[TASK] Do not set GeneralUtility::$container in unit tests
[Packages/TYPO3.CMS.git] / typo3 / sysext / indexed_search / Classes / Lexer.php
1 <?php
2 namespace TYPO3\CMS\IndexedSearch;
3
4 /*
5 * This file is part of the TYPO3 CMS project.
6 *
7 * It is free software; you can redistribute it and/or modify it under
8 * the terms of the GNU General Public License, either version 2
9 * of the License, or any later version.
10 *
11 * For the full copyright and license information, please read the
12 * LICENSE.txt file that was distributed with this source code.
13 *
14 * The TYPO3 project - inspiring people to share!
15 */
16
17 use TYPO3\CMS\Core\Charset\CharsetConverter;
18 use TYPO3\CMS\Core\Utility\GeneralUtility;
19
20 /**
21 * Lexer class for indexed_search
22 * A lexer splits the text into words
23 * @internal
24 */
25 class Lexer
26 {
27
28 /**
29 * Debugging options:
30 *
31 * @var bool
32 */
33 public $debug = false;
34
35 /**
36 * If set, the debugString is filled with HTML output highlighting search / non-search words (for backend display)
37 *
38 * @var string
39 */
40 public $debugString = '';
41
42 /**
43 * Configuration of the lexer:
44 *
45 * @var array
46 */
47 public $lexerConf = [
48 //Characters: . - _ : / '
49 'printjoins' => [46, 45, 95, 58, 47, 39],
50 'casesensitive' => false,
51 // Set, if case sensitive indexing is wanted.
52 'removeChars' => [45]
53 ];
54
55 /**
56 * Splitting string into words.
57 * Used for indexing, can also be used to find words in query.
58 *
59 * @param string String with UTF-8 content to process.
60 * @return array Array of words in utf-8
61 */
62 public function split2Words($wordString)
63 {
64 // Reset debug string:
65 $this->debugString = '';
66 // Then convert the string to lowercase:
67 if (!$this->lexerConf['casesensitive']) {
68 $wordString = mb_strtolower($wordString, 'utf-8');
69 }
70 // Now, splitting words:
71 $pos = 0;
72 $words = [];
73 $this->debugString = '';
74 while (1) {
75 list($start, $len) = $this->get_word($wordString, $pos);
76 if ($len) {
77 $this->addWords($words, $wordString, $start, $len);
78 if ($this->debug) {
79 $this->debugString .= '<span style="color:red">' . htmlspecialchars(substr(
80 $wordString,
81 $pos,
82 $start - $pos
83 )) . '</span>' . htmlspecialchars(substr($wordString, $start, $len));
84 }
85 $pos = $start + $len;
86 } else {
87 break;
88 }
89 }
90 return $words;
91 }
92
93 /**********************************
94 *
95 * Helper functions
96 *
97 ********************************/
98 /**
99 * Add word to word-array
100 * This function should be used to make sure CJK sequences are split up in the right way
101 *
102 * @param array $words Array of accumulated words
103 * @param string $wordString Complete Input string from where to extract word
104 * @param int $start Start position of word in input string
105 * @param int $len The Length of the word string from start position
106 */
107 public function addWords(&$words, &$wordString, $start, $len)
108 {
109 // Get word out of string:
110 $theWord = substr($wordString, $start, $len);
111 // Get next chars unicode number and find type:
112 $bc = 0;
113 $cp = $this->utf8_ord($theWord, $bc);
114 list($cType) = $this->charType($cp);
115 // If string is a CJK sequence we follow this algorithm:
116 /*
117 DESCRIPTION OF (CJK) ALGORITHMContinuous letters and numbers make up words. Spaces and symbols
118 separate letters and numbers into words. This is sufficient for
119 all western text.CJK doesn't use spaces or separators to separate words, so the only
120 way to really find out what constitutes a word would be to have a
121 dictionary and advanced heuristics. Instead, we form pairs from
122 consecutive characters, in such a way that searches will find only
123 characters that appear more-or-less the right sequence. For example:ABCDE => AB BC CD DEThis works okay since both the index and the search query is split
124 in the same manner, and since the set of characters is huge so the
125 extra matches are not significant.(Hint taken from ZOPEs chinese user group)[Kasper: As far as I can see this will only work well with or-searches!]
126 */
127 if ($cType === 'cjk') {
128 // Find total string length:
129 $strlen = mb_strlen($theWord, 'utf-8');
130 // Traverse string length and add words as pairs of two chars:
131 for ($a = 0; $a < $strlen; $a++) {
132 if ($strlen == 1 || $a < $strlen - 1) {
133 $words[] = mb_substr($theWord, $a, 2, 'utf-8');
134 }
135 }
136 } else {
137 // Normal "single-byte" chars:
138 // Remove chars:
139 $charsetConverter = GeneralUtility::makeInstance(CharsetConverter::class);
140 foreach ($this->lexerConf['removeChars'] as $skipJoin) {
141 $theWord = str_replace($charsetConverter->UnumberToChar($skipJoin), '', $theWord);
142 }
143 // Add word:
144 $words[] = $theWord;
145 }
146 }
147
148 /**
149 * Get the first word in a given utf-8 string (initial non-letters will be skipped)
150 *
151 * @param string $str Input string (reference)
152 * @param int $pos Starting position in input string
153 * @return array 0: start, 1: len or FALSE if no word has been found
154 */
155 public function get_word(&$str, $pos = 0)
156 {
157 $len = 0;
158 // If return is TRUE, a word was found starting at this position, so returning position and length:
159 if ($this->utf8_is_letter($str, $len, $pos)) {
160 return [$pos, $len];
161 }
162 // If the return value was FALSE it means a sequence of non-word chars were found (or blank string) - so we will start another search for the word:
163 $pos += $len;
164 if ($str[$pos] == '') {
165 // Check end of string before looking for word of course.
