xref: /dokuwiki/inc/fulltext.php (revision fd49f8df8d347af8d355ec7b8cb2261a17e2ae5c) 
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',realpath(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 */
19function ft_pageSearch($query,&$poswords){
20    $q = ft_queryParser($query);
21    // use this for higlighting later:
22    $poswords = str_replace('*','',join(' ',$q['and']));
23
24    // lookup all words found in the query
25    $words  = array_merge($q['and'],$q['not']);
26    if(!count($words)) return array();
27    $result = idx_lookup($words);
28
29    // merge search results with query
30    foreach($q['and'] as $pos => $w){
31        $q['and'][$pos] = $result[$w];
32    }
33    // create a list of unwanted docs
34    $not = array();
35    foreach($q['not'] as $pos => $w){
36        $not = array_merge($not,array_keys($result[$w]));
37    }
38
39    // combine and-words
40    if(count($q['and']) > 1){
41        $docs = ft_resultCombine($q['and']);
42    }else{
43        $docs = $q['and'][0];
44    }
45    if(!count($docs)) return array();
46
47    // create a list of hidden pages in the result
48    $hidden = array();
49    $hidden = array_filter(array_keys($docs),'isHiddenPage');
50    $not = array_merge($not,$hidden);
51
52    // filter unmatched namespaces
53    if(!empty($q['ns'])) {
54        $pattern = implode('|^',$q['ns']);
55        foreach($docs as $key => $val) {
56            if(!preg_match('/^'.$pattern.'/',$key)) {
57                unset($docs[$key]);
58            }
59        }
60    }
61
62    // remove negative matches
63    foreach($not as $n){
64        unset($docs[$n]);
65    }
66
67    if(!count($docs)) return array();
68    // handle phrases
69    if(count($q['phrases'])){
70        //build a regexp
71        $q['phrases'] = array_map('utf8_strtolower',$q['phrases']);
72        $q['phrases'] = array_map('preg_quote',$q['phrases']);
73        $regex = '('.join('|',$q['phrases']).')';
74        // check the source of all documents for the exact phrases
75        foreach(array_keys($docs) as $id){
76            $text  = utf8_strtolower(rawWiki($id));
77            if(!preg_match('/'.$regex.'/usi',$text)){
78                unset($docs[$id]); // no hit - remove
79            }
80        }
81    }
82
83    if(!count($docs)) return array();
84
85    // check ACL permissions
86    foreach(array_keys($docs) as $doc){
87        if(auth_quickaclcheck($doc) < AUTH_READ){
88            unset($docs[$doc]);
89        }
90    }
91
92    if(!count($docs)) return array();
93
94    // if there are any hits left, sort them by count
95    arsort($docs);
96
97    return $docs;
98}
99
100/**
101 * Returns the backlinks for a given page
102 *
103 * Does a quick lookup with the fulltext index, then
104 * evaluates the instructions of the found pages
105 */
106function ft_backlinks($id){
107    global $conf;
108    $result = array();
109
110    // quick lookup of the pagename
111    $page    = noNS($id);
112    $sw      = array(); // we don't use stopwords here
113    $matches = idx_lookup(idx_tokenizer($page,$sw));  // pagename may contain specials (_ or .)
