Annotation of parser3/src/main/pa_string.C, revision 1.81
1.45 paf 1: /** @file
1.55 paf 2: Parser: string class. @see untasize_t.C.
1.46 paf 3:
1.36 paf 4: Copyright (c) 2001 ArtLebedev Group (http://www.artlebedev.com)
1.46 paf 5:
1.37 paf 6: Author: Alexander Petrosyan <paf@design.ru> (http://design.ru/paf)
1.36 paf 7:
1.81 ! paf 8: $Id: pa_string.C,v 1.80 2001/04/26 11:37:40 paf Exp $
1.4 paf 9: */
10:
1.48 paf 11: #include "pa_config_includes.h"
1.1 paf 12:
1.70 paf 13: #include "pcre.h"
14:
1.13 paf 15: #include "pa_pool.h"
1.12 paf 16: #include "pa_string.h"
1.5 paf 17: #include "pa_hash.h"
1.22 paf 18: #include "pa_exception.h"
1.53 paf 19: #include "pa_common.h"
1.60 paf 20: #include "pa_array.h"
21: #include "pa_globals.h"
1.61 paf 22: #include "pa_table.h"
1.62 paf 23: #include "pa_threads.h"
1.60 paf 24:
25: //#include "pa_sapi.h"
1.1 paf 26:
1.18 paf 27: // String
28:
1.75 paf 29: String::String(Pool& apool, const char *src, size_t src_size, bool tainted) :
1.17 paf 30: Pooled(apool) {
1.28 paf 31: last_chunk=&head;
32: head.count=CR_PREALLOCATED_COUNT;
1.5 paf 33: append_here=head.rows;
1.2 paf 34: head.preallocated_link=0;
1.28 paf 35: link_row=&head.rows[head.count];
1.8 paf 36: fused_rows=fsize=0;
1.41 paf 37:
38: if(src)
1.75 paf 39: if(tainted)
40: APPEND_TAINTED(src, src_size, 0, 0);
1.41 paf 41: else
1.75 paf 42: APPEND_CLEAN(src, src_size, 0, 0);
1.1 paf 43: }
44:
45: void String::expand() {
1.55 paf 46: size_t new_chunk_count=last_chunk->count+last_chunk->count*CR_GROW_PERCENT/100;
1.28 paf 47: last_chunk=static_cast<Chunk *>(
1.55 paf 48: malloc(sizeof(size_t)+sizeof(Chunk::Row)*new_chunk_count+sizeof(Chunk *)));
1.28 paf 49: last_chunk->count=new_chunk_count;
50: link_row->link=last_chunk;
51: append_here=last_chunk->rows;
52: link_row=&last_chunk->rows[last_chunk->count];
1.8 paf 53: link_row->link=0;
1.1 paf 54: }
55:
1.40 paf 56: String::String(const String& src) : Pooled(src.pool()) {
1.8 paf 57: head.count=CR_PREALLOCATED_COUNT;
58:
1.55 paf 59: size_t src_used_rows=src.fused_rows;
1.8 paf 60: if(src_used_rows<=head.count) {
1.55 paf 61: // all new rows fit size_to preallocated area
62: size_t curr_chunk_rows=head.count;
1.8 paf 63: memcpy(head.rows, src.head.rows, sizeof(Chunk::Row)*src_used_rows);
64: append_here=&head.rows[src_used_rows];
65: link_row=&head.rows[curr_chunk_rows];
66: } else {
67: // warning:
1.10 paf 68: // heavily relies on the fact
69: // "preallocated area is the same for all strings"
1.8 paf 70: //
71: // info:
72: // allocating only enough mem to fit src string rows
73: // next append would allocate a new chunk
74: //
1.55 paf 75: // new rows don't fit size_to preallocated area: splitting size_to two chunks
1.8 paf 76: // preallocated chunk src to constructing head
77: memcpy(head.rows, src.head.rows, sizeof(Chunk::Row)*head.count);
1.55 paf 78: // remaining rows size_to new_chunk
79: size_t curr_chunk_rows=src_used_rows-head.count;
1.8 paf 80: Chunk *new_chunk=static_cast<Chunk *>(
1.55 paf 81: malloc(sizeof(size_t)+sizeof(Chunk::Row)*curr_chunk_rows+sizeof(Chunk *)));