166 return false;
167 }
168 $this->utf8_is_letter($str, $len, $pos);
169 return [$pos, $len];
170 }
171
172 /**
173 * See if a character is a letter (or a string of letters or non-letters).
174 *
175 * @param string $str Input string (reference)
176 * @param int $len Byte-length of character sequence (reference, return value)
177 * @param int $pos Starting position in input string
178 * @return bool letter (or word) found
179 */
180 public function utf8_is_letter(&$str, &$len, $pos = 0)
181 {
182 $len = 0;
183 $bc = 0;
184 $cp = 0;
185 $printJoinLgd = 0;
186 $cType = ($cType_prev = false);
187 // Letter type
188 $letter = true;
189 // looking for a letter?
190 if ($str[$pos] == '') {
191 // Return FALSE on end-of-string at this stage
192 return false;
193 }
194 while (1) {
195 // If characters has been obtained we will know whether the string starts as a sequence of letters or not:
196 if ($len) {
197 if ($letter) {
198 // We are in a sequence of words
199 if (
200 !$cType
201 || $cType_prev === 'cjk' && ($cType === 'num' || $cType === 'alpha')
202 || $cType === 'cjk' && ($cType_prev === 'num' || $cType_prev === 'alpha')
203 ) {
204 // Check if the non-letter char is NOT a print-join char because then it signifies the end of the word.
205 if (!in_array($cp, $this->lexerConf['printjoins'])) {
206 // If a printjoin start length has been recorded, set that back now so the length is right (filtering out multiple end chars)
207 if ($printJoinLgd) {
208 $len = $printJoinLgd;
209 }
210 return true;
211 }
212 // If a printJoin char is found, record the length if it has not been recorded already:
213 if (!$printJoinLgd) {
214 $printJoinLgd = $len;
215 }
216 } else {
217 // When a true letter is found, reset printJoinLgd counter:
218 $printJoinLgd = 0;
219 }
220 } elseif (!$letter && $cType) {
221 // end of non-word reached
222 return false;
223 }
224 }
225 $len += $bc;
226 // add byte-length of last found character
227 if ($str[$pos] == '') {
228 // End of string; return status of string till now
229 return $letter;
230 }
231 // Get next chars unicode number:
232 $cp = $this->utf8_ord($str, $bc, $pos);
233 $pos += $bc;
234 // Determine the type:
235 $cType_prev = $cType;
236 list($cType) = $this->charType($cp);
237 if ($cType) {
238 continue;
239 }
240 // Setting letter to FALSE if the first char was not a letter!
241 if (!$len) {
242 $letter = false;
243 }
244 }
245 return false;
246 }
247
248 /**
249 * Determine the type of character
250 *
251 * @param int $cp Unicode number to evaluate
252 * @return array Type of char; index-0: the main type: num, alpha or CJK (Chinese / Japanese / Korean)
253 */
254 public function charType($cp)
255 {
256 // Numeric?
257 if ($cp >= 48 && $cp <= 57) {
258 return ['num'];
259 }
260 // LOOKING for Alpha chars (Latin, Cyrillic, Greek, Hebrew and Arabic):
261 if ($cp >= 65 && $cp <= 90 || $cp >= 97 && $cp <= 122 || $cp >= 192 && $cp <= 255 && $cp != 215 && $cp != 247 || $cp >= 256 && $cp < 640 || ($cp == 902 || $cp >= 904 && $cp < 1024) || ($cp >= 1024 && $cp < 1154 || $cp >= 1162 && $cp < 1328) || ($cp >= 1424 && $cp < 1456 || $cp >= 1488 && $cp < 1523) || ($cp >= 1569 && $cp <= 1624 || $cp >= 1646 && $cp <= 1747) || $cp >= 7680 && $cp < 8192) {
262 return ['alpha'];
263 }
264 // Looking for CJK (Chinese / Japanese / Korean)
265 // Ranges are not certain - deducted from the translation tables in typo3/sysext/core/Resources/Private/Charsets/csconvtbl/
266 // Verified with http://www.unicode.org/charts/ (16/2) - may still not be complete.
267 if ($cp >= 12352 && $cp <= 12543 || $cp >= 12592 && $cp <= 12687 || $cp >= 13312 && $cp <= 19903 || $cp >= 19968 && $cp <= 40879 || $cp >= 44032 && $cp <= 55215 || $cp >= 131072 && $cp <= 195103) {
268 return ['cjk'];
269 }
270 }
271
272 /**
273 * Converts a UTF-8 multibyte character to a UNICODE codepoint
274 *
275 * @param string $str UTF-8 multibyte character string (reference)
276 * @param int $len The length of the character (reference, return value)
277 * @param int $pos Starting position in input string
278 * @param bool $hex If set, then a hex. number is returned
279 * @return int UNICODE codepoint
280 */
281 public function utf8_ord(&$str, &$len, $pos = 0, $hex = false)
282 {
283 $ord = ord($str[$pos]);
284 $len = 1;
285 if ($ord > 128) {
286 for ($bc = -1, $mbs = $ord; $mbs & 128; $mbs = $mbs << 1) {
287 // calculate number of extra bytes
288 $bc++;
289 }
290 $len += $bc;
291 $ord = $ord & (1 << 6 - $bc) - 1;
292 // mask utf-8 lead-in bytes
293 // "bring in" data bytes
294 for ($i = $pos + 1; $bc; $bc--, $i++) {
295 $ord = $ord << 6 | ord($str[$i]) & 63;
296 }
297 }
298 return $hex ? 'x' . dechex($ord) : $ord;
299 }
300 }