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