1.8 paf 82: new_chunk->count=curr_chunk_rows;
83: head.preallocated_link=new_chunk;
1.28 paf 84: append_here=link_row=&new_chunk->rows[new_chunk->count];
1.8 paf 85:
86: Chunk *old_chunk=src.head.preallocated_link;
87: Chunk::Row *new_rows=new_chunk->rows;
1.55 paf 88: size_t rows_left_to_copy=new_chunk->count;
1.8 paf 89: while(true) {
1.55 paf 90: size_t old_count=old_chunk->count;
1.8 paf 91: Chunk *next_chunk=old_chunk->rows[old_count].link;
92: if(next_chunk) {
93: // not last source chunk
94: // taking it all
95: memcpy(new_rows, old_chunk->rows, sizeof(Chunk::Row)*old_count);
96: new_rows+=old_count;
97: rows_left_to_copy-=old_count;
98:
99: old_chunk=next_chunk;
100: } else {
101: // the last source chunk
102: // taking only those rows of chunk that _left_to_copy
103: memcpy(new_rows, old_chunk->rows, sizeof(Chunk::Row)*rows_left_to_copy);
104: break;
105: }
106: }
1.5 paf 107: }
1.8 paf 108: link_row->link=0;
109: fused_rows=src_used_rows;
110: fsize=src.fsize;
1.5 paf 111: }
1.28 paf 112:
1.42 paf 113: String& String::append(const String& src, Untaint_lang lang, bool forced) {
1.60 paf 114: const Chunk *chunk=&src.head;
1.40 paf 115: do {
1.60 paf 116: const Chunk::Row *row=chunk->rows;
117: for(size_t i=0; i<chunk->count; i++, row++) {
118: if(row==src.append_here)
1.40 paf 119: goto break2;
1.60 paf 120:
121: APPEND(row->item.ptr, row->item.size,
122: (lang!=UL_PASS_APPENDED && (row->item.lang==UL_TAINTED || forced))?lang:row->item.lang,
123: row->item.origin.file, row->item.origin.line);
1.40 paf 124: }
125: chunk=row->link;
126: } while(chunk);
127: break2:
1.60 paf 128: return *this;
1.34 paf 129: }
1.60 paf 130:
1.13 paf 131: String& String::real_append(STRING_APPEND_PARAMS) {
1.9 paf 132: if(!src)
133: return *this;
1.26 paf 134: if(!size)
135: size=strlen(src);
136: if(!size)
1.9 paf 137: return *this;
138:
1.1 paf 139: if(chunk_is_full())
140: expand();
141:
142: append_here->item.ptr=src;
1.26 paf 143: fsize+=append_here->item.size=size;
1.52 paf 144: append_here->item.lang=lang;
1.13 paf 145: #ifndef NO_STRING_ORIGIN
1.14 paf 146: append_here->item.origin.file=file;
147: append_here->item.origin.line=line;
1.13 paf 148: #endif
1.8 paf 149: append_here++; fused_rows++;
1.1 paf 150:
151: return *this;
152: }
153:
1.16 paf 154: uint String::hash_code() const {
1.7 paf 155: uint result=0;
1.5 paf 156:
1.16 paf 157: const Chunk *chunk=&head;
1.5 paf 158: do {
1.16 paf 159: const Chunk::Row *row=chunk->rows;
1.55 paf 160: for(size_t i=0; i<chunk->count; i++) {
1.5 paf 161: if(row==append_here)
162: goto break2;
163:
1.6 paf 164: result=Hash::generic_code(result, row->item.ptr, row->item.size);
1.5 paf 165: row++;
166: }
167: chunk=row->link;
168: } while(chunk);
169: break2:
170: return result;
171: }
172:
1.60 paf 173: /// @todo move 'lang' skipping to pos
174: int String::cmp(int& partial, const String& src,
175: size_t this_offset, Untaint_lang lang) const {
1.59 paf 176: partial=-1;
1.55 paf 177: this_offset=min(this_offset, size()-1);
178:
1.16 paf 179: const Chunk *a_chunk=&head;
180: const Chunk *b_chunk=&src.head;
181: const Chunk::Row *a_row=a_chunk->rows;
182: const Chunk::Row *b_row=b_chunk->rows;
1.55 paf 183: size_t a_offset=this_offset;
184: size_t b_offset=0;
1.9 paf 185: Chunk::Row *a_end=append_here;
186: Chunk::Row *b_end=src.append_here;
1.55 paf 187: size_t a_countdown=a_chunk->count;
188: size_t b_countdown=b_chunk->count;
1.9 paf 189: bool a_break=false;
190: bool b_break=false;
1.55 paf 191: size_t result;
1.60 paf 192: size_t pos=0;
193: while(true) {
1.33 paf 194: a_break=a_row==a_end;
195: b_break=b_row==b_end;
196: if(a_break || b_break)
197: break;
198:
1.55 paf 199: if(pos+a_row->item.size > this_offset) {
1.71 paf 200: if(lang!=UL_UNSPECIFIED && a_row->item.lang!=lang)
1.60 paf 201: return -1; // wrong lang -- bail out
202:
1.55 paf 203: int size_diff=
204: (a_row->item.size-a_offset)-
205: (b_row->item.size-b_offset);
206:
207: if(size_diff==0) { // a has same size as b
1.60 paf 208: result=memcmp(a_row->item.ptr+a_offset, b_row->item.ptr+b_offset,
209: a_row->item.size-a_offset);
1.55 paf 210: if(result)
211: return result;
1.60 paf 212: pos+=a_row->item.size;
1.55 paf 213: a_row++; a_countdown--; a_offset=0;
214: b_row++; b_countdown--; b_offset=0;
215: } else if (size_diff>0) { // a longer
1.60 paf 216: result=memcmp(a_row->item.ptr+a_offset, b_row->item.ptr+b_offset,
217: b_row->item.size-b_offset);
1.55 paf 218: if(result)
219: return result;
220: a_offset+=b_row->item.size-b_offset;
221: b_row++; b_countdown--; b_offset=0;
222: } else { // b longer
1.60 paf 223: result=memcmp(a_row->item.ptr+a_offset, b_row->item.ptr+b_offset,
224: a_row->item.size-a_offset);
1.55 paf 225: if(result)
226: return result;
227: b_offset+=a_row->item.size-a_offset;
1.60 paf 228: pos+=a_row->item.size;
1.55 paf 229: a_row++; a_countdown--; a_offset=0;
230: }
1.60 paf 231:
1.55 paf 232: if(!b_countdown) {
233: b_chunk=b_row->link;
234: b_row=b_chunk->rows;
235: b_countdown=b_chunk->count;
236: }
237: } else {
1.60 paf 238: a_offset-=a_row->item.size;
239: pos+=a_row->item.size;
240: a_row++; a_countdown--;
1.9 paf 241: }
242:
1.11 paf 243: if(!a_countdown) {
1.9 paf 244: a_chunk=a_row->link;
245: a_row=a_chunk->rows;
1.11 paf 246: a_countdown=a_chunk->count;
1.9 paf 247: }
1.27 paf 248: }
1.55 paf 249: if(a_break==b_break) { // ended simultaneously
250: partial=0; return 0;
251: } else if(a_break) { // first bytes equal, but a ended before b
252: partial=1; return -1;
253: } else {
254: partial=2; return +1;
255: }
1.27 paf 256: }
257:
1.60 paf 258: /// @todo move 'lang' skipping to pos
1.59 paf 259: int String::cmp(int& partial, const char* b_ptr, size_t src_size,
1.60 paf 260: size_t this_offset, Untaint_lang lang) const {
1.59 paf 261: partial=-1;
1.50 paf 262: size_t b_size=src_size?src_size:b_ptr?strlen(b_ptr):0;
1.59 paf 263: this_offset=min(this_offset, size()-1);
1.27 paf 264:
265: const Chunk *a_chunk=&head;
266: const Chunk::Row *a_row=a_chunk->rows;
1.59 paf 267: size_t a_offset=this_offset;
1.55 paf 268: size_t b_offset=0;
1.27 paf 269: Chunk::Row *a_end=append_here;
1.55 paf 270: size_t a_countdown=a_chunk->count;
1.27 paf 271: bool a_break=false;
