Commit 2ece32e6 authored by Dries's avatar Dries

- Patch #326016 by jhodgdon: PREG_CLASS_CJK doesn't include all CJK characters.

parent daede057
......@@ -103,12 +103,25 @@
'\x{ff65}');
/**
* Matches all CJK characters that are candidates for auto-splitting
* (Chinese, Japanese, Korean).
* Contains kana and BMP ideographs.
* Matches CJK (Chinese, Japanese, Korean) letter-like characters.
*
* This list is derived from the "East Asian Scripts" section of
* http://www.unicode.org/charts/index.html, as well as a comment on
* http://unicode.org/reports/tr11/tr11-11.html listing some character
* ranges that are reserved for additional CJK ideographs.
*
* The character ranges do not include numbers, punctuation, or symbols, since
* these are handled separately in search. Note that radicals and strokes are
* considered symbols. (See
* http://www.unicode.org/Public/UNIDATA/extracted/DerivedGeneralCategory.txt)
*
* @see search_expand_cjk()
*/
define('PREG_CLASS_CJK', '\x{3041}-\x{30ff}\x{31f0}-\x{31ff}\x{3400}-\x{4db5}' .
'\x{4e00}-\x{9fbb}\x{f900}-\x{fad9}');
define('PREG_CLASS_CJK', '\x{1100}-\x{11FF}\x{3040}-\x{309F}\x{30A1}-\x{318E}' .
'\x{31A0}-\x{31B7}\x{31F0}-\x{31FF}\x{3400}-\x{4DBF}\x{4E00}-\x{9FCF}' .
'\x{A000}-\x{A48F}\x{A4D0}-\x{A4FD}\x{A960}-\x{A97F}\x{AC00}-\x{D7FF}' .
'\x{F900}-\x{FAFF}\x{FF21}-\x{FF3A}\x{FF41}-\x{FF5A}\x{FF66}-\x{FFDC}' .
'\x{20000}-\x{2FFFD}\x{30000}-\x{3FFFD}');
/**
* Implements hook_help().
......@@ -447,28 +460,45 @@ function search_simplify($text) {
}
/**
* Basic CJK tokenizer. Simply splits a string into consecutive, overlapping
* sequences of characters ('minimum_word_size' long).
* Splits CJK (Chinese, Japanese, Korean) text into tokens.
*
* The Search module matches exact words, where a word is defined to be a
* sequence of characters delimited by spaces or punctuation. CJK languages are
* written in long strings of characters, though, not split up into words. So
* in order to allow search matching, we split up CJK text into tokens
* consisting of consecutive, overlapping sequences of characters whose length
* is equal to the 'minimum_word_size' variable. This tokenizing is only done if
* the 'overlap_cjk' variable is TRUE.
*
* @param $matches
* This function is a callback for preg_replace_callback(), which is called
* from search_simplify(). So, $matches is an array of regular expression
* matches, which means that $matches[0] contains the matched text -- a string
* of CJK characters to tokenize.
*
* @return
* Tokenized text, starting and ending with a space character.
*/
function search_expand_cjk($matches) {
$min = variable_get('minimum_word_size', 3);
$str = $matches[0];
$l = drupal_strlen($str);
// Passthrough short words
if ($l <= $min) {
$length = drupal_strlen($str);
// If the text is shorter than the minimum word size, don't tokenize it.
if ($length <= $min) {
return ' ' . $str . ' ';
}
$tokens = ' ';
// FIFO queue of characters
// Build a FIFO queue of characters.
$chars = array();
// Begin loop
for ($i = 0; $i < $l; ++$i) {
// Grab next character
for ($i = 0; $i < $length; $i++) {
// Add the next character off the beginning of the string to the queue.
$current = drupal_substr($str, 0, 1);
$str = substr($str, strlen($current));
$chars[] = $current;
if ($i >= $min - 1) {
// Make a token of $min characters, and add it to the token string.
$tokens .= implode('', $chars) . ' ';
// Shift out the first character in the queue.
array_shift($chars);
}
}
......
......@@ -872,3 +872,153 @@ class SearchConfigSettingsForm extends DrupalWebTestCase {
}
}
/**
* Test the CJK tokenizer.
*/
class SearchTokenizerTestCase extends DrupalWebTestCase {
public static function getInfo() {
return array(
'name' => 'CJK tokenizer',
'description' => 'Check that CJK tokenizer works as intended.',
'group' => 'Search',
);
}
function setUp() {
parent::setUp('search');
}
/**
* Verifies that strings of CJK characters are tokenized.
