xref: /dokuwiki/inc/fulltext.php (revision 4cab3ca7bcb592c65b93e21e7528f0c2a02deaf3) 
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 * Wrapper around preg_quote adding the default delimiter
15 */
16function ft_preg_quote_cb($string){
17    return preg_quote($string,'/');
18}
19
20/**
21 * The fulltext search
22 *
23 * Returns a list of matching documents for the given query
24 *
25 */
26function ft_pageSearch($query,&$poswords){
27    $q = ft_queryParser($query);
28    // use this for higlighting later:
29    $poswords = str_replace('*','',join(' ',$q['and']));
30
31    // lookup all words found in the query
32    $words  = array_merge($q['and'],$q['not']);
33    if(!count($words)) return array();
34    $result = idx_lookup($words);
35    if(!count($result)) return array();
36
37    // merge search results with query
38    foreach($q['and'] as $pos => $w){
39        $q['and'][$pos] = $result[$w];
40    }
41    // create a list of unwanted docs
42    $not = array();
43    foreach($q['not'] as $pos => $w){
44        $not = array_merge($not,array_keys($result[$w]));
45    }
46
47    // combine and-words
48    if(count($q['and']) > 1){
49        $docs = ft_resultCombine($q['and']);
50    }else{
51        $docs = $q['and'][0];
52    }
53    if(!count($docs)) return array();
54
55    // create a list of hidden pages in the result
56    $hidden = array();
57    $hidden = array_filter(array_keys($docs),'isHiddenPage');
58    $not = array_merge($not,$hidden);
59
60    // filter unmatched namespaces
61    if(!empty($q['ns'])) {
62        $pattern = implode('|^',$q['ns']);
63        foreach($docs as $key => $val) {
64            if(!preg_match('/^'.$pattern.'/',$key)) {
65                unset($docs[$key]);
66            }
67        }
68    }
69
70    // remove negative matches
71    foreach($not as $n){
72        unset($docs[$n]);
73    }
74
75    if(!count($docs)) return array();
76    // handle phrases
77    if(count($q['phrases'])){
78        //build a regexp
79        $q['phrases'] = array_map('utf8_strtolower',$q['phrases']);
80        $q['phrases'] = array_map('ft_preg_quote_cb',$q['phrases']);
81        $regex = '('.join('|',$q['phrases']).')';
82        // check the source of all documents for the exact phrases
83        foreach(array_keys($docs) as $id){
84            $text  = utf8_strtolower(rawWiki($id));
85            if(!preg_match('/'.$regex.'/usi',$text)){
86                unset($docs[$id]); // no hit - remove
87            }
88        }
89    }
90
91    if(!count($docs)) return array();
92
93    // check ACL permissions
94    foreach(array_keys($docs) as $doc){
95        if(auth_quickaclcheck($doc) < AUTH_READ){
96            unset($docs[$doc]);
97        }
98    }
99
100    if(!count($docs)) return array();
101
102    // if there are any hits left, sort them by count
103    arsort($docs);
104
105    return $docs;
106}
107
108/**
109 * Returns the backlinks for a given page
110 *
111 * Does a quick lookup with the fulltext index, then
112 * evaluates the instructions of the found pages
113 */
114function ft_backlinks($id){
115    global $conf;
116    $swfile   = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
117    $stopwords = @file_exists($swfile) ? file($swfile) : array();
118
119    $result = array();
120
121    // quick lookup of the pagename
122    $page    = noNS($id);
123    $matches = idx_lookup(idx_tokenizer($page,$stopwords));  // pagename may contain specials (_ or .)
124    $docs    = array_keys(ft_resultCombine(array_values($matches)));
125    $docs    = array_filter($docs,'isVisiblePage'); // discard hidden pages
126    if(!count($docs)) return $result;
127    require_once(DOKU_INC.'inc/parserutils.php');
128
129    // check metadata for matching links
130    foreach($docs as $match){
131        // metadata relation reference links are already resolved
132        $links = p_get_metadata($match,'relation references');
133        if (isset($links[$id])) $result[] = $match;
134    }
135
136    if(!count($result)) return $result;
137
138    // check ACL permissions
139    foreach(array_keys($result) as $idx){
140        if(auth_quickaclcheck($result[$idx]) < AUTH_READ){
141            unset($result[$idx]);
142        }
143    }
144
145    sort($result);
146    return $result;
147}
148
149/**
150 * Quicksearch for pagenames
151 *
152 * By default it only matches the pagename and ignores the
153 * namespace. This can be changed with the second parameter
154 *
155 * @author Andreas Gohr <andi@splitbrain.org>
156 */
157function ft_pageLookup($id,$pageonly=true){
158    global $conf;
159    $id    = preg_quote($id,'/');
160    $pages = file($conf['indexdir'].'/page.idx');
161    if($id) $pages = array_values(preg_grep('/'.$id.'/',$pages));
162
163    $cnt = count($pages);
164    for($i=0; $i<$cnt; $i++){
165        if($pageonly){
166            if(!preg_match('/'.$id.'/',noNS($pages[$i]))){
167                unset($pages[$i]);
168                continue;
169            }
170        }
171        if(!@file_exists(wikiFN($pages[$i]))){
172            unset($pages[$i]);
173            continue;
174        }
175    }
176
177    $pages = array_filter($pages,'isVisiblePage'); // discard hidden pages
178    if(!count($pages)) return array();
179
180    // check ACL permissions
181    foreach(array_keys($pages) as $idx){
182        if(auth_quickaclcheck($pages[$idx]) < AUTH_READ){
183            unset($pages[$idx]);
184        }
185    }
186
187    $pages = array_map('trim',$pages);
188    sort($pages);
189    return $pages;
190}
191
192/**
193 * Creates a snippet extract
194 *
