1 | #include <string.h>
|
---|
2 | #include <stdio.h>
|
---|
3 | #include <stdlib.h>
|
---|
4 | #include <ctype.h>
|
---|
5 |
|
---|
6 | #ifdef __GNUG__
|
---|
7 | # include <iostream.h>
|
---|
8 | # include <gdbm.h>
|
---|
9 |
|
---|
10 | #else
|
---|
11 | # ifndef USE_OBJECTSPACE
|
---|
12 | # include <iostream>
|
---|
13 | # else
|
---|
14 | # include <ospace\std\iostream>
|
---|
15 | # endif
|
---|
16 |
|
---|
17 | // gdbm stuff
|
---|
18 | # include "autoconf.h"
|
---|
19 | # include "systems.h"
|
---|
20 | # include "gdbmconst.h"
|
---|
21 | # include "gdbm.h"
|
---|
22 | #endif
|
---|
23 |
|
---|
24 | #include <assert.h>
|
---|
25 |
|
---|
26 | #include "mgq.h"
|
---|
27 | #include "mgsearch.h"
|
---|
28 | #include "locateinfo.h"
|
---|
29 | #include "gsdlunicode.h"
|
---|
30 | #include "unitool.h"
|
---|
31 |
|
---|
32 |
|
---|
33 | /////////////
|
---|
34 | // globals //
|
---|
35 | /////////////
|
---|
36 |
|
---|
37 | static char *quotedquery = NULL;
|
---|
38 |
|
---|
39 |
|
---|
40 |
|
---|
41 | ////////////////////////
|
---|
42 | // callback functions //
|
---|
43 | ////////////////////////
|
---|
44 |
|
---|
45 | // This routine is called for each document found in a search
|
---|
46 | // it assumes that cache_num is set up correctly to point to
|
---|
47 | // a suitable result cache
|
---|
48 | int ourquerycallback(char *UDoc, int ULen, int DocNum,
|
---|
49 | float Weight, void *info) {
|
---|
50 |
|
---|
51 |
|
---|
52 | queryresultsclass *queryresults = (queryresultsclass * )info;
|
---|
53 |
|
---|
54 | // check the returned document for the presence of the
|
---|
55 | // quoted part of the query, if there was one
|
---|
56 |
|
---|
57 | if (UDoc != NULL && quotedquery != NULL &&
|
---|
58 | quotedquery[0] != '\0' && strstr (UDoc, quotedquery) == NULL) return 0;
|
---|
59 |
|
---|
60 | // append this entry to the document results
|
---|
61 | docresultclass docresult;
|
---|
62 | docresult.docnum = DocNum;
|
---|
63 | docresult.docweight = Weight;
|
---|
64 |
|
---|
65 | queryresults->docs.push_back(docresult);
|
---|
66 |
|
---|
67 | return 0;
|
---|
68 | }
|
---|
69 |
|
---|
70 | // This callback is called once for each term in the query
|
---|
71 | int termfreqcallback(char *Word, int ULen, int Freq,
|
---|
72 | float Weight, void *info) {
|
---|
73 | queryresultsclass *queryresults = (queryresultsclass *)info;
|
---|
74 |
|
---|
75 | termfreqclass termfreq;
|
---|
76 | termfreq.termstr.setcarr(Word, ULen);
|
---|
77 | termfreq.termfreq = Freq;
|
---|
78 | queryresults->terms.push_back(termfreq);
|
---|
79 |
|
---|
80 | return 0;
|
---|
81 | }
|
---|
82 |
|
---|
83 | // this callback is called once for each variation of each term
|
---|
84 | int termscallback(char *Word, int ULen, int Freq,
|
---|
85 | float Weight, void *info) {
|
---|
86 |
|
---|
87 | // convert term from utf8 to unicode
|
---|
88 | text_t term;
|
---|
89 | utf8inconvertclass inconvert;
|
---|
90 | convertclass::status_t status;
|
---|
91 | inconvert.reset ();
|
---|
92 | inconvert.setinput (Word, ULen);
|
---|
93 | inconvert.convert (term, status);
|
---|
94 |
|
---|
95 | queryresultsclass *queryresults = (queryresultsclass *)info;
|
---|
96 | queryresults->termvariants.push_back(term);
|
---|
97 |