*
* The search_simplify() function does special things with numbers, symbols,
* and punctuation. So we only test that CJK characters that are not in these
* character classes are tokenized properly. See PREG_CLASS_CKJ for more
* information.
*/
function testTokenizer() {
// Set the minimum word size to 1 (to split all CJK characters) and make
// sure CJK tokenizing is turned on.
variable_set('minimum_word_size', 1);
variable_set('overlap_cjk', TRUE);
$this->refreshVariables();
// Create a string of CJK characters from various character ranges in
// the Unicode tables.
// Beginnings of the character ranges.
$starts = array(
'CJK unified' => 0x4e00,
'CJK Ext A' => 0x3400,
'CJK Compat' => 0xf900,
'Hangul Jamo' => 0x1100,
'Hangul Ext A' => 0xa960,
'Hangul Ext B' => 0xd7b0,
'Hangul Compat' => 0x3131,
'Half non-punct 1' => 0xff21,
'Half non-punct 2' => 0xff41,
'Half non-punct 3' => 0xff66,
'Hangul Syllables' => 0xac00,
'Hiragana' => 0x3040,
'Katakana' => 0x30a1,
'Katakana Ext' => 0x31f0,
'CJK Reserve 1' => 0x20000,
'CJK Reserve 2' => 0x30000,
'Bomofo' => 0x3100,
'Bomofo Ext' => 0x31a0,
'Lisu' => 0xa4d0,
'Yi' => 0xa000,
);
// Ends of the character ranges.
$ends = array(
'CJK unified' => 0x9fcf,
'CJK Ext A' => 0x4dbf,
'CJK Compat' => 0xfaff,
'Hangul Jamo' => 0x11ff,
'Hangul Ext A' => 0xa97f,
'Hangul Ext B' => 0xd7ff,
'Hangul Compat' => 0x318e,
'Half non-punct 1' => 0xff3a,
'Half non-punct 2' => 0xff5a,
'Half non-punct 3' => 0xffdc,
'Hangul Syllables' => 0xd7af,
'Hiragana' => 0x309f,
'Katakana' => 0x30ff,
'Katakana Ext' => 0x31ff,
'CJK Reserve 1' => 0x2fffd,
'CJK Reserve 2' => 0x3fffd,
'Bomofo' => 0x312f,
'Bomofo Ext' => 0x31b7,
'Lisu' => 0xa4fd,
'Yi' => 0xa48f,
);
// Generate characters consisting of starts, midpoints, and ends.
$chars = array();
$charcodes = array();
foreach ($starts as $key => $value) {
$charcodes[] = $starts[$key];
$chars[] = $this->code2utf($starts[$key]);
$mid = round(0.5 * ($starts[$key] + $ends[$key]));
$charcodes[] = $mid;
$chars[] = $this->code2utf($mid);
$charcodes[] = $ends[$key];
$chars[] = $this->code2utf($ends[$key]);
}
// Merge into a string and tokenize.
$string = implode('', $chars);
$out = trim(search_simplify($string));
$expected = drupal_strtolower(implode(' ', $chars));
// Verify that the output matches what we expect.
$this->assertEqual($out, $expected, 'CJK tokenizer worked on all supplied CJK characters');
}
/**
* Verifies that strings of non-CJK characters are not tokenized.
*
* This is just a sanity check - it verifies that strings of letters are
* not tokenized.
*/
function testNoTokenizer() {
// Set the minimum word size to 1 (to split all CJK characters) and make
// sure CJK tokenizing is turned on.
variable_set('minimum_word_size', 1);
variable_set('overlap_cjk', TRUE);
$this->refreshVariables();
$letters = 'abcdefghijklmnopqrstuvwxyz';
$out = trim(search_simplify($letters));
$this->assertEqual($letters, $out, 'Letters are not CJK tokenized');
}
/**
* Like PHP chr() function, but for unicode characters.
*
* chr() only works for ASCII characters up to character 255. This function
* converts a number to the corresponding unicode character. Adapted from
* functions supplied in comments on several functions on php.net.
*/
function code2utf($num) {
if ($num < 128) {
return chr($num);
}
if ($num < 2048) {
return chr(($num >> 6) + 192) . chr(($num & 63) + 128);
}
if ($num < 65536) {
return chr(($num >> 12) + 224) . chr((($num >> 6) & 63) + 128) . chr(($num & 63) + 128);
}
if ($num < 2097152) {
return chr(($num >> 18) + 240) . chr((($num >> 12) & 63) + 128) . chr((($num >> 6) & 63) + 128) . chr(($num & 63) + 128);
}
return '';
}
}
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment