xref: /dokuwiki/inc/Search/FulltextSearch.php (revision 5e9d26e3624fd22ca3c57447be9e16d0b502761e) 
1<?php
2
3namespace dokuwiki\Search;
4
5use dokuwiki\Extension\Event;
6use dokuwiki\Search\Collection\CollectionSearch;
7use dokuwiki\Search\Collection\PageFulltextCollection;
8use dokuwiki\Search\Exception\SearchException;
9use dokuwiki\Search\Query\QueryEvaluator;
10use dokuwiki\Search\Query\QueryParser;
11use dokuwiki\Utf8;
12
13/**
14 * DokuWiki Fulltext Search
15 *
16 * @license    GPL 2 (http://www.gnu.org/licenses/gpl.html)
17 * @author     Andreas Gohr <andi@splitbrain.org>
18 */
19class FulltextSearch
20{
21    /** @var int Maximum number of results to generate snippets for */
22    protected int $maxSnippets = 15;
23
24    /**
25     * @return int
26     */
27    public function getMaxSnippets(): int
28    {
29        return $this->maxSnippets;
30    }
31
32    /**
33     * @param int $maxSnippets
34     */
35    public function setMaxSnippets(int $maxSnippets): void
36    {
37        $this->maxSnippets = $maxSnippets;
38    }
39
40    /**
41     * The fulltext search
42     *
43     * Returns a list of matching documents for the given query
44     *
45     * @triggers SEARCH_QUERY_FULLPAGE
46     *
47     * @param string     $query   the search query string
48     * @param array      $highlight  will be filled with terms to highlight
49     * @param string     $sort    sort mode: 'hits' (default) or 'mtime'
50     * @param int|string $after   only show results with mtime after this date,
51     *                            accepts timestamp or strtotime arguments
52     * @param int|string $before  only show results with mtime before this date,
53     *                            accepts timestamp or strtotime arguments
54     *
55     * @return array matching documents as pageid => score
56     */
57    public function pageSearch($query, &$highlight, $sort = null, $after = null, $before = null)
58    {
59        if ($sort === null) {
60            $sort = 'hits';
61        }
62        $data = [
63            'query' => $query,
64            'sort' => $sort,
65            'after' => $after,
66            'before' => $before
67        ];
68        $data['highlight'] =& $highlight;
69        $action = [$this, 'pageSearchCallBack'];
70        return Event::createAndTrigger('SEARCH_QUERY_FULLPAGE', $data, $action);
71    }
72
73    /**
74     * Returns a list of matching documents for the given query
75     *
76     * @author Andreas Gohr <andi@splitbrain.org>
77     * @author Kazutaka Miyasaka <kazmiya@gmail.com>
78     *
79     * @param array $data  event data
80     * @return array       matching documents as pageid => score
81     */
82    public function pageSearchCallBack(&$data)
83    {
84        // parse the given query
85        $q = (new QueryParser)->convert($data['query']);
86        $data['highlight'] = $q['highlight'];
87
88        if (empty($q['parsed_ary'])) return [];
89
90        // look up all words via CollectionSearch
91        $collection = new PageFulltextCollection();
92        $search = new CollectionSearch($collection);
93        foreach ($q['words'] as $word) {
94            try {
95                $search->addTerm($word);
96            } catch (SearchException $e) {
97                // term too short or invalid, skip
98            }
99        }
100        $terms = $search->execute();
101
102        // evaluate the query
103        $evaluator = new QueryEvaluator($q['parsed_ary'], $terms);
104        $docs = $evaluator->evaluate();
105
106        if (empty($docs)) return [];
107
108        // filter by visibility, acls, existence, and time range
109        $docs = MetadataSearch::filterPages($docs, false, $data['after'], $data['before']);
110
111        if ($data['sort'] === 'mtime') {
112            uksort($docs, static function ($a, $b) {
113                return filemtime(wikiFN($b)) - filemtime(wikiFN($a));
114            });
115        } else {
116            arsort($docs);
117        }
118
119        return $docs;
120    }
121
122    /**
123     * Creates a snippet extract
124     *
125     * @author Andreas Gohr <andi@splitbrain.org>
126     * @triggers FULLTEXT_SNIPPET_CREATE
127     *
128     * @param string $id page id
129     * @param array $highlight
130     * @return mixed
131     */
132    public function snippet($id, $highlight)
133    {
134        $text = rawWiki($id);
135        $text = str_replace("\xC2\xAD",'',$text); // remove soft-hyphens
136        $evdata = array(
137            'id'        => $id,
138            'text'      => &$text,
139            'highlight' => &$highlight,
140            'snippet'   => '',
141        );
142
143        $evt = new Event('FULLTEXT_SNIPPET_CREATE', $evdata);
144        if ($evt->advise_before()) {
145            $match = array();
146            $snippets = array();
147            $utf8_offset = $offset = $end = 0;
148            $len = Utf8\PhpString::strlen($text);
149
150            // build a regexp from the phrases to highlight
151            $re1 = '(' .
152                join(
153                    '|',
154                    array_map(
155                        [$this, 'snippetRePreprocess'],
156                        array_map(
157                            'preg_quote_cb',
158                            array_filter((array) $highlight)
159                        )
160                    )
161                ) .
162                ')';
163            $re2 = "$re1.{0,75}(?!\\1)$re1";
164            $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1";
165
166            for ($cnt=4; $cnt--;) {
167                if (0) {
168                } elseif (preg_match('/'.$re3.'/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) {
169                } elseif (preg_match('/'.$re2.'/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) {
170                } elseif (preg_match('/'.$re1.'/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) {
171                } else {
172                    break;
173                }
174
175                list($str, $idx) = $match[0];
176
177                // convert $idx (a byte offset) into a utf8 character offset
178                $utf8_idx = Utf8\PhpString::strlen(substr($text, 0, $idx));
179                $utf8_len = Utf8\PhpString::strlen($str);
180
181                // establish context, 100 bytes surrounding the match string
182                // first look to see if we can go 100 either side,
183                // then drop to 50 adding any excess if the other side can't go to 50,
184                $pre = min($utf8_idx - $utf8_offset, 100);
185                $post = min($len - $utf8_idx - $utf8_len, 100);
186
187                if ($pre > 50 && $post > 50) {
188                    $pre = $post = 50;
189                } elseif ($pre > 50) {
190                    $pre = min($pre, 100 - $post);
191                } elseif ($post > 50) {
192                    $post = min($post, 100 - $pre);
193                } elseif ($offset == 0) {
194                    // both are less than 50, means the context is the whole string
195                    // make it so and break out of this loop - there is no need for the
196                    // complex snippet calculations
197                    $snippets = array($text);
198                    break;
199                }
200
201                // establish context start and end points, try to append to previous
202                // context if possible
203                $start = $utf8_idx - $pre;
204                $append = ($start < $end) ? $end : false;  // still the end of the previous context snippet
205                $end = $utf8_idx + $utf8_len + $post;      // now set it to the end of this context
206
207                if ($append) {
208                    $snippets[count($snippets)-1] .= Utf8\PhpString::substr($text, $append, $end-$append);
209                } else {
210                    $snippets[] = Utf8\PhpString::substr($text, $start, $end-$start);
211                }
212
213                // set $offset for next match attempt
214                // continue matching after the current match
215                // if the current match is not the longest possible match starting at the current offset
216                // this prevents further matching of this snippet but for possible matches of length
217                // smaller than match length + context (at least 50 characters) this match is part of the context
218                $utf8_offset = $utf8_idx + $utf8_len;
219                $offset = $idx + strlen(Utf8\PhpString::substr($text, $utf8_idx, $utf8_len));
220                $offset = Utf8\Clean::correctIdx($text, $offset);
221            }
222
223            $m = "\1";
224            $snippets = preg_replace('/'.$re1.'/iu', $m.'$1'.$m, $snippets);
225            $snippet = preg_replace(
226                '/' . $m . '([^' . $m . ']*?)' . $m . '/iu',
227                '<strong class="search_hit">$1</strong>',
228                hsc(join('... ', $snippets))
229            );
230
231            $evdata['snippet'] = $snippet;
232        }
233        $evt->advise_after();
234        unset($evt);
235
236        return $evdata['snippet'];
237    }
238
239    /**
240     * Wraps a search term in regex boundary checks.
241     *
242     * @param string $term
243     * @return string
244     */
245    public function snippetRePreprocess($term)
246    {
247        // do not process asian terms where word boundaries are not explicit
248        if (Utf8\Asian::isAsianWords($term)) return $term;
249
250        if (UTF8_PROPERTYSUPPORT) {
251            // unicode word boundaries
252            // see http://stackoverflow.com/a/2449017/172068
253            $BL = '(?<!\pL)';
254            $BR = '(?!\pL)';
255        } else {
256            // not as correct as above, but at least won't break
257            $BL = '\b';
258            $BR = '\b';
259        }
260
261        if (substr($term, 0, 2) == '\\*') {
262            $term = substr($term, 2);
263        } else {
264            $term = $BL.$term;
265        }
266
267        if (substr($term, -2, 2) == '\\*') {
268            $term = substr($term, 0, -2);
269        } else {
270            $term = $term.$BR;
271        }
272
273        if ($term == $BL || $term == $BR || $term == $BL.$BR) {
274            $term = '';
275        }
276        return $term;
277    }
278}
279