private <T> void assertExpectedTokens(LuceneAnalyzer analyzer, List<T> testData) {
JavaRDD<T> rdd = jsc.parallelize(testData);
Row[] pairs =
analyzer
.transform(jsql.createDataFrame(rdd, testData.get(0).getClass()))
.select("wantedTokens", "tokens")
.collect();
for (Row r : pairs) {
Assert.assertEquals(r.get(0), r.get(1));
}
}
@Test
public void testMissingValues() {
LuceneAnalyzer analyzer = new LuceneAnalyzer().setInputCol("rawText").setOutputCol("tokens");
assertExpectedTokens(analyzer, Arrays.asList(new TokenizerTestData(null, new String[] {})));
assertExpectedTokens(analyzer, Arrays.asList(new TokenizerTestData("", new String[] {})));
assertExpectedTokens(
analyzer,
Collections.singletonList(
new MV_TokenizerTestData(
new String[] {null, "Harold's not around.", null, "The dog's nose KNOWS!", ""},
new String[] {"harold's", "not", "around", "the", "dog's", "nose", "knows"})));
analyzer.setInputCols(new String[] {"rawText1", "rawText2", "rawText3"});
assertExpectedTokens(
analyzer,
Collections.singletonList(
new SV_SV_SV_TokenizerTestData(
"",
"The dog's nose KNOWS!",
null,
new String[] {"the", "dog's", "nose", "knows"})));
}
@Test
public void testPrefixTokensWithInputCol() {
String[] rawText1 = new String[] {"Harold's NOT around.", "Anymore, I mean."};
String[] tokens1 = new String[] {"harold's", "not", "around", "anymore", "i", "mean"};
String[] rawText2 = new String[] {"The dog's nose KNOWS!", "Good, fine, great..."};
String[] tokens2 = new String[] {"the", "dog's", "nose", "knows", "good", "fine", "great"};
List<String> tokenList = new ArrayList<>();
List<String> prefixedTokenList = new ArrayList<>();
for (String token : tokens1) {
tokenList.add(token);
prefixedTokenList.add("rawText1=" + token);
}
for (String token : tokens2) {
tokenList.add(token);
prefixedTokenList.add("rawText2=" + token);
}
String[] tokens = tokenList.toArray(new String[tokenList.size()]);
String[] prefixedTokens = prefixedTokenList.toArray(new String[prefixedTokenList.size()]);
// First transform without token prefixes
LuceneAnalyzer analyzer =
new LuceneAnalyzer()
.setInputCols(new String[] {"rawText1", "rawText2"})
.setOutputCol("tokens");
assertExpectedTokens(
analyzer,
Collections.singletonList(new MV_MV_TokenizerTestData(rawText1, rawText2, tokens)));
// Then transform with token prefixes
analyzer.setPrefixTokensWithInputCol(true);
assertExpectedTokens(
analyzer,
Collections.singletonList(new MV_MV_TokenizerTestData(rawText1, rawText2, prefixedTokens)));
}
@Test
public void testStandardTokenizer() {
LuceneAnalyzer analyzer1 =
new LuceneAnalyzer()
.setInputCol("rawText")
.setOutputCol("tokens"); // Default analysis schema: StandardTokenizer + LowerCaseFilter
assertExpectedTokens(
analyzer1,
Arrays.asList(
new TokenizerTestData(
"Test for tokenization.", new String[] {"test", "for", "tokenization"}),
