xref: /dokuwiki/inc/fulltext.php (revision 886f80e7a9d4651ee3524837f0f5e430687ba6a5) 
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 $result;
117    require_once(DOKU_INC.'inc/parserutils.php');
118
119    // check metadata for matching links
120    foreach($docs as $match){
121        // metadata relation reference links are already resolved
122        $links = p_get_metadata($match,"relation references");
123        if (isset($links[$id])) $result[] = $match;
124    }
125
126    if(!count($result)) return $result;
127
128    // check ACL permissions
129    foreach(array_keys($result) as $idx){
130        if(auth_quickaclcheck($result[$idx]) < AUTH_READ){
131            unset($result[$idx]);
132        }
133    }
134
135    sort($result);
136    return $result;
137}
138
139/**
140 * Quicksearch for pagenames
141 *
142 * By default it only matches the pagename and ignores the
143 * namespace. This can be changed with the second parameter
144 *
145 * @author Andreas Gohr <andi@splitbrain.org>
146 */
147function ft_pageLookup($id,$pageonly=true){
148    global $conf;
149    $id    = preg_quote($id,'/');
150    $pages = file($conf['cachedir'].'/page.idx');
151    $pages = array_values(preg_grep('/'.$id.'/',$pages));
152
153    $cnt = count($pages);
154    for($i=0; $i<$cnt; $i++){
155        if($pageonly){
156            if(!preg_match('/'.$id.'/',noNS($pages[$i]))){
157                unset($pages[$i]);
158                continue;
159            }
160        }
161        if(!@file_exists(wikiFN($pages[$i]))){
162            unset($pages[$i]);
163            continue;
164        }
165    }
166
167    $pages = array_filter($pages,'isVisiblePage'); // discard hidden pages
168    if(!count($pages)) return array();
169
170    // check ACL permissions
171    foreach(array_keys($pages) as $idx){
172        if(auth_quickaclcheck($pages[$idx]) < AUTH_READ){
173            unset($pages[$idx]);
174        }
175    }
176
177    sort($pages);
178    return $pages;
179}
180
181/**
182 * Creates a snippet extract
183 *
184 * @author Andreas Gohr <andi@splitbrain.org>
185 */
186function ft_snippet($id,$poswords){
187    $poswords = preg_quote($poswords,'#');
188    $re       = '('.str_replace(' ','|',$poswords).')';
189    $text     = rawWiki($id);
190
191    $match = array();
192    $snippets = array();
193    $utf8_offset = $offset = $end = 0;
194    $len = utf8_strlen($text);
195
196    for ($cnt=3; $cnt--;) {
197      if (!preg_match('#'.$re.'#iu',$text,$match,PREG_OFFSET_CAPTURE,$offset)) break;
198
199      list($str,$idx) = $match[0];
200
201      // convert $idx (a byte offset) into a utf8 character offset
202      $utf8_idx = utf8_strlen(substr($text,0,$idx));
203      $utf8_len = utf8_strlen($str);
204
205      // establish context, 100 bytes surrounding the match string
206      // first look to see if we can go 100 either side,
207      // then drop to 50 adding any excess if the other side can't go to 50,
208      $pre = min($utf8_idx-$utf8_offset,100);
209      $post = min($len-$utf8_idx-$utf8_len,100);
210
211      if ($pre>50 && $post>50) {
212        $pre = $post = 50;
213      } else if ($pre>50) {
214        $pre = min($pre,100-$post);
215      } else if ($post>50) {
216        $post = min($post, 100-$pre);
217      } else {
218        // both are less than 50, means the context is the whole string
219        // make it so and break out of this loop - there is no need for the
220        // complex snippet calculations
221        $snippets = array($text);
222        break;
223      }
224
225      // establish context start and end points, try to append to previous
226      // context if possible
227      $start = $utf8_idx - $pre;
228      $append = ($start < $end) ? $end : false;  // still the end of the previous context snippet
229      $end = $utf8_idx + $utf8_len + $post;      // now set it to the end of this context
230
231      if ($append) {
232        $snippets[count($snippets)-1] .= utf8_substr($text,$append,$end-$append);
233      } else {
234        $snippets[] = utf8_substr($text,$start,$end-$start);
235      }
236
237      // set $offset for next match attempt
238      //   substract strlen to avoid splitting a potential search success,
239      //   this is an approximation as the search pattern may match strings
240      //   of varying length and it will fail if the context snippet
241      //   boundary breaks a matching string longer than the current match
242      $utf8_offset = $utf8_idx + $post;
243      $offset = $idx + strlen(utf8_substr($text,$utf8_idx,$post));
244      $offset = utf8_correctIdx($text,$offset);
245    }
246
247    $m = "\1";
248    $snippets = preg_replace('#'.$re.'#iu',$m.'$1'.$m,$snippets);
249    $snippet = preg_replace('#'.$m.'([^'.$m.']*?)'.$m.'#iu','<span class="search_hit">$1</span>',hsc(join('... ',$snippets)));
250
251    return $snippet;
252}
253
254/**
255 * Combine found documents and sum up their scores
256 *
257 * This function is used to combine searched words with a logical
258 * AND. Only documents available in all arrays are returned.
259 *
260 * based upon PEAR's PHP_Compat function for array_intersect_key()
261 *
262 * @param array $args An array of page arrays
263 */
264function ft_resultCombine($args){
265    $array_count = count($args);
266    if($array_count == 1){
267        return $args[0];
268    }
269
270    $result = array();
271    foreach ($args[0] as $key1 => $value1) {
272        for ($i = 1; $i !== $array_count; $i++) {
273            foreach ($args[$i] as $key2 => $value2) {
274                if ((string) $key1 === (string) $key2) {
275                    if(!isset($result[$key1])) $result[$key1] = $value1;
276                    $result[$key1] += $value2;
277                }
278            }
279        }
280    }
281    return $result;
282}
283
284/**
285 * Builds an array of search words from a query
286 *
287 * @todo support OR and parenthesises?
288 */
289function ft_queryParser($query){
290    global $conf;
291    $swfile   = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
292    if(@file_exists($swfile)){
293        $stopwords = file($swfile);
294    }else{
295        $stopwords = array();
296    }
297
298    $q = array();
299    $q['query']   = $query;
300    $q['ns']      = array();
301    $q['phrases'] = array();
302    $q['and']     = array();
303    $q['not']     = array();
304
305    // strip namespace from query
306    if(preg_match('/([^@]*)@(.*)/',$query,$match))  {
307        $query = $match[1];
308        $q['ns'] = explode('@',preg_replace("/ /",'',$match[2]));
309    }
310
311    // handle phrase searches
312    while(preg_match('/"(.*?)"/',$query,$match)){
313        $q['phrases'][] = $match[1];
314        $q['and'] = array_merge(idx_tokenizer($match[0],$stopwords));
315        $query = preg_replace('/"(.*?)"/','',$query,1);
316    }
317
318    $words = explode(' ',$query);
319    foreach($words as $w){
320        if($w{0} == '-'){
321            $token = idx_tokenizer($w,$stopwords,true);
322            if(count($token)) $q['not'] = array_merge($q['not'],$token);
323        }else{
324            // asian "words" need to be searched as phrases
325            if(@preg_match_all('/('.IDX_ASIAN.'+)/u',$w,$matches)){
326                $q['phrases'] = array_merge($q['phrases'],$matches[1]);
327
328            }
329            $token = idx_tokenizer($w,$stopwords,true);
330            if(count($token)) $q['and'] = array_merge($q['and'],$token);
331        }
332    }
333
334    return $q;
335}
336
337//Setup VIM: ex: et ts=4 enc=utf-8 :
338