195 * @author Andreas Gohr <andi@splitbrain.org>
196 */
197function ft_snippet($id,$poswords){
198    $poswords = preg_quote($poswords,'#');
199    $re       = '('.str_replace(' ','|',$poswords).')';
200    $text     = rawWiki($id);
201
202    $match = array();
203    $snippets = array();
204    $utf8_offset = $offset = $end = 0;
205    $len = utf8_strlen($text);
206
207    for ($cnt=3; $cnt--;) {
208      if (!preg_match('#'.$re.'#iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) break;
209
210      list($str,$idx) = $match[0];
211
212      // convert $idx (a byte offset) into a utf8 character offset
213      $utf8_idx = utf8_strlen(substr($text,0,$idx));
214      $utf8_len = utf8_strlen($str);
215
216      // establish context, 100 bytes surrounding the match string
217      // first look to see if we can go 100 either side,
218      // then drop to 50 adding any excess if the other side can't go to 50,
219      $pre = min($utf8_idx-$utf8_offset,100);
220      $post = min($len-$utf8_idx-$utf8_len,100);
221
222      if ($pre>50 && $post>50) {
223        $pre = $post = 50;
224      } else if ($pre>50) {
225        $pre = min($pre,100-$post);
226      } else if ($post>50) {
227        $post = min($post, 100-$pre);
228      } else {
229        // both are less than 50, means the context is the whole string
230        // make it so and break out of this loop - there is no need for the
231        // complex snippet calculations
232        $snippets = array($text);
233        break;
234      }
235
236      // establish context start and end points, try to append to previous
237      // context if possible
238      $start = $utf8_idx - $pre;
239      $append = ($start < $end) ? $end : false;  // still the end of the previous context snippet
240      $end = $utf8_idx + $utf8_len + $post;      // now set it to the end of this context
241
242      if ($append) {
243        $snippets[count($snippets)-1] .= utf8_substr($text,$append,$end-$append);
244      } else {
245        $snippets[] = utf8_substr($text,$start,$end-$start);
246      }
247
248      // set $offset for next match attempt
249      //   substract strlen to avoid splitting a potential search success,
250      //   this is an approximation as the search pattern may match strings
251      //   of varying length and it will fail if the context snippet
252      //   boundary breaks a matching string longer than the current match
253      $utf8_offset = $utf8_idx + $post;
254      $offset = $idx + strlen(utf8_substr($text,$utf8_idx,$post));
255      $offset = utf8_correctIdx($text,$offset);
256    }
257
258    $m = "\1";
259    $snippets = preg_replace('#'.$re.'#iu',$m.'$1'.$m,$snippets);
260    $snippet = preg_replace('#'.$m.'([^'.$m.']*?)'.$m.'#iu','<strong class="search_hit">$1</strong>',hsc(join('... ',$snippets)));
261
262    return $snippet;
263}
264
265/**
266 * Combine found documents and sum up their scores
267 *
268 * This function is used to combine searched words with a logical
269 * AND. Only documents available in all arrays are returned.
270 *
271 * based upon PEAR's PHP_Compat function for array_intersect_key()
272 *
273 * @param array $args An array of page arrays
274 */
275function ft_resultCombine($args){
276    $array_count = count($args);
277    if($array_count == 1){
278        return $args[0];
279    }
280
281    $result = array();
282    if ($array_count > 1) {
283      foreach ($args[0] as $key1 => $value1) {
284        for ($i = 1; $i !== $array_count; $i++) {
285            foreach ($args[$i] as $key2 => $value2) {
286                if ((string) $key1 === (string) $key2) {
287                    if(!isset($result[$key1])) $result[$key1] = $value1;
288                    $result[$key1] += $value2;
289                }
290            }
291        }
292      }
293    }
294    return $result;
295}
296
297/**
298 * Builds an array of search words from a query
299 *
300 * @todo support OR and parenthesises?
301 */
302function ft_queryParser($query){
303    global $conf;
304    $swfile   = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
305    if(@file_exists($swfile)){
306        $stopwords = file($swfile);
307    }else{
308        $stopwords = array();
309    }
310
311    $q = array();
312    $q['query']   = $query;
313    $q['ns']      = array();
314    $q['phrases'] = array();
315    $q['and']     = array();
316    $q['not']     = array();
317
318    // strip namespace from query
319    if(preg_match('/([^@]*)@(.*)/',$query,$match))  {
320        $query = $match[1];
321        $q['ns'] = explode('@',preg_replace("/ /",'',$match[2]));
322    }
323
324    // handle phrase searches
325    while(preg_match('/"(.*?)"/',$query,$match)){
326        $q['phrases'][] = $match[1];
327        $q['and'] = array_merge(idx_tokenizer($match[0],$stopwords));
328        $query = preg_replace('/"(.*?)"/','',$query,1);
329    }
330
331    $words = explode(' ',$query);
332    foreach($words as $w){
333        if($w{0} == '-'){
334            $token = idx_tokenizer($w,$stopwords,true);
335            if(count($token)) $q['not'] = array_merge($q['not'],$token);
336        }else{
337            // asian "words" need to be searched as phrases
338            if(@preg_match_all('/('.IDX_ASIAN.'+)/u',$w,$matches)){
339                $q['phrases'] = array_merge($q['phrases'],$matches[1]);
340
341            }
342            $token = idx_tokenizer($w,$stopwords,true);
343            if(count($token)) $q['and'] = array_merge($q['and'],$token);
344        }
345    }
346
347    return $q;
348}
349
350//Setup VIM: ex: et ts=4 enc=utf-8 :
351