272: bool b_break=false;
1.60 paf 273: size_t pos=0;
274: while(true) {
1.52 paf 275: a_break=a_row==a_end;
276: if(a_break || b_break)
277: break;
278:
1.59 paf 279: if(pos+a_row->item.size > this_offset) {
1.71 paf 280: if(lang!=UL_UNSPECIFIED && a_row->item.lang!=lang)
1.60 paf 281: return -1; // wrong lang -- bail out
282:
1.59 paf 283: int size_diff=
284: (a_row->item.size-a_offset)-
285: (b_size-b_offset);
286:
287: if(size_diff==0) { // a has same size as b
288: if(size_t result=memcmp(a_row->item.ptr+a_offset, b_ptr+b_offset,
289: a_row->item.size-a_offset)!=0)
290: return result;
1.60 paf 291: pos+=a_row->item.size;
1.59 paf 292: a_row++; a_countdown--; a_offset=0;
293: b_break=true;
294: } else if (size_diff>0) { // a longer
295: if(size_t result=memcmp(a_row->item.ptr+a_offset, b_ptr+b_offset,
296: b_size-b_offset)!=0)
297: return result;
298: a_offset+=b_size-b_offset;
299: b_break=true;
300: } else { // b longer
301: if(size_t result=memcmp(a_row->item.ptr+a_offset, b_ptr+b_offset,
302: a_row->item.size-a_offset)!=0)
303: return result;
304: b_offset+=a_row->item.size-a_offset;
1.60 paf 305: pos+=a_row->item.size;
1.59 paf 306: a_row++; a_countdown--; a_offset=0;
307: }
308: } else {
1.60 paf 309: a_offset-=a_row->item.size;
310: pos+=a_row->item.size;
311: a_row++; a_countdown--;
1.27 paf 312: }
313:
314: if(!a_countdown) {
315: a_chunk=a_row->link;
316: a_row=a_chunk->rows;
317: a_countdown=a_chunk->count;
1.9 paf 318: }
319: }
1.55 paf 320: if(a_break==b_break) { // ended simultaneously
321: partial=0; return 0;
322: } else if(a_break) { // first bytes equal, but a ended before b
323: partial=1; return -1;
324: } else {
325: partial=2; return +1;
326: }
1.5 paf 327: }
1.46 paf 328:
329: #ifndef NO_STRING_ORIGIN
330: const Origin& String::origin() const {
331: if(!fused_rows)
332: THROW(0, 0,
1.50 paf 333: 0,
334: "String::origin() of empty string called");
1.46 paf 335:
1.49 paf 336: // determining origin by last appended piece
1.50 paf 337: // because first one frequently constant.
338: // ex: ^load[/file] "document_root" + "/file"
1.80 paf 339: // when last peice is constant,
340: // ex: parser_root_auto_path{dynamic} / auto.p{const}
341: // using first piece
342: Origin& last_origin=append_here[-1].item.origin;
343: return last_origin.file ? last_origin : head.rows[0].item.origin;
1.46 paf 344: }
345: #endif
1.53 paf 346:
1.69 paf 347: String& String::mid(size_t start, size_t finish) const {
1.53 paf 348: start=max(0, start);
349: finish=min(size(), finish);
1.60 paf 350: if(start==finish)
351: return *empty_string;
1.53 paf 352:
353: String& result=*NEW String(pool());
354:
355: size_t pos=0;
356: const Chunk *chunk=&head;
357: do {
358: const Chunk::Row *row=chunk->rows;
1.55 paf 359: for(size_t i=0; i<chunk->count; pos+=row->item.size, i++, row++) {
1.53 paf 360: if(row==append_here)
361: goto break2;
362:
1.60 paf 363: size_t item_finish=pos+row->item.size;
364: if(item_finish > start) { // started now or already?
365: bool started=result.size()==0; // started now?
366: bool finished=finish <= item_finish; // finished now?
