xref: /dokuwiki/inc/fulltext.php (revision 2674fdfba016552f6f843df0e812db75b904fcda) 
1<?php
2/**
3 * DokuWiki fulltextsearch functions using the index
4 *
5 * @license    GPL 2 (http://www.gnu.org/licenses/gpl.html)
6 * @author     Andreas Gohr <andi@splitbrain.org>
7 */
8
9  if(!defined('DOKU_INC')) define('DOKU_INC',fullpath(dirname(__FILE__).'/../').'/');
10  require_once(DOKU_INC.'inc/indexer.php');
11
12
13/**
14 * The fulltext search
15 *
16 * Returns a list of matching documents for the given query
17 *
18 * refactored into ft_pageSearch(), _ft_pageSearch() and trigger_event()
19 *
20 */
21function ft_pageSearch($query,&$highlight){
22
23  $data['query'] = $query;
24  $data['highlight'] =& $highlight;
25
26  return trigger_event('SEARCH_QUERY_FULLPAGE', $data, '_ft_pageSearch');
27}
28function _ft_pageSearch(&$data){
29    // split out original parameters
30    $query = $data['query'];
31    $highlight =& $data['highlight'];
32
33    $q = ft_queryParser($query);
34
35    $highlight = array();
36
37    // remember for hilighting later
38    foreach($q['words'] as $wrd){
39        $highlight[] =  str_replace('*','',$wrd);
40    }
41
42    // lookup all words found in the query
43    $words  = array_merge($q['and'],$q['not']);
44    if(!count($words)) return array();
45    $result = idx_lookup($words);
46    if(!count($result)) return array();
47
48    // merge search results with query
49    foreach($q['and'] as $pos => $w){
50        $q['and'][$pos] = $result[$w];
51    }
52    // create a list of unwanted docs
53    $not = array();
54    foreach($q['not'] as $pos => $w){
55        $not = array_merge($not,array_keys($result[$w]));
56    }
57
58    // combine and-words
59    if(count($q['and']) > 1){
60        $docs = ft_resultCombine($q['and']);
61    }else{
62        $docs = $q['and'][0];
63    }
64    if(!count($docs)) return array();
65
66    // create a list of hidden pages in the result
67    $hidden = array();
68    $hidden = array_filter(array_keys($docs),'isHiddenPage');
69    $not = array_merge($not,$hidden);
70
71    // filter unmatched namespaces
72    if(!empty($q['ns'])) {
73        $pattern = implode('|^',$q['ns']);
74        foreach($docs as $key => $val) {
75            if(!preg_match('/^'.$pattern.'/',$key)) {
76                unset($docs[$key]);
77            }
78        }
79    }
80
81    // remove negative matches
82    foreach($not as $n){
83        unset($docs[$n]);
84    }
85
86    if(!count($docs)) return array();
87    // handle phrases
88    if(count($q['phrases'])){
89        $q['phrases'] = array_map('utf8_strtolower',$q['phrases']);
90        // use this for higlighting later:
91        $highlight = array_merge($highlight,$q['phrases']);
92        $q['phrases'] = array_map('preg_quote_cb',$q['phrases']);
93        // check the source of all documents for the exact phrases
94        foreach(array_keys($docs) as $id){
95            $text  = utf8_strtolower(rawWiki($id));
96            foreach($q['phrases'] as $phrase){
97                if(!preg_match('/'.$phrase.'/usi',$text)){
98                    unset($docs[$id]); // no hit - remove
99                    break;
100                }
101            }
102        }
103    }
104
105    if(!count($docs)) return array();
106
107    // check ACL permissions
108    foreach(array_keys($docs) as $doc){
109        if(auth_quickaclcheck($doc) < AUTH_READ){
110            unset($docs[$doc]);
111        }
112    }
113
114    if(!count($docs)) return array();
115
116    // if there are any hits left, sort them by count
117    arsort($docs);
118
119    return $docs;
120}
121
122/**
123 * Returns the backlinks for a given page
124 *
125 * Does a quick lookup with the fulltext index, then
126 * evaluates the instructions of the found pages
127 */
128function ft_backlinks($id){
129    global $conf;
130    $swfile   = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
131    $stopwords = @file_exists($swfile) ? file($swfile) : array();
132
133    $result = array();
134
135    // quick lookup of the pagename
136    $page    = noNS($id);
137    $matches = idx_lookup(idx_tokenizer($page,$stopwords));  // pagename may contain specials (_ or .)
