Annotation of parser3/src/lib/cord/cordbscs.c, revision 1.1.2.3
1.1.2.1 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:
1.1.2.2 paf 147: CORD CORD_cat_char_star(CORD x, const char* y, size_t leny)
1.1.2.1 paf 148: {
149: register size_t result_len;
150: register size_t lenx;
151: register int depth;
152:
1.1.2.3 ! paf 153: if (!y || leny == 0) return(x); // PAF: bug fix, moved this check upper than X == CORD_EMPTY, added !y check
! 154: // paf: bug fix
! 155: result_len=strlen(y); // reusing var
! 156: if(result_len != leny) { // PAF: author's idea was that that would be ALWAYS, which is not true
! 157: // cloning piece
! 158: register char * new_y = GC_MALLOC_ATOMIC(leny+1);
! 159:
! 160: if (new_y == 0) OUT_OF_MEMORY;
! 161: memcpy(new_y, y, leny);
! 162: new_y[leny] = '\0';
! 163: y=new_y;
! 164: }
! 165:
1.1.2.1 paf 166: if (x == CORD_EMPTY) return(y);
167: if (CORD_IS_STRING(x)) {
168: lenx = strlen(x);
169: result_len = lenx + leny;
170: if (result_len <= SHORT_LIMIT) {
171: register char * result = GC_MALLOC_ATOMIC(result_len+1);
172:
173: if (result == 0) OUT_OF_MEMORY;
174: memcpy(result, x, lenx);
175: memcpy(result + lenx, y, leny);
176: result[result_len] = '\0';
177: return((CORD) result);
178: } else {
179: depth = 1;
180: }
181: } else {
182: register CORD right;
183: register CORD left;
184: register char * new_right;
185: register size_t right_len;
186:
187: lenx = LEN(x);
188:
189: if (leny <= SHORT_LIMIT/2
190: && IS_CONCATENATION(x)
191: && CORD_IS_STRING(right = ((CordRep *)x) -> concatenation.right)) {
192: /* Merge y into right part of x. */
193: if (!CORD_IS_STRING(left = ((CordRep *)x) -> concatenation.left)) {
194: right_len = lenx - LEN(left);
195: } else if (((CordRep *)x) -> concatenation.left_len != 0) {
196: right_len = lenx - ((CordRep *)x) -> concatenation.left_len;
197: } else {
198: right_len = strlen(right);
199: }
200: result_len = right_len + leny; /* length of new_right */
201: if (result_len <= SHORT_LIMIT) {
202: new_right = GC_MALLOC_ATOMIC(result_len + 1);
203: memcpy(new_right, right, right_len);
204: memcpy(new_right + right_len, y, leny);
205: new_right[result_len] = '\0';
206: y = new_right;
207: leny = result_len;
208: x = left;
209: lenx -= right_len;
210: /* Now fall through to concatenate the two pieces: */
211: }
212: if (CORD_IS_STRING(x)) {
213: depth = 1;
214: } else {
215: depth = DEPTH(x) + 1;
216: }
217: } else {
218: depth = DEPTH(x) + 1;
219: }
220: result_len = lenx + leny;
221: }
222: {
223: /* The general case; lenx, result_len is known: */
224: register struct Concatenation * result;
225:
226: result = GC_NEW(struct Concatenation);
227: if (result == 0) OUT_OF_MEMORY;
228: result->header = CONCAT_HDR;
229: result->depth = depth;
230: if (lenx <= MAX_LEFT_LEN) result->left_len = lenx;
231: result->len = result_len;
232: result->left = x;
233: result->right = y;
234: if (depth > MAX_DEPTH) {
235: return(CORD_balance((CORD)result));
236: } else {
237: return((CORD) result);
238: }
239: }
240: }
241:
242:
243: CORD CORD_cat(CORD x, CORD y)
244: {
245: register size_t result_len;
246: register int depth;
247: register size_t lenx;
248:
249: if (x == CORD_EMPTY) return(y);
250: if (y == CORD_EMPTY) return(x);
251: if (CORD_IS_STRING(y)) {
252: return(CORD_cat_char_star(x, y, strlen(y)));
