xref: /dokuwiki/inc/indexer.php (revision 8605afb1b4e2a6a9e11e21a7bf0775bbb0d5af03)

<?php
/**
 * Functions to create the fulltext search index
 *
 * @license    GPL 2 (http://www.gnu.org/licenses/gpl.html)
 * @author     Andreas Gohr <andi@splitbrain.org>
 * @author     Tom N Harris <tnharris@whoopdedo.org>
 */

if(!defined('DOKU_INC')) die('meh.');

// Version tag used to force rebuild on upgrade
define('INDEXER_VERSION', 3);

// set the minimum token length to use in the index (note, this doesn't apply to numeric tokens)
if (!defined('IDX_MINWORDLENGTH')) define('IDX_MINWORDLENGTH',2);

// Asian characters are handled as words. The following regexp defines the
// Unicode-Ranges for Asian characters
// Ranges taken from http://en.wikipedia.org/wiki/Unicode_block
// I'm no language expert. If you think some ranges are wrongly chosen or
// a range is missing, please contact me
define('IDX_ASIAN1','[\x{0E00}-\x{0E7F}]'); // Thai
define('IDX_ASIAN2','['.
                   '\x{2E80}-\x{3040}'.  // CJK -> Hangul
                   '\x{309D}-\x{30A0}'.
                   '\x{30FD}-\x{31EF}\x{3200}-\x{D7AF}'.
                   '\x{F900}-\x{FAFF}'.  // CJK Compatibility Ideographs
                   '\x{FE30}-\x{FE4F}'.  // CJK Compatibility Forms
                   ']');
define('IDX_ASIAN3','['.                // Hiragana/Katakana (can be two characters)
                   '\x{3042}\x{3044}\x{3046}\x{3048}'.
                   '\x{304A}-\x{3062}\x{3064}-\x{3082}'.
                   '\x{3084}\x{3086}\x{3088}-\x{308D}'.
                   '\x{308F}-\x{3094}'.
                   '\x{30A2}\x{30A4}\x{30A6}\x{30A8}'.
                   '\x{30AA}-\x{30C2}\x{30C4}-\x{30E2}'.
                   '\x{30E4}\x{30E6}\x{30E8}-\x{30ED}'.
                   '\x{30EF}-\x{30F4}\x{30F7}-\x{30FA}'.
                   ']['.
                   '\x{3041}\x{3043}\x{3045}\x{3047}\x{3049}'.
                   '\x{3063}\x{3083}\x{3085}\x{3087}\x{308E}\x{3095}-\x{309C}'.
                   '\x{30A1}\x{30A3}\x{30A5}\x{30A7}\x{30A9}'.
                   '\x{30C3}\x{30E3}\x{30E5}\x{30E7}\x{30EE}\x{30F5}\x{30F6}\x{30FB}\x{30FC}'.
                   '\x{31F0}-\x{31FF}'.
                   ']?');
define('IDX_ASIAN', '(?:'.IDX_ASIAN1.'|'.IDX_ASIAN2.'|'.IDX_ASIAN3.')');

/**
 * Version of the indexer taking into consideration the external tokenizer.
 * The indexer is only compatible with data written by the same version.
 *
 * @author Tom N Harris <tnharris@whoopdedo.org>
 * @author Michael Hamann <michael@content-space.de>
 */
function idx_get_version(){
    global $conf;
    if($conf['external_tokenizer'])
        $version = INDEXER_VERSION . '+' . trim($conf['tokenizer_cmd']);
    else
        $version = INDEXER_VERSION;

    $data = array($version);
    trigger_event('INDEXER_VERSION_GET', $data, null, false);
    return implode('+', $data);
}

/**
 * Measure the length of a string.
 * Differs from strlen in handling of asian characters.
 *
 * @author Tom N Harris <tnharris@whoopdedo.org>
 */
function wordlen($w){
    $l = strlen($w);
    // If left alone, all chinese "words" will get put into w3.idx
    // So the "length" of a "word" is faked
    if(preg_match_all('/[\xE2-\xEF]/',$w,$leadbytes)) {
        foreach($leadbytes[0] as $b)
            $l += ord($b) - 0xE1;
    }
    return $l;
}

/**
 * Class that encapsulates operations on the indexer database.
 *
 * @author Tom N Harris <tnharris@whoopdedo.org>
 */
class Doku_Indexer {

    /**
     * Adds the contents of a page to the fulltext index
     *
     * The added text replaces previous words for the same page.
     * An empty value erases the page.
     *
     * @param string    $page   a page name
     * @param string    $text   the body of the page
     * @return boolean          the function completed successfully
     * @author Tom N Harris <tnharris@whoopdedo.org>
     * @author Andreas Gohr <andi@splitbrain.org>
     */
    public function addPageWords($page, $text) {
        if (!$this->_lock())
            return "locked";

