xref: /dokuwiki/inc/fulltext.php (revision 0cb422a21416da0999421c4e3697cc074157d7a4) 
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
22    // use this for higlighting later:
23    $poswords = join(' ',$q['and']);
24
25    // lookup all words found in the query
26    $words  = array_merge($q['and'],$q['not']);
27    if(!count($words)) return array();
28    $result = idx_lookup($words);
29
30    // merge search results with query
31    foreach($q['and'] as $pos => $w){
32        $q['and'][$pos] = $result[$w];
33    }
34    // create a list of unwanted docs
35    $not = array();
36    foreach($q['not'] as $pos => $w){
37        $not = array_merge($not,array_keys($result[$w]));
38    }
39
40    // combine and-words
41    if(count($q['and']) > 1){
42        $docs = ft_resultCombine($q['and']);
43    }else{
44        $docs = $q['and'][0];
45    }
46    if(!count($docs)) return array();
47
48    // remove negative matches
49    foreach($not as $n){
50        unset($docs[$n]);
51    }
52
53    if(!count($docs)) return array();
54
55    // handle phrases
56    if(count($q['phrases'])){
57        //build a regexp
58        $q['phrases'] = array_map('utf8_strtolower',$q['phrases']);
59        $q['phrases'] = array_map('preg_quote',$q['phrases']);
60        $regex = '('.join('|',$q['phrases']).')';
61
62        // check the source of all documents for the exact phrases
63        foreach(array_keys($docs) as $id){
64            $text  = utf8_strtolower(rawWiki($id));
65            if(!preg_match('/'.$regex.'/usi',$text)){
66                unset($docs[$id]); // no hit - remove
67            }
68        }
69    }
70
71    if(!count($docs)) return array();
72
73    // check ACL permissions
74    foreach(array_keys($docs) as $doc){
75        if(auth_quickaclcheck($doc) < AUTH_READ){
76            unset($docs[$doc]);
77        }
78    }
79
80    if(!count($docs)) return array();
81
82    // if there are any hits left, sort them by count
83    arsort($docs);
84
85    return $docs;
86}
87
88/**
89 * Returns the backlinks for a given page
90 *
91 * Does a quick lookup with the fulltext index, then
92 * evaluates the instructions of the found pages
93 */
94function ft_backlinks($id){
95    global $conf;
96    $result = array();
97
98    // quick lookup of the pagename
99    $page    = noNS($id);
100    $matches = idx_lookup(array($page));
101
102    if(!count($matches)) return $result;
103    require_once(DOKU_INC.'inc/parserutils.php');
104
105
106    // check instructions for matching links
107    foreach(array_keys($matches[$page]) as $match){
108        $instructions = p_cached_instructions(wikiFN($match),true);
109        if(is_null($instructions)) continue;
110
111        $match_ns =  getNS($match);
112
113        foreach($instructions as $ins){
114            if($ins[0] == 'internallink' || ($conf['camelcase'] && $ins[0] == 'camelcaselink') ){
115                $link = $ins[1][0];
116                resolve_pageid($match_ns,$link,$exists); //exists is not used
117                if($link == $id){
118                    //we have a match - finish
119                    $result[] = $match;
120                    break;
121                }
122            }
123        }
124    }
125
126    if(!count($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[$i]);
163            continue;
164        }
165    }
166
167    if(!count($pages)) return array();
168
169    // check ACL permissions
170    foreach(array_keys($pages) as $idx){
171        if(auth_quickaclcheck($pages[$idx]) < AUTH_READ){
172            unset($pages[$idx]);
173        }
174    }
175
176    sort($pages);
177    return $pages;
178}
179
180/**
181 * Creates a snippet extract
182 *
183 * @author Andreas Gohr <andi@splitbrain.org>
184 */
185function ft_snippet($id,$poswords){
186    $poswords = preg_quote($poswords,'#');
187    $re       = '('.str_replace(' ','|',$poswords).')';
188    $text     = rawWiki($id);
189    //FIXME caseinsensitive matching doesn't work with UTF-8!?
190    preg_match_all('#(.{0,50})'.$re.'(.{0,50})#iu',$text,$matches,PREG_SET_ORDER);
191
192    $cnt = 0;
193    $snippet = '';
194    foreach($matches as $match){
195        $snippet .= '...'.htmlspecialchars($match[1]);
196        $snippet .= '<span class="search_hit">';
197        $snippet .= htmlspecialchars($match[2]);
198        $snippet .= '</span>';
199        $snippet .= htmlspecialchars($match[3]).'... ';
200        if($cnt++ == 2) break;
201    }
202
203    return $snippet;
204}
205
206/**
207 * Combine found documents and sum up their scores
208 *
209 * This function is used to combine searched words with a logical
210 * AND. Only documents available in all arrays are returned.
211 *
212 * based upon PEAR's PHP_Compat function for array_intersect_key()
213 *
214 * @param array $args An array of page arrays
215 */
216function ft_resultCombine($args){
217    $array_count = count($args);
218    $result = array();
219    foreach ($args[0] as $key1 => $value1) {
220        for ($i = 1; $i !== $array_count; $i++) {
221            foreach ($args[$i] as $key2 => $value2) {
222                if ((string) $key1 === (string) $key2) {
223                    if(!isset($result[$key1])) $result[$key1] = $value1;
224                    $result[$key1] += $value2;
225                }
226            }
227        }
228    }
229    return $result;
230}
231
232/**
233 * Builds an array of search words from a query
234 *
235 * @todo support OR and parenthesises?
236 */
237function ft_queryParser($query){
238    global $conf;
239    $swfile   = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
240    if(@file_exists($swfile)){
241        $stopwords = file($swfile);
242    }else{
243        $stopwords = array();
244    }
245
246    $q = array();
247    $q['query']   = $query;
248    $q['phrases'] = array();
249    $q['and']     = array();
250    $q['not']     = array();
251
252    // handle phrase searches
253    while(preg_match('/"(.*?)"/',$query,$match)){
254        $q['phrases'][] = $match[0];
255        $q['and'] = array_merge(idx_tokenizer($match[0],$stopwords));
256        $query = preg_replace('/"(.*?)"/','',$query,1);
257    }
258
259    $words = explode(' ',$query);
260    foreach($words as $w){
261        if($w{0} == '-'){
262            $token = idx_tokenizer($w,$stopwords);
263            if(count($token)) $q['not'] = array_merge($q['not'],$token);
264        }else{
265            $token = idx_tokenizer($w,$stopwords);
266            if(count($token)) $q['and'] = array_merge($q['and'],$token);
267        }
268    }
269
270    return $q;
271}
272
273//Setup VIM: ex: et ts=4 enc=utf-8 :
274