253: } else if (CORD_IS_STRING(x)) {
254: lenx = strlen(x);
255: depth = DEPTH(y) + 1;
256: } else {
257: register int depthy = DEPTH(y);
258:
259: lenx = LEN(x);
260: depth = DEPTH(x) + 1;
261: if (depthy >= depth) depth = depthy + 1;
262: }
263: result_len = lenx + LEN(y);
264: {
265: register struct Concatenation * result;
266:
267: result = GC_NEW(struct Concatenation);
268: if (result == 0) OUT_OF_MEMORY;
269: result->header = CONCAT_HDR;
270: result->depth = depth;
271: if (lenx <= MAX_LEFT_LEN) result->left_len = lenx;
272: result->len = result_len;
273: result->left = x;
274: result->right = y;
275: return((CORD) result);
276: }
277: }
278:
279:
280:
281: CORD CORD_from_fn(CORD_fn fn, void * client_data, size_t len)
282: {
283: if (len <= 0) return(0);
284: if (len <= SHORT_LIMIT) {
285: register char * result;
286: register size_t i;
287: char buf[SHORT_LIMIT+1];
288: register char c;
289:
290: for (i = 0; i < len; i++) {
291: c = (*fn)(i, client_data);
292: if (c == '\0') goto gen_case;
293: buf[i] = c;
294: }
295: buf[i] = '\0';
296: result = GC_MALLOC_ATOMIC(len+1);
297: if (result == 0) OUT_OF_MEMORY;
298: strcpy(result, buf);
299: result[len] = '\0';
300: return((CORD) result);
301: }
302: gen_case:
303: {
304: register struct Function * result;
305:
306: result = GC_NEW(struct Function);
307: if (result == 0) OUT_OF_MEMORY;
308: result->header = FN_HDR;
309: /* depth is already 0 */
310: result->len = len;
311: result->fn = fn;
312: result->client_data = client_data;
313: return((CORD) result);
314: }
315: }
316:
317: size_t CORD_len(CORD x)
318: {
319: if (x == 0) {
320: return(0);
321: } else {
322: return(GEN_LEN(x));
323: }
324: }
325:
326: struct substr_args {
327: CordRep * sa_cord;
328: size_t sa_index;
329: };
330:
331: char CORD_index_access_fn(size_t i, void * client_data)
332: {
333: register struct substr_args *descr = (struct substr_args *)client_data;
334:
335: return(((char *)(descr->sa_cord))[i + descr->sa_index]);
336: }
337:
338: char CORD_apply_access_fn(size_t i, void * client_data)
339: {
340: register struct substr_args *descr = (struct substr_args *)client_data;
341: register struct Function * fn_cord = &(descr->sa_cord->function);
342:
343: return((*(fn_cord->fn))(i + descr->sa_index, fn_cord->client_data));
344: }
345:
346: /* A version of CORD_substr that simply returns a function node, thus */
347: /* postponing its work. The fourth argument is a function that may */
348: /* be used for efficient access to the ith character. */
349: /* Assumes i >= 0 and i + n < length(x). */
350: CORD CORD_substr_closure(CORD x, size_t i, size_t n, CORD_fn f)
351: {
352: register struct substr_args * sa = GC_NEW(struct substr_args);
353: CORD result;
354:
355: if (sa == 0) OUT_OF_MEMORY;
356: sa->sa_cord = (CordRep *)x;
357: sa->sa_index = i;
358: result = CORD_from_fn(f, (void *)sa, n);
359: ((CordRep *)result) -> function.header = SUBSTR_HDR;
360: return (result);
361: }
362:
363: # define SUBSTR_LIMIT (10 * SHORT_LIMIT)
364: /* Substrings of function nodes and flat strings shorter than */
365: /* this are flat strings. Othewise we use a functional */
366: /* representation, which is significantly slower to access. */
367:
368: /* A version of CORD_substr that assumes i >= 0, n > 0, and i + n < length(x).*/
369: CORD CORD_substr_checked(CORD x, size_t i, size_t n)
370: {
371: if (CORD_IS_STRING(x)) {
372: if (n > SUBSTR_LIMIT) {
373: return(CORD_substr_closure(x, i, n, CORD_index_access_fn));
374: } else {
375: register char * result = GC_MALLOC_ATOMIC(n+1);
376:
377: if (result == 0) OUT_OF_MEMORY;
378: strncpy(result, x+i, n);
379: result[n] = '\0';
380: return(result);
381: }
382: } else if (IS_CONCATENATION(x)) {
383: register struct Concatenation * conc
384: = &(((CordRep *)x) -> concatenation);
385: register size_t left_len;
386: register size_t right_len;
387:
388: left_len = LEFT_LEN(conc);
389: right_len = conc -> len - left_len;
390: if (i >= left_len) {
391: if (n == right_len) return(conc -> right);
392: return(CORD_substr_checked(conc -> right, i - left_len, n));
393: } else if (i+n <= left_len) {
394: if (n == left_len) return(conc -> left);
395: return(CORD_substr_checked(conc -> left, i, n));
396: } else {
397: /* Need at least one character from each side. */
398: register CORD left_part;
399: register CORD right_part;
400: register size_t left_part_len = left_len - i;
401:
402: if (i == 0) {
403: left_part = conc -> left;
404: } else {
405: left_part = CORD_substr_checked(conc -> left, i, left_part_len);
406: }
407: if (i + n == right_len + left_len) {
408: right_part = conc -> right;
409: } else {
410: right_part = CORD_substr_checked(conc -> right, 0,
411: n - left_part_len);
412: }
413: return(CORD_cat(left_part, right_part));
414: }
415: } else /* function */ {
416: if (n > SUBSTR_LIMIT) {
417: if (IS_SUBSTR(x)) {
418: /* Avoid nesting substring nodes. */
419: register struct Function * f = &(((CordRep *)x) -> function);
420: register struct substr_args *descr =
421: (struct substr_args *)(f -> client_data);
422:
423: return(CORD_substr_closure((CORD)descr->sa_cord,
424: i + descr->sa_index,
425: n, f -> fn));
426: } else {
427: return(CORD_substr_closure(x, i, n, CORD_apply_access_fn));
428: }
429: } else {
430: char * result;
431: register struct Function * f = &(((CordRep *)x) -> function);
432: char buf[SUBSTR_LIMIT+1];
433: register char * p = buf;
434: register char c;
435: register int j;
436: register int lim = i + n;
437:
438: for (j = i; j < lim; j++) {
439: c = (*(f -> fn))(j, f -> client_data);
440: if (c == '\0') {
441: return(CORD_substr_closure(x, i, n, CORD_apply_access_fn));
442: }
443: *p++ = c;
444: }
445: *p = '\0';
446: result = GC_MALLOC_ATOMIC(n+1);
447: if (result == 0) OUT_OF_MEMORY;
448: strcpy(result, buf);
449: return(result);
450: }
451: }
452: }
453:
454: CORD CORD_substr(CORD x, size_t i, size_t n)
455: {
456: register size_t len = CORD_len(x);
457:
458: if (i >= len || n <= 0) return(0);
459: /* n < 0 is impossible in a correct C implementation, but */
460: /* quite possible under SunOS 4.X. */
461: if (i + n > len) n = len - i;
462: # ifndef __STDC__
463: if (i < 0) ABORT("CORD_substr: second arg. negative");
464: /* Possible only if both client and C implementation are buggy. */
465: /* But empirically this happens frequently. */
466: # endif
467: return(CORD_substr_checked(x, i, n));
468: }
469:
470: /* See cord.h for definition. We assume i is in range. */
471: int CORD_iter5(CORD x, size_t i, CORD_iter_fn f1,
472: CORD_batched_iter_fn f2, void * client_data)
473: {
474: if (x == 0) return(0);
475: if (CORD_IS_STRING(x)) {
1.1.2.2 paf 476: register const char* p = x+i;
1.1.2.1 paf 477:
478: if (*p == '\0') ABORT("2nd arg to CORD_iter5 too big");
479: if (f2 != CORD_NO_FN) {
480: return((*f2)(p, client_data));
481: } else {
482: while (*p) {
483: if ((*f1)(*p, client_data)) return(1);
484: p++;
485: }
486: return(0);
487: }
488: } else if (IS_CONCATENATION(x)) {
489: register struct Concatenation * conc
490: = &(((CordRep *)x) -> concatenation);
491:
492:
493: if (i > 0) {
494: register size_t left_len = LEFT_LEN(conc);
495:
496: if (i >= left_len) {
497: return(CORD_iter5(conc -> right, i - left_len, f1, f2,
498: client_data));
499: }
500: }
501: if (CORD_iter5(conc -> left, i, f1, f2, client_data)) {
502: return(1);
503: }
504: return(CORD_iter5(conc -> right, 0, f1, f2, client_data));
505: } else /* function */ {
506: register struct Function * f = &(((CordRep *)x) -> function);
507: register size_t j;
508: register size_t lim = f -> len;
509:
510: for (j = i; j < lim; j++) {
511: if ((*f1)((*(f -> fn))(j, f -> client_data), client_data)) {
512: return(1);
513: }
514: }
515: return(0);
516: }
517: }
518:
519: #undef CORD_iter
520: int CORD_iter(CORD x, CORD_iter_fn f1, void * client_data)
521: {
522: return(CORD_iter5(x, 0, f1, CORD_NO_FN, client_data));
523: }
524:
525: int CORD_riter4(CORD x, size_t i, CORD_iter_fn f1, void * client_data)
526: {
527: if (x == 0) return(0);
528: if (CORD_IS_STRING(x)) {
1.1.2.2 paf 529: register const char* p = x + i;
1.1.2.1 paf 530: register char c;
531:
532: for(;;) {
533: c = *p;
534: if (c == '\0') ABORT("2nd arg to CORD_riter4 too big");
535: if ((*f1)(c, client_data)) return(1);
536: if (p == x) break;
537: p--;
538: }
539: return(0);
540: } else if (IS_CONCATENATION(x)) {
541: register struct Concatenation * conc
542: = &(((CordRep *)x) -> concatenation);
543: register CORD left_part = conc -> left;
544: register size_t left_len;
545:
546: left_len = LEFT_LEN(conc);
547: if (i >= left_len) {
548: if (CORD_riter4(conc -> right, i - left_len, f1, client_data)) {
549: return(1);
550: }
551: return(CORD_riter4(left_part, left_len - 1, f1, client_data));
552: } else {
553: return(CORD_riter4(left_part, i, f1, client_data));
554: }
555: } else /* function */ {
556: register struct Function * f = &(((CordRep *)x) -> function);
557: register size_t j;
558:
559: for (j = i; ; j--) {
560: if ((*f1)((*(f -> fn))(j, f -> client_data), client_data)) {
561: return(1);
562: }
563: if (j == 0) return(0);
564: }
565: }
566: }
567:
568: int CORD_riter(CORD x, CORD_iter_fn f1, void * client_data)
569: {
570: return(CORD_riter4(x, CORD_len(x) - 1, f1, client_data));
571: }
572:
573: /*
574: * The following functions are concerned with balancing cords.
575: * Strategy:
576: * Scan the cord from left to right, keeping the cord scanned so far
577: * as a forest of balanced trees of exponentialy decreasing length.
578: * When a new subtree needs to be added to the forest, we concatenate all
579: * shorter ones to the new tree in the appropriate order, and then insert
580: * the result into the forest.
581: * Crucial invariants:
582: * 1. The concatenation of the forest (in decreasing order) with the
583: * unscanned part of the rope is equal to the rope being balanced.
584: * 2. All trees in the forest are balanced.