        // load known documents
        $page_idx = $this->_addIndexKey('page', '', $page);
        if ($page_idx === false) {
            $this->_unlock();
            return false;
        }

        $pagewords = array();
        // get word usage in page
        $words = $this->_getPageWords($text);
        if ($words === false) {
            $this->_unlock();
            return false;
        }

        if (!empty($words)) {
            foreach (array_keys($words) as $wlen) {
                $index = $this->_getIndex('i', $wlen);
                foreach ($words[$wlen] as $wid => $freq) {
                    $idx = ($wid<count($index)) ? $index[$wid] : '';
                    $index[$wid] = $this->_updateTuple($idx, $pid, $freq);
                    $pagewords[] = "$wlen*$wid";
                }
                if (!$this->_saveIndex('i', $wlen, $index)) {
                    $this->_unlock();
                    return false;
                }
            }
        }

        // Remove obsolete index entries
        $pageword_idx = $this->_getIndexKey('pageword', '', $pid);
        if ($pageword_idx !== '') {
            $oldwords = explode(':',$pageword_idx);
            $delwords = array_diff($oldwords, $pagewords);
            $upwords = array();
            foreach ($delwords as $word) {
                if ($word != '') {
                    list($wlen,$wid) = explode('*', $word);
                    $wid = (int)$wid;
                    $upwords[$wlen][] = $wid;
                }
            }
            foreach ($upwords as $wlen => $widx) {
                $index = $this->_getIndex('i', $wlen);
                foreach ($widx as $wid) {
                    $index[$wid] = $this->_updateTuple($index[$wid], $pid, 0);
                }
                $this->_saveIndex('i', $wlen, $index);
            }
        }
        // Save the reverse index
        $pageword_idx = join(':', $pagewords);
        if (!$this->_saveIndexKey('pageword', '', $pid, $pageword_idx)) {
            $this->_unlock();
            return false;
        }

        $this->_unlock();
        return true;
    }

    /**
     * Split the words in a page and add them to the index.
     *
     * @author Andreas Gohr <andi@splitbrain.org>
     * @author Christopher Smith <chris@jalakai.co.uk>
     * @author Tom N Harris <tnharris@whoopdedo.org>
     */
    private function _getPageWords($text) {
        global $conf;

        $tokens = $this->tokenizer($text);
        $tokens = array_count_values($tokens);  // count the frequency of each token

        $words = array();
        foreach ($tokens as $w=>$c) {
            $l = wordlen($w);
            if (isset($words[$l])){
                $words[$l][$w] = $c + (isset($words[$l][$w]) ? $words[$l][$w] : 0);
            }else{
                $words[$l] = array($w => $c);
            }
        }

        // arrive here with $words = array(wordlen => array(word => frequency))
        $word_idx_modified = false;
        $index = array();   //resulting index
        foreach (array_keys($words) as $wlen) {
            $word_idx = $this->_getIndex('w', $wlen);
            foreach ($words[$wlen] as $word => $freq) {
                $wid = array_search($word, $word_idx);
                if ($wid === false) {
                    $wid = count($word_idx);
                    $word_idx[] = $word;
                    $word_idx_modified = true;
                }
                if (!isset($index[$wlen]))
                    $index[$wlen] = array();
                $index[$wlen][$wid] = $freq;
            }
            // save back the word index
            if ($word_idx_modified && !$this->_saveIndex('w', $wlen, $word_idx))
                return false;
        }

        return $index;
    }

    /**
     * Add/update keys to/of the metadata index.
     *
     * Adding new keys does not remove other keys for the page.
     * An empty value will erase the key.
     * The $key parameter can be an array to add multiple keys. $value will
     * not be used if $key is an array.
     *
     * @param string    $page   a page name
     * @param mixed     $key    a key string or array of key=>value pairs
     * @param mixed     $value  the value or list of values
     * @return boolean          the function completed successfully
     * @author Tom N Harris <tnharris@whoopdedo.org>
     * @author Michael Hamann <michael@content-space.de>
     */
    public function addMetaKeys($page, $key, $value=null) {
        if (!is_array($key)) {
            $key = array($key => $value);
        } elseif (!is_null($value)) {
            // $key is array, but $value is not null
            trigger_error("array passed to addMetaKeys but value is not null", E_USER_WARNING);
        }

        if (!$this->_lock())
            return "locked";

        // load known documents
        $pid = $this->_addIndexKey('page', '', $page);
        if ($pid === false) {
            $this->_unlock();
            return false;
        }

        foreach ($key as $name => $values) {
            $metaname = idx_cleanName($name);
            $metaidx = $this->_getIndex($metaname, '_i');
            $metawords = $this->_getIndex($metaname, '_w');
            $addwords = false;

            if (!is_array($values)) $values = array($values);

            $val_idx = $this->_getIndexKey($metaname, '_p', $pid);
            if ($val_idx != '') {
                $val_idx = explode(':', $val_idx);
                // -1 means remove, 0 keep, 1 add
                $val_idx = array_combine($val_idx, array_fill(0, count($val_idx), -1));
            } else {
                $val_idx = array();
            }


            foreach ($values as $val) {
                $val = (string)$val;
                if ($val !== "") {
                    $id = array_search($val, $metawords);
                    if ($id === false) {
                        $id = count($metawords);
                        $metawords[$id] = $val;
                        $addwords = true;
                    }
                    // test if value is already in the index
                    if (isset($val_idx[$id]) && $val_idx[$id] <= 0)
                        $val_idx[$id] = 0;
                    else // else add it
                        $val_idx[$id] = 1;
                }
            }

            if ($addwords)
                $this->_saveIndex($metaname.'_w', '', $metawords);
            $vals_changed = false;
            foreach ($val_idx as $id => $action) {
                if ($action == -1) {
                    $metaidx[$id] = $this->_updateTuple($metaidx[$id], $pid, 0);
                    $vals_changed = true;
                    unset($val_idx[$id]);
                } elseif ($action == 1) {
                    $metaidx[$id] = $this->_updateTuple($metaidx[$id], $pid, 1);
                    $vals_changed = true;
                }
            }

            if ($vals_changed) {
                $this->_saveIndex($metaname.'_i', '', $metaidx);
                $val_idx = implode(':', array_keys($val_idx));
                $this->_saveIndexKey($metaname.'_p', '', $pid, $val_idx);
            }

            unset($metaidx);
            unset($metawords);
        }

        $this->_unlock();
        return true;
    }

    /**
     * Remove a page from the index
     *
     * Erases entries in all known indexes.
     *
     * @param string    $page   a page name
     * @return boolean          the function completed successfully
     * @author Tom N Harris <tnharris@whoopdedo.org>
     */
    public function deletePage($page) {
    }

    /**
     * Split the text into words for fulltext search
     *
     * TODO: does this also need &$stopwords ?
     *
     * @param string    $text   plain text
     * @param boolean   $wc     are wildcards allowed?
     * @return array            list of words in the text
     * @author Tom N Harris <tnharris@whoopdedo.org>
     * @author Andreas Gohr <andi@splitbrain.org>
     */
    public function tokenizer($text, $wc=false) {
        global $conf;
        $words = array();
        $wc = ($wc) ? '' : '\*';
        $stopwords =& idx_get_stopwords();

        if ($conf['external_tokenizer'] && $conf['tokenizer_cmd'] != '') {
            if (0 == io_exec($conf['tokenizer_cmd'], $text, $output))
                $text = $output;
        } else {
            if (preg_match('/[^0-9A-Za-z ]/u', $text)) {
                // handle asian chars as single words (may fail on older PHP version)
                $asia = @preg_replace('/('.IDX_ASIAN.')/u', ' \1 ', $text);
                if (!is_null($asia)) $text = $asia; // recover from regexp falure
            }
        }
        $text = strtr($text, "\r\n\t", '   ');
        if (preg_match('/[^0-9A-Za-z ]/u', $text))
            $text = utf8_stripspecials($text, ' ', '\._\-:'.$wc);

        $wordlist = explode(' ', $text);
        foreach ($wordlist as $word) {
            $word = (preg_match('/[^0-9A-Za-z]/u', $word)) ?
                utf8_strtolower($word) : strtolower($word);
            if (!is_numeric($word) && strlen($word) < IDX_MINWORDLENGTH) continue;
            if (array_search($word, $stopwords) !== false) continue;
            $words[] = $word;
        }
        return $words;
    }

    /**
     * Find pages in the fulltext index containing the words,
     *
     * The search words must be pre-tokenized, meaning only letters and
     * numbers with an optional wildcard
     *
     * The returned array will have the original tokens as key. The values
     * in the returned list is an array with the page names as keys and the
     * number of times that token appeas on the page as value.
     *
     * @param arrayref  $tokens list of words to search for
     * @return array            list of page names with usage counts
     * @author Tom N Harris <tnharris@whoopdedo.org>
     * @author Andreas Gohr <andi@splitbrain.org>
     */
    public function lookup(&$tokens) {
        $result = array();
        $wids = $this->_getIndexWords($tokens, $result);
        if (empty($wids)) return array();
        // load known words and documents
        $page_idx = $this->_getIndex('page', '');
        $docs = array();
        foreach (array_keys($wids) as $wlen) {
            $wids[$wlen] = array_unique($wids[$wlen]);
            $index = $this->_getIndex('i', $wlen);
            foreach($wids[$wlen] as $ixid) {
                if ($ixid < count($index))
                    $docs["$wlen*$ixid"] = $this->_parseTuples($page_idx, $index[$ixid]);
            }
        }
        // merge found pages into final result array
        $final = array();
        foreach ($result as $word => $res) {
            $final[$word] = array();
            foreach ($res as $wid) {
                $hits = &$docs[$wid];
                foreach ($hits as $hitkey => $hitcnt) {
                    // make sure the document still exists
                    if (!page_exists($hitkey, '', false)) continue;
                    if (!isset($final[$word][$hitkey]))
                        $final[$word][$hitkey] = $hitcnt;
                    else
                        $final[$word][$hitkey] += $hitcnt;
                }
            }
        }
        return $final;
    }

    /**
     * Find pages containing a metadata key.
     *
     * The metadata values are compared as case-sensitive strings. Pass a
     * callback function that returns true or false to use a different
     * comparison function
     *
     * @param string    $key    name of the metadata key to look for
     * @param string    $value  search term to look for
     * @param callback  $func   comparison function
     * @return array            lists with page names, keys are query values
     * @author Tom N Harris <tnharris@whoopdedo.org>
     * @author Michael Hamann <michael@content-space.de>
     */
    public function lookupKey($key, $value, $func=null) {
        $metaname = idx_cleanName($key);

        // get all words in order to search the matching ids
        $words = $this->_getIndex($metaname, '_w');

        // the matching ids for the provided value(s)
        $value_ids = array();

        if (!is_array($value)) $value = array($value);

        foreach ($value as $val) {
            if (is_null($func)) {
                if (($i = array_search($val, $words)) !== false)
                    $value_ids[$i] = $val;
            } else {
                foreach ($words as $i => $word) {
                    if (call_user_func_array($func, array($word, $value)))
                        $value_ids[$i] = $val;
                }
            }
        }

        unset($words); // free the used memory

        // load all lines and pages so the used lines can be taken and matched with the pages
        $lines = $this->_getIndex($metaname, '_i');
        $page_idx = $this->_getIndex('page', '');

        $result = array();
        foreach ($value_ids as $value_id => $val) {
            // parse the tuples of the form page_id*1:page2_id*1 and so on, return value
            // is an array with page_id => 1, page2_id => 1 etc. so take the keys only
            $result[$val] = array_keys($this->_parseTuples($page_idx, $lines[$value_id]));
        }
        return $result;
    }

    /**
     * Find the index ID of each search term.
     *
     * The query terms should only contain valid characters, with a '*' at
     * either the beginning or end of the word (or both).
     * The $result parameter can be used to merge the index locations with
     * the appropriate query term.
     *
     * @param arrayref  $words  The query terms.
     * @param arrayref  $result Set to word => array("length*id" ...)
     * @return array            Set to length => array(id ...)
     * @author Tom N Harris <tnharris@whoopdedo.org>
     */
    private function _getIndexWords(&$words, &$result) {
        $tokens = array();
        $tokenlength = array();
        $tokenwild = array();
        foreach ($words as $word) {
            $result[$word] = array();
            $caret = false;
            $dollar = false;
            $xword = $word;
            $wlen = wordlen($word);

            // check for wildcards
            if (substr($xword, 0, 1) == '*') {
                $xword = substr($xword, 1);
                $caret = true;
                $wlen -= 1;
            }
            if (substr($xword, -1, 1) == '*') {
                $xword = substr($xword, 0, -1);
                $dollar = true;
                $wlen -= 1;
            }
            if ($wlen < IDX_MINWORDLENGTH && !$caret && !$dollar && !is_numeric($xword))
                continue;
            if (!isset($tokens[$xword]))
                $tokenlength[$wlen][] = $xword;
            if ($caret || $dollar) {
                $re = preg_quote($xword, '/');
                if ($caret) $re = '^'.$re;
                if ($dollar) $re = $re.'$';
                $tokens[$xword][] = array($word, '/'.$re.'/');
                if (!isset($tokenwild[$xword]))
                    $tokenwild[$xword] = $wlen;
            } else {
                $tokens[$xword][] = array($word, null);
            }
        }
        asort($tokenwild);
        // $tokens = array( base word => array( [ query term , regexp ] ... ) ... )
        // $tokenlength = array( base word length => base word ... )
        // $tokenwild = array( base word => base word length ... )
        $length_filter = empty($tokenwild) ? $tokenlength : min(array_keys($tokenlength));
        $indexes_known = $this->_indexLengths($length_filter);
        if (!empty($tokenwild)) sort($indexes_known);
        // get word IDs
        $wids = array();
        foreach ($indexes_known as $ixlen) {
            $word_idx = $this->_getIndex('w', $ixlen);
            // handle exact search
            if (isset($tokenlength[$ixlen])) {
                foreach ($tokenlength[$ixlen] as $xword) {
                    $wid = array_search($xword, $word_idx);
                    if ($wid !== false) {
                        $wids[$ixlen][] = $wid;
                        foreach ($tokens[$xword] as $w)
                            $result[$w[0]][] = "$ixlen*$wid";
                    }
                }
            }
            // handle wildcard search
            foreach ($tokenwild as $xword => $wlen) {
                if ($wlen >= $ixlen) break;
                foreach ($tokens[$xword] as $w) {
                    if (is_null($w[1])) continue;
                    foreach(array_keys(preg_grep($w[1], $word_idx)) as $wid) {
                        $wids[$ixlen][] = $wid;
                        $result[$w[0]][] = "$ixlen*$wid";
                    }
                }
            }
        }
        return $wids;
    }

    /**
     * Return a list of all pages
     *
     * @param string    $key    list only pages containing the metadata key (optional)
     * @return array            list of page names
     * @author Tom N Harris <tnharris@whoopdedo.org>
     */
    public function getPages($key=null) {
        $page_idx = $this->_getIndex('page', '');
        if (is_null($key)) return $page_idx;
    }

    /**
     * Return a list of words sorted by number of times used
     *
     * @param int       $min    bottom frequency threshold
     * @param int       $max    upper frequency limit. No limit if $max<$min
     * @param string    $key    metadata key to list. Uses the fulltext index if not given
     * @return array            list of words as the keys and frequency as values
     * @author Tom N Harris <tnharris@whoopdedo.org>
     */
    public function histogram($min=1, $max=0, $key=null) {
    }

    /**
     * Lock the indexer.
     *
     * @author Tom N Harris <tnharris@whoopdedo.org>
     */
    private function _lock() {
        global $conf;
        $status = true;
        $lock = $conf['lockdir'].'/_indexer.lock';
        while (!@mkdir($lock, $conf['dmode'])) {
            usleep(50);
            if (time() - @filemtime($lock) > 60*5) {
                // looks like a stale lock, remove it
                @rmdir($lock);
                $status = "stale lock removed";
            } else {
                return false;
            }
        }
        if ($conf['dperm'])
            chmod($lock, $conf['dperm']);
        return $status;
    }

    /**
     * Release the indexer lock.
     *
     * @author Tom N Harris <tnharris@whoopdedo.org>
     */
    private function _unlock() {
        global $conf;
        @rmdir($conf['lockdir'].'/_indexer.lock');
        return true;
    }

    /**
     * Retrieve the entire index.
     *
     * @author Tom N Harris <tnharris@whoopdedo.org>
     */
    private function _getIndex($idx, $suffix) {
        global $conf;
        $fn = $conf['indexdir'].'/'.$idx.$suffix.'.idx';
        if (!@file_exists($fn)) return array();
        return file($fn, FILE_IGNORE_NEW_LINES);
    }

    /**
     * Replace the contents of the index with an array.
     *
     * @author Tom N Harris <tnharris@whoopdedo.org>
     */
    private function _saveIndex($idx, $suffix, &$lines) {
        global $conf;
        $fn = $conf['indexdir'].'/'.$idx.$suffix;
        $fh = @fopen($fn.'.tmp', 'w');
        if (!$fh) return false;
        fwrite($fh, join("\n", $lines));
        fclose($fh);
        if (isset($conf['fperm']))
            chmod($fn.'.tmp', $conf['fperm']);
        io_rename($fn.'.tmp', $fn.'.idx');
        if ($suffix !== '')
            $this->_cacheIndexDir($idx, $suffix, empty($lines));
        return true;
    }

    /**
     * Retrieve a line from the index.
     *
     * @author Tom N Harris <tnharris@whoopdedo.org>
     */
    private function _getIndexKey($idx, $suffix, $id) {
        global $conf;
        $fn = $conf['indexdir'].'/'.$idx.$suffix.'.idx';
        if (!@file_exists($fn)) return '';
        $fh = @fopen($fn, 'r');
        if (!$fh) return '';
        $ln = -1;
        while (($line = fgets($fh)) !== false) {
            if (++$ln == $id) break;
        }
        fclose($fh);
        return rtrim((string)$line);
    }

