Annotation of libwww/Robot/src/HTRobot.c, revision 1.82
1.75 frystyk 1: /*
1.82 ! frystyk 2: ** @(#) $Id: HTRobot.c,v 1.81 1999/01/22 14:05:59 frystyk Exp $
1.75 frystyk 3: **
4: ** W3C Webbot can be found at "http://www.w3.org/Robot/"
5: **
6: ** Copyright 1995-1998 World Wide Web Consortium, (Massachusetts
7: ** Institute of Technology, Institut National de Recherche en
8: ** Informatique et en Automatique, Keio University). All Rights
9: ** Reserved. This program is distributed under the W3C's Software
10: ** Intellectual Property License. This program is distributed in the hope
11: ** that it will be useful, but WITHOUT ANY WARRANTY; without even the
12: ** implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
13: ** PURPOSE. See W3C License http://www.w3.org/Consortium/Legal/ for more
14: ** details.
1.1 frystyk 15: **
16: ** Authors:
17: ** HFN Henrik Frystyk Nielsen, (frystyk@w3.org)
1.75 frystyk 18: ** BR Bob Racko
19: ** JP John Punin
1.1 frystyk 20: **
21: ** History:
22: ** Dec 04 95 First version
1.75 frystyk 23: ** Oct 1998 Split into separate files
1.1 frystyk 24: */
25:
1.75 frystyk 26: #include "HTRobMan.h"
27: #include "HTQueue.h"
28: #include "HTAncMan.h"
1.51 frystyk 29:
1.62 frystyk 30: #define SHOW_QUIET(mr) ((mr) && !((mr)->flags & MR_QUIET))
31: #define SHOW_REAL_QUIET(mr) ((mr) && !((mr)->flags & MR_REAL_QUIET))
1.1 frystyk 32:
1.75 frystyk 33: PRIVATE HTErrorMessage HTErrors[HTERR_ELEMENTS] = {HTERR_ENGLISH_INITIALIZER};
1.58 frystyk 34:
35: /*
36: ** Some sorting algorithms
37: */
1.63 frystyk 38: PRIVATE HTComparer HitSort, FormatSort, LastModifiedSort, TitleSort;
1.58 frystyk 39:
1.80 frystyk 40: /*
41: ** Ths callbacks that we need from the libwww HTML parser
42: */
43: PRIVATE HText_new RHText_new;
44: PRIVATE HText_delete RHText_delete;
45: PRIVATE HText_foundLink RHText_foundLink;
1.1 frystyk 46:
47: /* ------------------------------------------------------------------------- */
48:
1.2 frystyk 49: /* Create a "HyperDoc" object
50: ** --------------------------
51: ** A HyperDoc object contains information about whether we have already
52: ** started checking the anchor and the depth in our search
53: */
1.75 frystyk 54: PUBLIC HyperDoc * HyperDoc_new (Robot * mr,HTParentAnchor * anchor, int depth)
1.2 frystyk 55: {
56: HyperDoc * hd;
1.14 frystyk 57: if ((hd = (HyperDoc *) HT_CALLOC(1, sizeof(HyperDoc))) == NULL)
58: HT_OUTOFMEM("HyperDoc_new");
1.2 frystyk 59: hd->depth = depth;
1.55 frystyk 60: hd->hits = 1;
1.75 frystyk 61:
62: hd->code = -1;
63: hd->index = ++mr->cindex;
64:
1.2 frystyk 65: /* Bind the HyperDoc object together with the Anchor Object */
66: hd->anchor = anchor;
67: HTAnchor_setDocument(anchor, (void *) hd);
68:
69: /* Add this HyperDoc object to our list */
70: if (!mr->hyperdoc) mr->hyperdoc = HTList_new();
71: HTList_addObject(mr->hyperdoc, (void *) hd);
72: return hd;
73: }
74:
75: /* Delete a "HyperDoc" object
76: ** --------------------------
77: */
1.75 frystyk 78: PUBLIC BOOL HyperDoc_delete (HyperDoc * hd)
1.2 frystyk 79: {
80: if (hd) {
1.11 frystyk 81: HT_FREE (hd);
1.2 frystyk 82: return YES;
83: }
84: return NO;
85: }
86:
1.55 frystyk 87: /*
88: ** Sort the anchor array and log reference count
89: */
90: PRIVATE BOOL calculate_hits (Robot * mr, HTArray * array)
91: {
92: if (mr && array) {
93: HTLog * log = HTLog_open(mr->hitfile, YES, YES);
94: if (log) {
95: void ** data = NULL;
96: HTParentAnchor * anchor = NULL;
97: HTArray_sort(array, HitSort);
98: anchor = (HTParentAnchor *) HTArray_firstObject(array, data);
99: while (anchor) {
100: char * uri = HTAnchor_address((HTAnchor *) anchor);
101: HyperDoc * hd = (HyperDoc *) HTAnchor_document(anchor);
1.63 frystyk 102: if (uri && hd) HTLog_addText(log, "%8d %s\n", hd->hits, uri);
1.55 frystyk 103: HT_FREE(uri);
104: anchor = (HTParentAnchor *) HTArray_nextObject(array, data);
105: }
106: }
107: HTLog_close(log);
108: return YES;
109: }
110: return NO;
111: }
112:
113: PRIVATE int HitSort (const void * a, const void * b)
114: {
115: HyperDoc * aa = HTAnchor_document(*(HTParentAnchor **) a);
116: HyperDoc * bb = HTAnchor_document(*(HTParentAnchor **) b);
117: if (aa && bb) return (bb->hits - aa->hits);
118: return bb - aa;
119: }
120:
1.58 frystyk 121: /*
1.64 frystyk 122: ** Sort the anchor array and log link relations
123: */
124: PRIVATE BOOL calculate_linkRelations (Robot * mr, HTArray * array)
125: {
126: if (mr && array) {
1.68 frystyk 127: HTLog * log = mr->relfile ? HTLog_open(mr->relfile, YES, YES) : NULL;
128: void ** data = NULL;
129: HTParentAnchor * anchor = NULL;
130: anchor = (HTParentAnchor *) HTArray_firstObject(array, data);
131: while (anchor) {
132:
133: /*
134: ** If we have a specific link relation to look for then do this.
135: ** Otherwise look for all link relations.
136: */
137: if (mr->relation) {
138: HTLink * link = HTAnchor_findLinkType((HTAnchor *) anchor, mr->relation);
139: if (link) {
140: HTParentAnchor * dest = HTAnchor_parent(HTLink_destination(link));
141: char * src_uri = HTAnchor_address((HTAnchor *) anchor);
142: char * dest_uri = HTAnchor_address((HTAnchor *) dest);
143: if (src_uri && dest_uri) {
144: #ifdef HT_MYSQL
145: if (mr->sqllog) {
146: HTSQLLog_addLinkRelationship (mr->sqllog,
147: src_uri, dest_uri,
148: HTAtom_name(mr->relation),
149: NULL);
150: }
151: #endif
152: if (log) {
153: HTFormat format = HTAnchor_format(dest);
154: HTLog_addText(log, "%s %s %s --> %s\n",
155: HTAtom_name(mr->relation),
156: format != WWW_UNKNOWN ?
157: HTAtom_name(format) : "<unknown>",
158: src_uri, dest_uri);
159: }
160:
161: /* Cleanup */
162: HT_FREE(src_uri);
163: HT_FREE(dest_uri);
164: }
165: }
166: } else {
167: HTLink * link = HTAnchor_mainLink((HTAnchor *) anchor);
168: HTList * sublinks = HTAnchor_subLinks((HTAnchor *) anchor);
169: char * src_uri = HTAnchor_address((HTAnchor *) anchor);
170: HTLinkType linktype;
171:
172: /* First look in the main link */
173: if (link && (linktype = HTLink_type(link))) {
174: HTParentAnchor * dest = HTAnchor_parent(HTLink_destination(link));
175: char * dest_uri = HTAnchor_address((HTAnchor *) dest);
176: if (src_uri && dest_uri) {
177: #ifdef HT_MYSQL
178: if (mr->sqllog) {
179: HTSQLLog_addLinkRelationship (mr->sqllog,
180: src_uri, dest_uri,
181: HTAtom_name(linktype),
182: NULL);
183: }
184: #endif
185: if (log) {
186: HTFormat format = HTAnchor_format(dest);
187: HTLog_addText(log, "%s %s %s --> %s\n",
188: HTAtom_name(linktype),
189: format != WWW_UNKNOWN ?
190: HTAtom_name(format) : "<unknown>",
191: src_uri, dest_uri);
192: }
193: }
194: HT_FREE(dest_uri);
195: }
196:
197: /* and then in any sublinks */
198: if (sublinks) {
199: HTLink * pres;
200: while ((pres = (HTLink *) HTList_nextObject(sublinks))) {
201: if ((linktype = HTLink_type(pres))) {
202: HTParentAnchor * dest = HTAnchor_parent(HTLink_destination(pres));
1.64 frystyk 203: char * dest_uri = HTAnchor_address((HTAnchor *) dest);
1.68 frystyk 204: if (src_uri && dest_uri) {
205: #ifdef HT_MYSQL
206: if (mr->sqllog) {
207: HTSQLLog_addLinkRelationship (mr->sqllog,
208: src_uri, dest_uri,
209: HTAtom_name(linktype),
210: NULL);
211: }
212: #endif
213: if (log) {
214: HTFormat format = HTAnchor_format(dest);
215: HTLog_addText(log, "%s %s %s --> %s\n",
216: HTAtom_name(linktype),
217: format != WWW_UNKNOWN ?
218: HTAtom_name(format) : "<unknown>",
219: src_uri, dest_uri);
220: }
1.64 frystyk 221: HT_FREE(dest_uri);
222: }
223: }
224: }
225: }
1.68 frystyk 226:
227: /* Cleanup */
228: HT_FREE(src_uri);
229: }
230: anchor = (HTParentAnchor *) HTArray_nextObject(array, data);
1.64 frystyk 231: }
1.68 frystyk 232: if (log) HTLog_close(log);
1.64 frystyk 233: return YES;
234: }
235: return NO;
236: }
237:
238: /*
1.63 frystyk 239: ** Sort the anchor array and log last modified date
240: */
241: PRIVATE BOOL calculate_lm (Robot * mr, HTArray * array)
242: {
243: if (mr && array) {
244: HTLog * log = HTLog_open(mr->lmfile, YES, YES);
245: if (log) {
246: void ** data = NULL;
247: HTParentAnchor * anchor = NULL;
248: HTArray_sort(array, LastModifiedSort);
249: anchor = (HTParentAnchor *) HTArray_firstObject(array, data);
250: while (anchor) {
251: char * uri = HTAnchor_address((HTAnchor *) anchor);
252: time_t lm = HTAnchor_lastModified(anchor);
253: if (uri && lm > 0)
254: HTLog_addText(log, "%s %s\n", HTDateTimeStr(&lm, NO), uri);
255: HT_FREE(uri);
256: anchor = (HTParentAnchor *) HTArray_nextObject(array, data);
257: }
258: }
259: HTLog_close(log);
260: return YES;
261: }
262: return NO;
263: }
264:
265: PRIVATE int LastModifiedSort (const void * a, const void * b)
266: {
267: time_t aa = HTAnchor_lastModified(*(HTParentAnchor **) a);
268: time_t bb = HTAnchor_lastModified(*(HTParentAnchor **) b);
269: return bb - aa;
270: }
271:
272: /*
273: ** Sort the anchor array and log the document title
274: */
275: PRIVATE BOOL calculate_title (Robot * mr, HTArray * array)
276: {
277: if (mr && array) {
278: HTLog * log = HTLog_open(mr->titlefile, YES, YES);
279: if (log) {
280: void ** data = NULL;
281: HTParentAnchor * anchor = NULL;
282: HTArray_sort(array, TitleSort);
283: anchor = (HTParentAnchor *) HTArray_firstObject(array, data);
284: while (anchor) {
285: char * uri = HTAnchor_address((HTAnchor *) anchor);
286: const char * title = HTAnchor_title(anchor);
287: HTCharset charset = HTAnchor_charset(anchor);
288: if (uri) HTLog_addText(log, "%s `%s\' %s\n",
289: charset ? HTAtom_name(charset) : "<none>",
290: title ? title : "<none>",
291: uri);
292: HT_FREE(uri);
293: anchor = (HTParentAnchor *) HTArray_nextObject(array, data);
294: }
295: }
296: HTLog_close(log);
297: return YES;
298: }
299: return NO;
300: }
301:
302: PRIVATE int TitleSort (const void * a, const void * b)
303: {
304: const char * aa = HTAnchor_title(*(HTParentAnchor **) a);
305: const char * bb = HTAnchor_title(*(HTParentAnchor **) b);
306: return strcasecomp(bb?bb:"", aa?aa:"");
307: }
308:
309: /*
1.58 frystyk 310: ** Calculate distributions for media types. The same mechanism
311: ** can be used for other characteristics with relatively
312: ** few outcomes.
313: */
314: PRIVATE HTList * mediatype_distribution (HTArray * array)
315: {
316: if (array) {
317: HTList * mt = HTList_new();
318: MetaDist * pres = NULL;
319: void ** data = NULL;
320: HTParentAnchor * anchor = NULL;
321: anchor = (HTParentAnchor *) HTArray_firstObject(array, data);
322: while (anchor) {
323: HTFormat format = HTAnchor_format(anchor);
324: if (format && format != WWW_UNKNOWN) {
325: HTList * cur = mt;
326:
327: /* If found then increase counter */
328: while ((pres = (MetaDist *) HTList_nextObject(cur))) {
329: if (pres->name == format) {
330: pres->hits++;
331: break;
332: }
333: }
334:
335: /* If not found then add new format to list */
336: if (!pres) {
337: if ((pres = (MetaDist *) HT_CALLOC(1, sizeof(MetaDist))) == NULL)
338: HT_OUTOFMEM("mediatype_distribution");
339: pres->name = format;
340: pres->hits = 1;
341: HTList_addObject(mt, pres);
342: HTList_insertionSort(mt, FormatSort);
343: }
344: }
345:
346: /* Find next anchor in array */
347: anchor = (HTParentAnchor *) HTArray_nextObject(array, data);
348: }
349: return mt;
350: }
351: return NULL;
352: }
353:
1.60 frystyk 354: /*
355: ** Calculate distributions for charsets. The same mechanism
356: ** can be used for other characteristics with relatively
357: ** few outcomes.
358: */
359: PRIVATE HTList * charset_distribution (HTArray * array)
360: {
361: if (array) {
362: HTList * cs = HTList_new();
363: MetaDist * pres = NULL;
364: void ** data = NULL;
365: HTParentAnchor * anchor = NULL;
366: anchor = (HTParentAnchor *) HTArray_firstObject(array, data);
367: while (anchor) {
368: HTCharset charset = HTAnchor_charset(anchor);
369: if (charset) {
370: HTList * cur = cs;
371:
372: /* If found then increase counter */
373: while ((pres = (MetaDist *) HTList_nextObject(cur))) {
374: if (pres->name == charset) {
375: pres->hits++;
376: break;
377: }
378: }
379:
380: /* If not found then add new format to list */
381: if (!pres) {
382: if ((pres = (MetaDist *) HT_CALLOC(1, sizeof(MetaDist))) == NULL)
383: HT_OUTOFMEM("charset_distribution");
384: pres->name = charset;
385: pres->hits = 1;
386: HTList_addObject(cs, pres);
387: HTList_insertionSort(cs, FormatSort);
388: }
389: }
390:
391: /* Find next anchor in array */
392: anchor = (HTParentAnchor *) HTArray_nextObject(array, data);
393: }
394: return cs;
395: }
396: return NULL;
397: }
398:
1.58 frystyk 399: PRIVATE int FormatSort (const void * a, const void * b)
400: {
401: MetaDist * aa = (MetaDist *) a;
402: MetaDist * bb = (MetaDist *) b;
403: return strcmp(HTAtom_name(bb->name), HTAtom_name(aa->name));
404: }
405:
406: PRIVATE BOOL log_meta_distribution (const char * logfile, HTList * distribution)
407: {
408: if (logfile && distribution) {
409: HTLog * log = HTLog_open(logfile, YES, YES);
410: if (log) {
411: HTList * cur = distribution;
412: MetaDist * pres;
413: while ((pres = (MetaDist *) HTList_nextObject(cur))) {
414: if (pres->name) {
1.60 frystyk 415: HTLog_addText(log, "%8d %s\n", pres->hits, HTAtom_name(pres->name));
1.58 frystyk 416: }
417: }
418: HTLog_close(log);
419: }
420: }
421: return NO;
422: }
423:
424: PRIVATE BOOL delete_meta_distribution (HTList * distribution)
425: {
426: if (distribution) {
427: HTList * cur = distribution;
428: MetaDist * pres;
429: while ((pres = (MetaDist *) HTList_nextObject(cur)))
430: HT_FREE(pres);
431: HTList_delete(distribution);
432: return YES;
433: }
434: return NO;
435: }
436:
437:
1.55 frystyk 438: /* Statistics
439: ** ----------
440: ** Calculates a bunch of statistics for the anchors traversed
441: */
442: PRIVATE BOOL calculate_statistics (Robot * mr)
443: {
1.59 frystyk 444: long total_docs = mr->get_docs + mr->head_docs + mr->other_docs;
1.55 frystyk 445: if (!mr) return NO;
446:
447: /* Calculate efficiency */
1.59 frystyk 448: if (mr->time > 0) {
1.56 frystyk 449: ms_t t = HTGetTimeInMillis() - mr->time;
450: if (t > 0) {
1.60 frystyk 451: double loadfactor = (mr->get_bytes / (t * 0.001));
452: double reqprsec = (total_docs / (t * 0.001));
1.56 frystyk 453: double secs = t / 1000.0;
1.55 frystyk 454: char bytes[50];
1.62 frystyk 455: if (SHOW_REAL_QUIET(mr))
1.82 ! frystyk 456: HTPrint("\nAccessed %ld documents in %.2f seconds (%.2f requests pr sec)\n",
1.62 frystyk 457: total_docs, secs, reqprsec);
1.59 frystyk 458:
459: HTNumToStr(mr->get_bytes, bytes, 50);
1.62 frystyk 460: if (SHOW_REAL_QUIET(mr))
1.82 ! frystyk 461: HTPrint("\tDid a GET on %ld document(s) and downloaded %s bytes of document bodies (%2.1f bytes/sec)\n",
1.62 frystyk 462: mr->get_docs, bytes, loadfactor);
1.59 frystyk 463:
464: HTNumToStr(mr->head_bytes, bytes, 50);
1.62 frystyk 465: if (SHOW_REAL_QUIET(mr))
1.82 ! frystyk 466: HTPrint("\tDid a HEAD on %ld document(s) with a total of %s bytes\n",
1.62 frystyk 467: mr->head_docs, bytes);
1.55 frystyk 468: }
469: }
470:
471: /* Create an array of existing anchors */
1.59 frystyk 472: if (total_docs > 1) {
473: HTArray * array = HTAnchor_getArray(total_docs);
1.55 frystyk 474: if (array) {
475:
1.63 frystyk 476: /* Distributions */
477: if (mr->flags & MR_DISTRIBUTIONS) {
1.82 ! frystyk 478: if (SHOW_REAL_QUIET(mr)) HTPrint("\nDistributions:\n");
1.63 frystyk 479: }
480:
1.55 frystyk 481: /* Sort after hit counts */
1.63 frystyk 482: if (mr->hitfile) {
483: if (SHOW_REAL_QUIET(mr))
1.82 ! frystyk 484: HTPrint("\tLogged hit count distribution in file `%s\'\n",
1.63 frystyk 485: mr->hitfile);
486: calculate_hits(mr, array);
487: }
488:
1.64 frystyk 489: /* Sort after link relations */
1.68 frystyk 490: #ifdef HT_MYSQL
491: if (mr->relfile || mr->sqllog) {
1.69 frystyk 492: #else
493: if (mr->relfile) {
494: #endif
1.68 frystyk 495: if (mr->relfile && SHOW_REAL_QUIET(mr))
1.82 ! frystyk 496: HTPrint("\tLogged link relationship distribution in file `%s\'\n",
1.64 frystyk 497: mr->relfile);
498: calculate_linkRelations(mr, array);
499: }
500:
1.63 frystyk 501: /* Sort after modified date */
502: if (mr->lmfile) {
503: if (SHOW_REAL_QUIET(mr))
1.82 ! frystyk 504: HTPrint("\tLogged last modified distribution in file `%s\'\n",
1.63 frystyk 505: mr->lmfile);
506: calculate_lm(mr, array);
507: }
508:
509: /* Sort after title */
510: if (mr->titlefile) {
511: if (SHOW_REAL_QUIET(mr))
1.82 ! frystyk 512: HTPrint("\tLogged title distribution in file `%s\'\n",
1.63 frystyk 513: mr->titlefile);
514: calculate_title(mr, array);
515: }
1.55 frystyk 516:
1.58 frystyk 517: /* Find mediatype distribution */
518: if (mr->mtfile) {
519: HTList * mtdist = mediatype_distribution(array);
520: if (mtdist) {
1.63 frystyk 521: if (SHOW_REAL_QUIET(mr))
1.82 ! frystyk 522: HTPrint("\tLogged media type distribution in file `%s\'\n",
1.63 frystyk 523: mr->mtfile);
1.58 frystyk 524: log_meta_distribution(mr->mtfile, mtdist);
525: delete_meta_distribution(mtdist);
526: }
527: }
1.55 frystyk 528:
1.60 frystyk 529: /* Find charset distribution */
530: if (mr->charsetfile) {
531: HTList * charsetdist = charset_distribution(array);
532: if (charsetdist) {
1.63 frystyk 533: if (SHOW_REAL_QUIET(mr))
1.82 ! frystyk 534: HTPrint("\tLogged charset distribution in file `%s\'\n",
1.63 frystyk 535: mr->charsetfile);
1.60 frystyk 536: log_meta_distribution(mr->charsetfile, charsetdist);
537: delete_meta_distribution(charsetdist);
538: }
539: }
540:
1.55 frystyk 541: /* Add as may other stats here as you like */
1.60 frystyk 542: /* ... */
1.58 frystyk 543:
544: /* Delete the array */
1.55 frystyk 545: HTArray_delete(array);
546: }
547: }
548: return YES;
549: }
550:
1.75 frystyk 551: PRIVATE HTParentAnchor *
552: get_last_parent(HTParentAnchor *anchor)
553: {
554: HTAnchor *anc;
555: HTList *sources = anchor->sources;
556:
557: while((anc = (HTAnchor *) HTList_nextObject(sources)) != NULL)
558: {
559: HTParentAnchor *panchor = HTAnchor_parent(anc);
560: return panchor;
561: }
562: return NULL;
563: }
564:
565: PRIVATE void
566: set_error_state_hyperdoc(HyperDoc * hd, HTRequest *request)
567: {
568: HTList * cur = HTRequest_error(request);
569: HTError *pres;
570:
571: while((pres = (HTError *) HTList_nextObject(cur)) != NULL)
572: {
573: int code =HTErrors[HTError_index(pres)].code;
574:
575: hd->code = code;
576: }
577: }
578:
579:
580: PRIVATE int
581: test_for_blank_spaces(char *uri)
582: {
583: char *ptr = uri;
584: for(;*ptr!='\0';ptr++)
585: if(*ptr == ' ')
586: return 1;
587: return 0;
588: }
589:
590:
1.1 frystyk 591: /* Create a Command Line Object
592: ** ----------------------------
593: */
1.75 frystyk 594: PUBLIC Robot * Robot_new (void)
1.1 frystyk 595: {
596: Robot * me;
1.41 frystyk 597: if ((me = (Robot *) HT_CALLOC(1, sizeof(Robot))) == NULL)
1.14 frystyk 598: HT_OUTOFMEM("Robot_new");
1.2 frystyk 599: me->hyperdoc = HTList_new();
1.4 frystyk 600: me->htext = HTList_new();
1.74 frystyk 601: me->timer = DEFAULT_TIMEOUT*MILLIES;
1.75 frystyk 602: me->waits = 0;
1.25 frystyk 603: me->cwd = HTGetCurrentDirectoryURL();
1.1 frystyk 604: me->output = OUTPUT;
1.35 eric 605: me->cnt = 0;
1.75 frystyk 606: me->ndoc = -1;
1.34 eric 607: me->fingers = HTList_new();
1.75 frystyk 608:
609: /* This is new */
610: me->queue = HTQueue_new();
611: me->cq = 0;
612: me->furl = NULL;
613:
1.1 frystyk 614: return me;
615: }
616:
617: /* Delete a Command Line Object
618: ** ----------------------------
619: */
1.62 frystyk 620: PRIVATE BOOL Robot_delete (Robot * mr)
1.1 frystyk 621: {
1.62 frystyk 622: if (mr) {
623: HTList_delete(mr->fingers);
1.55 frystyk 624:
625: /* Calculate statistics */
1.62 frystyk 626: calculate_statistics(mr);
1.55 frystyk 627:
1.62 frystyk 628: if (mr->hyperdoc) {
629: HTList * cur = mr->hyperdoc;
1.2 frystyk 630: HyperDoc * pres;
631: while ((pres = (HyperDoc *) HTList_nextObject(cur)))
632: HyperDoc_delete(pres);
1.62 frystyk 633: HTList_delete(mr->hyperdoc);
1.2 frystyk 634: }
1.62 frystyk 635: if (mr->htext) {
636: HTList * cur = mr->htext;
1.4 frystyk 637: HText * pres;
638: while ((pres = (HText *) HTList_nextObject(cur)))
1.80 frystyk 639: RHText_delete(pres);
1.62 frystyk 640: HTList_delete(mr->htext);
1.4 frystyk 641: }
1.62 frystyk 642:
643: /* Close all the log files */
1.63 frystyk 644: if (mr->flags & MR_LOGGING) {
1.82 ! frystyk 645: if (SHOW_REAL_QUIET(mr)) HTPrint("\nRaw Log files:\n");
1.63 frystyk 646: }
647:
1.62 frystyk 648: if (mr->log) {
649: if (SHOW_REAL_QUIET(mr))
1.82 ! frystyk 650: HTPrint("\tLogged %5d entries in general log file `%s\'\n",
1.62 frystyk 651: HTLog_accessCount(mr->log), mr->logfile);
652: HTLog_close(mr->log);
653: }
654: if (mr->ref) {
655: if (SHOW_REAL_QUIET(mr))
1.82 ! frystyk 656: HTPrint("\tLogged %5d entries in referer log file `%s\'\n",
1.62 frystyk 657: HTLog_accessCount(mr->ref), mr->reffile);
658: HTLog_close(mr->ref);
659: }
660: if (mr->reject) {
661: if (SHOW_REAL_QUIET(mr))
1.82 ! frystyk 662: HTPrint("\tLogged %5d entries in rejected log file `%s\'\n",
1.62 frystyk 663: HTLog_accessCount(mr->reject), mr->rejectfile);
664: HTLog_close(mr->reject);
665: }
666: if (mr->notfound) {
667: if (SHOW_REAL_QUIET(mr))
1.82 ! frystyk 668: HTPrint("\tLogged %5d entries in not found log file `%s\'\n",
1.62 frystyk 669: HTLog_accessCount(mr->notfound), mr->notfoundfile);
670: HTLog_close(mr->notfound);
671: }
672: if (mr->conneg) {
673: if (SHOW_REAL_QUIET(mr))
1.82 ! frystyk 674: HTPrint("\tLogged %5d entries in content negotiation log file `%s\'\n",
1.62 frystyk 675: HTLog_accessCount(mr->conneg), mr->connegfile);
676: HTLog_close(mr->conneg);
677: }
678: if (mr->noalttag) {
679: if (SHOW_REAL_QUIET(mr))
1.82 ! frystyk 680: HTPrint("\tLogged %5d entries in missing alt tag log file `%s\'\n",
1.62 frystyk 681: HTLog_accessCount(mr->noalttag), mr->noalttagfile);
682: HTLog_close(mr->noalttag);
683: }
684:
685: if (mr->output && mr->output != STDOUT) fclose(mr->output);
686:
687: if (mr->flags & MR_TIME) {
1.12 frystyk 688: time_t local = time(NULL);
1.62 frystyk 689: if (SHOW_REAL_QUIET(mr))
1.82 ! frystyk 690: HTPrint("\nRobot terminated %s\n", HTDateTimeStr(&local, YES));
1.12 frystyk 691: }
1.55 frystyk 692:
1.75 frystyk 693: /* This is new */
694: if(mr->cdepth)
695: HT_FREE(mr->cdepth);
696: if(mr->furl) HT_FREE(mr->furl);
697:
1.58 frystyk 698: #ifdef HT_POSIX_REGEX
1.62 frystyk 699: if (mr->include) {
700: regfree(mr->include);
701: HT_FREE(mr->include);
702: }
703: if (mr->exclude) {
704: regfree(mr->exclude);
705: HT_FREE(mr->exclude);
706: }
1.75 frystyk 707: if (mr->exc_robot) {
708: regfree(mr->exc_robot);
709: HT_FREE(mr->exc_robot);
710: }
1.62 frystyk 711: if (mr->check) {
712: regfree(mr->check);
713: HT_FREE(mr->check);
1.58 frystyk 714: }
715: #endif
716:
1.68 frystyk 717: #ifdef HT_MYSQL
718: if (mr->sqllog) {
719: HTSQLLog_close(mr->sqllog);
720: mr->sqllog = NULL;
721: }
722: #endif
723:
1.81 frystyk 724: if (mr->queue) HTQueue_delete(mr->queue);
1.62 frystyk 725: HT_FREE(mr->cwd);
726: HT_FREE(mr->prefix);
727: HT_FREE(mr->img_prefix);
728: HT_FREE(mr);
1.1 frystyk 729: return YES;
730: }
731: return NO;
732: }
733:
1.2 frystyk 734: /*
1.34 eric 735: ** This function creates a new finger object and initializes it with a new request
1.2 frystyk 736: */
1.75 frystyk 737: PUBLIC Finger * Finger_new (Robot * robot, HTParentAnchor * dest, HTMethod method)
1.2 frystyk 738: {
1.34 eric 739: Finger * me;
740: HTRequest * request = HTRequest_new();
741: if ((me = (Finger *) HT_CALLOC(1, sizeof(Finger))) == NULL)
742: HT_OUTOFMEM("Finger_new");
743: me->robot = robot;
744: me->request = request;
745: me->dest = dest;
746: HTList_addObject(robot->fingers, (void *)me);
747:
1.48 frystyk 748: /* Set the context for this request */
1.34 eric 749: HTRequest_setContext (request, me);
1.48 frystyk 750:
751: /* Check the various flags to customize the request */
752: if (robot->flags & MR_PREEMPTIVE)
753: HTRequest_setPreemptive(request, YES);
754: if (robot->flags & MR_VALIDATE)
755: HTRequest_setReloadMode(request, HT_CACHE_VALIDATE);
756: if (robot->flags & MR_END_VALIDATE)
757: HTRequest_setReloadMode(request, HT_CACHE_END_VALIDATE);
758:
759: /* We wanna make sure that we are sending a Host header (default) */
1.34 eric 760: HTRequest_addRqHd(request, HT_C_HOST);
1.48 frystyk 761:
762: /* Set the method for this request */
1.34 eric 763: HTRequest_setMethod(request, method);
764: robot->cnt++;
765: return me;
1.2 frystyk 766: }
767:
1.34 eric 768: PRIVATE int Finger_delete (Finger * me)
1.2 frystyk 769: {
1.34 eric 770: HTList_removeObject(me->robot->fingers, (void *)me);
771: me->robot->cnt--;
1.37 frystyk 772:
773: /*
774: ** If we are down at one request then flush the output buffer
775: */
776: if (me->request) {
777: if (me->robot->cnt == 1) HTRequest_forceFlush(me->request);
1.34 eric 778: HTRequest_delete(me->request);
1.37 frystyk 779: }
780:
781: /*
782: ** Delete the request and free myself
783: */
1.34 eric 784: HT_FREE(me);
785: return YES;
1.2 frystyk 786: }
787:
788: /*
789: ** Cleanup and make sure we close all connections including the persistent
790: ** ones
791: */
1.75 frystyk 792: PUBLIC void Cleanup (Robot * me, int status)
1.1 frystyk 793: {
1.81 frystyk 794: HTProfile_delete();
1.1 frystyk 795: Robot_delete(me);
1.50 frystyk 796: #ifdef HT_MEMLOG
1.39 eric 797: HTMemLog_close();
1.47 frystyk 798: #endif
799:
1.1 frystyk 800: #ifdef VMS
801: exit(status ? status : 1);
802: #else
803: exit(status ? status : 0);
804: #endif
805: }
806:
1.58 frystyk 807: #ifdef HT_POSIX_REGEX
808: PRIVATE char * get_regerror (int errcode, regex_t * compiled)
809: {
810: size_t length = regerror (errcode, compiled, NULL, 0);
811: char * str = NULL;
812: if ((str = (char *) HT_MALLOC(length+1)) == NULL)
813: HT_OUTOFMEM("get_regerror");
814: (void) regerror (errcode, compiled, str, length);
815: return str;
816: }
817:
1.75 frystyk 818: PUBLIC regex_t * get_regtype (Robot * mr, const char * regex_str, int cflags)
1.58 frystyk 819: {
820: regex_t * regex = NULL;
821: if (regex_str && *regex_str) {
822: int status;
823: if ((regex = (regex_t *) HT_CALLOC(1, sizeof(regex_t))) == NULL)
824: HT_OUTOFMEM("get_regtype");
1.60 frystyk 825: if ((status = regcomp(regex, regex_str, cflags))) {
1.58 frystyk 826: char * err_msg = get_regerror(status, regex);
1.62 frystyk 827: if (SHOW_REAL_QUIET(mr))
1.82 ! frystyk 828: HTPrint("Regular expression error: %s\n", err_msg);
1.58 frystyk 829: HT_FREE(err_msg);
830: Cleanup(mr, -1);
831: }
832: }
833: return regex;
834: }
835: #endif
836:
1.75 frystyk 837: PUBLIC void VersionInfo (void)
1.1 frystyk 838: {
1.82 ! frystyk 839: HTPrint("\nW3C OpenSource Software");
! 840: HTPrint("\n-----------------------\n\n");
! 841: HTPrint("\tWebbot version %s\n", APP_VERSION);
! 842: HTPrint("\tusing the W3C libwww library version %s.\n\n",HTLib_version());
! 843: HTPrint("\tSee \"%s\" for help\n", COMMAND_LINE);
! 844: HTPrint("\tSee \"http://www.w3.org/Robot/User/\" for user information\n");
! 845: HTPrint("\tSee \"http://www.w3.org/Robot/\" for general information\n\n");
! 846: HTPrint("\tPlease send feedback to the <www-lib@w3.org> mailing list,\n");
! 847: HTPrint("\tsee \"http://www.w3.org/Library/#Forums\" for details\n\n");
1.1 frystyk 848: }
849:
850: /* terminate_handler
851: ** -----------------
1.2 frystyk 852: ** This function is registered to handle the result of the request.
853: ** If no more requests are pending then terminate program
1.1 frystyk 854: */
1.75 frystyk 855: PUBLIC int terminate_handler (HTRequest * request, HTResponse * response,
1.32 frystyk 856: void * param, int status)
1.1 frystyk 857: {
1.34 eric 858: Finger * finger = (Finger *) HTRequest_context(request);
1.46 eric 859: Robot * mr = finger->robot;
1.82 ! frystyk 860: if (SHOW_QUIET(mr)) HTPrint("Robot....... done with %s\n", HTAnchor_physical(finger->dest));
1.55 frystyk 861:
1.68 frystyk 862: #ifdef HT_MYSQL
863: if (mr->sqllog) HTSQLLog_addEntry(mr->sqllog, request, status);
864: #endif
865:
1.58 frystyk 866: /* Check if negotiated resource and whether we should log that*/
867: if (mr->conneg) {
868: HTAssocList * cur = HTResponse_variant(response);
869: if (cur) {
870: BOOL first = YES;
871: HTChunk * buffer = HTChunk_new(128);
872: char * uri = HTAnchor_address((HTAnchor *) finger->dest);
873: HTAssoc * pres;
1.60 frystyk 874: HTChunk_puts(buffer, uri);
1.58 frystyk 875: while ((pres = (HTAssoc *) HTAssocList_nextObject(cur))) {
876: char * value = HTAssoc_value(pres);
877: if (first) {
1.60 frystyk 878: HTChunk_puts(buffer, "\t(");
1.58 frystyk 879: first = NO;
880: } else
881: HTChunk_puts(buffer, ", ");
882:
883: /* Output the name */
884: HTChunk_puts(buffer, HTAssoc_name(pres));
885:
886: /* Only output the value if not empty string */
1.60 frystyk 887: if (value && *value) {
1.58 frystyk 888: HTChunk_puts(buffer, "=");
889: HTChunk_puts(buffer, value);
890: }
891: }
1.60 frystyk 892: if (!first) HTChunk_puts(buffer, ")");
893: HTLog_addLine(mr->conneg, HTChunk_data(buffer));
1.58 frystyk 894: HTChunk_delete(buffer);
895: HT_FREE(uri);
896: }
897: }
898:
1.55 frystyk 899: /* Count the amount of body data that we have read */
1.59 frystyk 900: if (HTRequest_method(request) == METHOD_GET) {
901: int length = HTAnchor_length(HTRequest_anchor(request));
902: if (length > 0) mr->get_bytes += length;
903: mr->get_docs++;
904: } else if (HTRequest_method(request) == METHOD_HEAD) {
1.56 frystyk 905: int length = HTAnchor_length(HTRequest_anchor(request));
1.59 frystyk 906: if (length > 0) mr->head_bytes += length;
907: mr->head_docs++;
908: } else {
909: mr->other_docs++;
1.55 frystyk 910: }
911:
1.78 frystyk 912: if (!(mr->flags & MR_BFS)) {
913:
914: /* Delete this thread */
915: Finger_delete(finger);
916:
917: /* Should we stop? */
918: if (mr->cnt <= 0) {
1.82 ! frystyk 919: if (SHOW_QUIET(mr)) HTPrint(" Everything is finished...\n");
1.78 frystyk 920: Cleanup(mr, 0); /* No way back from here */
921: }
922: }
923:
1.82 ! frystyk 924: if (SHOW_QUIET(mr)) HTPrint(" %d outstanding request%s\n", mr->cnt, mr->cnt == 1 ? "" : "s");
1.75 frystyk 925: return HT_OK;
926:
927: }
928: PUBLIC int my_terminate_handler (HTRequest * request, HTResponse * response,
929: void * param, int status)
930: {
931: Finger * finger = (Finger *) HTRequest_context(request);
932: Robot * mr = finger->robot;
933: HTParentAnchor * dest = finger->dest;
934: HyperDoc * hd = HTAnchor_document(dest);
935: int depth = (hd ? hd->depth : -1);
936:
937: if (hd) set_error_state_hyperdoc(hd,request);
938:
939: if(hd && (HTRequest_method(request)== METHOD_HEAD) &&
940: (depth < mr->depth))
941: {
942: hd->method = METHOD_GET;
943: HTQueue_append(mr->queue, (void *)hd); (mr->cq)++;
944: }
1.58 frystyk 945:
1.34 eric 946: Finger_delete(finger);
1.55 frystyk 947:
1.75 frystyk 948: if(!(mr->flags & MR_PREEMPTIVE))
949: Serving_queue(mr);
950:
951: return HT_OK;
952: }
953:
954:
955: PUBLIC void Serving_queue(Robot *mr)
956: {
957: BOOL abort = NO;
958: Finger *nfinger;
959:
960: while(!abort)
961: {
962: if(!HTQueue_isEmpty(mr->queue))
963: {
964: HTRequest *newreq;
965:
966: HyperDoc *nhd = (HyperDoc *)HTQueue_headOfQueue(mr->queue);
967:
968: if(nhd)
969: {
970: char *uri = HTAnchor_address((HTAnchor *)nhd->anchor);
971: HTQueue_dequeue(mr->queue); (mr->cq)--;
972:
973: nfinger = Finger_new(mr, nhd->anchor, nhd->method);
974:
975: newreq = nfinger->request;
976:
1.82 ! frystyk 977: if(SHOW_QUIET(mr)) HTPrint("Request from QUEUE %s\n",uri);
1.75 frystyk 978: HT_FREE(uri);
1.82 ! frystyk 979: if(SHOW_QUIET(mr)) HTPrint("%d elements in queue \n", mr->cq);
1.75 frystyk 980:
981: HTRequest_setParent(newreq,get_last_parent(nhd->anchor));
982:
1.76 frystyk 983: /* @@@ Should be done using a timer and not sleep! @@@ */
984: #if 0
1.75 frystyk 985: if(mr->waits)
986: sleep(mr->waits);
1.76 frystyk 987: #endif
1.75 frystyk 988:
989: if (HTLoadAnchor((HTAnchor *)nhd->anchor , newreq) != YES)
990: {
1.82 ! frystyk 991: if (SHOW_QUIET(mr)) HTPrint("not tested!\n");
1.75 frystyk 992: Finger_delete(nfinger);
993: }
994: }
995: else
996: abort = YES;
997: }
998: else
999: abort = YES;
1000: }
1001:
1.82 ! frystyk 1002: if(SHOW_QUIET(mr)) HTPrint("Queue size: %d \n", mr->cq);
1.75 frystyk 1003:
1004: if (mr->cnt <= 0 || (abort && (mr->flags & MR_PREEMPTIVE)))
1005: {
1006: if(mr->cnt > 0)
1.82 ! frystyk 1007: if(SHOW_QUIET(mr)) HTPrint("%d requests were not served\n", mr->cnt);
1.75 frystyk 1008:
1.82 ! frystyk 1009: if (SHOW_QUIET(mr)) HTPrint(" Everything is finished...\n");
1.46 eric 1010: Cleanup(mr, 0); /* No way back from here */
1.75 frystyk 1011: }
1.1 frystyk 1012: }
1013:
1014: /* ------------------------------------------------------------------------- */
1015: /* HTEXT INTERFACE */
1016: /* ------------------------------------------------------------------------- */
1017:
1.80 frystyk 1018: PUBLIC BOOL Robot_registerHTMLParser (void)
1019: {
1020: HText_registerCDCallback(RHText_new, RHText_delete);
1021: HText_registerLinkCallback(RHText_foundLink);
1022: return YES;
1023: }
1024:
1025: PRIVATE HText * RHText_new (HTRequest * request, HTParentAnchor * anchor,
