1 package org.apache.lucene.analysis;
3 import org.apache.lucene.analysis.standard.UAX29URLEmailTokenizer;
4 import org.apache.lucene.analysis.tokenattributes.TypeAttribute;
6 import java.io.BufferedReader;
7 import java.io.IOException;
8 import java.io.InputStreamReader;
10 import java.io.StringReader;
11 import java.util.ArrayList;
12 import java.util.Arrays;
13 import java.util.List;
16 * Licensed to the Apache Software Foundation (ASF) under one or more
17 * contributor license agreements. See the NOTICE file distributed with
18 * this work for additional information regarding copyright ownership.
19 * The ASF licenses this file to You under the Apache License, Version 2.0
20 * (the "License"); you may not use this file except in compliance with
21 * the License. You may obtain a copy of the License at
23 * http://www.apache.org/licenses/LICENSE-2.0
25 * Unless required by applicable law or agreed to in writing, software
26 * distributed under the License is distributed on an "AS IS" BASIS,
27 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
28 * See the License for the specific language governing permissions and
29 * limitations under the License.
32 public class TestUAX29URLEmailTokenizer extends BaseTokenStreamTestCase {
34 public void testHugeDoc() throws IOException {
35 StringBuilder sb = new StringBuilder();
36 char whitespace[] = new char[4094];
37 Arrays.fill(whitespace, ' ');
38 sb.append(whitespace);
39 sb.append("testing 1234");
40 String input = sb.toString();
41 UAX29URLEmailTokenizer tokenizer = new UAX29URLEmailTokenizer(new StringReader(input));
42 BaseTokenStreamTestCase.assertTokenStreamContents(tokenizer, new String[] { "testing", "1234" });
45 private Analyzer a = new ReusableAnalyzerBase() {
47 protected TokenStreamComponents createComponents
48 (String fieldName, Reader reader) {
50 Tokenizer tokenizer = new UAX29URLEmailTokenizer(reader);
51 return new TokenStreamComponents(tokenizer);
56 /** Passes through tokens with type "<URL>" and blocks all other types. */
57 private class URLFilter extends TokenFilter {
58 private final TypeAttribute typeAtt = addAttribute(TypeAttribute.class);
59 public URLFilter(TokenStream in) {
63 public final boolean incrementToken() throws java.io.IOException {
64 boolean isTokenAvailable = false;
65 while (input.incrementToken()) {
66 if (typeAtt.type() == UAX29URLEmailTokenizer.URL_TYPE) {
67 isTokenAvailable = true;
71 return isTokenAvailable;
75 /** Passes through tokens with type "<EMAIL>" and blocks all other types. */
76 private class EmailFilter extends TokenFilter {
77 private final TypeAttribute typeAtt = addAttribute(TypeAttribute.class);
78 public EmailFilter(TokenStream in) {
82 public final boolean incrementToken() throws java.io.IOException {
83 boolean isTokenAvailable = false;
84 while (input.incrementToken()) {
85 if (typeAtt.type() == UAX29URLEmailTokenizer.EMAIL_TYPE) {
86 isTokenAvailable = true;
90 return isTokenAvailable;
94 private Analyzer urlAnalyzer = new ReusableAnalyzerBase() {
96 protected TokenStreamComponents createComponents(String fieldName, Reader reader) {
97 UAX29URLEmailTokenizer tokenizer = new UAX29URLEmailTokenizer(reader);
98 tokenizer.setMaxTokenLength(Integer.MAX_VALUE); // Tokenize arbitrary length URLs
99 TokenFilter filter = new URLFilter(tokenizer);
100 return new TokenStreamComponents(tokenizer, filter);
104 private Analyzer emailAnalyzer = new ReusableAnalyzerBase() {
106 protected TokenStreamComponents createComponents(String fieldName, Reader reader) {
107 UAX29URLEmailTokenizer tokenizer = new UAX29URLEmailTokenizer(reader);
108 TokenFilter filter = new EmailFilter(tokenizer);
109 return new TokenStreamComponents(tokenizer, filter);
114 public void testArmenian() throws Exception {
115 BaseTokenStreamTestCase.assertAnalyzesTo(a, "Վիքիպեդիայի 13 միլիոն հոդվածները (4,600` հայերեն վիքիպեդիայում) գրվել են կամավորների կողմից ու համարյա բոլոր հոդվածները կարող է խմբագրել ցանկաց մարդ ով կարող է բացել Վիքիպեդիայի կայքը։",
116 new String[] { "Վիքիպեդիայի", "13", "միլիոն", "հոդվածները", "4,600", "հայերեն", "վիքիպեդիայում", "գրվել", "են", "կամավորների", "կողմից",
117 "ու", "համարյա", "բոլոր", "հոդվածները", "կարող", "է", "խմբագրել", "ցանկաց", "մարդ", "ով", "կարող", "է", "բացել", "Վիքիպեդիայի", "կայքը" } );
120 public void testAmharic() throws Exception {
121 BaseTokenStreamTestCase.assertAnalyzesTo(a, "ዊኪፔድያ የባለ ብዙ ቋንቋ የተሟላ ትክክለኛና ነጻ መዝገበ ዕውቀት (ኢንሳይክሎፒዲያ) ነው። ማንኛውም",
122 new String[] { "ዊኪፔድያ", "የባለ", "ብዙ", "ቋንቋ", "የተሟላ", "ትክክለኛና", "ነጻ", "መዝገበ", "ዕውቀት", "ኢንሳይክሎፒዲያ", "ነው", "ማንኛውም" } );
125 public void testArabic() throws Exception {
126 BaseTokenStreamTestCase.assertAnalyzesTo(a, "الفيلم الوثائقي الأول عن ويكيبيديا يسمى \"الحقيقة بالأرقام: قصة ويكيبيديا\" (بالإنجليزية: Truth in Numbers: The Wikipedia Story)، سيتم إطلاقه في 2008.",
127 new String[] { "الفيلم", "الوثائقي", "الأول", "عن", "ويكيبيديا", "يسمى", "الحقيقة", "بالأرقام", "قصة", "ويكيبيديا",
128 "بالإنجليزية", "Truth", "in", "Numbers", "The", "Wikipedia", "Story", "سيتم", "إطلاقه", "في", "2008" } );
131 public void testAramaic() throws Exception {
132 BaseTokenStreamTestCase.assertAnalyzesTo(a, "ܘܝܩܝܦܕܝܐ (ܐܢܓܠܝܐ: Wikipedia) ܗܘ ܐܝܢܣܩܠܘܦܕܝܐ ܚܐܪܬܐ ܕܐܢܛܪܢܛ ܒܠܫܢ̈ܐ ܣܓܝܐ̈ܐ܂ ܫܡܗ ܐܬܐ ܡܢ ܡ̈ܠܬܐ ܕ\"ܘܝܩܝ\" ܘ\"ܐܝܢܣܩܠܘܦܕܝܐ\"܀",
133 new String[] { "ܘܝܩܝܦܕܝܐ", "ܐܢܓܠܝܐ", "Wikipedia", "ܗܘ", "ܐܝܢܣܩܠܘܦܕܝܐ", "ܚܐܪܬܐ", "ܕܐܢܛܪܢܛ", "ܒܠܫܢ̈ܐ", "ܣܓܝܐ̈ܐ", "ܫܡܗ",
134 "ܐܬܐ", "ܡܢ", "ܡ̈ܠܬܐ", "ܕ", "ܘܝܩܝ", "ܘ", "ܐܝܢܣܩܠܘܦܕܝܐ"});
137 public void testBengali() throws Exception {
138 BaseTokenStreamTestCase.assertAnalyzesTo(a, "এই বিশ্বকোষ পরিচালনা করে উইকিমিডিয়া ফাউন্ডেশন (একটি অলাভজনক সংস্থা)। উইকিপিডিয়ার শুরু ১৫ জানুয়ারি, ২০০১ সালে। এখন পর্যন্ত ২০০টিরও বেশী ভাষায় উইকিপিডিয়া রয়েছে।",
139 new String[] { "এই", "বিশ্বকোষ", "পরিচালনা", "করে", "উইকিমিডিয়া", "ফাউন্ডেশন", "একটি", "অলাভজনক", "সংস্থা", "উইকিপিডিয়ার",
140 "শুরু", "১৫", "জানুয়ারি", "২০০১", "সালে", "এখন", "পর্যন্ত", "২০০টিরও", "বেশী", "ভাষায়", "উইকিপিডিয়া", "রয়েছে" });
143 public void testFarsi() throws Exception {
144 BaseTokenStreamTestCase.assertAnalyzesTo(a, "ویکی پدیای انگلیسی در تاریخ ۲۵ دی ۱۳۷۹ به صورت مکملی برای دانشنامهٔ تخصصی نوپدیا نوشته شد.",
145 new String[] { "ویکی", "پدیای", "انگلیسی", "در", "تاریخ", "۲۵", "دی", "۱۳۷۹", "به", "صورت", "مکملی",
146 "برای", "دانشنامهٔ", "تخصصی", "نوپدیا", "نوشته", "شد" });
149 public void testGreek() throws Exception {
150 BaseTokenStreamTestCase.assertAnalyzesTo(a, "Γράφεται σε συνεργασία από εθελοντές με το λογισμικό wiki, κάτι που σημαίνει ότι άρθρα μπορεί να προστεθούν ή να αλλάξουν από τον καθένα.",
151 new String[] { "Γράφεται", "σε", "συνεργασία", "από", "εθελοντές", "με", "το", "λογισμικό", "wiki", "κάτι", "που",
152 "σημαίνει", "ότι", "άρθρα", "μπορεί", "να", "προστεθούν", "ή", "να", "αλλάξουν", "από", "τον", "καθένα" });
155 public void testThai() throws Exception {
156 BaseTokenStreamTestCase.assertAnalyzesTo(a, "การที่ได้ต้องแสดงว่างานดี. แล้วเธอจะไปไหน? ๑๒๓๔",
157 new String[] { "การที่ได้ต้องแสดงว่างานดี", "แล้วเธอจะไปไหน", "๑๒๓๔" });
160 public void testLao() throws Exception {
161 BaseTokenStreamTestCase.assertAnalyzesTo(a, "ສາທາລະນະລັດ ປະຊາທິປະໄຕ ປະຊາຊົນລາວ",
162 new String[] { "ສາທາລະນະລັດ", "ປະຊາທິປະໄຕ", "ປະຊາຊົນລາວ" });
165 public void testTibetan() throws Exception {
166 BaseTokenStreamTestCase.assertAnalyzesTo(a, "སྣོན་མཛོད་དང་ལས་འདིས་བོད་ཡིག་མི་ཉམས་གོང་འཕེལ་དུ་གཏོང་བར་ཧ་ཅང་དགེ་མཚན་མཆིས་སོ། །",
167 new String[] { "སྣོན", "མཛོད", "དང", "ལས", "འདིས", "བོད", "ཡིག",
168 "མི", "ཉམས", "གོང", "འཕེལ", "དུ", "གཏོང", "བར",
169 "ཧ", "ཅང", "དགེ", "མཚན", "མཆིས", "སོ" });
173 * For chinese, tokenize as char (these can later form bigrams or whatever)
175 public void testChinese() throws Exception {
176 BaseTokenStreamTestCase.assertAnalyzesTo(a, "我是中国人。 1234 Tests ",
177 new String[] { "我", "是", "中", "国", "人", "1234", "Tests"});
180 public void testEmpty() throws Exception {
181 BaseTokenStreamTestCase.assertAnalyzesTo(a, "", new String[] {});
182 BaseTokenStreamTestCase.assertAnalyzesTo(a, ".", new String[] {});
183 BaseTokenStreamTestCase.assertAnalyzesTo(a, " ", new String[] {});
186 /* test various jira issues this analyzer is related to */
188 public void testLUCENE1545() throws Exception {
190 * Standard analyzer does not correctly tokenize combining character U+0364 COMBINING LATIN SMALL LETTRE E.
191 * The word "moͤchte" is incorrectly tokenized into "mo" "chte", the combining character is lost.
192 * Expected result is only on token "moͤchte".
194 BaseTokenStreamTestCase.assertAnalyzesTo(a, "moͤchte", new String[] { "moͤchte" });
197 /* Tests from StandardAnalyzer, just to show behavior is similar */
198 public void testAlphanumericSA() throws Exception {
199 // alphanumeric tokens
200 BaseTokenStreamTestCase.assertAnalyzesTo(a, "B2B", new String[]{"B2B"});
201 BaseTokenStreamTestCase.assertAnalyzesTo(a, "2B", new String[]{"2B"});
204 public void testDelimitersSA() throws Exception {
205 // other delimiters: "-", "/", ","
206 BaseTokenStreamTestCase.assertAnalyzesTo(a, "some-dashed-phrase", new String[]{"some", "dashed", "phrase"});
207 BaseTokenStreamTestCase.assertAnalyzesTo(a, "dogs,chase,cats", new String[]{"dogs", "chase", "cats"});
208 BaseTokenStreamTestCase.assertAnalyzesTo(a, "ac/dc", new String[]{"ac", "dc"});
211 public void testApostrophesSA() throws Exception {
212 // internal apostrophes: O'Reilly, you're, O'Reilly's
213 BaseTokenStreamTestCase.assertAnalyzesTo(a, "O'Reilly", new String[]{"O'Reilly"});
214 BaseTokenStreamTestCase.assertAnalyzesTo(a, "you're", new String[]{"you're"});
215 BaseTokenStreamTestCase.assertAnalyzesTo(a, "she's", new String[]{"she's"});
216 BaseTokenStreamTestCase.assertAnalyzesTo(a, "Jim's", new String[]{"Jim's"});
217 BaseTokenStreamTestCase.assertAnalyzesTo(a, "don't", new String[]{"don't"});
218 BaseTokenStreamTestCase.assertAnalyzesTo(a, "O'Reilly's", new String[]{"O'Reilly's"});
221 public void testNumericSA() throws Exception {
222 // floating point, serial, model numbers, ip addresses, etc.
223 BaseTokenStreamTestCase.assertAnalyzesTo(a, "21.35", new String[]{"21.35"});
224 BaseTokenStreamTestCase.assertAnalyzesTo(a, "R2D2 C3PO", new String[]{"R2D2", "C3PO"});
225 BaseTokenStreamTestCase.assertAnalyzesTo(a, "216.239.63.104", new String[]{"216.239.63.104"});
228 public void testTextWithNumbersSA() throws Exception {
230 BaseTokenStreamTestCase.assertAnalyzesTo(a, "David has 5000 bones", new String[]{"David", "has", "5000", "bones"});
233 public void testVariousTextSA() throws Exception {
235 BaseTokenStreamTestCase.assertAnalyzesTo(a, "C embedded developers wanted", new String[]{"C", "embedded", "developers", "wanted"});
236 BaseTokenStreamTestCase.assertAnalyzesTo(a, "foo bar FOO BAR", new String[]{"foo", "bar", "FOO", "BAR"});
237 BaseTokenStreamTestCase.assertAnalyzesTo(a, "foo bar . FOO <> BAR", new String[]{"foo", "bar", "FOO", "BAR"});
238 BaseTokenStreamTestCase.assertAnalyzesTo(a, "\"QUOTED\" word", new String[]{"QUOTED", "word"});
241 public void testKoreanSA() throws Exception {
243 BaseTokenStreamTestCase.assertAnalyzesTo(a, "안녕하세요 한글입니다", new String[]{"안녕하세요", "한글입니다"});
246 public void testOffsets() throws Exception {
247 BaseTokenStreamTestCase.assertAnalyzesTo(a, "David has 5000 bones",
248 new String[] {"David", "has", "5000", "bones"},
249 new int[] {0, 6, 10, 15},
250 new int[] {5, 9, 14, 20});
253 public void testTypes() throws Exception {
254 BaseTokenStreamTestCase.assertAnalyzesTo(a, "David has 5000 bones",
255 new String[] {"David", "has", "5000", "bones"},
256 new String[] { "<ALPHANUM>", "<ALPHANUM>", "<NUM>", "<ALPHANUM>" });
259 public void testWikiURLs() throws Exception {
260 Reader reader = null;
261 String luceneResourcesWikiPage;
263 reader = new InputStreamReader(getClass().getResourceAsStream
264 ("LuceneResourcesWikiPage.html"), "UTF-8");
265 StringBuilder builder = new StringBuilder();
266 char[] buffer = new char[1024];
268 while (-1 != (numCharsRead = reader.read(buffer))) {
269 builder.append(buffer, 0, numCharsRead);
271 luceneResourcesWikiPage = builder.toString();
273 if (null != reader) {
277 assertTrue(null != luceneResourcesWikiPage
278 && luceneResourcesWikiPage.length() > 0);
279 BufferedReader bufferedReader = null;
282 List<String> urlList = new ArrayList<String>();
283 bufferedReader = new BufferedReader(new InputStreamReader
284 (getClass().getResourceAsStream("LuceneResourcesWikiPageURLs.txt"), "UTF-8"));
286 while (null != (line = bufferedReader.readLine())) {
288 if (line.length() > 0) {
292 urls = urlList.toArray(new String[urlList.size()]);
294 if (null != bufferedReader) {
295 bufferedReader.close();
298 assertTrue(null != urls && urls.length > 0);
299 BaseTokenStreamTestCase.assertAnalyzesTo
300 (urlAnalyzer, luceneResourcesWikiPage, urls);
303 public void testEmails() throws Exception {
304 Reader reader = null;
305 String randomTextWithEmails;
307 reader = new InputStreamReader(getClass().getResourceAsStream
308 ("random.text.with.email.addresses.txt"), "UTF-8");
309 StringBuilder builder = new StringBuilder();
310 char[] buffer = new char[1024];
312 while (-1 != (numCharsRead = reader.read(buffer))) {
313 builder.append(buffer, 0, numCharsRead);
315 randomTextWithEmails = builder.toString();
317 if (null != reader) {
321 assertTrue(null != randomTextWithEmails
322 && randomTextWithEmails.length() > 0);
323 BufferedReader bufferedReader = null;
326 List<String> emailList = new ArrayList<String>();
327 bufferedReader = new BufferedReader(new InputStreamReader
328 (getClass().getResourceAsStream
329 ("email.addresses.from.random.text.with.email.addresses.txt"), "UTF-8"));
331 while (null != (line = bufferedReader.readLine())) {
333 if (line.length() > 0) {
337 emails = emailList.toArray(new String[emailList.size()]);
339 if (null != bufferedReader) {
340 bufferedReader.close();
343 assertTrue(null != emails && emails.length > 0);
344 BaseTokenStreamTestCase.assertAnalyzesTo
345 (emailAnalyzer, randomTextWithEmails, emails);
348 public void testURLs() throws Exception {
349 Reader reader = null;
350 String randomTextWithURLs;
352 reader = new InputStreamReader(getClass().getResourceAsStream
353 ("random.text.with.urls.txt"), "UTF-8");
354 StringBuilder builder = new StringBuilder();
355 char[] buffer = new char[1024];
357 while (-1 != (numCharsRead = reader.read(buffer))) {
358 builder.append(buffer, 0, numCharsRead);
360 randomTextWithURLs = builder.toString();
362 if (null != reader) {
366 assertTrue(null != randomTextWithURLs
367 && randomTextWithURLs.length() > 0);
368 BufferedReader bufferedReader = null;
371 List<String> urlList = new ArrayList<String>();
372 bufferedReader = new BufferedReader(new InputStreamReader
373 (getClass().getResourceAsStream
374 ("urls.from.random.text.with.urls.txt"), "UTF-8"));
376 while (null != (line = bufferedReader.readLine())) {
378 if (line.length() > 0) {
382 urls = urlList.toArray(new String[urlList.size()]);
384 if (null != bufferedReader) {
385 bufferedReader.close();
388 assertTrue(null != urls && urls.length > 0);
389 BaseTokenStreamTestCase.assertAnalyzesTo
390 (urlAnalyzer, randomTextWithURLs, urls);
393 public void testUnicodeWordBreaks() throws Exception {
394 WordBreakTestUnicode_6_0_0 wordBreakTest = new WordBreakTestUnicode_6_0_0();
395 wordBreakTest.test(a);
398 public void testSupplementary() throws Exception {
399 BaseTokenStreamTestCase.assertAnalyzesTo(a, "𩬅艱鍟䇹愯瀛",
400 new String[] {"𩬅", "艱", "鍟", "䇹", "愯", "瀛"},
401 new String[] { "<IDEOGRAPHIC>", "<IDEOGRAPHIC>", "<IDEOGRAPHIC>", "<IDEOGRAPHIC>", "<IDEOGRAPHIC>", "<IDEOGRAPHIC>" });
404 public void testKorean() throws Exception {
405 BaseTokenStreamTestCase.assertAnalyzesTo(a, "훈민정음",
406 new String[] { "훈민정음" },
407 new String[] { "<HANGUL>" });
410 public void testJapanese() throws Exception {
411 BaseTokenStreamTestCase.assertAnalyzesTo(a, "仮名遣い カタカナ",
412 new String[] { "仮", "名", "遣", "い", "カタカナ" },
413 new String[] { "<IDEOGRAPHIC>", "<IDEOGRAPHIC>", "<IDEOGRAPHIC>", "<HIRAGANA>", "<KATAKANA>" });
416 /** blast some random strings through the analyzer */
417 public void testRandomStrings() throws Exception {
418 checkRandomData(random, a, 10000*RANDOM_MULTIPLIER);