    /**
     * Write a line into the index.
     *
     * @author Tom N Harris <tnharris@whoopdedo.org>
     */
    private function _saveIndexKey($idx, $suffix, $id, $line) {
        global $conf;
        if (substr($line, -1) != "\n")
            $line .= "\n";
        $fn = $conf['indexdir'].'/'.$idx.$suffix;
        $fh = @fopen($fn.'.tmp', 'w');
        if (!fh) return false;
        $ih = @fopen($fn.'.idx', 'r');
        if ($ih) {
            $ln = -1;
            while (($curline = fgets($ih)) !== false) {
                fwrite($fh, (++$ln == $id) ? $line : $curline);
            }
            if ($id > $ln) {
                while ($id > ++$ln)
                    fwrite($fh, "\n");
                fwrite($fh, $line);
            }
            fclose($ih);
        } else {
            $ln = -1;
            while ($id > ++$ln)
                fwrite($fh, "\n");
            fwrite($fh, $line);
        }
        fclose($fh);
        if (isset($conf['fperm']))
            chmod($fn.'.tmp', $conf['fperm']);
        io_rename($fn.'.tmp', $fn.'.idx');
        if ($suffix !== '')
            $this->_cacheIndexDir($idx, $suffix);
        return true;
    }

    /**
     * Retrieve or insert a value in the index.
     *
     * @author Tom N Harris <tnharris@whoopdedo.org>
     */
    private function _addIndexKey($idx, $suffix, $value) {
        $index = $this->_getIndex($idx, $suffix);
        $id = array_search($value, $index);
        if ($id === false) {
            $id = count($index);
            $index[$id] = $value;
            if (!$this->_saveIndex($idx, $suffix, $index)) {
                trigger_error("Failed to write $idx index", E_USER_ERROR);
                return false;
            }
        }
        return $id;
    }

    private function _cacheIndexDir($idx, $suffix, $delete=false) {
        global $conf;
        if ($idx == 'i')
            $cachename = $conf['indexdir'].'/lengths';
        else
            $cachename = $conf['indexdir'].'/'.$idx.'lengths';
        $lengths = @file($cachename.'.idx', FILE_IGNORE_NEW_LINES | FILE_SKIP_EMPTY_LINES);
        if ($lengths === false) $lengths = array();
        $old = array_search((string)$suffix, $lengths);
        if (empty($lines)) {
            if ($old === false) return;
            unset($lengths[$old]);
        } else {
            if ($old !== false) return;
            $lengths[] = $suffix;
            sort($lengths);
        }
        $fh = @fopen($cachename.'.tmp', 'w');
        if (!$fh) {
            trigger_error("Failed to write index cache", E_USER_ERROR);
            return;
        }
        @fwrite($fh, implode("\n", $lengths));
        @fclose($fh);
        if (isset($conf['fperm']))
            chmod($cachename.'.tmp', $conf['fperm']);
        io_rename($cachename.'.tmp', $cachename.'.idx');
    }

    /**
     * Get the list of lengths indexed in the wiki.
     *
     * Read the index directory or a cache file and returns
     * a sorted array of lengths of the words used in the wiki.
     *
     * @author YoBoY <yoboy.leguesh@gmail.com>
     */
    private function _listIndexLengths() {
        global $conf;
        $cachename = $conf['indexdir'].'/lengths';
        clearstatcache();
        if (@file_exists($cachename.'.idx')) {
            $lengths = @file($cachename.'.idx', FILE_IGNORE_NEW_LINES | FILE_SKIP_EMPTY_LINES);
            if ($lengths !== false) {
                $idx = array();
                foreach ($lengths as $length)
                    $idx[] = (int)$length;
                return $idx;
            }
        }

        $dir = @opendir($conf['indexdir']);
        if ($dir === false)
            return array();
        $lengths[] = array();
        while (($f = readdir($dir)) !== false) {
            if (substr($f, 0, 1) == 'i' && substr($f, -4) == '.idx') {
                $i = substr($f, 1, -4);
                if (is_numeric($i))
                    $lengths[] = (int)$i;
            }
        }
        closedir($dir);
        sort($lengths);
        // save this in a file
        $fh = @fopen($cachename.'.tmp', 'w');
        if (!$fh) {
            trigger_error("Failed to write index cache", E_USER_ERROR);
            return;
        }
        @fwrite($fh, implode("\n", $lengths));
        @fclose($fh);
        if (isset($conf['fperm']))
            chmod($cachename.'.tmp', $conf['fperm']);
        io_rename($cachename.'.tmp', $cachename.'.idx');

        return $lengths;
    }

    /**
     * Get the word lengths that have been indexed.
     *
     * Reads the index directory and returns an array of lengths
     * that there are indices for.
     *
     * @author YoBoY <yoboy.leguesh@gmail.com>
     */
    private function _indexLengths($filter) {
        global $conf;
        $idx = array();
        if (is_array($filter)) {
            // testing if index files exist only
            $path = $conf['indexdir']."/i";
            foreach ($filter as $key => $value) {
                if (@file_exists($path.$key.'.idx'))
                    $idx[] = $key;
            }
        } else {
            $lengths = idx_listIndexLengths();
            foreach ($lengths as $key => $length) {
                // keep all the values equal or superior
                if ((int)$length >= (int)$filter)
                    $idx[] = $length;
            }
        }
        return $idx;
    }

    /**
     * Insert or replace a tuple in a line.
     *
     * @author Tom N Harris <tnharris@whoopdedo.org>
     */
    private function _updateTuple($line, $id, $count) {
        $newLine = $line;
        if ($newLine !== '')
            $newLine = preg_replace('/(^|:)'.preg_quote($id,'/').'\*\d*/', '', $newLine);
        $newLine = trim($newLine, ':');
        if ($count) {
            if (strlen($newLine) > 0)
                return "$id*$count:".$newLine;
            else
                return "$id*$count".$newLine;
        }
        return $newLine;
    }

    /**
     * Split a line into an array of tuples.
     *
     * @author Tom N Harris <tnharris@whoopdedo.org>
     * @author Andreas Gohr <andi@splitbrain.org>
     */
    private function _parseTuples(&$keys, $line) {
        $result = array();
        if ($line == '') return $result;
        $parts = explode(':', $line);
        foreach ($parts as $tuple) {
            if ($tuple == '') continue;
            list($key, $cnt) = explode('*', $tuple);
            if (!$cnt) continue;
            $key = $keys[$key];
            if (!$key) continue;
            $result[$key] = $cnt;
        }
        return $result;
    }
}

/**
 * Create an instance of the indexer.
 *
 * @return object               a Doku_Indexer
 * @author Tom N Harris <tnharris@whoopdedo.org>
 */
function idx_get_indexer() {
    static $Indexer = null;
    if (is_null($Indexer)) {
        $Indexer = new Doku_Indexer();
    }
    return $Indexer;
}

/**
 * Returns words that will be ignored.
 *
 * @return array                list of stop words
 * @author Tom N Harris <tnharris@whoopdedo.org>
 */
function & idx_get_stopwords() {
    static $stopwords = null;
    if (is_null($stopwords)) {
        global $conf;
        $swfile = DOKU_INC.'inc/lang/'.$conf['lang'].'/stopwords.txt';
        if(@file_exists($swfile)){
            $stopwords = file($swfile, FILE_IGNORE_NEW_LINES);
        }else{
            $stopwords = array();
        }
    }
    return $stopwords;
}

/**
 * Adds/updates the search index for the given page
 *
 * Locking is handled internally.
 *
 * @param string        $page   name of the page to index
 * @param boolean       $verbose    print status messages
 * @return boolean              the function completed successfully
 * @author Tom N Harris <tnharris@whoopdedo.org>
 */
function idx_addPage($page, $verbose=false) {
    // check if indexing needed
    $idxtag = metaFN($page,'.indexed');
    if(@file_exists($idxtag)){
        if(trim(io_readFile($idxtag)) == idx_get_version()){
            $last = @filemtime($idxtag);
            if($last > @filemtime(wikiFN($ID))){
                if ($verbose) print("Indexer: index for $page up to date".DOKU_LF);
                return false;
            }
        }
    }

    $body = '';
    $data = array($page, $body);
    $evt = new Doku_Event('INDEXER_PAGE_ADD', $data);
    if ($evt->advise_before()) $data[1] = $data[1] . " " . rawWiki($page);
    $evt->advise_after();
    unset($evt);
    list($page,$body) = $data;

    $Indexer = idx_get_indexer();
    $result = $Indexer->addPageWords($page, $body);
    if ($result === "locked") {
        if ($verbose) print("Indexer: locked".DOKU_LF);
        return false;
    }

    if ($result) {
        $data = array('page' => $page, 'metadata' => array());

        if (($references = p_get_metadata($page, 'relation references')) !== null)
            $data['metadata']['relation_references'] = array_keys($references);

        $evt = new Doku_Event('INDEXER_METADATA_INDEX', $data);
        if ($evt->advise_before()) {
            $result = $Indexer->addMetaKeys($page, $data['metadata']);
            if ($result === "locked") {
                if ($verbose) print("Indexer: locked".DOKU_LF);
                return false;
            }
        }
        $evt->advise_after();
        unset($evt);
    }

    if ($result)
        io_saveFile(metaFN($page,'.indexed'), idx_get_version());
    if ($verbose) {
        print("Indexer: finished".DOKU_LF);
        return true;
    }
    return $result;
}

/**
 * Find tokens in the fulltext index
 *
 * Takes an array of words and will return a list of matching
 * pages for each one.
 *
 * Important: No ACL checking is done here! All results are
 *            returned, regardless of permissions
 *
 * @param arrayref      $words  list of words to search for
 * @return array                list of pages found, associated with the search terms
 */
function idx_lookup(&$words) {
    $Indexer = idx_get_indexer();
    return $Indexer->lookup($words);
}

/**
 * Split a string into tokens
 *
 */
function idx_tokenizer($string, $wc=false) {
    $Indexer = idx_get_indexer();
    return $Indexer->tokenizer($string, $wc);
}

/* For compatibility */

/**
 * Read the list of words in an index (if it exists).
 *
 * @author Tom N Harris <tnharris@whoopdedo.org>
 */
function idx_getIndex($idx, $suffix) {
    global $conf;
    $fn = $conf['indexdir'].'/'.$idx.$suffix.'.idx';
    if (!@file_exists($fn)) return array();
    return file($fn);
}

/**
 * Get the list of lengths indexed in the wiki.
 *
 * Read the index directory or a cache file and returns
 * a sorted array of lengths of the words used in the wiki.
 *
 * @author YoBoY <yoboy.leguesh@gmail.com>
 */
function idx_listIndexLengths() {
    global $conf;
    // testing what we have to do, create a cache file or not.
    if ($conf['readdircache'] == 0) {
        $docache = false;
    } else {
        clearstatcache();
        if (@file_exists($conf['indexdir'].'/lengths.idx')
        && (time() < @filemtime($conf['indexdir'].'/lengths.idx') + $conf['readdircache'])) {
            if (($lengths = @file($conf['indexdir'].'/lengths.idx', FILE_IGNORE_NEW_LINES | FILE_SKIP_EMPTY_LINES)) !== false) {
                $idx = array();
                foreach ($lengths as $length) {
                    $idx[] = (int)$length;
                }
                return $idx;
            }
        }
        $docache = true;
    }

    if ($conf['readdircache'] == 0 || $docache) {
        $dir = @opendir($conf['indexdir']);
        if ($dir === false)
            return array();
        $idx[] = array();
        while (($f = readdir($dir)) !== false) {
            if (substr($f, 0, 1) == 'i' && substr($f, -4) == '.idx') {
                $i = substr($f, 1, -4);
                if (is_numeric($i))
                    $idx[] = (int)$i;
            }
        }
        closedir($dir);
        sort($idx);
        // save this in a file
        if ($docache) {
            $handle = @fopen($conf['indexdir'].'/lengths.idx', 'w');
            @fwrite($handle, implode("\n", $idx));
            @fclose($handle);
        }
        return $idx;
    }

    return array();
}

/**
 * Get the word lengths that have been indexed.
 *
 * Reads the index directory and returns an array of lengths
 * that there are indices for.
 *
 * @author YoBoY <yoboy.leguesh@gmail.com>
 */
function idx_indexLengths($filter) {
    global $conf;
    $idx = array();
    if (is_array($filter)) {
        // testing if index files exist only
        $path = $conf['indexdir']."/i";
        foreach ($filter as $key => $value) {
            if (@file_exists($path.$key.'.idx'))
                $idx[] = $key;
        }
    } else {
        $lengths = idx_listIndexLengths();
        foreach ($lengths as $key => $length) {
            // keep all the values equal or superior
            if ((int)$length >= (int)$filter)
                $idx[] = $length;
        }
    }
    return $idx;
}

/**
 * Clean a name of a key for use as a file name.
 *
 * Romanizes non-latin characters, then strips away anything that's
 * not a letter, number, or underscore.
 *
 * @author Tom N Harris <tnharris@whoopdedo.org>
 */
function idx_cleanName($name) {
    $name = utf8_romanize(trim((string)$name));
    $name = preg_replace('#[ \./\\:-]+#', '_', $name);
    $name = preg_replace('/[^A-Za-z0-9_]/', '', $name);
    return strtolower($name);
}

//Setup VIM: ex: et ts=4 :