1.53 paf 367: size_t offset=started?start-pos:0;
368: size_t size=finished?finish-pos:row->item.size;
369: result.APPEND(
370: row->item.ptr+offset, size-offset,
371: row->item.lang,
372: row->item.origin.file, row->item.origin.line);
373: if(finished)
374: goto break2;
375: }
376: }
377: chunk=row->link;
378: } while(chunk);
379: break2:
1.60 paf 380: // SAPI::log(pool(), "piece of '%s' from %d to %d is '%s'",
381: //cstr(), start, finish, result.cstr());
1.53 paf 382: return result;
1.54 paf 383: }
384:
1.60 paf 385: int String::pos(const String& substr,
386: size_t result, Untaint_lang lang) const {
1.58 paf 387: for(; result<size(); result++) {
1.60 paf 388: int partial; cmp(partial, substr, result, lang);
1.58 paf 389: if(
390: partial==0 || // full match
391: partial==2) // 'substr' starts 'this'+'result'
392: return result;
393: }
394:
395: return -1;
396: }
397:
1.60 paf 398: int String::pos(const char *substr, size_t substr_size,
399: size_t result, Untaint_lang lang) const {
1.57 paf 400: for(; result<size(); result++) {
1.60 paf 401: int partial; cmp(partial, substr, substr_size, result, lang);
1.55 paf 402: if(
403: partial==0 || // full match
404: partial==2) // 'substr' starts 'this'+'result'
405: return result;
406: }
407:
408: return -1;
1.60 paf 409: }
410:
411: void String::split(Array& result,
412: size_t* pos_after_ref,
413: const char *delim, size_t delim_size,
414: Untaint_lang lang, int limit) const {
415: if(delim_size) {
416: size_t pos_after=pos_after_ref?*pos_after_ref:0;
417: int pos_before;
418: // while we have 'delim'...
419: for(; (pos_before=pos(delim, delim_size, pos_after, lang))>=0 && limit; limit--) {
1.69 paf 420: result+=&mid(pos_after, pos_before);
1.60 paf 421: pos_after=pos_before+delim_size;
422: }
423: // last piece
424: if(pos_after<size() && limit) {
1.69 paf 425: result+=&mid(pos_after, size());
1.60 paf 426: pos_after=size();
427: }
428: if(pos_after_ref)
429: *pos_after_ref=pos_after;
430: } else { // empty delim
431: result+=this;
432: if(pos_after_ref)
433: *pos_after_ref+=size();
434: }
435: }
436:
437: void String::split(Array& result,
438: size_t* pos_after_ref,
439: const String& delim, Untaint_lang lang,
440: int limit) const {
441: if(delim.size()) {
442: size_t pos_after=pos_after_ref?*pos_after_ref:0;
443: int pos_before;
444: // while we have 'delim'...
445: for(; (pos_before=pos(delim, pos_after, lang))>=0 && limit; limit--) {
1.69 paf 446: result+=&mid(pos_after, pos_before);
1.60 paf 447: pos_after=pos_before+delim.size();
448: }
449: // last piece
450: if(pos_after<size() && limit) {
1.69 paf 451: result+=&mid(pos_after, size());
1.60 paf 452: pos_after=size();
453: }
454: if(pos_after_ref)
455: *pos_after_ref=pos_after;
456: } else { // empty delim
457: result+=this;
458: if(pos_after_ref)
459: *pos_after_ref+=size();
460: }
1.61 paf 461: }
462:
1.63 paf 463: static void regex_options(char *options, int *result){
464: struct Regex_option {
465: char key;
466: int clear, set;
467: int *result;
468: } regex_option[]={
469: {'i', 0, PCRE_CASELESS, result}, // a=A
1.79 paf 470: {'s', 0, PCRE_DOTALL, result}, // \n\n$ [default]
1.63 paf 471: {'x', 0, PCRE_EXTENDED, result}, // whitespace in regex ignored
472: {'m', PCRE_DOTALL, PCRE_MULTILINE, result}, // ^aaa\n$^bbb\n$
473: {'g', 0, true, result+1}, // many rows
474: {0},
475: };
476: result[0]=PCRE_EXTRA | PCRE_DOTALL;
477: result[1]=0;
478:
479: if(options)
480: for(Regex_option *o=regex_option; o->key; o++)
481: if(
482: strchr(options, o->key) ||
483: strchr(options, toupper(o->key))) {
484: *(o->result)&=~o->clear;
485: *(o->result)|=o->set;
486: }
487: }
488:
1.66 paf 489: /**
490: returns true if fills table.
491: table format is defined and fixed[can be used by others]:
492: @verbatim
1.81 ! paf 493: prematch/match/postmatch/1/2/3/...
1.66 paf 494: @endverbatim
495: */
1.77 paf 496: bool String::match(const unsigned char *pcre_tables,
497: const String *aorigin,
1.62 paf 498: const String& regexp,
1.63 paf 499: const String *options,
1.64 paf 500: Table **table,
501: Row_action row_action, void *info) const {
502:
1.73 paf 503: if(!regexp.size())
504: THROW(0, 0,
505: aorigin,
506: "regexp is empty");
1.68 paf 507: const char *pattern=regexp.cstr(UL_AS_IS);
1.62 paf 508: const char *errptr;
509: int erroffset;
1.63 paf 510: int option_bits[2]; regex_options(options?options->cstr():0, option_bits);
511: pcre *code=pcre_compile(pattern, option_bits[0],
1.62 paf 512: &errptr, &erroffset,
1.74 paf 513: pcre_tables);
1.62 paf 514:
1.67 paf 515: if(!code)
1.62 paf 516: THROW(0, 0,
1.69 paf 517: ®exp.mid(erroffset, regexp.size()),
1.74 paf 518: "regular expression syntax error - %s", errptr);
1.62 paf 519:
1.63 paf 520: int info_substrings=pcre_info(code, 0, 0);
521: if(info_substrings<0) {
522: (*pcre_free)(code);
523: THROW(0, 0,
1.73 paf 524: aorigin,
1.76 paf 525: "pcre_info error (%d)",
1.73 paf 526: info_substrings);
1.63 paf 527: }
528:
529: int startoffset=0;
1.68 paf 530: const char *subject=cstr(UL_AS_IS);
1.62 paf 531: int length=strlen(subject);
1.63 paf 532: int ovecsize;
533: int *ovector=(int *)malloc(sizeof(int)*
1.65 paf 534: (ovecsize=(1/*match*/+info_substrings)*3));
1.62 paf 535:
1.64 paf 536: { // create table
537: Array& columns=*NEW Array(pool());
538: columns+=string_pre_match_name;
539: columns+=string_match_name;
540: columns+=string_post_match_name;
541: for(int i=1; i<=info_substrings; i++) {
542: char *column=(char *)malloc(MAX_NUMBER);
543: snprintf(column, MAX_NUMBER, "%d", i);
544: columns+=NEW String(pool(), column); // .i column name
545: }
546: *table=NEW Table(pool(), aorigin, &columns);
1.62 paf 547: }
1.63 paf 548:
1.64 paf 549: int exec_option_bits=0;
1.63 paf 550: while(true) {
551: int exec_substrings=pcre_exec(code, 0,
552: subject, length, startoffset,
1.64 paf 553: exec_option_bits, ovector, ovecsize);
1.63 paf 554:
555: if(exec_substrings==PCRE_ERROR_NOMATCH) {
556: (*pcre_free)(code);
1.67 paf 557: (*row_action)(**table, 0/*last time, no row*/, 0, 0, info);
1.63 paf 558: return option_bits[1]!=0; // global=true+table, not global=false
559: }
560:
561: if(exec_substrings<0) {
562: (*pcre_free)(code);
563: THROW(0, 0,
564: aorigin,
1.76 paf 565: "regular expression execute error (%d)",
1.63 paf 566: exec_substrings);
567: }
568:
569: Array& row=*NEW Array(pool());
1.81 ! paf 570: row+=&mid(0, ovector[0]); // .prematch column value
1.69 paf 571: row+=&mid(ovector[0], ovector[1]); // .match
1.81 ! paf 572: row+=&mid(ovector[1], size()); // .postmatch
1.63 paf 573:
574: for(int i=1; i<exec_substrings; i++) {
1.69 paf 575: // -1:-1 case handled peacefully by mid() itself
576: row+=&mid(ovector[i*2+0], ovector[i*2+1]); // .i column value
1.63 paf 577: }
578:
1.67 paf 579: (*row_action)(**table, &row, startoffset, ovector[0], info);
1.63 paf 580:
1.67 paf 581: if(!option_bits[1] || !(startoffset=ovector[1])) { // not global | going to hang
1.63 paf 582: (*pcre_free)(code);
1.67 paf 583: (*row_action)(**table, 0/*last time, no row*/, 0, 0, info);
1.63 paf 584: return true;
585: }
586:
587: /*
588: if(option_bits[0] & PCRE_MULTILINE)
1.64 paf 589: exec_option_bits|=PCRE_NOTBOL; // start of subject+startoffset not BOL
1.63 paf 590: */
591: }
1.61 paf 592: }
E-mail: