xref: /dokuwiki/inc/fulltext.php (revision 58b6f612a14a40fc14fffa363d6f272b36cd9f30) 
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($docs)) return array();
72
73    // if there are any hits left, sort them by count
74    arsort($docs);
75
76    return $docs;
77}
78
79/**
80 * Quicksearch for pagenames
81 *
82 * By default it only matches the pagename and ignores the
83 * namespace. This can be changed with the second parameter
84 *
85 * @author Andreas Gohr <andi@splitbrain.org>
86 */
87function ft_pageLookup($id,$pageonly=true){
88    global $conf;
89    $id    = preg_quote($id,'/');
90    $pages = file($conf['cachedir'].'/page.idx');
91    $pages = array_values(preg_grep('/'.$id.'/',$pages));
92
93    $cnt = count($pages);
94    for($i=0; $i<$cnt; $i++){
95        if($pageonly){
96            if(!preg_match('/'.$id.'/',noNS($pages[$i]))){
97                unset($pages[$i]);
98                continue;
99            }
100        }
101        if(!@file_exists(wikiFN($pages[$i]))){
102            unset($pages[$i]);
103            continue;
104        }
105    }
106    sort($pages);
107    return $pages;
108}
109
110/**
111 * Creates a snippet extract
112 *
113 * @author Andreas Gohr <andi@splitbrain.org>
114 */
115function ft_snippet($id,$poswords){
116    $poswords = preg_quote($poswords,'#');
117    $re       = '('.str_replace(' ','|',$poswords).')';
118    $text     = rawWiki($id);
119    //FIXME caseinsensitive matching doesn't work with UTF-8!?
120    preg_match_all('#(.{0,50})'.$re.'(.{0,50})#iu',$text,$matches,PREG_SET_ORDER);
121
122    $cnt = 0;
123    $snippet = '';
124    foreach($matches as $match){
125        $snippet .= '...'.htmlspecialchars($match[1]);
126        $snippet .= '<span class="search_hit">';
127        $snippet .= htmlspecialchars($match[2]);
128        $snippet .= '</span>';
129        $snippet .= htmlspecialchars($match[3]).'... ';
130        if($cnt++ == 2) break;
131    }
132
133    return $snippet;
134}
135
136/**
137 * Combine found documents and sum up their scores
138 *
139 * This function is used to combine searched words with a logical
140 * AND. Only documents available in all arrays are returned.
141 *
142 * based upon PEAR's PHP_Compat function for array_intersect_key()
143 *
144 * @param array $args An array of page arrays
145 */
146function ft_resultCombine($args){
147    $array_count = count($args);
148    $result = array();
149    foreach ($args[0] as $key1 => $value1) {
150        for ($i = 1; $i !== $array_count; $i++) {
151            foreach ($args[$i] as $key2 => $value2) {
152                if ((string) $key1 === (string) $key2) {
153                    if(!isset($result[$key1])) $result[$key1] = $value1;
154                    $result[$key1] += $value2;
155                }
156            }
157        }
158    }
159    return $result;
160}
161
162/**
163 * Builds an array of search words from a query
164 *
165 * @todo support OR and parenthesises?
166 */
167function ft_queryParser($query){
168    global $conf;
169    $swfile   = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
170    if(@file_exists($swfile)){
171        $stopwords = file($swfile);
172    }else{
173        $stopwords = array();
174    }
175
176    $q = array();
177    $q['query']   = $query;
178    $q['phrases'] = array();
179    $q['and']     = array();
180    $q['not']     = array();
181
182    // handle phrase searches
183    while(preg_match('/"(.*?)"/',$query,$match)){
184        $q['phrases'][] = $match[0];
185        $q['and'] = array_merge(idx_tokenizer($match[0],$stopwords));
186        $query = preg_replace('/"(.*?)"/','',$query,1);
187    }
188
189    $words = explode(' ',$query);
190    foreach($words as $w){
191        if($w{0} == '-'){
192            $token = idx_tokenizer($w,$stopwords);
193            if(count($token)) $q['not'] = array_merge($q['not'],$token);
194        }else{
195            $token = idx_tokenizer($w,$stopwords);
196            if(count($token)) $q['and'] = array_merge($q['and'],$token);
197        }
198    }
199
200    return $q;
201}
202
203//Setup VIM: ex: et ts=4 enc=utf-8 :
204