|
---|
98 | return 0;
|
---|
99 | }
|
---|
100 |
|
---|
101 | // This callback is for getting document text
|
---|
102 | int doctextcallback(char *Word, int ULen, int Freq,
|
---|
103 | float Weight, void *info) {
|
---|
104 | text_t *output = (text_t *)info;
|
---|
105 | if (output == NULL) return 0;
|
---|
106 | output->clear();
|
---|
107 |
|
---|
108 | utf8inconvertclass inconvert;
|
---|
109 | convertclass::status_t status;
|
---|
110 | inconvert.reset ();
|
---|
111 | inconvert.setinput (Word, ULen);
|
---|
112 | inconvert.convert (*output, status);
|
---|
113 |
|
---|
114 | // replace all control-Cs with spaces
|
---|
115 | text_t::iterator here = output->begin();
|
---|
116 | text_t::iterator end = output->end();
|
---|
117 | while (here != end) {
|
---|
118 | if (*here == '\x3') *here = ' ';
|
---|
119 | here++;
|
---|
120 | }
|
---|
121 |
|
---|
122 | return 0;
|
---|
123 | }
|
---|
124 |
|
---|
125 |
|
---|
126 |
|
---|
127 | ////////////////////
|
---|
128 | // mgsearch class //
|
---|
129 | ////////////////////
|
---|
130 |
|
---|
131 | mgsearchclass::mgsearchclass ()
|
---|
132 | {
|
---|
133 | cache = new querycache (RESULTCACHESIZE);
|
---|
134 | }
|
---|
135 |
|
---|
136 | mgsearchclass::~mgsearchclass ()
|
---|
137 | {
|
---|
138 | if (cache != NULL)
|
---|
139 | {
|
---|
140 | delete cache;
|
---|
141 | cache = NULL;
|
---|
142 | }
|
---|
143 | }
|
---|
144 |
|
---|
145 |
|
---|
146 | void mgsearchclass::setcollectdir (const text_t &thecollectdir)
|
---|
147 | {
|
---|
148 | collectdir = thecollectdir;
|
---|
149 | }
|
---|
150 |
|
---|
151 |
|
---|
152 | bool mgsearchclass::search(const queryparamclass &queryparams,
|
---|
153 | queryresultsclass &queryresults)
|
---|
154 | {
|
---|
155 | bool databaseloaded = true;
|
---|
156 |
|
---|
157 | assert (cache != NULL);
|
---|
158 |
|
---|
159 | queryresults.clear();
|
---|
160 |
|
---|
161 | // first check the cache
|
---|
162 | if (cache->find(queryparams, queryresults))
|
---|
163 | return true;
|
---|
164 |
|
---|
165 | // make sure there is a query to be processed
|
---|
166 | text_t::const_iterator queryhere = queryparams.querystring.begin();
|
---|
167 | text_t::const_iterator queryend = queryparams.querystring.end();
|
---|
168 | while (queryhere != queryend) {
|
---|
169 | if (is_unicode_letdig (*queryhere)) break;
|
---|
170 | queryhere++;
|
---|
171 | }
|
---|
172 |
|
---|
173 | // if we reached the end of the query string without finding
|
---|
174 | // any alphanumeric characters then return no results (and say
|
---|
175 | // the database was loaded)
|
---|
176 | if (queryhere == queryend) return true;
|
---|
177 |
|
---|
178 |
|
---|
179 | // get the names of the index and text suffixes
|
---|
180 | char *idxsuffix = (getindexsuffix (queryparams.collection,
|
---|
181 | queryparams.search_index)).getcstr();
|
---|
182 | assert (idxsuffix != NULL);
|
---|
183 | char *txtsuffix = (gettextsuffix (queryparams.collection)).getcstr();
|
---|
184 | assert (txtsuffix != NULL);
|
---|
185 |
|
---|
186 | #ifdef __WIN32__
|
---|
187 | char *ccollectdir = (collectdir+"\\").getcstr(); assert (ccollectdir != NULL);
|
---|
188 | #else
|
---|
189 | char *ccollectdir = collectdir.getcstr(); assert (ccollectdir != NULL);
|
---|
