private void verify(List<String> tokens, boolean[] isHashtag) {
for (int i = 0; i < tokens.size(); i++) {
assertTrue(stream.incrementToken());
assertEquals(tokens.get(i), termAttr.getTermString());
assertEquals(isHashtag[i], TokenType.HASHTAG.equals(typeAttr.getType()));
}
assertFalse(stream.incrementToken());
}
@Before
public void setup() {
stream =
new HashtagTokenCombiner(
// This toknizes text into alphabet-only, number-only tokens
// and '#', '_'.
new RegexExtractor.Builder()
.setRegexPattern(Pattern.compile("([0-9]+|[a-zA-Z]+|\\p{InKatakana}+|#|_)"))
.build());
termAttr = stream.getAttribute(CharSequenceTermAttribute.class);
typeAttr = stream.getAttribute(TokenTypeAttribute.class);
}
@Test
public void testMultipleHashtags() {
String text = "#this #is #a #hashtag";
stream.reset(text);
verify(
ImmutableList.of("#this", "#is", "#a", "#hashtag"), new boolean[] {true, true, true, true});
}
@Test
public void testNotHashtag() {
String text = "this is also not#hashtag";
stream.reset(text);
verify(
ImmutableList.of("this", "is", "also", "not", "#", "hashtag"),
new boolean[] {false, false, false, false, false, false});
}
@Test
public void testJAHashtag() {
String text = "this is now #ハッシュタグ";
stream.reset(text);
verify(
ImmutableList.of("this", "is", "now", "#ハッシュタグ"),
new boolean[] {false, false, false, true});
}
@Test
public void testSingleHashtag() {
String text = "this is a #hashtag";
stream.reset(text);
verify(
ImmutableList.of("this", "is", "a", "#hashtag"), new boolean[] {false, false, false, true});
}
@Test
public void testHashtagWithMultipleTokens() {
String text = "#hash_tag_123";
stream.reset(text);
verify(ImmutableList.of("#hash_tag_123"), new boolean[] {true});
}