new TokenizerTestData("Te,st. punct", new String[] {"te", "st", "punct"})));
assertExpectedTokens(
analyzer1,
Arrays.asList(
new TokenizerTestData(
"Test for tokenization.", new String[] {"test", "for", "tokenization"}),
new TokenizerTestData("Te,st. punct", new String[] {"te", "st", "punct"})));
String analysisSchema1 =
json(
"{\n"
+ "'schemaType': 'LuceneAnalyzerSchema.v1',\n"
+ "'analyzers': [{\n"
+ " 'name': 'StdTok_max10',\n"
+ " 'tokenizer': {\n"
+ " 'type': 'standard',\n"
+ " 'maxTokenLength': '10'\n"
+ " }\n"
+ "}],\n"
+ "'inputColumns': [{\n"
+ " 'regex': '.+',\n"
+ " 'analyzer': 'StdTok_max10'\n"
+ "}]}\n");
analyzer1.setAnalysisSchema(analysisSchema1);
assertExpectedTokens(
analyzer1,
Arrays.asList(
new TokenizerTestData(
"我是中国人。 １２３４ Ｔｅｓｔｓ ", new String[] {"我", "是", "中", "国", "人", "１２３４", "Ｔｅｓｔｓ"}),
new TokenizerTestData(
"some-dashed-phrase", new String[] {"some", "dashed", "phrase"})));
String analysisSchema2 =
json(
"{\n"
+ "'schemaType': 'LuceneAnalyzerSchema.v1',\n"
+ "'defaultLuceneMatchVersion': '4.10.4',\n"
+ "'analyzers': [{\n"
+ " 'name': 'StdTok_max3',\n"
+ " 'tokenizer': {\n"
+ " 'type': 'standard',\n"
+ " 'maxTokenLength': '3'\n"
+ " }\n"
+ "}],\n"
+ "'inputColumns': [{\n"
+ " 'regex': '.+',\n"
+ " 'analyzer': 'StdTok_max3'\n"
+ "}]}\n");
LuceneAnalyzer analyzer2 =
new LuceneAnalyzer()
.setAnalysisSchema(analysisSchema2)
.setInputCol("rawText")
.setOutputCol("tokens");
assertExpectedTokens(
analyzer2,
Arrays.asList(
new TokenizerTestData(
"Test for tokenization.",
new String[] {"Tes", "t", "for", "tok", "eni", "zat", "ion"}),
new TokenizerTestData("Te,st. punct", new String[] {"Te", "st", "pun", "ct"})));
}
@Test
public void testMultipleInputCols() {
LuceneAnalyzer analyzer1 =
new LuceneAnalyzer()
.setInputCols(new String[] {"rawText1", "rawText2"})
.setOutputCol("tokens");
assertExpectedTokens(
analyzer1,
Collections.singletonList(
new SV_SV_TokenizerTestData(
"Harold's not around.",
"The dog's nose KNOWS!",
new String[] {"harold's", "not", "around", "the", "dog's", "nose", "knows"})));
String analysisSchema =
json(
"{\n"
+ "'schemaType': 'LuceneAnalyzerSchema.v1',\n"
+ "'analyzers': [{\n"
+ " 'name': 'std_tok_lower',\n"
+ " 'tokenizer': { 'type': 'standard' },\n"
+ " 'filters':[{ 'type': 'lowercase' }]\n"
+ " }, {\n"
+ " 'name': 'std_tok',\n"
+ " 'tokenizer': { 'type': 'standard' }\n"
+ " }, {\n"
+ " 'name': 'htmlstrip_std_tok_lower',\n"
+ " 'charFilters': [{ 'type': 'htmlstrip' }],\n"
+ " 'tokenizer': { 'type': 'standard' },\n"
+ " 'filters': [{ 'type': 'lowercase' }]\n"
+ "}],\n"
+ "'inputColumns': [{\n"
+ " 'name': 'rawText1',\n"
+ " 'analyzer': 'std_tok_lower'\n"
+ " }, {\n"
+ " 'name': 'rawText2',\n"
+ " 'analyzer': 'std_tok'\n"
+ " }, {\n"
+ " 'regex': '.+',\n"
+ " 'analyzer': 'htmlstrip_std_tok_lower'\n"
+ "}]}\n");
LuceneAnalyzer analyzer2 =
new LuceneAnalyzer()
.setAnalysisSchema(analysisSchema)
.setInputCols(new String[] {"rawText1", "rawText2"})
.setOutputCol("tokens");
assertExpectedTokens(
analyzer2,
Collections.singletonList(
new SV_SV_TokenizerTestData(
"Harold's NOT around.",
"The dog's nose KNOWS!",
new String[] {"harold's", "not", "around", "The", "dog's", "nose", "KNOWS"})));
assertExpectedTokens(
analyzer2,
Collections.singletonList(
new SV_MV_TokenizerTestData(
"Harold's NOT around.",
new String[] {"The dog's nose KNOWS!", "Good, fine, great..."},
new String[] {
"harold's",
"not",
"around",
"The",
"dog's",
"nose",
"KNOWS",
"Good",
"fine",
"great"
})));
assertExpectedTokens(
analyzer2,
Collections.singletonList(
new MV_MV_TokenizerTestData(
new String[] {"Harold's NOT around.", "Anymore, I mean."},
new String[] {"The dog's nose KNOWS!", "Good, fine, great..."},
new String[] {
"harold's",
"not",
"around",
"anymore",
"i",
"mean",
"The",
"dog's",
"nose",
"KNOWS",
"Good",
"fine",
"great"
})));
analyzer2.setInputCols(new String[] {"rawText1", "rawText2", "rawText3"});
assertExpectedTokens(
analyzer2,
Collections.singletonList(
new SV_SV_SV_TokenizerTestData(
"Harold's NOT around.",
"The dog's nose KNOWS!",
"<html><body>Content</body></html>",
new String[] {
"harold's", "not", "around", "The", "dog's", "nose", "KNOWS", "content"
})));
}
@Test
@SuppressWarnings("unchecked")
public void testCharFilters() {
String analysisSchema1 =
json(
"{\n"
+ "'schemaType': 'LuceneAnalyzerSchema.v1',\n"
+ "'analyzers': [{\n"
+ " 'name': 'strip_alpha_std_tok',\n"
+ " 'charFilters': [{\n"
+ " 'type': 'patternreplace',\n"
+ " 'pattern': '[A-Za-z]+',\n"
+ " 'replacement': ''\n"
+ " }],\n"
+ " 'tokenizer': {\n"
+ " 'type': 'standard'\n"
+ " }\n"
+ "}],\n"
+ "'inputColumns': [{\n"
+ " 'regex': '.+',\n"
+ " 'analyzer': 'strip_alpha_std_tok'\n"
+ "}]}\n");
LuceneAnalyzer analyzer =
new LuceneAnalyzer()
.setAnalysisSchema(analysisSchema1)
.setInputCol("rawText")
.setOutputCol("tokens");
assertExpectedTokens(
analyzer,
Arrays.asList(
new TokenizerTestData("Test for 9983, tokenization.", new String[] {"9983"}),
new TokenizerTestData("Te,st. punct", new String[] {})));
String analysisSchema2 =
json(
"{\n"
+ "'schemaType': 'LuceneAnalyzerSchema.v1',\n"
+ "'analyzers': [{\n"
+ " 'name': 'htmlstrip_drop_removeme_std_tok',\n"
+ " 'charFilters': [{\n"
+ " 'type': 'htmlstrip'\n"
+ " }, {\n"
+ " 'type': 'patternreplace',\n"
+ " 'pattern': 'removeme',\n"
+ " 'replacement': ''\n"
+ " }],\n"
+ " 'tokenizer': {\n"
+ " 'type': 'standard'\n"
+ " }\n"
+ "}],\n"
+ "'inputColumns': [{\n"
+ " 'name': 'rawText',\n"
+ " 'analyzer': 'htmlstrip_drop_removeme_std_tok'\n"
+ "}]}\n");
analyzer.setAnalysisSchema(analysisSchema2);
assertExpectedTokens(
analyzer,
Collections.singletonList(
new TokenizerTestData(
"<html><body>remove<b>me</b> but leave<div>the rest.</div></body></html>",
new String[] {"but", "leave", "the", "rest"})));
}