138    $docs    = array_keys(ft_resultCombine(array_values($matches)));
139    $docs    = array_filter($docs,'isVisiblePage'); // discard hidden pages
140    if(!count($docs)) return $result;
141    require_once(DOKU_INC.'inc/parserutils.php');
142
143    // check metadata for matching links
144    foreach($docs as $match){
145        // metadata relation reference links are already resolved
146        $links = p_get_metadata($match,'relation references');
147        if (isset($links[$id])) $result[] = $match;
148    }
149
150    if(!count($result)) return $result;
151
152    // check ACL permissions
153    foreach(array_keys($result) as $idx){
154        if(auth_quickaclcheck($result[$idx]) < AUTH_READ){
155            unset($result[$idx]);
156        }
157    }
158
159    sort($result);
160    return $result;
161}
162
163/**
164 * Returns the pages that use a given media file
165 *
166 * Does a quick lookup with the fulltext index, then
167 * evaluates the instructions of the found pages
168 *
169 * Aborts after $max found results
170 */
171function ft_mediause($id,$max){
172    global $conf;
173    $swfile   = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
174    $stopwords = @file_exists($swfile) ? file($swfile) : array();
175
176    if(!$max) $max = 1; // need to find at least one
177
178    $result = array();
179
180    // quick lookup of the mediafile
181    $media   = noNS($id);
182    $matches = idx_lookup(idx_tokenizer($media,$stopwords));
183    $docs    = array_keys(ft_resultCombine(array_values($matches)));
184    if(!count($docs)) return $result;
185
186    // go through all found pages
187    $found = 0;
188    $pcre  = preg_quote($media,'/');
189    foreach($docs as $doc){
190        $ns = getNS($doc);
191        preg_match_all('/\{\{([^|}]*'.$pcre.'[^|}]*)(|[^}]+)?\}\}/i',rawWiki($doc),$matches);
192        foreach($matches[1] as $img){
193            $img = trim($img);
194            if(preg_match('/^https?:\/\//i',$img)) continue; // skip external images
195            list($img) = explode('?',$img);                  // remove any parameters
196            resolve_mediaid($ns,$img,$exists);               // resolve the possibly relative img
197
198            if($img == $id){                                 // we have a match
199                $result[] = $doc;
200                $found++;
201                break;
202            }
203        }
204        if($found >= $max) break;
205    }
206
207    sort($result);
208    return $result;
209}
210
211
212
213/**
214 * Quicksearch for pagenames
215 *
216 * By default it only matches the pagename and ignores the
217 * namespace. This can be changed with the second parameter
218 *
219 * refactored into ft_pageLookup(), _ft_pageLookup() and trigger_event()
220 *
221 * @author Andreas Gohr <andi@splitbrain.org>
222 */
223function ft_pageLookup($id,$pageonly=true){
224    $data = array('id' => $id, 'pageonly' => $pageonly);
225    return trigger_event('SEARCH_QUERY_PAGELOOKUP',$data,'_ft_pageLookup');
226}
227
228function _ft_pageLookup(&$data){
229    // split out original parameterrs
230    $id = $data['id'];
231    $pageonly = $data['pageonly'];
232
233    global $conf;
234    $id    = preg_quote($id,'/');
235    $pages = file($conf['indexdir'].'/page.idx');
236    if($id) $pages = array_values(preg_grep('/'.$id.'/',$pages));
237
238    $cnt = count($pages);
239    for($i=0; $i<$cnt; $i++){
240        if($pageonly){
241            if(!preg_match('/'.$id.'/',noNS($pages[$i]))){
242                unset($pages[$i]);
243                continue;
244            }
245        }
246        if(!page_exists($pages[$i])){
247            unset($pages[$i]);
248            continue;
249        }
250    }
251
252    $pages = array_filter($pages,'isVisiblePage'); // discard hidden pages
253    if(!count($pages)) return array();
254
255    // check ACL permissions
256    foreach(array_keys($pages) as $idx){
257        if(auth_quickaclcheck($pages[$idx]) < AUTH_READ){
258            unset($pages[$idx]);
259        }
260    }
261
262    $pages = array_map('trim',$pages);
263    usort($pages,'ft_pagesorter');
264    return $pages;
265}
266
267/**
268 * Sort pages based on their namespace level first, then on their string
269 * values. This makes higher hierarchy pages rank higher than lower hierarchy
270 * pages.
271 */
272function ft_pagesorter($a, $b){
273    $ac = count(explode(':',$a));
274    $bc = count(explode(':',$b));
275    if($ac < $bc){
276        return -1;
277    }elseif($ac > $bc){
278        return 1;
279    }
280    return strcmp ($a,$b);
281}
282
283/**
284 * Creates a snippet extract
285 *
286 * @author Andreas Gohr <andi@splitbrain.org>
287 */
288function ft_snippet($id,$highlight){
289    $text     = rawWiki($id);
290    $match = array();
291    $snippets = array();
292    $utf8_offset = $offset = $end = 0;
293    $len = utf8_strlen($text);
294
295    // build a regexp from the phrases to highlight
296    $re = join('|',array_map('preg_quote_cb',array_filter((array) $highlight)));
297
298    for ($cnt=3; $cnt--;) {
299      if (!preg_match('#('.$re.')#iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) break;
300
301      list($str,$idx) = $match[0];
302
303      // convert $idx (a byte offset) into a utf8 character offset
304      $utf8_idx = utf8_strlen(substr($text,0,$idx));
305      $utf8_len = utf8_strlen($str);
306
307      // establish context, 100 bytes surrounding the match string
308      // first look to see if we can go 100 either side,
309      // then drop to 50 adding any excess if the other side can't go to 50,
310      $pre = min($utf8_idx-$utf8_offset,100);
311      $post = min($len-$utf8_idx-$utf8_len,100);
312
313      if ($pre>50 && $post>50) {
314        $pre = $post = 50;
315      } else if ($pre>50) {
316        $pre = min($pre,100-$post);
317      } else if ($post>50) {
318        $post = min($post, 100-$pre);
319      } else {
320        // both are less than 50, means the context is the whole string
321        // make it so and break out of this loop - there is no need for the
322        // complex snippet calculations
323        $snippets = array($text);
324        break;
325      }
326
327      // establish context start and end points, try to append to previous
328      // context if possible
329      $start = $utf8_idx - $pre;
330      $append = ($start < $end) ? $end : false;  // still the end of the previous context snippet
331      $end = $utf8_idx + $utf8_len + $post;      // now set it to the end of this context
332
333      if ($append) {
334        $snippets[count($snippets)-1] .= utf8_substr($text,$append,$end-$append);
335      } else {
336        $snippets[] = utf8_substr($text,$start,$end-$start);
337      }
338
339      // set $offset for next match attempt
340      //   substract strlen to avoid splitting a potential search success,
341      //   this is an approximation as the search pattern may match strings
342      //   of varying length and it will fail if the context snippet
343      //   boundary breaks a matching string longer than the current match
344      $utf8_offset = $utf8_idx + $post;
345      $offset = $idx + strlen(utf8_substr($text,$utf8_idx,$post));
346      $offset = utf8_correctIdx($text,$offset);
347    }
348
349    $m = "\1";
350    $snippets = preg_replace('#('.$re.')#iu',$m.'$1'.$m,$snippets);
351    $snippet = preg_replace('#'.$m.'([^'.$m.']*?)'.$m.'#iu','<strong class="search_hit">$1</strong>',hsc(join('... ',$snippets)));
352
353    return $snippet;
354}
355
356/**
357 * Combine found documents and sum up their scores
358 *
359 * This function is used to combine searched words with a logical
360 * AND. Only documents available in all arrays are returned.
361 *
362 * based upon PEAR's PHP_Compat function for array_intersect_key()
363 *
364 * @param array $args An array of page arrays
365 */
366function ft_resultCombine($args){
367    $array_count = count($args);
368    if($array_count == 1){
369        return $args[0];
370    }
371
372    $result = array();
373    if ($array_count > 1) {
374      foreach ($args[0] as $key => $value) {
375        $result[$key] = $value;
376        for ($i = 1; $i !== $array_count; $i++) {
377            if (!isset($args[$i][$key])) {
378                unset($result[$key]);
379                break;
380            }
381            $result[$key] += $args[$i][$key];
382        }
383      }
384    }
385    return $result;
386}
387
388/**
389 * Builds an array of search words from a query
390 *
391 * @todo support OR and parenthesises?
392 */
393function ft_queryParser($query){
394    global $conf;
395    $swfile   = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
396    if(@file_exists($swfile)){
397        $stopwords = file($swfile);
398    }else{
399        $stopwords = array();
400    }
401
402    $q = array();
403    $q['query']   = $query;
404    $q['ns']      = array();
405    $q['phrases'] = array();
406    $q['words']   = array();
407    $q['and']     = array();
408    $q['not']     = array();
409
410    // strip namespace from query
411    if(preg_match('/([^@]*)@(.*)/',$query,$match))  {
412        $query = $match[1];
413        $q['ns'] = explode('@',preg_replace("/ /",'',$match[2]));
414    }
415
416    // handle phrase searches
417    while(preg_match('/"(.*?)"/',$query,$match)){
418        $q['phrases'][] = $match[1];
419        $q['and'] = array_merge($q['and'], idx_tokenizer($match[0],$stopwords));
420        $query = preg_replace('/"(.*?)"/','',$query,1);
421    }
422
423    $words = explode(' ',$query);
424    foreach($words as $w){
425        if($w{0} == '-'){
426            $token = idx_tokenizer($w,$stopwords,true);
427            if(count($token)) $q['not'] = array_merge($q['not'],$token);
428        }else{
429            // asian "words" need to be searched as phrases
430            if(@preg_match_all('/(('.IDX_ASIAN.')+)/u',$w,$matches)){
431                $q['phrases'] = array_merge($q['phrases'],$matches[1]);
432
433            }
434            $token = idx_tokenizer($w,$stopwords,true);
435            if(count($token)){
436                $q['and']   = array_merge($q['and'],$token);
437                $q['words'] = array_merge($q['words'],$token);
438            }
439        }
440    }
441
442    return $q;
443}
444
445//Setup VIM: ex: et ts=4 enc=utf-8 :
446