1026: HTStream * stream)
1.1 frystyk 1027: {
1028: HText * me;
1.34 eric 1029: Finger * finger = (Finger *) HTRequest_context(request);
1030: Robot * mr = finger->robot;
1.65 frystyk 1031: char * robots = NULL;
1032:
1.14 frystyk 1033: if ((me = (HText *) HT_CALLOC(1, sizeof(HText))) == NULL)
1034: HT_OUTOFMEM("HText_new2");
1.4 frystyk 1035:
1036: /* Bind the HText object together with the Request Object */
1.1 frystyk 1037: me->request = request;
1.65 frystyk 1038: me->follow = YES;
1039:
1040: /* Check to see if we have any meta tags */
1.77 frystyk 1041: if (!(mr->flags & MR_NOMETATAGS) && (robots = HTAnchor_robots(anchor)) != NULL) {
1.65 frystyk 1042: char * strval = NULL;
1043: char * ptr = NULL;
1044: char * token = NULL;
1045: StrAllocCopy(strval, robots);
1046: ptr = strval;
1047: while ((token = HTNextField(&ptr)) != NULL) {
1048: if (!strcasecomp(token, "nofollow")) {
1049: me->follow = NO;
1050: break;
1051: }
1052: }
1053: HT_FREE(strval);
1054: }
1.4 frystyk 1055:
1056: /* Add this HyperDoc object to our list */
1057: if (!mr->htext) mr->htext = HTList_new();
1058: HTList_addObject(mr->htext, (void *) me);
1.1 frystyk 1059: return me;
1060: }
1061:
1.80 frystyk 1062: PRIVATE BOOL RHText_delete (HText * me) {
1.81 frystyk 1063: if (me) {
1064: HT_FREE(me);
1065: return YES;
1066: }
1067: return NO;
1.4 frystyk 1068: }
1069:
1.80 frystyk 1070: PRIVATE void RHText_foundAnchor (HText * text, HTChildAnchor * anchor)
1.1 frystyk 1071: {
1072: if (text && anchor) {
1.34 eric 1073: Finger * finger = (Finger *) HTRequest_context(text->request);
1074: Robot * mr = finger->robot;
1.1 frystyk 1075: HTAnchor * dest = HTAnchor_followMainLink((HTAnchor *) anchor);
1076: HTParentAnchor * dest_parent = HTAnchor_parent(dest);
1.7 frystyk 1077: char * uri = HTAnchor_address((HTAnchor *) dest_parent);
1.1 frystyk 1078: HyperDoc * hd = HTAnchor_document(dest_parent);
1.60 frystyk 1079: HTParentAnchor * referer = HTRequest_anchor(text->request);
1.65 frystyk 1080: BOOL match = text->follow;
1.58 frystyk 1081: BOOL check = NO;
1.1 frystyk 1082:
1.75 frystyk 1083: /* These are new variables */
1084: HyperDoc * nhd = NULL;
1085: BOOL follow = YES;
1086:
1087: /* These three variables were moved */
1088: /*HTParentAnchor * last_anchor = HTRequest_parent(text->request);*/
1089: HTParentAnchor * last_anchor = HTRequest_anchor(text->request);
1090: HyperDoc * last_doc = HTAnchor_document(last_anchor);
1091: int depth = last_doc ? last_doc->depth+1 : 0;
1092:
1.55 frystyk 1093: if (!uri) return;
1.82 ! frystyk 1094: if (SHOW_QUIET(mr)) HTPrint("Robot....... Found `%s\' - \n", uri ? uri : "NULL\n");
1.55 frystyk 1095:
1096: if (hd) {
1.82 ! frystyk 1097: if (SHOW_QUIET(mr)) HTPrint("............ Already checked\n");
1.55 frystyk 1098: hd->hits++;
1.68 frystyk 1099: #ifdef HT_MYSQL
1100: if (mr->sqllog) {
1101: char * ref_addr = HTAnchor_address((HTAnchor *) referer);
1102: if (ref_addr) {
1103: HTSQLLog_addLinkRelationship(mr->sqllog, ref_addr, uri,
1104: "referer", NULL);
1105: HT_FREE(ref_addr);
1106: }
1107: }
1108: #endif
1.58 frystyk 1109: HT_FREE(uri);
1110: return;
1111: }
1.70 frystyk 1112:
1.58 frystyk 1113: /* Check for prefix match */
1.65 frystyk 1114: if (match && mr->prefix) {
1115: match = HTStrMatch(mr->prefix, uri) ? YES : NO;
1116: }
1.58 frystyk 1117:
1118: #ifdef HT_POSIX_REGEX
1.69 frystyk 1119: /*
1120: ** Check for any regular expression. The include may override
1121: ** the prefix matching
1122: */
1123: if (mr->include) {
1.58 frystyk 1124: match = regexec(mr->include, uri, 0, NULL, 0) ? NO : YES;
1125: }
1.75 frystyk 1126: if (match && mr->exc_robot) {
1127: match = regexec(mr->exc_robot, uri, 0, NULL, 0) ? YES : NO;
1128: }
1.58 frystyk 1129: if (match && mr->exclude) {
1130: match = regexec(mr->exclude, uri, 0, NULL, 0) ? YES : NO;
1131: }
1132: if (match && mr->check) {
1133: check = regexec(mr->check, uri, 0, NULL, 0) ? NO : YES;
1134: }
1135: #endif
1.75 frystyk 1136: if(uri && test_for_blank_spaces(uri))
1137: follow = NO;
1138: else if (mr->ndoc == 0) /* Number of Documents is reached */
1139: follow = NO;
1140:
1141: /* Test whether we already have a hyperdoc for this document */
1142: if(!hd && dest_parent)
1143: {
1144: nhd = HyperDoc_new(mr, dest_parent, depth);
1145: mr->cdepth[depth]++;
1146: }
1.58 frystyk 1147:
1148: /* Test whether we already have a hyperdoc for this document */
1.78 frystyk 1149: if (mr->flags & MR_LINK && match && dest_parent && follow && !hd) {
1150: if (mr->flags & MR_BFS) {
1151: nhd->method = METHOD_HEAD;
1152: HTQueue_enqueue(mr->queue, (void *) nhd);
1153: (mr->cq)++;
1154: if(mr->ndoc > 0) mr->ndoc--;
1155: } else {
1156: Finger * newfinger = Finger_new(mr, dest_parent, METHOD_GET);
1157: HTRequest * newreq = newfinger->request;
1158: HTRequest_setParent(newreq, referer);
1159: if (check || depth >= mr->depth) {
1.82 ! frystyk 1160: if (SHOW_QUIET(mr)) HTPrint("loading at depth %d using HEAD\n", depth);
1.78 frystyk 1161: HTRequest_setMethod(newreq, METHOD_HEAD);
1162: } else {
1.82 ! frystyk 1163: if (SHOW_QUIET(mr)) HTPrint("loading at depth %d\n", depth);
1.78 frystyk 1164: }
1165: if (HTLoadAnchor((HTAnchor *) dest_parent, newreq) != YES) {
1.82 ! frystyk 1166: if (SHOW_QUIET(mr)) HTPrint("not tested!\n");
1.78 frystyk 1167: Finger_delete(newfinger);
1168: }
1169: }
1.75 frystyk 1170:
1.7 frystyk 1171: } else {
1.82 ! frystyk 1172: if (SHOW_QUIET(mr)) HTPrint("............ does not fulfill constraints\n");
1.68 frystyk 1173: #ifdef HT_MYSQL
1174: if (mr->reject || mr->sqllog) {
1175: #else
1.60 frystyk 1176: if (mr->reject) {
1.68 frystyk 1177: #endif
1.60 frystyk 1178: if (referer) {
1179: char * ref_addr = HTAnchor_address((HTAnchor *) referer);
1.68 frystyk 1180: if (mr->reject && ref_addr)
1181: HTLog_addText(mr->reject, "%s --> %s\n", ref_addr, uri);
1182: #ifdef HT_MYSQL
1183: if (mr->sqllog && mr->sqlexternals && ref_addr)
1184: HTSQLLog_addLinkRelationship(mr->sqllog,
1185: ref_addr, uri,
1186: "referer", NULL);
1187: #endif
1188:
1.60 frystyk 1189: HT_FREE(ref_addr);
1190: }
1191: }
1.2 frystyk 1192: }
1.11 frystyk 1193: HT_FREE(uri);
1.2 frystyk 1194: }
1195: }
1196:
1.80 frystyk 1197: PRIVATE void RHText_foundImage (HText * text, HTChildAnchor * anchor,
1198: const char *alt, const char * align, BOOL isMap)
1.2 frystyk 1199: {
1200: if (text && anchor) {
1.34 eric 1201: Finger * finger = (Finger *) HTRequest_context(text->request);
1202: Robot * mr = finger->robot;
1.75 frystyk 1203:
1.59 frystyk 1204: if (mr->flags & MR_IMG) {
1.60 frystyk 1205: HTAnchor * dest = HTAnchor_followMainLink((HTAnchor *) anchor);
1206: HTParentAnchor * dest_parent = HTAnchor_parent(dest);
1207: char * uri = HTAnchor_address((HTAnchor *) dest_parent);
1208: HyperDoc * hd = HTAnchor_document(dest_parent);
1209: HTParentAnchor * referer = HTRequest_anchor(text->request);
1.59 frystyk 1210: BOOL match = YES;
1211:
1.72 frystyk 1212: if (!uri) return;
1.59 frystyk 1213: if (hd) {
1.82 ! frystyk 1214: if (SHOW_QUIET(mr)) HTPrint("............ Already checked\n");
1.59 frystyk 1215: hd->hits++;
1.68 frystyk 1216: #ifdef HT_MYSQL
1217: if (mr->sqllog) {
1218: char * ref_addr = HTAnchor_address((HTAnchor *) referer);
1219: if (ref_addr) {
1220: HTSQLLog_addLinkRelationship(mr->sqllog,
1221: ref_addr, uri,
1222: "image", alt);
1223: HT_FREE(ref_addr);
1224: }
1225: }
1226: #endif
1.11 frystyk 1227: HT_FREE(uri);
1.59 frystyk 1228: return;
1.2 frystyk 1229: }
1.59 frystyk 1230:
1231: /* Check for prefix match */
1232: if (mr->img_prefix) match = HTStrMatch(mr->img_prefix, uri) ? YES : NO;
1233:
1.79 br334 1234: #ifdef HT_POSIX_REGEX
1235: /*
1236: ** Check for any regular expression. The include may override
1237: ** the prefix matching
1238: */
1239: if (mr->include) {
1240: match = regexec(mr->include, uri, 0, NULL, 0) ? NO : YES;
1241: }
1242: if (match && mr->exc_robot) {
1243: match = regexec(mr->exc_robot, uri, 0, NULL, 0) ? YES : NO;
1244: }
1245: if (match && mr->exclude) {
1246: match = regexec(mr->exclude, uri, 0, NULL, 0) ? YES : NO;
1247: }
1248: #endif
1.59 frystyk 1249: /* Test whether we already have a hyperdoc for this document */
1250: if (match && dest) {
1.60 frystyk 1251: Finger * newfinger = Finger_new(mr, dest_parent,
1.59 frystyk 1252: mr->flags & MR_SAVE ?
1253: METHOD_GET : METHOD_HEAD);
1254: HTRequest * newreq = newfinger->request;
1.60 frystyk 1255: HyperDoc_new(mr, dest_parent, 1);
1256: HTRequest_setParent(newreq, referer);
1257:
1258: /* Check whether we should report missing ALT tags */
1259: if (mr->noalttag && (alt==NULL || *alt=='\0')) {
1260: if (referer) {
1261: char * ref_addr = HTAnchor_address((HTAnchor *) referer);
1262: if (ref_addr) HTLog_addText(mr->noalttag, "%s --> %s\n", ref_addr, uri);
1263: HT_FREE(ref_addr);
1264: }
1265: }
1266:
1.82 ! frystyk 1267: if (SHOW_QUIET(mr)) HTPrint("Robot....... Checking Image `%s\'\n", uri);
1.59 frystyk 1268: if (HTLoadAnchor((HTAnchor *) dest, newreq) != YES) {
1.82 ! frystyk 1269: if (SHOW_QUIET(mr)) HTPrint("Robot....... Image not tested!\n");
1.59 frystyk 1270: Finger_delete(newfinger);
1271: }
1272: } else {
1.82 ! frystyk 1273: if (SHOW_QUIET(mr)) HTPrint("............ does not fulfill constraints\n");
1.68 frystyk 1274: #ifdef HT_MYSQL
1275: if (mr->reject || mr->sqllog) {
1276: #else
1.60 frystyk 1277: if (mr->reject) {
1.68 frystyk 1278: #endif
1.60 frystyk 1279: if (referer) {
1280: char * ref_addr = HTAnchor_address((HTAnchor *) referer);
1.68 frystyk 1281: if (mr->reject && ref_addr)
1282: HTLog_addText(mr->reject, "%s --> %s\n", ref_addr, uri);
1283: #ifdef HT_MYSQL
1284: if (mr->sqllog && mr->sqlexternals && ref_addr)
1285: HTSQLLog_addLinkRelationship(mr->sqllog,
1286: ref_addr, uri,
1287: "image", alt);
1288: #endif
1289:
1.60 frystyk 1290: HT_FREE(ref_addr);
1291: }
1292: }
1.1 frystyk 1293: }
1.59 frystyk 1294: HT_FREE(uri);
1.72 frystyk 1295: }
1296: }
1297: }
1298:
1.80 frystyk 1299: PRIVATE void RHText_foundLink (HText * text,
1300: int element_number, int attribute_number,
1301: HTChildAnchor * anchor,
1302: const BOOL * present, const char ** value)
1.72 frystyk 1303: {
1304: if (text && anchor) {
1305: Finger * finger = (Finger *) HTRequest_context(text->request);
1306: Robot * mr = finger->robot;
1307: if (SHOW_QUIET(mr))
1.82 ! frystyk 1308: HTPrint("Robot....... Received element %d, attribute %d with anchor %p\n",
1.80 frystyk 1309: element_number, attribute_number, anchor);
1310: if ((element_number==HTML_IMG && attribute_number==HTML_IMG_SRC) ||
1311: (element_number==HTML_BODY && attribute_number==HTML_BODY_BACKGROUND))
1312: RHText_foundImage(text, anchor, NULL, NULL, NO);
1313: else
1314: RHText_foundAnchor(text, anchor);
1.72 frystyk 1315: }
1316: }
1317:
1.80 frystyk 1318: PUBLIC char * get_robots_txt(char * uri)
1.48 frystyk 1319: {
1.75 frystyk 1320: char *str = NULL;
1321: HTChunk * chunk;
1322: HTParentAnchor *anchor = HTAnchor_parent(HTAnchor_findAddress(uri));
1323: HTRequest *request = HTRequest_new();
1324: HTRequest_setOutputFormat(request, WWW_SOURCE);
1325: HTRequest_setPreemptive(request, YES);
1326: HTRequest_setMethod(request, METHOD_GET);
1327: chunk = HTLoadAnchorToChunk ((HTAnchor *)anchor, request);
1328: str = HTChunk_toCString(chunk);
1329: HTRequest_delete(request);
1330: return str;
1.48 frystyk 1331: }
1332:
Webmaster
![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to HTRobot.c CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [Public] / libwww / Robot / src