599: register char * p = buf;
600: register char c;
601: register int j;
602: register int lim = i + n;
603:
604: for (j = i; j < lim; j++) {
605: c = (*(f -> fn))(j, f -> client_data);
606: if (c == '\0') {
607: return(CORD_substr_closure(x, i, n, CORD_apply_access_fn));
608: }
609: *p++ = c;
610: }
611: *p = '\0';
612: result = GC_MALLOC_ATOMIC(n+1);
613: if (result == 0) OUT_OF_MEMORY;
614: strcpy(result, buf);
615: return(result);
616: }
617: }
618: }
619:
1.9 misha 620: CORD CORD_substr(CORD x, size_t i, size_t n, size_t len)
1.2 paf 621: {
1.9 misha 622: if(0 == len) len = CORD_len(x);
1.2 paf 623: if (i >= len || n <= 0) return(0);
624: /* n < 0 is impossible in a correct C implementation, but */
625: /* quite possible under SunOS 4.X. */
626: if (i + n > len) n = len - i;
627: # ifndef __STDC__
628: if (i < 0) ABORT("CORD_substr: second arg. negative");
629: /* Possible only if both client and C implementation are buggy. */
630: /* But empirically this happens frequently. */
631: # endif
632: return(CORD_substr_checked(x, i, n));
633: }
634:
635: /* See cord.h for definition. We assume i is in range. */
636: int CORD_iter5(CORD x, size_t i, CORD_iter_fn f1,
637: CORD_batched_iter_fn f2, void * client_data)
638: {
1.3 paf 639: int result;
1.2 paf 640: if (x == 0) return(0);
641: if (CORD_IS_STRING(x)) {
642: register const char* p = x+i;
643:
644: if (*p == '\0') ABORT("2nd arg to CORD_iter5 too big");
645: if (f2 != CORD_NO_FN) {
646: return((*f2)(p, client_data));
647: } else {
648: while (*p) {
1.3 paf 649: if (result=(*f1)(*p, client_data))
650: return result;
1.2 paf 651: p++;
652: }
653: return(0);
654: }
655: } else if (IS_CONCATENATION(x)) {
1.3 paf 656: register struct Concatenation * conc = &(((CordRep *)x) -> concatenation);
1.2 paf 657: if (i > 0) {
658: register size_t left_len = LEFT_LEN(conc);
659:
660: if (i >= left_len) {
661: return(CORD_iter5(conc -> right, i - left_len, f1, f2,
662: client_data));
663: }
664: }
1.3 paf 665: result=CORD_iter5(conc -> left, i, f1, f2, client_data);
666: if (result) return result;
1.2 paf 667: return(CORD_iter5(conc -> right, 0, f1, f2, client_data));
668: } else /* function */ {
669: register struct Function * f = &(((CordRep *)x) -> function);
670: register size_t j;
671: register size_t lim = f -> len;
672:
673: for (j = i; j < lim; j++) {
1.3 paf 674: if (result=(*f1)((*(f -> fn))(j, f -> client_data), client_data))
675: return result;
1.2 paf 676: }
677: return(0);
678: }
679: }
1.3 paf 680:
681: /* See cord.h for definition. We assume i is in range. */
682: int CORD_block_iter(CORD x, size_t i, CORD_block_iter_fn fb, void * client_data)
683: {
684: int result;
685: if (x == 0) return(0);
686: if (CORD_IS_STRING(x)) {
687: register const char* p = x+i;
688: const char *b=p;
689: char bc=*b;
690: int pc;
691:
692: if (bc == '\0') ABORT("2nd arg to CORD_iter5 too big");
693: do {
694: pc=*++p;
695: if(pc!=bc) {
696: if (result=fb(bc, p-b, client_data))
697: return result;
698: b=p; bc=pc;
699: }
700: } while (pc);
701: return(0);
702: } else if (IS_CONCATENATION(x)) {
703: register struct Concatenation * conc= &(((CordRep *)x) -> concatenation);
704: if (i > 0) {
705: register size_t left_len = LEFT_LEN(conc);
706:
707: if (i >= left_len) {
708: return(CORD_block_iter(conc -> right, i - left_len, fb, client_data));
709: }
710: }
711: result=CORD_block_iter(conc -> left, i, fb, client_data);
712: if (result) return result;
713: return(CORD_block_iter(conc -> right, 0, fb, client_data));
714: } else /* function */ {
715: register struct Function * f = &(((CordRep *)x) -> function);
716: register size_t lim = f -> len;
717:
718: if(f->fn == CORD_nul_func ) {
719: if (result=fb((char)(unsigned long)f -> client_data, f -> len-i, client_data)) return result;
720: } else if(f->fn == CORD_apply_access_fn) {
721: register struct substr_args *descr = (struct substr_args *)f->client_data;
722: register struct Function * fn_cord = &(descr->sa_cord->function);
723:
724: if(fn_cord->fn == CORD_nul_func ) {
725: if (result=fb((char)(unsigned long)fn_cord->client_data, f -> len-i, client_data))
726: return result;
727: } else
728: ABORT("CORD_block_iter:CORD_apply_access_fn:unknown_fn should not happen");
729: } else {
730: if(f->fn == CORD_index_access_fn)
731: ABORT("CORD_block_iter:CORD_index_access_fn should not happen");
732: ABORT("CORD_block_iter:unknown_fn should not happen");
733: }
734: }
735: return(0);
736: }
1.2 paf 737:
1.3 paf 738:
1.2 paf 739: #undef CORD_iter
740: int CORD_iter(CORD x, CORD_iter_fn f1, void * client_data)
741: {
742: return(CORD_iter5(x, 0, f1, CORD_NO_FN, client_data));
743: }
744:
745: int CORD_riter4(CORD x, size_t i, CORD_iter_fn f1, void * client_data)
746: {
747: if (x == 0) return(0);
748: if (CORD_IS_STRING(x)) {
749: register const char* p = x + i;
750: register char c;
751:
752: for(;;) {
753: c = *p;
754: if (c == '\0') ABORT("2nd arg to CORD_riter4 too big");
755: if ((*f1)(c, client_data)) return(1);
756: if (p == x) break;
757: p--;
758: }
759: return(0);
760: } else if (IS_CONCATENATION(x)) {
761: register struct Concatenation * conc
762: = &(((CordRep *)x) -> concatenation);
763: register CORD left_part = conc -> left;
764: register size_t left_len;
765:
766: left_len = LEFT_LEN(conc);
767: if (i >= left_len) {
768: if (CORD_riter4(conc -> right, i - left_len, f1, client_data)) {
769: return(1);
770: }
771: return(CORD_riter4(left_part, left_len - 1, f1, client_data));
772: } else {
773: return(CORD_riter4(left_part, i, f1, client_data));
774: }
775: } else /* function */ {
776: register struct Function * f = &(((CordRep *)x) -> function);
777: register size_t j;
778:
779: for (j = i; ; j--) {
780: if ((*f1)((*(f -> fn))(j, f -> client_data), client_data)) {
781: return(1);
782: }
783: if (j == 0) return(0);
784: }
785: }
786: }
787:
788: int CORD_riter(CORD x, CORD_iter_fn f1, void * client_data)
789: {
790: return(CORD_riter4(x, CORD_len(x) - 1, f1, client_data));
791: }
792:
793: /*
794: * The following functions are concerned with balancing cords.
795: * Strategy:
796: * Scan the cord from left to right, keeping the cord scanned so far
797: * as a forest of balanced trees of exponentialy decreasing length.
798: * When a new subtree needs to be added to the forest, we concatenate all
799: * shorter ones to the new tree in the appropriate order, and then insert
800: * the result into the forest.
801: * Crucial invariants:
802: * 1. The concatenation of the forest (in decreasing order) with the
803: * unscanned part of the rope is equal to the rope being balanced.
804: * 2. All trees in the forest are balanced.
805: * 3. forest[i] has depth at most i.
806: */
807:
808: typedef struct {
809: CORD c;
810: size_t len; /* Actual length of c */
811: } ForestElement;
812:
813: static size_t min_len [ MAX_DEPTH ];
814:
815: static int min_len_init = 0;
816:
817: int CORD_max_len;
818:
819: typedef ForestElement Forest [ MAX_DEPTH ];
820: /* forest[i].len >= fib(i+1) */
821: /* The string is the concatenation */
822: /* of the forest in order of DECREASING */
823: /* indices. */
824:
825: void CORD_init_min_len()
826: {
827: register int i;
828: register size_t last, previous, current;
829:
830: min_len[0] = previous = 1;
831: min_len[1] = last = 2;
832: for (i = 2; i < MAX_DEPTH; i++) {
833: current = last + previous;
834: if (current < last) /* overflow */ current = last;
835: min_len[i] = current;
836: previous = last;
837: last = current;
838: }
839: CORD_max_len = last - 1;
840: min_len_init = 1;
841: }
842:
843:
844: void CORD_init_forest(ForestElement * forest, size_t max_len)
845: {
846: register int i;
847:
848: for (i = 0; i < MAX_DEPTH; i++) {
849: forest[i].c = 0;
850: if (min_len[i] > max_len) return;
851: }
852: ABORT("Cord too long");
853: }
854:
855: /* Add a leaf to the appropriate level in the forest, cleaning */
856: /* out lower levels as necessary. */
857: /* Also works if x is a balanced tree of concatenations; however */
858: /* in this case an extra concatenation node may be inserted above x; */
859: /* This node should not be counted in the statement of the invariants. */
860: void CORD_add_forest(ForestElement * forest, CORD x, size_t len)
861: {
862: register int i = 0;
863: register CORD sum = CORD_EMPTY;
864: register size_t sum_len = 0;
865:
866: while (len > min_len[i + 1]) {
867: if (forest[i].c != 0) {
868: sum = CORD_cat(forest[i].c, sum);
869: sum_len += forest[i].len;
870: forest[i].c = 0;
871: }
872: i++;
873: }
874: /* Sum has depth at most 1 greter than what would be required */
875: /* for balance. */
876: sum = CORD_cat(sum, x);
877: sum_len += len;
878: /* If x was a leaf, then sum is now balanced. To see this */
879: /* consider the two cases in which forest[i-1] either is or is */
880: /* not empty. */
881: while (sum_len >= min_len[i]) {
882: if (forest[i].c != 0) {
883: sum = CORD_cat(forest[i].c, sum);
884: sum_len += forest[i].len;
885: /* This is again balanced, since sum was balanced, and has */
886: /* allowable depth that differs from i by at most 1. */
887: forest[i].c = 0;
888: }
889: i++;
890: }
891: i--;
892: forest[i].c = sum;
893: forest[i].len = sum_len;
894: }
895:
896: CORD CORD_concat_forest(ForestElement * forest, size_t expected_len)
897: {
898: register int i = 0;
899: CORD sum = 0;
900: size_t sum_len = 0;
901:
902: while (sum_len != expected_len) {
903: if (forest[i].c != 0) {
904: sum = CORD_cat(forest[i].c, sum);
905: sum_len += forest[i].len;
906: }
907: i++;
908: }
909: return(sum);
910: }
911:
912: /* Insert the frontier of x into forest. Balanced subtrees are */
913: /* treated as leaves. This potentially adds one to the depth */
914: /* of the final tree. */
915: void CORD_balance_insert(CORD x, size_t len, ForestElement * forest)
916: {
917: register int depth;
918:
919: if (CORD_IS_STRING(x)) {
920: CORD_add_forest(forest, x, len);
921: } else if (IS_CONCATENATION(x)
922: && ((depth = DEPTH(x)) >= MAX_DEPTH
923: || len < min_len[depth])) {
924: register struct Concatenation * conc
925: = &(((CordRep *)x) -> concatenation);
926: size_t left_len = LEFT_LEN(conc);
927:
928: CORD_balance_insert(conc -> left, left_len, forest);
929: CORD_balance_insert(conc -> right, len - left_len, forest);
930: } else /* function or balanced */ {
931: CORD_add_forest(forest, x, len);
932: }
933: }
934:
935:
936: CORD CORD_balance(CORD x)
937: {
938: Forest forest;
939: register size_t len;
940:
941: if (x == 0) return(0);
942: if (CORD_IS_STRING(x)) return(x);
943: if (!min_len_init) CORD_init_min_len();
944: len = LEN(x);
945: CORD_init_forest(forest, len);
946: CORD_balance_insert(x, len, forest);
947: return(CORD_concat_forest(forest, len));
948: }
949:
950:
951: /* Position primitives */
952:
953: /* Private routines to deal with the hard cases only: */
954:
955: /* P contains a prefix of the path to cur_pos. Extend it to a full */
956: /* path and set up leaf info. */
957: /* Return 0 if past the end of cord, 1 o.w. */
958: void CORD__extend_path(register CORD_pos p)
959: {
960: register struct CORD_pe * current_pe = &(p[0].path[p[0].path_len]);
961: register CORD top = current_pe -> pe_cord;
962: register size_t pos = p[0].cur_pos;
963: register size_t top_pos = current_pe -> pe_start_pos;
964: register size_t top_len = GEN_LEN(top);
965:
966: /* Fill in the rest of the path. */
967: while(!CORD_IS_STRING(top) && IS_CONCATENATION(top)) {
968: register struct Concatenation * conc =
969: &(((CordRep *)top) -> concatenation);
970: register size_t left_len;
971:
972: left_len = LEFT_LEN(conc);
973: current_pe++;
974: if (pos >= top_pos + left_len) {
975: current_pe -> pe_cord = top = conc -> right;
976: current_pe -> pe_start_pos = top_pos = top_pos + left_len;
977: top_len -= left_len;
978: } else {
979: current_pe -> pe_cord = top = conc -> left;
980: current_pe -> pe_start_pos = top_pos;
981: top_len = left_len;
982: }
983: p[0].path_len++;
984: }
985: /* Fill in leaf description for fast access. */
986: if (CORD_IS_STRING(top)) {
987: p[0].cur_leaf = top;
988: p[0].cur_start = top_pos;
989: p[0].cur_end = top_pos + top_len;
990: } else {
991: p[0].cur_end = 0;
992: }
993: if (pos >= top_pos + top_len) p[0].path_len = CORD_POS_INVALID;
994: }
995:
996: char CORD__pos_fetch(register CORD_pos p)
997: {
998: /* Leaf is a function node */
999: struct CORD_pe * pe = &((p)[0].path[(p)[0].path_len]);
1000: CORD leaf = pe -> pe_cord;
1001: register struct Function * f = &(((CordRep *)leaf) -> function);
1002:
1003: if (!IS_FUNCTION(leaf)) ABORT("CORD_pos_fetch: bad leaf");
1004: return ((*(f -> fn))(p[0].cur_pos - pe -> pe_start_pos, f -> client_data));
1005: }
1006:
1007: void CORD__next(register CORD_pos p)
1008: {
1009: register size_t cur_pos = p[0].cur_pos + 1;
1010: register struct CORD_pe * current_pe = &((p)[0].path[(p)[0].path_len]);
1011: register CORD leaf = current_pe -> pe_cord;
1012:
1013: /* Leaf is not a string or we're at end of leaf */
1014: p[0].cur_pos = cur_pos;
1015: if (!CORD_IS_STRING(leaf)) {
1016: /* Function leaf */
1017: register struct Function * f = &(((CordRep *)leaf) -> function);
1018: register size_t start_pos = current_pe -> pe_start_pos;
1019: register size_t end_pos = start_pos + f -> len;
1020:
1021: if (cur_pos < end_pos) {
1022: /* Fill cache and return. */
1023: register size_t i;
1024: register size_t limit = cur_pos + FUNCTION_BUF_SZ;
1025: register CORD_fn fn = f -> fn;
1026: register void * client_data = f -> client_data;
1027:
1028: if (limit > end_pos) {
1029: limit = end_pos;
1030: }
1031: for (i = cur_pos; i < limit; i++) {
1032: p[0].function_buf[i - cur_pos] =
1033: (*fn)(i - start_pos, client_data);
1034: }
1035: p[0].cur_start = cur_pos;
1036: p[0].cur_leaf = p[0].function_buf;
1037: p[0].cur_end = limit;
1038: return;
1039: }
1040: }
1041: /* End of leaf */
1042: /* Pop the stack until we find two concatenation nodes with the */
1043: /* same start position: this implies we were in left part. */
1044: {
1045: while (p[0].path_len > 0
1046: && current_pe[0].pe_start_pos != current_pe[-1].pe_start_pos) {
1047: p[0].path_len--;
1048: current_pe--;
1049: }
1050: if (p[0].path_len == 0) {
1051: p[0].path_len = CORD_POS_INVALID;
1052: return;
1053: }
1054: }
1055: p[0].path_len--;
1056: CORD__extend_path(p);
1057: }
1058:
1059: void CORD__prev(register CORD_pos p)
1060: {
1061: register struct CORD_pe * pe = &(p[0].path[p[0].path_len]);
1062:
1063: if (p[0].cur_pos == 0) {
1064: p[0].path_len = CORD_POS_INVALID;
1065: return;
1066: }
1067: p[0].cur_pos--;
1068: if (p[0].cur_pos >= pe -> pe_start_pos) return;
1069:
1070: /* Beginning of leaf */
1071:
1072: /* Pop the stack until we find two concatenation nodes with the */
1073: /* different start position: this implies we were in right part. */
1074: {
1075: register struct CORD_pe * current_pe = &((p)[0].path[(p)[0].path_len]);
1076:
1077: while (p[0].path_len > 0
1078: && current_pe[0].pe_start_pos == current_pe[-1].pe_start_pos) {
1079: p[0].path_len--;
1080: current_pe--;
1081: }
1082: }
1083: p[0].path_len--;
1084: CORD__extend_path(p);
1085: }
1086:
1087: #undef CORD_pos_fetch
1088: #undef CORD_next
1089: #undef CORD_prev
1090: #undef CORD_pos_to_index
1091: #undef CORD_pos_to_cord
1092: #undef CORD_pos_valid
1093:
1094: char CORD_pos_fetch(register CORD_pos p)
1095: {
1096: if (p[0].cur_start <= p[0].cur_pos && p[0].cur_pos < p[0].cur_end) {
1097: return(p[0].cur_leaf[p[0].cur_pos - p[0].cur_start]);
1098: } else {
1099: return(CORD__pos_fetch(p));
1100: }
1101: }
1102:
1103: void CORD_next(CORD_pos p)
1104: {
1105: if (p[0].cur_pos < p[0].cur_end - 1) {
1106: p[0].cur_pos++;
1107: } else {
1108: CORD__next(p);
1109: }
1110: }
1111:
1112: void CORD_prev(CORD_pos p)
1113: {
1114: if (p[0].cur_end != 0 && p[0].cur_pos > p[0].cur_start) {
1115: p[0].cur_pos--;
1116: } else {
1117: CORD__prev(p);
1118: }
1119: }
1120:
1121: size_t CORD_pos_to_index(CORD_pos p)
1122: {
1123: return(p[0].cur_pos);
1124: }
1125:
1126: CORD CORD_pos_to_cord(CORD_pos p)
1127: {
1128: return(p[0].path[0].pe_cord);
1129: }
1130:
1131: int CORD_pos_valid(CORD_pos p)
1132: {
1133: return(p[0].path_len != CORD_POS_INVALID);
1134: }
1135:
1136: void CORD_set_pos(CORD_pos p, CORD x, size_t i)
1137: {
1138: if (x == CORD_EMPTY) {
1139: p[0].path_len = CORD_POS_INVALID;
1140: return;
1141: }
1142: p[0].path[0].pe_cord = x;
1143: p[0].path[0].pe_start_pos = 0;
1144: p[0].path_len = 0;
1145: p[0].cur_pos = i;
1146: CORD__extend_path(p);
1147: }
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