190 | #endif
|
---|
191 |
|
---|
192 | if (load_database(ccollectdir, idxsuffix, txtsuffix))
|
---|
193 | {
|
---|
194 | setsearchmode (queryparams);
|
---|
195 | submitquery (queryparams);
|
---|
196 | getresults (queryresults);
|
---|
197 | }
|
---|
198 | else databaseloaded = false;
|
---|
199 |
|
---|
200 | // free up the c strings
|
---|
201 | delete idxsuffix;
|
---|
202 | delete txtsuffix;
|
---|
203 | delete ccollectdir;
|
---|
204 |
|
---|
205 | return databaseloaded;
|
---|
206 | }
|
---|
207 |
|
---|
208 |
|
---|
209 | void mgsearchclass::setsearchmode (const queryparamclass &queryparams)
|
---|
210 | {
|
---|
211 | mgq_ask(".set expert true");
|
---|
212 | mgq_ask(".set accumulator_method list");
|
---|
213 | mgq_ask(".set max_accumulators 50000");
|
---|
214 | mgq_ask(".set verbatim true");
|
---|
215 | mgq_ask(".unset skip_dump");
|
---|
216 | mgq_ask(".set mode docnums");
|
---|
217 |
|
---|
218 | switch (queryparams.search_type)
|
---|
219 | {
|
---|
220 | case 0: mgq_ask(".set query boolean"); break;
|
---|
221 | case 1: mgq_ask(".set query ranked"); break;
|
---|
222 | }
|
---|
223 | switch (queryparams.casefolding)
|
---|
224 | {
|
---|
225 | case 1: mgq_ask(".set casefold on"); break;
|
---|
226 | case 0: mgq_ask(".set casefold off"); break;
|
---|
227 | }
|
---|
228 | switch (queryparams.stemming)
|
---|
229 | {
|
---|
230 | case 1: mgq_ask(".set stem on"); break;
|
---|
231 | case 0: mgq_ask(".set stem off"); break;
|
---|
232 | }
|
---|
233 | mgq_ask(".set heads_length 150");
|
---|
234 |
|
---|
235 | char maxdocstr[32];
|
---|
236 | sprintf(maxdocstr, ".set maxdocs %i", queryparams.maxdocs);
|
---|
237 | mgq_ask(maxdocstr);
|
---|
238 | }
|
---|
239 |
|
---|
240 |
|
---|
241 | void mgsearchclass::submitquery (const queryparamclass &queryparams)
|
---|
242 | {
|
---|
243 | // sort out the query string
|
---|
244 | text_t ttquerystring = queryparams.querystring;
|
---|
245 | text_t ttquotedquery;
|
---|
246 | extractquoted (ttquerystring, ttquotedquery);
|
---|
247 | filterquery (ttquerystring);
|
---|
248 |
|
---|
249 | // turn the strings into c strings for mg
|
---|
250 | if (quotedquery != NULL) // quotedquery is a global
|
---|
251 | {
|
---|
252 | delete quotedquery;
|
---|
253 | quotedquery = NULL;
|
---|
254 | }
|
---|
255 |
|
---|
256 | // quotedquery will be deleted on the next call to this function
|
---|
257 | quotedquery = ttquotedquery.getcstr ();
|
---|
258 | char *querystring = ttquerystring.getcstr();
|
---|
259 |
|
---|
260 | // submit the query
|
---|
261 | mgq_ask(querystring);
|
---|
262 |
|
---|
263 | delete querystring;
|
---|
264 | }
|
---|
265 |
|
---|
266 |
|
---|
267 | void mgsearchclass::getresults (queryresultsclass &queryresults)
|
---|
268 | {
|
---|
269 | if (quotedquery[0] == '\0')
|
---|
270 | {
|
---|
271 | // don't need the text
|
---|
272 | mgq_results(result_docnums, 0, MAXNUMDOCS,
|
---|
273 | ourquerycallback, (void *)(&queryresults));
|
---|
274 | }
|
---|
275 | else
|
---|
276 | {
|
---|
277 | // we need the text for this one
|
---|
278 | mgq_results(result_docs, 0, MAXNUMDOCS,
|
---|
279 | ourquerycallback, (void *)(&queryresults));
|
---|
280 | }
|
---|
281 |
|
---|
282 | // get the term frequencies
|
---|
283 | mgq_results(result_termfreqs, 0, MAXNUMTERMS,
|
---|
284 | termfreqcallback, (void *)(&queryresults));
|
---|
285 | mgq_results(result_terms, 0, MAXNUMTERMS,
|
---|
286 | termscallback, (void *)(&queryresults));
|
---|
287 | queryresults.sortqueryterms();
|
---|
288 | queryresults.uniqqueryterms();
|
---|
289 | }
|
---|
290 |
|
---|
291 |
|
---|
292 | void mgsearchclass::extractquoted (text_t &ttquerystring, text_t &ttquotedquery)
|
---|
293 | {
|
---|
294 | ttquotedquery.clear();
|
---|
295 |
|
---|
296 | text_t::iterator ithere = ttquerystring.begin ();
|
---|
297 | text_t::iterator itend = ttquerystring.end ();
|
---|
298 |
|
---|
299 | bool inquote = false;
|
---|
300 |
|
---|
301 | while (ithere != itend)
|
---|
302 | {
|
---|
303 | if ((*ithere) == '\"')
|
---|
304 | {
|
---|
305 | if (!inquote) ttquotedquery.clear ();
|
---|
306 | inquote = !inquote;
|
---|
307 | *ithere = ' '; // delete the quote
|
---|
308 | }
|
---|
309 | else if (inquote)
|
---|
310 | {
|
---|
311 | ttquotedquery.push_back(*ithere);
|
---|
312 | *ithere = ' ';
|
---|
313 | }
|
---|
314 |
|
---|
315 | ithere++;
|
---|
316 | }
|
---|
317 | }
|
---|
318 |
|
---|
319 |
|
---|
320 | void mgsearchclass::filterquery (text_t &ttquerystring)
|
---|
321 | {
|
---|
322 |
|
---|
323 | text_t::iterator ithere = ttquerystring.begin ();
|
---|
324 | text_t::iterator itend = ttquerystring.end ();
|
---|
325 | unsigned short c;
|
---|
326 |
|
---|
327 | // remove all non alphanumeric characters below 127
|
---|
328 | while (ithere != itend)
|
---|
329 | {
|
---|
330 | c = *ithere;
|
---|
331 |
|
---|
332 | // if ((c <= 127) && !((c >= '0' && c <= '9') ||
|
---|
333 | // (c >= 'A' && c <= 'Z') ||
|
---|
334 | // (c >= 'a' && c <= 'z')))
|
---|
335 | if (!(((c >= 65) && (c <= 90)) ||
|
---|
336 | ((c >= 97) && (c <= 122)) ||
|
---|
337 | ((c >= 192) && (c <= 214)) ||
|
---|
338 | ((c >= 216) && (c <= 246)) ||
|
---|
339 | ((c >= 248) && (c <= 255)) ||
|
---|
340 | ((c >= '0') && (c <= '9')) ||
|
---|
341 | (c == 176)))
|
---|
342 | (*ithere) = ' ';
|
---|
343 |
|
---|
344 | ithere++;
|
---|
345 | }
|
---|
346 | }
|
---|
347 |
|
---|
348 |
|
---|
349 | // the document text for 'docnum' is placed in 'output'
|
---|
350 | // docTargetDocument returns 'true' if it was able to
|
---|
351 | // try to get a document
|
---|
352 | // collection is needed to see if an index from the
|
---|
353 | // collection is loaded. If no index has been loaded
|
---|
354 | // defaultindex is needed to load one
|
---|
355 | bool mgsearchclass::docTargetDocument(const text_t &defaultindex,
|
---|
356 | const text_t &collection,
|
---|
357 | int docnum,
|
---|
358 | text_t &output)
|
---|
359 | {
|
---|
360 | bool databaseloaded = true;
|
---|
361 |
|
---|
362 | output.clear();
|
---|
363 |
|
---|
364 |
|
---|
365 | // make sure index is level 2
|
---|
366 |
|
---|
367 | ////// this changed with new naming scheme in new building software
|
---|
368 | ///// i.e paragraph level index no longer contain number '3' but begin
|
---|
369 | ///// with letter 'p'
|
---|
370 |
|
---|
371 | text_t db_loaded = db_loaded_name;
|
---|
372 |
|
---|
373 | if (!db_loaded.empty()) {
|
---|
374 | text_t::const_iterator here = db_loaded.begin();
|
---|
375 | text_t::const_iterator end = db_loaded.end();
|
---|
376 |
|
---|
377 |
|
---|
378 | //while (here != end) {
|
---|
379 | // if (*here == '3')
|
---|
380 | // databaseloaded = false;
|
---|
381 | // here ++;
|
---|
382 | //}
|
---|
383 |
|
---|
384 | char separator = '/';
|
---|
385 | text_t db;
|
---|
386 | int found = 0;
|
---|
387 | #ifdef __WIN32__
|
---|
388 | separator = '\\';
|
---|
389 | #endif;
|
---|
390 | // strip away path to db and following collection name
|
---|
391 | end --;
|
---|
392 | while (end != here) {
|
---|
393 | if (*end == separator) {
|
---|
394 | if (found) break;
|
---|
395 | else {db.clear(); found = 1; end--; continue;}
|
---|
396 | }
|
---|
397 | db.push_back(*end);
|
---|
398 | end --;
|
---|
399 | }
|
---|
400 |
|
---|
401 | // string will have been reversed above so see if last
|
---|
402 | // character is 'p'
|
---|
403 | if (db[db.size()-1] == 'p') databaseloaded = false;
|
---|
404 | }
|
---|
405 |
|
---|
406 | // find out if the database is already loaded
|
---|
407 | // this is needed because a different index (but valid one)
|
---|
408 | // might be already loaded.
|
---|
409 | // this comparison is needed because 'load_database'
|
---|
410 | // is now more oriented towards indexes
|
---|
411 | if (databaseloaded == true) {
|
---|
412 | text_t::const_iterator here = collection.begin();
|
---|
413 | text_t::const_iterator end = collection.end();
|
---|
414 | char *dbhere = &db_loaded_name[strlen(db_loaded_name) - collection.size()]; // assumes collection shorter than db_loaded_name
|
---|
415 | while (here != end)
|
---|
416 | {
|
---|
417 | if (*here != *dbhere)
|
---|
418 | {
|
---|
419 | databaseloaded = false;
|
---|
420 | break;
|
---|
421 | }
|
---|
422 | here++;
|
---|
423 | dbhere++;
|
---|
424 | }
|
---|
425 | }
|
---|
426 |
|
---|
427 | // try and load the database
|
---|
428 | if (!databaseloaded)
|
---|
429 | {
|
---|
430 | // get the names of the index and text suffixes
|
---|
431 | char *idxsuffix = (getindexsuffix (collection,
|
---|
432 | defaultindex)).getcstr();
|
---|
433 | assert (idxsuffix != NULL);
|
---|
434 | char *txtsuffix = (gettextsuffix (collection)).getcstr();
|
---|
435 | assert (txtsuffix != NULL);
|
---|
436 |
|
---|
437 | #ifdef __WIN32__
|
---|
438 | char *ccollectdir = (collectdir+"\\").getcstr(); assert (ccollectdir != NULL);
|
---|
439 | #else
|
---|
440 | char *ccollectdir = collectdir.getcstr(); assert (ccollectdir != NULL);
|
---|
441 | #endif
|
---|
442 |
|
---|
443 | if (load_database(ccollectdir, idxsuffix, txtsuffix))
|
---|
444 | databaseloaded = true;
|
---|
445 | else
|
---|
446 | databaseloaded = false;
|
---|
447 |
|
---|
448 | // free up the c strings
|
---|
449 | delete idxsuffix;
|
---|
450 | delete txtsuffix;
|
---|
451 | delete ccollectdir;
|
---|
452 | }
|
---|
453 |
|
---|
454 | if (databaseloaded)
|
---|
455 | {
|
---|
456 | // retrieve the document from mg
|
---|
457 | char docstr[32];
|
---|
458 | sprintf(docstr, "%i", docnum);
|
---|
459 |
|
---|
460 | mgq_ask(".set mode text");
|
---|
461 | mgq_ask(".set query docnums");
|
---|
462 | mgq_ask(docstr);
|
---|
463 | mgq_results (result_docs, 0, 1, doctextcallback, (void *)&output);
|
---|
464 | }
|
---|
465 |
|
---|
466 | return databaseloaded;
|
---|
467 | }
|
---|
468 |
|
---|