114    $docs    = array_keys(ft_resultCombine(array_values($matches)));
115    $docs    = array_filter($docs,'isVisiblePage'); // discard hidden pages
116    if(!count($docs)) return $result;
117    require_once(DOKU_INC.'inc/parserutils.php');
118
119    // check instructions for matching links
120    foreach($docs as $match){
121        $instructions = p_cached_instructions(wikiFN($match),true);
122        if(is_null($instructions)) continue;
123
124        $match_ns =  getNS($match);
125
126        foreach($instructions as $ins){
127            if($ins[0] == 'internallink' || ($conf['camelcase'] && $ins[0] == 'camelcaselink') ){
128                $link = $ins[1][0];
129                resolve_pageid($match_ns,$link,$exists); //exists is not used
130                if($link == $id){
131                    //we have a match - finish
132                    $result[] = $match;
133                    break;
134                }
135            }
136        }
137    }
138
139    if(!count($result)) return $result;
140
141    // check ACL permissions
142    foreach(array_keys($result) as $idx){
143        if(auth_quickaclcheck($result[$idx]) < AUTH_READ){
144            unset($result[$idx]);
145        }
146    }
147
148    sort($result);
149    return $result;
150}
151
152/**
153 * Quicksearch for pagenames
154 *
155 * By default it only matches the pagename and ignores the
156 * namespace. This can be changed with the second parameter
157 *
158 * @author Andreas Gohr <andi@splitbrain.org>
159 */
160function ft_pageLookup($id,$pageonly=true){
161    global $conf;
162    $id    = preg_quote($id,'/');
163    $pages = file($conf['cachedir'].'/page.idx');
164    $pages = array_values(preg_grep('/'.$id.'/',$pages));
165
166    $cnt = count($pages);
167    for($i=0; $i<$cnt; $i++){
168        if($pageonly){
169            if(!preg_match('/'.$id.'/',noNS($pages[$i]))){
170                unset($pages[$i]);
171                continue;
172            }
173        }
174        if(!@file_exists(wikiFN($pages[$i]))){
175            unset($pages[$i]);
176            continue;
177        }
178    }
179
180    $pages = array_filter($pages,'isVisiblePage'); // discard hidden pages
181    if(!count($pages)) return array();
182
183    // check ACL permissions
184    foreach(array_keys($pages) as $idx){
185        if(auth_quickaclcheck($pages[$idx]) < AUTH_READ){
186            unset($pages[$idx]);
187        }
188    }
189
190    sort($pages);
191    return $pages;
192}
193
194/**
195 * Creates a snippet extract
196 *
197 * @author Andreas Gohr <andi@splitbrain.org>
198 */
199function ft_snippet($id,$poswords){
200    $poswords = preg_quote($poswords,'#');
201    $re       = '('.str_replace(' ','|',$poswords).')';
202    $text     = rawWiki($id);
203
204// extra code to allow selection of search algorithm - remove before release
205global $conf;
206$algorithm = '';
207if ($conf['allowdebug']) {
208  if (!empty($_REQUEST['_search'])) $algorithm = $_REQUEST['_search'];
209}
210
211switch ($algorithm) {
212  case 'orig' :
213// original code ... dokuwiki
214
215    //FIXME caseinsensitive matching doesn't work with UTF-8!?
216    preg_match_all('#(.{0,50})'.$re.'(.{0,50})#iu',$text,$matches,PREG_SET_ORDER);
217
218    $cnt = 0;
219    $snippet = '';
220    foreach($matches as $match){
221        $snippet .= '...'.htmlspecialchars($match[1]);
222        $snippet .= '<span class="search_hit">';
223        $snippet .= htmlspecialchars($match[2]);
224        $snippet .= '</span>';
225        $snippet .= htmlspecialchars($match[3]).'... ';
226        if($cnt++ == 2) break;
227    }
228
229  break;
230
231  case 'opt1' :
232// my snippet algorithm, first cut ... CS 2006-08-25
233// reduce the impact of the original regex
234    $matches = array();
235    preg_match_all('#'.$re.'#iu',$text,$matches,PREG_OFFSET_CAPTURE|PREG_SET_ORDER);
236
237    $cnt = 3;
238    $snippets = array();
239    $len = strlen($text);
240    foreach ($matches as $match) {
241      list($str,$idx) = $match[0];
242      if ($idx < $end) continue;
243
244      $pre = min($idx,50);
245      $start = $idx - $pre;
246      $end = min($idx+100+strlen($str)-$pre,$len);
247      $snippets[] = substr($text,$start,$end-$start);
248      if (!($cnt--)) break;
249    }
250
251    $m = "\1";
252    $snippets = preg_replace('#'.$re.'#iu',$m.'$1'.$m,$snippets);
253    $snippet = preg_replace('#'.$m.'([^'.$m.']*?)'.$m.'#iu','<span class="search_hit">$1</span>',hsc(join('... ',$snippets)));
254
255  break;
256
257  case 'opt2' :
258	default :
259// option 2 ... CS 2006-08-25
260// above + reduce amount of the file searched
261    $match = array();
262    $snippets = array();
263    $offset = 0;
264    $len = strlen($text);
265    for ($cnt=3; $cnt--;) {
266      if (!preg_match('#'.$re.'#iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) break;
267
268      list($str,$idx) = $match[0];
269
270      // establish context, 100 characters surrounding the match string
271      // first look to see if we can go 100 either side,
272      // then drop to 50 adding any excess if the other side can't go to 50.
273      $pre = min($idx-$offset,100);
274      $post = min($len-$idx-strlen($str),100);
275
276      if ($pre>50 && $post>50) {
277        $pre = $post = 50;
278      } else if ($pre>50) {
279        $pre = min($pre,100-$post);
280      } else if ($post>50) {
281        $post = min($post, 100-$pre);
282      }
283
284      // establish context start and end points, try to append to previous context if possible
285      $start = $idx - $pre;
286      $append = ($start < $end) ? $end : false;   // still the end of the previous context snippet
287      $end = $idx + strlen($str) + $post;         // now set it to the end of this context
288
289      if ($append) {
290        $snippets[count($snippets)-1] .= substr($text,$append,$end-$append);
291      } else {
292        $snippets[] = substr($text,$start,$end-$start);
293      }
294
295      // set $offset for next match attempt
296      //   substract strlen to avoid splitting a potential search success, this is an approximation as the
297      //   search pattern may match strings of varying length and it will fail if the context snippet
298      //   boundary breaks a matching string longer than the current match
299      $offset = $end - strlen($str);
300    }
301    $m = "\1";
302    $snippets = preg_replace('#'.$re.'#iu',$m.'$1'.$m,$snippets);
303    $snippet = preg_replace('#'.$m.'([^'.$m.']*?)'.$m.'#iu','<span class="search_hit">$1</span>',hsc(join('... ',$snippets)));
304
305  break;
306}
307
308    return $snippet;
309}
310
311/**
312 * Combine found documents and sum up their scores
313 *
314 * This function is used to combine searched words with a logical
315 * AND. Only documents available in all arrays are returned.
316 *
317 * based upon PEAR's PHP_Compat function for array_intersect_key()
318 *
319 * @param array $args An array of page arrays
320 */
321function ft_resultCombine($args){
322    $array_count = count($args);
323    if($array_count == 1){
324        return $args[0];
325    }
326
327    $result = array();
328    foreach ($args[0] as $key1 => $value1) {
329        for ($i = 1; $i !== $array_count; $i++) {
330            foreach ($args[$i] as $key2 => $value2) {
331                if ((string) $key1 === (string) $key2) {
332                    if(!isset($result[$key1])) $result[$key1] = $value1;
333                    $result[$key1] += $value2;
334                }
335            }
336        }
337    }
338    return $result;
339}
340
341/**
342 * Builds an array of search words from a query
343 *
344 * @todo support OR and parenthesises?
345 * @todo add namespace handling
346 */
347function ft_queryParser($query){
348    global $conf;
349    $swfile   = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
350    if(@file_exists($swfile)){
351        $stopwords = file($swfile);
352    }else{
353        $stopwords = array();
354    }
355
356    $q = array();
357    $q['query']   = $query;
358    $q['ns']      = array();
359    $q['phrases'] = array();
360    $q['and']     = array();
361    $q['not']     = array();
362
363    // strip namespace from query
364    if(preg_match('/([^@]*)@(.*)/',$query,$match))  {
365        $query = $match[1];
366        $q['ns'] = explode('@',preg_replace("/ /",'',$match[2]));
367    }
368
369    // handle phrase searches
370    while(preg_match('/"(.*?)"/',$query,$match)){
371        $q['phrases'][] = $match[1];
372        $q['and'] = array_merge(idx_tokenizer($match[0],$stopwords));
373        $query = preg_replace('/"(.*?)"/','',$query,1);
374    }
375
376    $words = explode(' ',$query);
377    foreach($words as $w){
378        if($w{0} == '-'){
379            $token = idx_tokenizer($w,$stopwords,true);
380            if(count($token)) $q['not'] = array_merge($q['not'],$token);
381        }else{
382            // asian "words" need to be searched as phrases
383            if(@preg_match_all('/('.IDX_ASIAN.'+)/u',$w,$matches)){
384                $q['phrases'] = array_merge($q['phrases'],$matches[1]);
385
386            }
387            $token = idx_tokenizer($w,$stopwords,true);
388            if(count($token)) $q['and'] = array_merge($q['and'],$token);
389        }
390    }
391
392    return $q;
393}
394
395//Setup VIM: ex: et ts=4 enc=utf-8 :
396