use of com.linkedin.pinot.core.segment.creator.impl.SegmentDictionaryCreator in project pinot by linkedin.
the class DictionariesTest method testUTF8Characters.
/**
* Test for ensuring that Strings with special characters can be handled
* correctly.
*
* @throws Exception
*/
@Test
public void testUTF8Characters() throws Exception {
File indexDir = new File("/tmp/dict.test");
indexDir.deleteOnExit();
FieldSpec fieldSpec = new DimensionFieldSpec("test", DataType.STRING, true);
String[] inputStrings = new String[3];
char paddingChar = '%';
// "Café";
inputStrings[0] = new String(new byte[] { 67, 97, 102, -61, -87 });
// "François";
inputStrings[1] = new String(new byte[] { 70, 114, 97, 110, -61, -89, 111, 105, 115 });
// "Côte d'Ivoire";
inputStrings[2] = new String(new byte[] { 67, -61, -76, 116, 101, 32, 100, 39, 73, 118, 111, 105, 114, 101 });
Arrays.sort(inputStrings);
SegmentDictionaryCreator dictionaryCreator = new SegmentDictionaryCreator(false, inputStrings, fieldSpec, indexDir, paddingChar);
dictionaryCreator.build(new boolean[] { false });
for (String inputString : inputStrings) {
Assert.assertTrue(dictionaryCreator.indexOfSV(inputString) >= 0, "Value not found in dictionary " + inputString);
}
dictionaryCreator.close();
FileUtils.deleteQuietly(indexDir);
}
use of com.linkedin.pinot.core.segment.creator.impl.SegmentDictionaryCreator in project pinot by linkedin.
the class DictionariesTest method testStringsValuesWithPadding.
/**
* Tests DictionaryCreator for case when one value is a substring of another.
* For example, in case of sorted values {"abc", "abc def"} after padding,
* the sorted order would change to {"abc def%%%%", "abc%%%%%%%"}
*
* This test asserts that DictionaryCreator.indexOfSV("abc") returns 1 (ie index of "abc%%%%%%%"
* in actual padded dictionary), and not 0.
*
* @throws Exception
*/
@Test
public void testStringsValuesWithPadding() throws Exception {
File indexDir = new File("/tmp/dict.test");
indexDir.deleteOnExit();
FieldSpec fieldSpec = new DimensionFieldSpec("test", DataType.STRING, true);
String[] inputStrings = new String[2];
String[] paddedStrings = new String[2];
char paddingChar = '%';
inputStrings[0] = "abc def";
inputStrings[1] = "abc";
// Sorted order: {"abc", "abc def"}
Arrays.sort(inputStrings);
boolean[] isSorted = new boolean[1];
isSorted[0] = true;
SegmentDictionaryCreator dictionaryCreator = new SegmentDictionaryCreator(false, inputStrings, fieldSpec, indexDir, paddingChar);
dictionaryCreator.build(isSorted);
Assert.assertFalse(isSorted[0]);
// Get the padded string as stored in the dictionary.
int targetPaddedLength = dictionaryCreator.getStringColumnMaxLength();
for (int i = 0; i < inputStrings.length; i++) {
paddedStrings[i] = SegmentDictionaryCreator.getPaddedString(inputStrings[i], targetPaddedLength, paddingChar);
}
// Sorted Order: {"abc def%%%%", "abc%%%%%%%"}
Arrays.sort(paddedStrings);
// Assert that indexOfSV for un-padded string returns the index of the corresponding padded string.
for (int i = 0; i < inputStrings.length; i++) {
int paddedIndex = dictionaryCreator.indexOfSV(inputStrings[i]);
Assert.assertTrue(paddedStrings[paddedIndex].equals(SegmentDictionaryCreator.getPaddedString(inputStrings[i], targetPaddedLength, paddingChar)));
}
dictionaryCreator.close();
FileUtils.deleteQuietly(indexDir);
}
use of com.linkedin.pinot.core.segment.creator.impl.SegmentDictionaryCreator in project pinot by linkedin.
the class DictionariesTest method testSingleNullString.
/**
* Tests SegmentDictionaryCreator for case when there is only one string
* and it is "null"
*
* This test asserts that the padded length of the null string is 4
*
* @throws Exception
*/
@Test
public void testSingleNullString() throws Exception {
File indexDir = new File("/tmp/dict.test");
FieldSpec fieldSpec = new DimensionFieldSpec("test", DataType.STRING, true);
String[] inputStrings = new String[1];
String[] paddedStrings = new String[1];
inputStrings[0] = "null";
// Sorted order: {"null"}
Arrays.sort(inputStrings);
try {
SegmentDictionaryCreator dictionaryCreator = new SegmentDictionaryCreator(false, inputStrings, fieldSpec, indexDir, V1Constants.Str.DEFAULT_STRING_PAD_CHAR);
boolean[] isSorted = new boolean[1];
isSorted[0] = true;
dictionaryCreator.build(isSorted);
// Get the padded string as stored in the dictionary.
int targetPaddedLength = dictionaryCreator.getStringColumnMaxLength();
Assert.assertTrue(targetPaddedLength == 4);
for (int i = 0; i < inputStrings.length; i++) {
paddedStrings[i] = SegmentDictionaryCreator.getPaddedString(inputStrings[i], targetPaddedLength, V1Constants.Str.DEFAULT_STRING_PAD_CHAR);
}
// Sorted Order: {"null"}
Arrays.sort(paddedStrings);
// Assert that indexOfSV for un-padded string returns the index of the corresponding padded string.
for (int i = 0; i < inputStrings.length; i++) {
int paddedIndex = dictionaryCreator.indexOfSV(inputStrings[i]);
Assert.assertTrue(paddedStrings[paddedIndex].equals(SegmentDictionaryCreator.getPaddedString(inputStrings[i], targetPaddedLength, V1Constants.Str.DEFAULT_STRING_PAD_CHAR)));
}
// Verify that the string "null" did not get changed
Assert.assertTrue(paddedStrings[0].equals("null"));
dictionaryCreator.close();
} catch (Exception e) {
throw e;
} finally {
FileUtils.deleteQuietly(indexDir);
}
}
use of com.linkedin.pinot.core.segment.creator.impl.SegmentDictionaryCreator in project pinot by linkedin.
the class BaseDefaultColumnHandler method createColumnV1Indices.
/**
* Helper method to create the V1 indices (dictionary and forward index) for a column.
*
* @param column column name.
*/
protected void createColumnV1Indices(String column) throws Exception {
FieldSpec fieldSpec = _schema.getFieldSpecFor(column);
Preconditions.checkNotNull(fieldSpec);
// Generate column index creation information.
int totalDocs = _segmentMetadata.getTotalDocs();
int totalRawDocs = _segmentMetadata.getTotalRawDocs();
int totalAggDocs = totalDocs - totalRawDocs;
FieldSpec.DataType dataType = fieldSpec.getDataType();
Object defaultValue = fieldSpec.getDefaultNullValue();
boolean isSingleValue = fieldSpec.isSingleValueField();
int maxNumberOfMultiValueElements = isSingleValue ? 0 : 1;
int dictionaryElementSize = 0;
Object sortedArray;
switch(dataType) {
case STRING:
Preconditions.checkState(defaultValue instanceof String);
String stringDefaultValue = (String) defaultValue;
// Length of the UTF-8 encoded byte array.
dictionaryElementSize = stringDefaultValue.getBytes("UTF8").length;
sortedArray = new String[] { stringDefaultValue };
break;
case INT:
Preconditions.checkState(defaultValue instanceof Integer);
sortedArray = new int[] { (Integer) defaultValue };
break;
case LONG:
Preconditions.checkState(defaultValue instanceof Long);
sortedArray = new long[] { (Long) defaultValue };
break;
case FLOAT:
Preconditions.checkState(defaultValue instanceof Float);
sortedArray = new float[] { (Float) defaultValue };
break;
case DOUBLE:
Preconditions.checkState(defaultValue instanceof Double);
sortedArray = new double[] { (Double) defaultValue };
break;
default:
throw new UnsupportedOperationException("Unsupported data type: " + dataType + " for column: " + column);
}
ColumnIndexCreationInfo columnIndexCreationInfo = new ColumnIndexCreationInfo(true, /*createDictionary*/
defaultValue, /*min*/
defaultValue, /*max*/
sortedArray, ForwardIndexType.FIXED_BIT_COMPRESSED, InvertedIndexType.SORTED_INDEX, isSingleValue, /*isSortedColumn*/
false, /*hasNulls*/
totalDocs, /*totalNumberOfEntries*/
maxNumberOfMultiValueElements, -1, /* Unused max length*/
true, /*isAutoGenerated*/
defaultValue);
// Create dictionary.
// We will have only one value in the dictionary.
SegmentDictionaryCreator segmentDictionaryCreator = new SegmentDictionaryCreator(false, /*hasNulls*/
sortedArray, fieldSpec, _indexDir, V1Constants.Str.DEFAULT_STRING_PAD_CHAR);
segmentDictionaryCreator.build(new boolean[] { true });
segmentDictionaryCreator.close();
// Create forward index.
if (isSingleValue) {
// Single-value column.
SingleValueSortedForwardIndexCreator svFwdIndexCreator = new SingleValueSortedForwardIndexCreator(_indexDir, 1, /*cardinality*/
fieldSpec);
for (int docId = 0; docId < totalDocs; docId++) {
svFwdIndexCreator.add(0, /*dictionaryId*/
docId);
}
svFwdIndexCreator.close();
} else {
// Multi-value column.
MultiValueUnsortedForwardIndexCreator mvFwdIndexCreator = new MultiValueUnsortedForwardIndexCreator(fieldSpec, _indexDir, 1, /*cardinality*/
totalDocs, /*numDocs*/
totalDocs, /*totalNumberOfValues*/
false);
int[] dictionaryIds = { 0 };
for (int docId = 0; docId < totalDocs; docId++) {
mvFwdIndexCreator.index(docId, dictionaryIds);
}
mvFwdIndexCreator.close();
}
// Add the column metadata information to the metadata properties.
SegmentColumnarIndexCreator.addColumnMetadataInfo(_segmentProperties, column, columnIndexCreationInfo, totalDocs, totalRawDocs, totalAggDocs, fieldSpec, true, /*hasDictionary*/
dictionaryElementSize, true, /*hasInvertedIndex*/
null);
}
use of com.linkedin.pinot.core.segment.creator.impl.SegmentDictionaryCreator in project pinot by linkedin.
the class DictionariesTest method testPaddedConflict.
/**
* Tests SegmentDictionaryCreator for case when there is one empty string
* and a string with a single padding character
*
* This test asserts that the padded length of the empty string is 1
* in actual padded dictionary), and not 0.
*
* @throws Exception
*/
@Test
public void testPaddedConflict() throws Exception {
File indexDir = new File("/tmp/dict.test");
indexDir.deleteOnExit();
FieldSpec fieldSpec = new DimensionFieldSpec("test", DataType.STRING, true);
String[] inputStrings = new String[2];
String[] paddedStrings = new String[2];
char paddingChar = '%';
try {
inputStrings[0] = "";
inputStrings[1] = "%";
// Sorted order: {"", "%"}
Arrays.sort(inputStrings);
SegmentDictionaryCreator dictionaryCreator = new SegmentDictionaryCreator(false, inputStrings, fieldSpec, indexDir, paddingChar);
boolean[] isSorted = new boolean[1];
isSorted[0] = true;
dictionaryCreator.build(isSorted);
} catch (Exception e) {
Assert.assertEquals(e.getMessage(), "Number of entries in dictionary != number of unique values in the data in column test");
} finally {
FileUtils.deleteQuietly(indexDir);
}
}
Aggregations