Fix whitespace irregularities in code
[xapian.git] / xapian-applications / omega / omega.cc
blob545d44ed0f52b033271e76cd75d481900f2409f7
1 /* omega.cc: Main module for omega (example CGI frontend for Xapian)
3 * Copyright 1999,2000,2001 BrightStation PLC
4 * Copyright 2001 James Aylett
5 * Copyright 2001,2002 Ananova Ltd
6 * Copyright 2002,2003,2004,2006,2007,2008,2009,2010,2011,2014,2015,2016 Olly Betts
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License as
10 * published by the Free Software Foundation; either version 2 of the
11 * License, or (at your option) any later version.
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
18 * You should have received a copy of the GNU General Public License
19 * along with this program; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301
21 * USA
24 #include <config.h>
26 // If we're building against git after the expand API changed but before the
27 // version gets bumped to 1.3.2, we'll get a deprecation warning from
28 // get_eset() unless we suppress such warnings here.
29 #define XAPIAN_DEPRECATED(D) D
31 #include <cstdio>
32 #include <ctime>
34 #include <algorithm>
35 #include <cstring>
36 #include <iostream>
37 #include <set>
39 #include "safefcntl.h"
40 #include "safeunistd.h"
42 #include "omega.h"
43 #include "utils.h"
44 #include "cgiparam.h"
45 #include "query.h"
46 #include "str.h"
47 #include "stringutils.h"
48 #include "expand.h"
50 using namespace std;
52 static const char DEFAULT_STEM_LANGUAGE[] = "english";
54 // A character which doesn't require URL encoding, and isn't likely to appear
55 // in filter values.
56 const char filter_sep = '~';
58 // What we used for filter_sep in Omega < 1.3.4.
59 const char filter_sep_old = '-';
61 Xapian::Enquire * enquire;
62 Xapian::Database db;
63 Xapian::RSet rset;
65 map<string, string> option;
67 string date_start, date_end, date_span;
68 Xapian::valueno date_value_slot = Xapian::BAD_VALUENO;
70 bool set_content_type = false;
72 bool suppress_http_headers = false;
74 string dbname;
75 string fmtname;
76 string filters, old_filters;
78 Xapian::docid topdoc = 0;
79 Xapian::docid hits_per_page = 0;
80 Xapian::docid min_hits = 0;
82 // percentage cut-off
83 int threshold = 0;
85 Xapian::MultiValueKeyMaker* sort_keymaker = NULL;
86 Xapian::valueno sort_key = Xapian::BAD_VALUENO; // Don't sort.
87 bool reverse_sort = true;
88 bool sort_after = false;
89 Xapian::Enquire::docid_order docid_order = Xapian::Enquire::ASCENDING;
91 Xapian::valueno collapse_key = 0;
92 bool collapse = false;
94 static string
95 map_dbname_to_dir(const string &database_name)
97 return database_dir + database_name;
100 int main(int argc, char *argv[])
101 try {
102 read_config_file();
104 MCI val;
105 pair<MCI, MCI> g;
107 option["flag_default"] = "true";
109 // set default thousands and decimal separators: e.g. "16,729 hits" "1.4K"
110 option["decimal"] = ".";
111 option["thousand"] = ",";
113 // set the default stemming language
114 option["stemmer"] = DEFAULT_STEM_LANGUAGE;
116 // FIXME: set cout to linebuffered not stdout. Or just flush regularly...
117 //setvbuf(stdout, NULL, _IOLBF, 0);
119 const char * method = getenv("REQUEST_METHOD");
120 if (method == NULL) {
121 if (argc > 1 && (argv[1][0] != '-' || strchr(argv[1], '='))) {
122 // omega 'P=information retrieval' DB=papers
123 // check for a leading '-' on the first arg so "omega --version",
124 // "omega --help", and similar take the next branch
125 decode_argv(argv + 1);
126 } else {
127 // Seems we're running from the command line so give version
128 // and allow a query to be entered for testing
129 cout << PROGRAM_NAME " - " PACKAGE " " VERSION "\n";
130 if (argc > 1) exit(0);
131 cout << "Enter NAME=VALUE lines, end with blank line\n";
132 decode_test();
134 } else {
135 if (*method == 'P')
136 decode_post();
137 else
138 decode_get();
141 try {
142 // get database(s) to search
143 dbname.resize(0);
144 set<string> seen; // only add a repeated db once
145 g = cgi_params.equal_range("DB");
146 for (MCI i = g.first; i != g.second; ++i) {
147 const string & v = i->second;
148 if (!v.empty()) {
149 size_t p = 0, q;
150 while (true) {
151 q = v.find('/', p);
152 string s(v, p, q - p);
153 if (!s.empty() && seen.find(s) == seen.end()) {
154 // Translate DB parameter to path of database directory
155 if (!dbname.empty()) dbname += '/';
156 dbname += s;
157 db.add_database(Xapian::Database(map_dbname_to_dir(s)));
158 seen.insert(s);
160 if (q == string::npos) break;
161 p = q + 1;
165 if (dbname.empty()) {
166 dbname = default_db;
167 db.add_database(Xapian::Database(map_dbname_to_dir(dbname)));
169 enquire = new Xapian::Enquire(db);
170 } catch (const Xapian::Error &) {
171 enquire = NULL;
174 hits_per_page = 0;
175 val = cgi_params.find("HITSPERPAGE");
176 if (val != cgi_params.end()) hits_per_page = atol(val->second.c_str());
177 if (hits_per_page == 0) {
178 hits_per_page = 10;
179 } else if (hits_per_page > 1000) {
180 hits_per_page = 1000;
183 val = cgi_params.find("DEFAULTOP");
184 if (val != cgi_params.end()) {
185 const string & v = val->second;
186 if (v == "OR" || v == "or")
187 default_op = Xapian::Query::OP_OR;
190 val = cgi_params.find("FMT");
191 if (val != cgi_params.end()) {
192 const string & v = val->second;
193 if (!v.empty()) fmtname = v;
195 if (fmtname.empty())
196 fmtname = default_template;
198 val = cgi_params.find("MORELIKE");
199 if (enquire && val != cgi_params.end()) {
200 const string & v = val->second;
201 Xapian::docid docid = atol(v.c_str());
202 if (docid == 0) {
203 // Assume it's MORELIKE=Quid1138 and that Quid1138 is a UID
204 // from an external source - we just find the correspond docid
205 Xapian::PostingIterator p = db.postlist_begin(v);
206 if (p != db.postlist_end(v)) docid = *p;
209 if (docid != 0) {
210 Xapian::RSet tmprset;
211 tmprset.add_document(docid);
213 OmegaExpandDecider decider(db);
214 set_expansion_scheme(*enquire, option);
215 #if XAPIAN_AT_LEAST(1,3,2)
216 Xapian::ESet eset(enquire->get_eset(40, tmprset, &decider));
217 #else
218 Xapian::ESet eset(enquire->get_eset(40, tmprset, 0,
219 expand_param_k, &decider));
220 #endif
221 string morelike_query;
222 for (Xapian::ESetIterator i = eset.begin(); i != eset.end(); i++) {
223 if (!morelike_query.empty()) morelike_query += ' ';
224 morelike_query += pretty_term(*i);
226 set_probabilistic_query(string(), morelike_query);
228 } else {
229 // add expand/topterms terms if appropriate
230 string expand_terms;
231 if (cgi_params.find("ADD") != cgi_params.end()) {
232 g = cgi_params.equal_range("X");
233 for (MCI i = g.first; i != g.second; i++) {
234 const string & v = i->second;
235 if (!v.empty()) {
236 if (!expand_terms.empty())
237 expand_terms += ' ';
238 expand_terms += v;
243 // collect the unprefixed prob fields
244 g = cgi_params.equal_range("P");
245 for (MCI i = g.first; i != g.second; i++) {
246 const string & v = i->second;
247 if (!v.empty()) {
248 // If there are expand terms, append them to the first
249 // non-empty P parameter.
250 if (!expand_terms.empty()) {
251 string q = v;
252 q += ' ';
253 q += expand_terms;
254 set_probabilistic_query(string(), q);
255 expand_terms = string();
256 } else {
257 set_probabilistic_query(string(), v);
262 if (!expand_terms.empty()) {
263 set_probabilistic_query(string(), expand_terms);
267 g.first = cgi_params.lower_bound("P.");
268 g.second = cgi_params.lower_bound("P/"); // '/' is '.' + 1.
269 for (MCI i = g.first; i != g.second; i++) {
270 const string & v = i->second;
271 if (!v.empty()) {
272 string pfx(i->first, 2, string::npos);
273 set_probabilistic_query(pfx, v);
277 // set any boolean filters
278 g = cgi_params.equal_range("B");
279 if (g.first != g.second) {
280 vector<string> filter_v;
281 for (MCI i = g.first; i != g.second; i++) {
282 const string & v = i->second;
283 // we'll definitely get empty B fields from "-ALL-" options
284 if (!v.empty() && C_isalnum(v[0])) {
285 add_bterm(v);
286 filter_v.push_back(v);
289 sort(filter_v.begin(), filter_v.end());
290 vector<string>::const_iterator j;
291 for (j = filter_v.begin(); j != filter_v.end(); ++j) {
292 const string & bterm = *j;
293 string::size_type e = bterm.find(filter_sep);
294 if (usual(e == string::npos)) {
295 filters += bterm;
296 } else {
297 // If a filter contains filter_sep then double it to escape.
298 // Each filter must start with an alnum (checked above) and
299 // the value after the last filter is the default op, which
300 // is encoded as a non-alnum so filter_sep followed by
301 // something other than filter_sep must be separating filters.
302 string::size_type b = 0;
303 while (e != string::npos) {
304 filters.append(bterm, b, e + 1 - b);
305 b = e;
306 e = bterm.find(filter_sep, b + 1);
308 filters.append(bterm, b, string::npos);
310 filters += filter_sep;
311 old_filters += bterm;
312 old_filters += filter_sep_old;
316 // set any negated boolean filters
317 g = cgi_params.equal_range("N");
318 if (g.first != g.second) {
319 vector<string> filter_v;
320 for (MCI i = g.first; i != g.second; i++) {
321 const string & v = i->second;
322 // we'll definitely get empty N fields from "-ALL-" options
323 if (!v.empty() && C_isalnum(v[0])) {
324 add_nterm(v);
325 filter_v.push_back(v);
328 sort(filter_v.begin(), filter_v.end());
329 vector<string>::const_iterator j;
330 for (j = filter_v.begin(); j != filter_v.end(); ++j) {
331 const string & nterm = *j;
332 string::size_type e = nterm.find(filter_sep);
333 filters += '!';
334 if (usual(e == string::npos)) {
335 filters += nterm;
336 } else {
337 // If a filter contains filter_sep then double it to escape.
338 // Each filter must start with an alnum (checked above) and
339 // the value after the last filter is the default op, which
340 // is encoded as a non-alnum so filter_sep followed by
341 // something other than filter_sep must be separating filters.
342 string::size_type b = 0;
343 while (e != string::npos) {
344 filters.append(nterm, b, e + 1 - b);
345 b = e;
346 e = nterm.find(filter_sep, b + 1);
348 filters.append(nterm, b, string::npos);
350 filters += filter_sep;
351 // old_filters predates 'N' terms, so if there are 'N' terms this
352 // is definitely a different query.
353 old_filters.clear();
357 // date range filters
358 val = cgi_params.find("START");
359 if (val != cgi_params.end()) date_start = val->second;
360 val = cgi_params.find("END");
361 if (val != cgi_params.end()) date_end = val->second;
362 val = cgi_params.find("SPAN");
363 if (val != cgi_params.end()) date_span = val->second;
364 val = cgi_params.find("DATEVALUE");
365 if (val != cgi_params.end()) date_value_slot = string_to_int(val->second);
367 // If more default_op values are supported, encode them as non-alnums
368 // other than filter_sep or '!'.
369 filters += (default_op == Xapian::Query::OP_AND ? '.' : '-');
370 filters += date_start;
371 filters += filter_sep;
372 filters += date_end;
373 filters += filter_sep;
374 filters += date_span;
375 if (date_value_slot != Xapian::BAD_VALUENO) {
376 // This means we'll force the first page when reloading or changing
377 // page starting from existing URLs upon upgrade to 1.4.1, but the
378 // exact same existing URL could be for a search without the date
379 // filter where we want to force the first page, so there's an inherent
380 // ambiguity there. Forcing first page in this case seems the least
381 // problematic side-effect.
382 filters += filter_sep;
383 filters += str(date_value_slot);
386 if (!old_filters.empty()) {
387 old_filters += date_start;
388 old_filters += filter_sep_old;
389 old_filters += date_end;
390 old_filters += filter_sep_old;
391 old_filters += date_span;
392 old_filters += (default_op == Xapian::Query::OP_AND ? 'A' : 'O');
395 // Percentage relevance cut-off
396 val = cgi_params.find("THRESHOLD");
397 if (val != cgi_params.end()) {
398 threshold = atoi(val->second.c_str());
399 if (threshold < 0) threshold = 0;
400 if (threshold > 100) threshold = 100;
403 // collapsing
404 val = cgi_params.find("COLLAPSE");
405 if (val != cgi_params.end()) {
406 const string & v = val->second;
407 if (!v.empty()) {
408 collapse_key = atoi(v.c_str());
409 collapse = true;
410 filters += filter_sep;
411 filters += str(collapse_key);
412 if (!old_filters.empty()) {
413 old_filters += filter_sep_old;
414 old_filters += str(collapse_key);
418 if (!collapse && date_value_slot != Xapian::BAD_VALUENO) {
419 // We need to either omit filter_sep for both or neither, or else the
420 // encoding is ambiguous.
421 filters += filter_sep;
424 // docid order
425 val = cgi_params.find("DOCIDORDER");
426 if (val != cgi_params.end()) {
427 const string & v = val->second;
428 if (!v.empty()) {
429 char ch = v[0];
430 if (ch == 'D') {
431 docid_order = Xapian::Enquire::DESCENDING;
432 filters += 'D';
433 if (!old_filters.empty()) old_filters += 'D';
434 } else if (ch != 'A') {
435 docid_order = Xapian::Enquire::DONT_CARE;
436 } else {
437 // This is a bug (should add nothing here and 'X' in the (ch !=
438 // 'A') case, but the current "DONT_CARE" implementation
439 // actually always results in ascending docid order so it's not
440 // worth breaking compatibility to fix - let's just do it next
441 // time we change the encoding $filters uses.
442 filters += 'X';
443 if (!old_filters.empty()) old_filters += 'X';
448 // sorting
449 val = cgi_params.find("SORT");
450 if (val != cgi_params.end()) {
451 const char * base = val->second.c_str();
452 const char * p = base;
453 do {
454 bool rev = (*p != '+');
455 if (*p == '-' || *p == '+') {
456 // old_filters predates support for direction in SORT, so if
457 // there's a direction specified this is definitely a different
458 // query.
459 old_filters.clear();
460 ++p;
462 if (!C_isdigit(*p)) {
463 // Invalid.
464 break;
466 errno = 0;
467 char * q;
468 Xapian::valueno slot = strtoul(p, &q, 10);
469 p = q;
470 if (errno != 0) {
471 // Invalid.
472 break;
475 if (sort_key != Xapian::BAD_VALUENO) {
476 // Multiple sort keys specified, so we need a KeyMaker.
478 // Omit leading '+'.
479 if (reverse_sort) filters += '-';
480 filters += str(sort_key);
482 sort_keymaker = new Xapian::MultiValueKeyMaker;
483 sort_keymaker->add_value(sort_key, !reverse_sort);
484 sort_key = Xapian::BAD_VALUENO;
485 reverse_sort = true;
486 // old_filters predates multiple sort keys, so if there are
487 // multiple sort keys this is definitely a different query.
488 old_filters.clear();
491 if (sort_keymaker) {
492 filters += (rev ? '-' : '+');
493 filters += str(slot);
494 sort_keymaker->add_value(slot, !rev);
495 } else {
496 sort_key = slot;
497 reverse_sort = rev;
499 while (C_isspace(*p) || *p == ',') ++p;
500 } while (*p);
502 val = cgi_params.find("SORTREVERSE");
503 if (val != cgi_params.end() && atoi(val->second.c_str()) != 0) {
504 reverse_sort = !reverse_sort;
507 val = cgi_params.find("SORTAFTER");
508 if (val != cgi_params.end()) {
509 sort_after = (atoi(val->second.c_str()) != 0);
512 // Add the sorting related options to filters too.
514 // Note: old_filters really does encode a reversed sort as 'F' and a
515 // non-reversed sort as 'R' or 'r'.
517 // filters has them the other way around for sanity (except in
518 // development snapshot 1.3.4, which was when the new filter encoding
519 // was introduced).
520 if (!sort_keymaker) filters += str(sort_key);
521 if (!old_filters.empty()) old_filters += str(sort_key);
522 if (sort_after) {
523 if (reverse_sort) {
524 filters += 'R';
525 if (!old_filters.empty()) old_filters += 'F';
526 } else {
527 filters += 'F';
528 if (!old_filters.empty()) old_filters += 'R';
530 } else {
531 if (!reverse_sort) {
532 filters += 'f';
533 if (!old_filters.empty()) old_filters += 'r';
538 if (old_filters.empty()) old_filters = filters;
540 // min_hits (fill mset past topdoc+(hits_per_page+1) to
541 // topdoc+max(hits_per_page+1,min_hits)
542 val = cgi_params.find("MINHITS");
543 if (val != cgi_params.end()) {
544 min_hits = atol(val->second.c_str());
547 parse_omegascript();
548 } catch (const Xapian::Error &e) {
549 if (!set_content_type && !suppress_http_headers)
550 cout << "Content-Type: text/html\n\n";
551 cout << "Exception: " << html_escape(e.get_msg()) << endl;
552 } catch (const std::exception &e) {
553 if (!set_content_type && !suppress_http_headers)
554 cout << "Content-Type: text/html\n\n";
555 cout << "Exception: std::exception " << html_escape(e.what()) << endl;
556 } catch (const string &s) {
557 if (!set_content_type && !suppress_http_headers)
558 cout << "Content-Type: text/html\n\n";
559 cout << "Exception: " << html_escape(s) << endl;
560 } catch (const char *s) {
561 if (!set_content_type && !suppress_http_headers)
562 cout << "Content-Type: text/html\n\n";
563 cout << "Exception: " << html_escape(s) << endl;
564 } catch (...) {
565 if (!set_content_type && !suppress_http_headers)
566 cout << "Content-Type: text/html\n\n";
567 cout << "Caught unknown exception" << endl;