xref: /dokuwiki/inc/Search/FulltextSearch.php (revision ede4646658cf51245060332d97a319a39c788ea1) 
1<?php
2
3namespace dokuwiki\Search;
4
5use dokuwiki\Extension\Event;
6use dokuwiki\Search\Collection\FrequencyCollectionSearch;
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 FrequencyCollectionSearch
91        $collection = new PageFulltextCollection();
92        $search = new FrequencyCollectionSearch($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        // prepare time filters
109        $after = $data['after'] ? (is_int($data['after']) ? $data['after'] : strtotime($data['after'])) : null;
110        $before = $data['before'] ? (is_int($data['before']) ? $data['before'] : strtotime($data['before'])) : null;
111
112        // filter by settings, acls, existence, and time range
113        $docs = array_filter($docs, static function ($score, $id) use ($after, $before) {
114            if (isHiddenPage($id) || auth_quickaclcheck($id) < AUTH_READ || !page_exists($id, '', false)) {
115                return false;
116            }
117            if ($after || $before) {
118                $mTime = filemtime(wikiFN($id));
119                if ($after && $after > $mTime) return false;
120                if ($before && $before < $mTime) return false;
121            }
122            return true;
123        }, ARRAY_FILTER_USE_BOTH);
124
125        if ($data['sort'] === 'mtime') {
126            uksort($docs, static function ($a, $b) {
127                return filemtime(wikiFN($b)) - filemtime(wikiFN($a));
128            });
129        } else {
130            arsort($docs);
131        }
132
133        return $docs;
134    }
135
136    /**
137     * Creates a snippet extract
138     *
139     * @author Andreas Gohr <andi@splitbrain.org>
140     * @triggers FULLTEXT_SNIPPET_CREATE
141     *
142     * @param string $id page id
143     * @param array $highlight
144     * @return mixed
145     */
146    public function snippet($id, $highlight)
147    {
148        $text = rawWiki($id);
149        $text = str_replace("\xC2\xAD",'',$text); // remove soft-hyphens
150        $evdata = array(
151            'id'        => $id,
152            'text'      => &$text,
153            'highlight' => &$highlight,
154            'snippet'   => '',
155        );
156
157        $evt = new Event('FULLTEXT_SNIPPET_CREATE', $evdata);
158        if ($evt->advise_before()) {
159            $match = array();
160            $snippets = array();
161            $utf8_offset = $offset = $end = 0;
162            $len = Utf8\PhpString::strlen($text);
163
164            // build a regexp from the phrases to highlight
165            $re1 = '(' .
166                join(
167                    '|',
168                    array_map(
169                        [$this, 'snippetRePreprocess'],
170                        array_map(
171                            'preg_quote_cb',
172                            array_filter((array) $highlight)
173                        )
174                    )
175                ) .
176                ')';
177            $re2 = "$re1.{0,75}(?!\\1)$re1";
178            $re3 = "$re1.{0,45}(?!\\1)$re1.{0,45}(?!\\1)(?!\\2)$re1";
179
180            for ($cnt=4; $cnt--;) {
181                if (0) {
182                } elseif (preg_match('/'.$re3.'/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) {
183                } elseif (preg_match('/'.$re2.'/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) {
184                } elseif (preg_match('/'.$re1.'/iu', $text, $match, PREG_OFFSET_CAPTURE, $offset)) {
185                } else {
186                    break;
187                }
188
189                list($str, $idx) = $match[0];
190
191                // convert $idx (a byte offset) into a utf8 character offset
192                $utf8_idx = Utf8\PhpString::strlen(substr($text, 0, $idx));
193                $utf8_len = Utf8\PhpString::strlen($str);
194
195                // establish context, 100 bytes surrounding the match string
196                // first look to see if we can go 100 either side,
197                // then drop to 50 adding any excess if the other side can't go to 50,
198                $pre = min($utf8_idx - $utf8_offset, 100);
199                $post = min($len - $utf8_idx - $utf8_len, 100);
200
201                if ($pre > 50 && $post > 50) {
202                    $pre = $post = 50;
203                } elseif ($pre > 50) {
204                    $pre = min($pre, 100 - $post);
205                } elseif ($post > 50) {
206                    $post = min($post, 100 - $pre);
207                } elseif ($offset == 0) {
208                    // both are less than 50, means the context is the whole string
209                    // make it so and break out of this loop - there is no need for the
210                    // complex snippet calculations
211                    $snippets = array($text);
212                    break;
213                }
214
215                // establish context start and end points, try to append to previous
216                // context if possible
217                $start = $utf8_idx - $pre;
218                $append = ($start < $end) ? $end : false;  // still the end of the previous context snippet
219                $end = $utf8_idx + $utf8_len + $post;      // now set it to the end of this context
220
221                if ($append) {
222                    $snippets[count($snippets)-1] .= Utf8\PhpString::substr($text, $append, $end-$append);
223                } else {
224                    $snippets[] = Utf8\PhpString::substr($text, $start, $end-$start);
225                }
226
227                // set $offset for next match attempt
228                // continue matching after the current match
229                // if the current match is not the longest possible match starting at the current offset
230                // this prevents further matching of this snippet but for possible matches of length
231                // smaller than match length + context (at least 50 characters) this match is part of the context
232                $utf8_offset = $utf8_idx + $utf8_len;
233                $offset = $idx + strlen(Utf8\PhpString::substr($text, $utf8_idx, $utf8_len));
234                $offset = Utf8\Clean::correctIdx($text, $offset);
235            }
236
237            $m = "\1";
238            $snippets = preg_replace('/'.$re1.'/iu', $m.'$1'.$m, $snippets);
239            $snippet = preg_replace(
240                '/' . $m . '([^' . $m . ']*?)' . $m . '/iu',
241                '<strong class="search_hit">$1</strong>',
242                hsc(join('... ', $snippets))
243            );
244
245            $evdata['snippet'] = $snippet;
246        }
247        $evt->advise_after();
248        unset($evt);
249
250        return $evdata['snippet'];
251    }
252
253    /**
254     * Wraps a search term in regex boundary checks.
255     *
256     * @param string $term
257     * @return string
258     */
259    public function snippetRePreprocess($term)
260    {
261        // do not process asian terms where word boundaries are not explicit
262        if (Utf8\Asian::isAsianWords($term)) return $term;
263
264        if (UTF8_PROPERTYSUPPORT) {
265            // unicode word boundaries
266            // see http://stackoverflow.com/a/2449017/172068
267            $BL = '(?<!\pL)';
268            $BR = '(?!\pL)';
269        } else {
270            // not as correct as above, but at least won't break
271            $BL = '\b';
272            $BR = '\b';
273        }
274
275        if (substr($term, 0, 2) == '\\*') {
276            $term = substr($term, 2);
277        } else {
278            $term = $BL.$term;
279        }
280
281        if (substr($term, -2, 2) == '\\*') {
282            $term = substr($term, 0, -2);
283        } else {
284            $term = $term.$BR;
285        }
286
287        if ($term == $BL || $term == $BR || $term == $BL.$BR) {
288            $term = '';
289        }
290        return $term;
291    }
292}
293