585: * 3. forest[i] has depth at most i.
586: */
587:
588: typedef struct {
589: CORD c;
590: size_t len; /* Actual length of c */
591: } ForestElement;
592:
593: static size_t min_len [ MAX_DEPTH ];
594:
595: static int min_len_init = 0;
596:
597: int CORD_max_len;
598:
599: typedef ForestElement Forest [ MAX_DEPTH ];
600: /* forest[i].len >= fib(i+1) */
601: /* The string is the concatenation */
602: /* of the forest in order of DECREASING */
603: /* indices. */
604:
605: void CORD_init_min_len()
606: {
607: register int i;
608: register size_t last, previous, current;
609:
610: min_len[0] = previous = 1;
611: min_len[1] = last = 2;
612: for (i = 2; i < MAX_DEPTH; i++) {
613: current = last + previous;
614: if (current < last) /* overflow */ current = last;
615: min_len[i] = current;
616: previous = last;
617: last = current;
618: }
619: CORD_max_len = last - 1;
620: min_len_init = 1;
621: }
622:
623:
624: void CORD_init_forest(ForestElement * forest, size_t max_len)
625: {
626: register int i;
627:
628: for (i = 0; i < MAX_DEPTH; i++) {
629: forest[i].c = 0;
630: if (min_len[i] > max_len) return;
631: }
632: ABORT("Cord too long");
633: }
634:
635: /* Add a leaf to the appropriate level in the forest, cleaning */
636: /* out lower levels as necessary. */
637: /* Also works if x is a balanced tree of concatenations; however */
638: /* in this case an extra concatenation node may be inserted above x; */
639: /* This node should not be counted in the statement of the invariants. */
640: void CORD_add_forest(ForestElement * forest, CORD x, size_t len)
641: {
642: register int i = 0;
643: register CORD sum = CORD_EMPTY;
644: register size_t sum_len = 0;
645:
646: while (len > min_len[i + 1]) {
647: if (forest[i].c != 0) {
648: sum = CORD_cat(forest[i].c, sum);
649: sum_len += forest[i].len;
650: forest[i].c = 0;
651: }
652: i++;
653: }
654: /* Sum has depth at most 1 greter than what would be required */
655: /* for balance. */
656: sum = CORD_cat(sum, x);
657: sum_len += len;
658: /* If x was a leaf, then sum is now balanced. To see this */
659: /* consider the two cases in which forest[i-1] either is or is */
660: /* not empty. */
661: while (sum_len >= min_len[i]) {
662: if (forest[i].c != 0) {
663: sum = CORD_cat(forest[i].c, sum);
664: sum_len += forest[i].len;
665: /* This is again balanced, since sum was balanced, and has */
666: /* allowable depth that differs from i by at most 1. */
667: forest[i].c = 0;
668: }
669: i++;
670: }
671: i--;
672: forest[i].c = sum;
673: forest[i].len = sum_len;
674: }
675:
676: CORD CORD_concat_forest(ForestElement * forest, size_t expected_len)
677: {
678: register int i = 0;
679: CORD sum = 0;
680: size_t sum_len = 0;
681:
682: while (sum_len != expected_len) {
683: if (forest[i].c != 0) {
684: sum = CORD_cat(forest[i].c, sum);
685: sum_len += forest[i].len;
686: }
687: i++;
688: }
689: return(sum);
690: }
691:
692: /* Insert the frontier of x into forest. Balanced subtrees are */
693: /* treated as leaves. This potentially adds one to the depth */
694: /* of the final tree. */
695: void CORD_balance_insert(CORD x, size_t len, ForestElement * forest)
696: {
697: register int depth;
698:
699: if (CORD_IS_STRING(x)) {
700: CORD_add_forest(forest, x, len);
701: } else if (IS_CONCATENATION(x)
702: && ((depth = DEPTH(x)) >= MAX_DEPTH
703: || len < min_len[depth])) {
704: register struct Concatenation * conc
705: = &(((CordRep *)x) -> concatenation);
706: size_t left_len = LEFT_LEN(conc);
707:
708: CORD_balance_insert(conc -> left, left_len, forest);
709: CORD_balance_insert(conc -> right, len - left_len, forest);
710: } else /* function or balanced */ {
711: CORD_add_forest(forest, x, len);
712: }
713: }
714:
715:
716: CORD CORD_balance(CORD x)
717: {
718: Forest forest;
719: register size_t len;
720:
721: if (x == 0) return(0);
722: if (CORD_IS_STRING(x)) return(x);
723: if (!min_len_init) CORD_init_min_len();
724: len = LEN(x);
725: CORD_init_forest(forest, len);
726: CORD_balance_insert(x, len, forest);
727: return(CORD_concat_forest(forest, len));
728: }
729:
730:
731: /* Position primitives */
732:
733: /* Private routines to deal with the hard cases only: */
734:
735: /* P contains a prefix of the path to cur_pos. Extend it to a full */
736: /* path and set up leaf info. */
737: /* Return 0 if past the end of cord, 1 o.w. */
738: void CORD__extend_path(register CORD_pos p)
739: {
740: register struct CORD_pe * current_pe = &(p[0].path[p[0].path_len]);
741: register CORD top = current_pe -> pe_cord;
742: register size_t pos = p[0].cur_pos;
743: register size_t top_pos = current_pe -> pe_start_pos;
744: register size_t top_len = GEN_LEN(top);
745:
746: /* Fill in the rest of the path. */
747: while(!CORD_IS_STRING(top) && IS_CONCATENATION(top)) {
748: register struct Concatenation * conc =
749: &(((CordRep *)top) -> concatenation);
750: register size_t left_len;
751:
752: left_len = LEFT_LEN(conc);
753: current_pe++;
754: if (pos >= top_pos + left_len) {
755: current_pe -> pe_cord = top = conc -> right;
756: current_pe -> pe_start_pos = top_pos = top_pos + left_len;
757: top_len -= left_len;
758: } else {
759: current_pe -> pe_cord = top = conc -> left;
760: current_pe -> pe_start_pos = top_pos;
761: top_len = left_len;
762: }
763: p[0].path_len++;
764: }
765: /* Fill in leaf description for fast access. */
766: if (CORD_IS_STRING(top)) {
767: p[0].cur_leaf = top;
768: p[0].cur_start = top_pos;
769: p[0].cur_end = top_pos + top_len;
770: } else {
771: p[0].cur_end = 0;
772: }
773: if (pos >= top_pos + top_len) p[0].path_len = CORD_POS_INVALID;
774: }
775:
776: char CORD__pos_fetch(register CORD_pos p)
777: {
778: /* Leaf is a function node */
779: struct CORD_pe * pe = &((p)[0].path[(p)[0].path_len]);
780: CORD leaf = pe -> pe_cord;
781: register struct Function * f = &(((CordRep *)leaf) -> function);
782:
783: if (!IS_FUNCTION(leaf)) ABORT("CORD_pos_fetch: bad leaf");
784: return ((*(f -> fn))(p[0].cur_pos - pe -> pe_start_pos, f -> client_data));
785: }
786:
787: void CORD__next(register CORD_pos p)
788: {
789: register size_t cur_pos = p[0].cur_pos + 1;
790: register struct CORD_pe * current_pe = &((p)[0].path[(p)[0].path_len]);
791: register CORD leaf = current_pe -> pe_cord;
792:
793: /* Leaf is not a string or we're at end of leaf */
794: p[0].cur_pos = cur_pos;
795: if (!CORD_IS_STRING(leaf)) {
796: /* Function leaf */
797: register struct Function * f = &(((CordRep *)leaf) -> function);
798: register size_t start_pos = current_pe -> pe_start_pos;
799: register size_t end_pos = start_pos + f -> len;
800:
801: if (cur_pos < end_pos) {
802: /* Fill cache and return. */
803: register size_t i;
804: register size_t limit = cur_pos + FUNCTION_BUF_SZ;
805: register CORD_fn fn = f -> fn;
806: register void * client_data = f -> client_data;
807:
808: if (limit > end_pos) {
809: limit = end_pos;
810: }
811: for (i = cur_pos; i < limit; i++) {
812: p[0].function_buf[i - cur_pos] =
813: (*fn)(i - start_pos, client_data);
814: }
815: p[0].cur_start = cur_pos;
816: p[0].cur_leaf = p[0].function_buf;
817: p[0].cur_end = limit;
818: return;
819: }
820: }
821: /* End of leaf */
822: /* Pop the stack until we find two concatenation nodes with the */
823: /* same start position: this implies we were in left part. */
824: {
825: while (p[0].path_len > 0
826: && current_pe[0].pe_start_pos != current_pe[-1].pe_start_pos) {
827: p[0].path_len--;
828: current_pe--;
829: }
830: if (p[0].path_len == 0) {
831: p[0].path_len = CORD_POS_INVALID;
832: return;
833: }
834: }
835: p[0].path_len--;
836: CORD__extend_path(p);
837: }
838:
839: void CORD__prev(register CORD_pos p)
840: {
841: register struct CORD_pe * pe = &(p[0].path[p[0].path_len]);
842:
843: if (p[0].cur_pos == 0) {
844: p[0].path_len = CORD_POS_INVALID;
845: return;
846: }
847: p[0].cur_pos--;
848: if (p[0].cur_pos >= pe -> pe_start_pos) return;
849:
850: /* Beginning of leaf */
851:
852: /* Pop the stack until we find two concatenation nodes with the */
853: /* different start position: this implies we were in right part. */
854: {
855: register struct CORD_pe * current_pe = &((p)[0].path[(p)[0].path_len]);
856:
857: while (p[0].path_len > 0
858: && current_pe[0].pe_start_pos == current_pe[-1].pe_start_pos) {
859: p[0].path_len--;
860: current_pe--;
861: }
862: }
863: p[0].path_len--;
864: CORD__extend_path(p);
865: }
866:
867: #undef CORD_pos_fetch
868: #undef CORD_next
869: #undef CORD_prev
870: #undef CORD_pos_to_index
871: #undef CORD_pos_to_cord
872: #undef CORD_pos_valid
873:
874: char CORD_pos_fetch(register CORD_pos p)
875: {
876: if (p[0].cur_start <= p[0].cur_pos && p[0].cur_pos < p[0].cur_end) {
877: return(p[0].cur_leaf[p[0].cur_pos - p[0].cur_start]);
878: } else {
879: return(CORD__pos_fetch(p));
880: }
881: }
882:
883: void CORD_next(CORD_pos p)
884: {
885: if (p[0].cur_pos < p[0].cur_end - 1) {
886: p[0].cur_pos++;
887: } else {
888: CORD__next(p);
889: }
890: }
891:
892: void CORD_prev(CORD_pos p)
893: {
894: if (p[0].cur_end != 0 && p[0].cur_pos > p[0].cur_start) {
895: p[0].cur_pos--;
896: } else {
897: CORD__prev(p);
898: }
899: }
900:
901: size_t CORD_pos_to_index(CORD_pos p)
902: {
903: return(p[0].cur_pos);
904: }
905:
906: CORD CORD_pos_to_cord(CORD_pos p)
907: {
908: return(p[0].path[0].pe_cord);
909: }
910:
911: int CORD_pos_valid(CORD_pos p)
912: {
913: return(p[0].path_len != CORD_POS_INVALID);
914: }
915:
916: void CORD_set_pos(CORD_pos p, CORD x, size_t i)
917: {
918: if (x == CORD_EMPTY) {
919: p[0].path_len = CORD_POS_INVALID;
920: return;
921: }
922: p[0].path[0].pe_cord = x;
923: p[0].path[0].pe_start_pos = 0;
924: p[0].path_len = 0;
925: p[0].cur_pos = i;
926: CORD__extend_path(p);
927: }
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