use of org.apache.hadoop.hive.serde2.fast.DeserializeRead in project hive by apache.
the class VerifyFast method verifyDeserializeRead.
public static void verifyDeserializeRead(DeserializeRead deserializeRead, PrimitiveTypeInfo primitiveTypeInfo, Writable writable) throws IOException {
boolean isNull;
isNull = !deserializeRead.readNextField();
if (isNull) {
if (writable != null) {
TestCase.fail("Field reports null but object is not null (class " + writable.getClass().getName() + ", " + writable.toString() + ")");
}
return;
} else if (writable == null) {
TestCase.fail("Field report not null but object is null");
}
switch(primitiveTypeInfo.getPrimitiveCategory()) {
case BOOLEAN:
{
boolean value = deserializeRead.currentBoolean;
if (!(writable instanceof BooleanWritable)) {
TestCase.fail("Boolean expected writable not Boolean");
}
boolean expected = ((BooleanWritable) writable).get();
if (value != expected) {
TestCase.fail("Boolean field mismatch (expected " + expected + " found " + value + ")");
}
}
break;
case BYTE:
{
byte value = deserializeRead.currentByte;
if (!(writable instanceof ByteWritable)) {
TestCase.fail("Byte expected writable not Byte");
}
byte expected = ((ByteWritable) writable).get();
if (value != expected) {
TestCase.fail("Byte field mismatch (expected " + (int) expected + " found " + (int) value + ")");
}
}
break;
case SHORT:
{
short value = deserializeRead.currentShort;
if (!(writable instanceof ShortWritable)) {
TestCase.fail("Short expected writable not Short");
}
short expected = ((ShortWritable) writable).get();
if (value != expected) {
TestCase.fail("Short field mismatch (expected " + expected + " found " + value + ")");
}
}
break;
case INT:
{
int value = deserializeRead.currentInt;
if (!(writable instanceof IntWritable)) {
TestCase.fail("Integer expected writable not Integer");
}
int expected = ((IntWritable) writable).get();
if (value != expected) {
TestCase.fail("Int field mismatch (expected " + expected + " found " + value + ")");
}
}
break;
case LONG:
{
long value = deserializeRead.currentLong;
if (!(writable instanceof LongWritable)) {
TestCase.fail("Long expected writable not Long");
}
Long expected = ((LongWritable) writable).get();
if (value != expected) {
TestCase.fail("Long field mismatch (expected " + expected + " found " + value + ")");
}
}
break;
case FLOAT:
{
float value = deserializeRead.currentFloat;
if (!(writable instanceof FloatWritable)) {
TestCase.fail("Float expected writable not Float");
}
float expected = ((FloatWritable) writable).get();
if (value != expected) {
TestCase.fail("Float field mismatch (expected " + expected + " found " + value + ")");
}
}
break;
case DOUBLE:
{
double value = deserializeRead.currentDouble;
if (!(writable instanceof DoubleWritable)) {
TestCase.fail("Double expected writable not Double");
}
double expected = ((DoubleWritable) writable).get();
if (value != expected) {
TestCase.fail("Double field mismatch (expected " + expected + " found " + value + ")");
}
}
break;
case STRING:
{
byte[] stringBytes = Arrays.copyOfRange(deserializeRead.currentBytes, deserializeRead.currentBytesStart, deserializeRead.currentBytesStart + deserializeRead.currentBytesLength);
Text text = new Text(stringBytes);
String string = text.toString();
String expected = ((Text) writable).toString();
if (!string.equals(expected)) {
TestCase.fail("String field mismatch (expected '" + expected + "' found '" + string + "')");
}
}
break;
case CHAR:
{
byte[] stringBytes = Arrays.copyOfRange(deserializeRead.currentBytes, deserializeRead.currentBytesStart, deserializeRead.currentBytesStart + deserializeRead.currentBytesLength);
Text text = new Text(stringBytes);
String string = text.toString();
HiveChar hiveChar = new HiveChar(string, ((CharTypeInfo) primitiveTypeInfo).getLength());
HiveChar expected = ((HiveCharWritable) writable).getHiveChar();
if (!hiveChar.equals(expected)) {
TestCase.fail("Char field mismatch (expected '" + expected + "' found '" + hiveChar + "')");
}
}
break;
case VARCHAR:
{
byte[] stringBytes = Arrays.copyOfRange(deserializeRead.currentBytes, deserializeRead.currentBytesStart, deserializeRead.currentBytesStart + deserializeRead.currentBytesLength);
Text text = new Text(stringBytes);
String string = text.toString();
HiveVarchar hiveVarchar = new HiveVarchar(string, ((VarcharTypeInfo) primitiveTypeInfo).getLength());
HiveVarchar expected = ((HiveVarcharWritable) writable).getHiveVarchar();
if (!hiveVarchar.equals(expected)) {
TestCase.fail("Varchar field mismatch (expected '" + expected + "' found '" + hiveVarchar + "')");
}
}
break;
case DECIMAL:
{
HiveDecimal value = deserializeRead.currentHiveDecimalWritable.getHiveDecimal();
if (value == null) {
TestCase.fail("Decimal field evaluated to NULL");
}
HiveDecimal expected = ((HiveDecimalWritable) writable).getHiveDecimal();
if (!value.equals(expected)) {
DecimalTypeInfo decimalTypeInfo = (DecimalTypeInfo) primitiveTypeInfo;
int precision = decimalTypeInfo.getPrecision();
int scale = decimalTypeInfo.getScale();
TestCase.fail("Decimal field mismatch (expected " + expected.toString() + " found " + value.toString() + ") precision " + precision + ", scale " + scale);
}
}
break;
case DATE:
{
Date value = deserializeRead.currentDateWritable.get();
Date expected = ((DateWritable) writable).get();
if (!value.equals(expected)) {
TestCase.fail("Date field mismatch (expected " + expected.toString() + " found " + value.toString() + ")");
}
}
break;
case TIMESTAMP:
{
Timestamp value = deserializeRead.currentTimestampWritable.getTimestamp();
Timestamp expected = ((TimestampWritable) writable).getTimestamp();
if (!value.equals(expected)) {
TestCase.fail("Timestamp field mismatch (expected " + expected.toString() + " found " + value.toString() + ")");
}
}
break;
case INTERVAL_YEAR_MONTH:
{
HiveIntervalYearMonth value = deserializeRead.currentHiveIntervalYearMonthWritable.getHiveIntervalYearMonth();
HiveIntervalYearMonth expected = ((HiveIntervalYearMonthWritable) writable).getHiveIntervalYearMonth();
if (!value.equals(expected)) {
TestCase.fail("HiveIntervalYearMonth field mismatch (expected " + expected.toString() + " found " + value.toString() + ")");
}
}
break;
case INTERVAL_DAY_TIME:
{
HiveIntervalDayTime value = deserializeRead.currentHiveIntervalDayTimeWritable.getHiveIntervalDayTime();
HiveIntervalDayTime expected = ((HiveIntervalDayTimeWritable) writable).getHiveIntervalDayTime();
if (!value.equals(expected)) {
TestCase.fail("HiveIntervalDayTime field mismatch (expected " + expected.toString() + " found " + value.toString() + ")");
}
}
break;
case BINARY:
{
byte[] byteArray = Arrays.copyOfRange(deserializeRead.currentBytes, deserializeRead.currentBytesStart, deserializeRead.currentBytesStart + deserializeRead.currentBytesLength);
BytesWritable bytesWritable = (BytesWritable) writable;
byte[] expected = Arrays.copyOfRange(bytesWritable.getBytes(), 0, bytesWritable.getLength());
if (byteArray.length != expected.length) {
TestCase.fail("Byte Array field mismatch (expected " + Arrays.toString(expected) + " found " + Arrays.toString(byteArray) + ")");
}
for (int b = 0; b < byteArray.length; b++) {
if (byteArray[b] != expected[b]) {
TestCase.fail("Byte Array field mismatch (expected " + Arrays.toString(expected) + " found " + Arrays.toString(byteArray) + ")");
}
}
}
break;
default:
throw new Error("Unknown primitive category " + primitiveTypeInfo.getPrimitiveCategory());
}
}
use of org.apache.hadoop.hive.serde2.fast.DeserializeRead in project hive by apache.
the class VectorDeserializeOrcWriter method writeOneRow.
@Override
public void writeOneRow(Writable row) throws IOException {
if (sourceBatch.size == VectorizedRowBatch.DEFAULT_SIZE) {
flushBatch();
}
BinaryComparable binComp = (BinaryComparable) row;
deserializeRead.set(binComp.getBytes(), 0, binComp.getLength());
// Deserialize and append new row using the current batch size as the index.
try {
// Not using ByRef now since it's unsafe for text readers. Might be safe for others.
vectorDeserializeRow.deserialize(sourceBatch, sourceBatch.size++);
} catch (Exception e) {
throw new IOException("DeserializeRead detail: " + vectorDeserializeRow.getDetailedReadPositionString(), e);
}
}
use of org.apache.hadoop.hive.serde2.fast.DeserializeRead in project hive by apache.
the class TestBinarySortableFast method testBinarySortableFast.
private void testBinarySortableFast(SerdeRandomRowSource source, Object[][] rows, boolean[] columnSortOrderIsDesc, byte[] columnNullMarker, byte[] columnNotNullMarker, AbstractSerDe serde, StructObjectInspector rowOI, AbstractSerDe serde_fewer, StructObjectInspector writeRowOI, boolean ascending, PrimitiveTypeInfo[] primitiveTypeInfos, boolean useIncludeColumns, boolean doWriteFewerColumns, Random r) throws Throwable {
int rowCount = rows.length;
int columnCount = primitiveTypeInfos.length;
boolean[] columnsToInclude = null;
if (useIncludeColumns) {
columnsToInclude = new boolean[columnCount];
for (int i = 0; i < columnCount; i++) {
columnsToInclude[i] = r.nextBoolean();
}
}
int writeColumnCount = columnCount;
if (doWriteFewerColumns) {
writeColumnCount = writeRowOI.getAllStructFieldRefs().size();
}
BinarySortableSerializeWrite binarySortableSerializeWrite = new BinarySortableSerializeWrite(columnSortOrderIsDesc, columnNullMarker, columnNotNullMarker);
// Try to serialize
// One Writable per row.
BytesWritable[] serializeWriteBytes = new BytesWritable[rowCount];
int[][] perFieldWriteLengthsArray = new int[rowCount][];
for (int i = 0; i < rowCount; i++) {
Object[] row = rows[i];
Output output = new Output();
binarySortableSerializeWrite.set(output);
int[] perFieldWriteLengths = new int[columnCount];
for (int index = 0; index < writeColumnCount; index++) {
Writable writable = (Writable) row[index];
VerifyFast.serializeWrite(binarySortableSerializeWrite, primitiveTypeInfos[index], writable);
perFieldWriteLengths[index] = output.getLength();
}
perFieldWriteLengthsArray[i] = perFieldWriteLengths;
BytesWritable bytesWritable = new BytesWritable();
bytesWritable.set(output.getData(), 0, output.getLength());
serializeWriteBytes[i] = bytesWritable;
if (i > 0) {
int compareResult = serializeWriteBytes[i - 1].compareTo(serializeWriteBytes[i]);
if ((compareResult < 0 && !ascending) || (compareResult > 0 && ascending)) {
System.out.println("Test failed in " + (ascending ? "ascending" : "descending") + " order with " + (i - 1) + " and " + i);
System.out.println("serialized data [" + (i - 1) + "] = " + TestBinarySortableSerDe.hexString(serializeWriteBytes[i - 1]));
System.out.println("serialized data [" + i + "] = " + TestBinarySortableSerDe.hexString(serializeWriteBytes[i]));
fail("Sort order of serialized " + (i - 1) + " and " + i + " are reversed!");
}
}
}
// Try to deserialize using DeserializeRead our Writable row objects created by SerializeWrite.
for (int i = 0; i < rowCount; i++) {
Object[] row = rows[i];
BinarySortableDeserializeRead binarySortableDeserializeRead = new BinarySortableDeserializeRead(primitiveTypeInfos, /* useExternalBuffer */
false, columnSortOrderIsDesc);
BytesWritable bytesWritable = serializeWriteBytes[i];
binarySortableDeserializeRead.set(bytesWritable.getBytes(), 0, bytesWritable.getLength());
for (int index = 0; index < columnCount; index++) {
if (useIncludeColumns && !columnsToInclude[index]) {
binarySortableDeserializeRead.skipNextField();
} else if (index >= writeColumnCount) {
// Should come back a null.
VerifyFast.verifyDeserializeRead(binarySortableDeserializeRead, primitiveTypeInfos[index], null);
} else {
Writable writable = (Writable) row[index];
VerifyFast.verifyDeserializeRead(binarySortableDeserializeRead, primitiveTypeInfos[index], writable);
}
}
if (writeColumnCount == columnCount) {
TestCase.assertTrue(binarySortableDeserializeRead.isEndOfInputReached());
}
/*
* Clip off one byte and expect to get an EOFException on the write field.
*/
BinarySortableDeserializeRead binarySortableDeserializeRead2 = new BinarySortableDeserializeRead(primitiveTypeInfos, /* useExternalBuffer */
false, columnSortOrderIsDesc);
binarySortableDeserializeRead2.set(bytesWritable.getBytes(), 0, // One fewer byte.
bytesWritable.getLength() - 1);
for (int index = 0; index < writeColumnCount; index++) {
Writable writable = (Writable) row[index];
if (index == writeColumnCount - 1) {
boolean threw = false;
try {
VerifyFast.verifyDeserializeRead(binarySortableDeserializeRead2, primitiveTypeInfos[index], writable);
} catch (EOFException e) {
// debugDetailedReadPositionString = binarySortableDeserializeRead2.getDetailedReadPositionString();
// debugStackTrace = e.getStackTrace();
threw = true;
}
TestCase.assertTrue(threw);
} else {
if (useIncludeColumns && !columnsToInclude[index]) {
binarySortableDeserializeRead2.skipNextField();
} else {
VerifyFast.verifyDeserializeRead(binarySortableDeserializeRead2, primitiveTypeInfos[index], writable);
}
}
}
}
// Try to deserialize using SerDe class our Writable row objects created by SerializeWrite.
for (int i = 0; i < rowCount; i++) {
BytesWritable bytesWritable = serializeWriteBytes[i];
// Note that regular SerDe doesn't tolerate fewer columns.
List<Object> deserializedRow;
if (doWriteFewerColumns) {
deserializedRow = (List<Object>) serde_fewer.deserialize(bytesWritable);
} else {
deserializedRow = (List<Object>) serde.deserialize(bytesWritable);
}
Object[] row = rows[i];
for (int index = 0; index < writeColumnCount; index++) {
Object expected = row[index];
Object object = deserializedRow.get(index);
if (expected == null || object == null) {
if (expected != null || object != null) {
fail("SerDe deserialized NULL column mismatch");
}
} else {
if (!object.equals(expected)) {
fail("SerDe deserialized value does not match (expected " + expected.getClass().getName() + " " + expected.toString() + ", actual " + object.getClass().getName() + " " + object.toString() + ")");
}
}
}
}
// One Writable per row.
BytesWritable[] serdeBytes = new BytesWritable[rowCount];
// Serialize using the SerDe, then below deserialize using DeserializeRead.
for (int i = 0; i < rowCount; i++) {
Object[] row = rows[i];
// Since SerDe reuses memory, we will need to make a copy.
BytesWritable serialized;
if (doWriteFewerColumns) {
serialized = (BytesWritable) serde_fewer.serialize(row, rowOI);
} else {
serialized = (BytesWritable) serde.serialize(row, rowOI);
;
}
BytesWritable bytesWritable = new BytesWritable();
bytesWritable.set(serialized);
byte[] serDeOutput = Arrays.copyOfRange(bytesWritable.getBytes(), 0, bytesWritable.getLength());
byte[] serializeWriteExpected = Arrays.copyOfRange(serializeWriteBytes[i].getBytes(), 0, serializeWriteBytes[i].getLength());
if (!Arrays.equals(serDeOutput, serializeWriteExpected)) {
int mismatchPos = -1;
if (serDeOutput.length != serializeWriteExpected.length) {
for (int b = 0; b < Math.min(serDeOutput.length, serializeWriteExpected.length); b++) {
if (serDeOutput[b] != serializeWriteExpected[b]) {
mismatchPos = b;
break;
}
}
fail("Different byte array lengths: serDeOutput.length " + serDeOutput.length + ", serializeWriteExpected.length " + serializeWriteExpected.length + " mismatchPos " + mismatchPos + " perFieldWriteLengths " + Arrays.toString(perFieldWriteLengthsArray[i]));
}
List<Integer> differentPositions = new ArrayList();
for (int b = 0; b < serDeOutput.length; b++) {
if (serDeOutput[b] != serializeWriteExpected[b]) {
differentPositions.add(b);
}
}
if (differentPositions.size() > 0) {
List<String> serializeWriteExpectedFields = new ArrayList<String>();
List<String> serDeFields = new ArrayList<String>();
int f = 0;
int lastBegin = 0;
for (int b = 0; b < serDeOutput.length; b++) {
int writeLength = perFieldWriteLengthsArray[i][f];
if (b + 1 == writeLength) {
serializeWriteExpectedFields.add(displayBytes(serializeWriteExpected, lastBegin, writeLength - lastBegin));
serDeFields.add(displayBytes(serDeOutput, lastBegin, writeLength - lastBegin));
f++;
lastBegin = b + 1;
}
}
fail("SerializeWrite and SerDe serialization does not match at positions " + differentPositions.toString() + "\n(SerializeWrite: " + serializeWriteExpectedFields.toString() + "\nSerDe: " + serDeFields.toString() + "\nperFieldWriteLengths " + Arrays.toString(perFieldWriteLengthsArray[i]) + "\nprimitiveTypeInfos " + Arrays.toString(primitiveTypeInfos) + "\nrow " + Arrays.toString(row));
}
}
serdeBytes[i] = bytesWritable;
}
// Try to deserialize using DeserializeRead our Writable row objects created by SerDe.
for (int i = 0; i < rowCount; i++) {
Object[] row = rows[i];
BinarySortableDeserializeRead binarySortableDeserializeRead = new BinarySortableDeserializeRead(primitiveTypeInfos, /* useExternalBuffer */
false, columnSortOrderIsDesc);
BytesWritable bytesWritable = serdeBytes[i];
binarySortableDeserializeRead.set(bytesWritable.getBytes(), 0, bytesWritable.getLength());
for (int index = 0; index < columnCount; index++) {
if (useIncludeColumns && !columnsToInclude[index]) {
binarySortableDeserializeRead.skipNextField();
} else if (index >= writeColumnCount) {
// Should come back a null.
VerifyFast.verifyDeserializeRead(binarySortableDeserializeRead, primitiveTypeInfos[index], null);
} else {
Writable writable = (Writable) row[index];
VerifyFast.verifyDeserializeRead(binarySortableDeserializeRead, primitiveTypeInfos[index], writable);
}
}
if (writeColumnCount == columnCount) {
TestCase.assertTrue(binarySortableDeserializeRead.isEndOfInputReached());
}
}
}
use of org.apache.hadoop.hive.serde2.fast.DeserializeRead in project hive by apache.
the class VectorDeserializeRow method convertRowColumn.
/**
* Convert one row column value that is the current value in deserializeRead.
*
* We deserialize into a writable and then pass that writable to an instance of VectorAssignRow
* to convert the writable to the target data type and assign it into the VectorizedRowBatch.
*
* @param batch
* @param batchIndex
* @param logicalColumnIndex
* @throws IOException
*/
private void convertRowColumn(VectorizedRowBatch batch, int batchIndex, int logicalColumnIndex) throws IOException {
final int projectionColumnNum = projectionColumnNums[logicalColumnIndex];
Writable convertSourceWritable = convertSourceWritables[logicalColumnIndex];
switch(sourceCategories[logicalColumnIndex]) {
case PRIMITIVE:
{
switch(sourcePrimitiveCategories[logicalColumnIndex]) {
case VOID:
convertSourceWritable = null;
break;
case BOOLEAN:
((BooleanWritable) convertSourceWritable).set(deserializeRead.currentBoolean);
break;
case BYTE:
((ByteWritable) convertSourceWritable).set(deserializeRead.currentByte);
break;
case SHORT:
((ShortWritable) convertSourceWritable).set(deserializeRead.currentShort);
break;
case INT:
((IntWritable) convertSourceWritable).set(deserializeRead.currentInt);
break;
case LONG:
((LongWritable) convertSourceWritable).set(deserializeRead.currentLong);
break;
case TIMESTAMP:
((TimestampWritable) convertSourceWritable).set(deserializeRead.currentTimestampWritable);
break;
case DATE:
((DateWritable) convertSourceWritable).set(deserializeRead.currentDateWritable);
break;
case FLOAT:
((FloatWritable) convertSourceWritable).set(deserializeRead.currentFloat);
break;
case DOUBLE:
((DoubleWritable) convertSourceWritable).set(deserializeRead.currentDouble);
break;
case BINARY:
if (deserializeRead.currentBytes == null) {
LOG.info("null binary entry: batchIndex " + batchIndex + " projection column num " + projectionColumnNum);
}
((BytesWritable) convertSourceWritable).set(deserializeRead.currentBytes, deserializeRead.currentBytesStart, deserializeRead.currentBytesLength);
break;
case STRING:
if (deserializeRead.currentBytes == null) {
throw new RuntimeException("null string entry: batchIndex " + batchIndex + " projection column num " + projectionColumnNum);
}
// Use org.apache.hadoop.io.Text as our helper to go from byte[] to String.
((Text) convertSourceWritable).set(deserializeRead.currentBytes, deserializeRead.currentBytesStart, deserializeRead.currentBytesLength);
break;
case VARCHAR:
{
// that does not use Java String objects.
if (deserializeRead.currentBytes == null) {
throw new RuntimeException("null varchar entry: batchIndex " + batchIndex + " projection column num " + projectionColumnNum);
}
int adjustedLength = StringExpr.truncate(deserializeRead.currentBytes, deserializeRead.currentBytesStart, deserializeRead.currentBytesLength, maxLengths[logicalColumnIndex]);
((HiveVarcharWritable) convertSourceWritable).set(new String(deserializeRead.currentBytes, deserializeRead.currentBytesStart, adjustedLength, Charsets.UTF_8), -1);
}
break;
case CHAR:
{
// that does not use Java String objects.
if (deserializeRead.currentBytes == null) {
throw new RuntimeException("null char entry: batchIndex " + batchIndex + " projection column num " + projectionColumnNum);
}
int adjustedLength = StringExpr.rightTrimAndTruncate(deserializeRead.currentBytes, deserializeRead.currentBytesStart, deserializeRead.currentBytesLength, maxLengths[logicalColumnIndex]);
((HiveCharWritable) convertSourceWritable).set(new String(deserializeRead.currentBytes, deserializeRead.currentBytesStart, adjustedLength, Charsets.UTF_8), -1);
}
break;
case DECIMAL:
((HiveDecimalWritable) convertSourceWritable).set(deserializeRead.currentHiveDecimalWritable);
break;
case INTERVAL_YEAR_MONTH:
((HiveIntervalYearMonthWritable) convertSourceWritable).set(deserializeRead.currentHiveIntervalYearMonthWritable);
break;
case INTERVAL_DAY_TIME:
((HiveIntervalDayTimeWritable) convertSourceWritable).set(deserializeRead.currentHiveIntervalDayTimeWritable);
break;
default:
throw new RuntimeException("Primitive category " + sourcePrimitiveCategories[logicalColumnIndex] + " not supported");
}
}
break;
default:
throw new RuntimeException("Category " + sourceCategories[logicalColumnIndex] + " not supported");
}
/*
* Convert our source object we just read into the target object and store that in the
* VectorizedRowBatch.
*/
convertVectorAssignRow.assignConvertRowColumn(batch, batchIndex, logicalColumnIndex, convertSourceWritable);
}
use of org.apache.hadoop.hive.serde2.fast.DeserializeRead in project hive by apache.
the class TestVectorSerDeRow method testVectorDeserializeRow.
void testVectorDeserializeRow(Random r, SerializationType serializationType, boolean alternate1, boolean alternate2, boolean useExternalBuffer) throws HiveException, IOException, SerDeException {
String[] emptyScratchTypeNames = new String[0];
VectorRandomRowSource source = new VectorRandomRowSource();
source.init(r);
VectorizedRowBatchCtx batchContext = new VectorizedRowBatchCtx();
batchContext.init(source.rowStructObjectInspector(), emptyScratchTypeNames);
VectorizedRowBatch batch = batchContext.createVectorizedRowBatch();
// junk the destination for the 1st pass
for (ColumnVector cv : batch.cols) {
Arrays.fill(cv.isNull, true);
}
PrimitiveTypeInfo[] primitiveTypeInfos = source.primitiveTypeInfos();
int fieldCount = source.typeNames().size();
DeserializeRead deserializeRead;
SerializeWrite serializeWrite;
switch(serializationType) {
case BINARY_SORTABLE:
boolean useColumnSortOrderIsDesc = alternate1;
if (!useColumnSortOrderIsDesc) {
deserializeRead = new BinarySortableDeserializeRead(source.primitiveTypeInfos(), useExternalBuffer);
serializeWrite = new BinarySortableSerializeWrite(fieldCount);
} else {
boolean[] columnSortOrderIsDesc = new boolean[fieldCount];
for (int i = 0; i < fieldCount; i++) {
columnSortOrderIsDesc[i] = r.nextBoolean();
}
deserializeRead = new BinarySortableDeserializeRead(source.primitiveTypeInfos(), useExternalBuffer, columnSortOrderIsDesc);
byte[] columnNullMarker = new byte[fieldCount];
byte[] columnNotNullMarker = new byte[fieldCount];
for (int i = 0; i < fieldCount; i++) {
if (columnSortOrderIsDesc[i]) {
// Descending
// Null last (default for descending order)
columnNullMarker[i] = BinarySortableSerDe.ZERO;
columnNotNullMarker[i] = BinarySortableSerDe.ONE;
} else {
// Ascending
// Null first (default for ascending order)
columnNullMarker[i] = BinarySortableSerDe.ZERO;
columnNotNullMarker[i] = BinarySortableSerDe.ONE;
}
}
serializeWrite = new BinarySortableSerializeWrite(columnSortOrderIsDesc, columnNullMarker, columnNotNullMarker);
}
boolean useBinarySortableCharsNeedingEscape = alternate2;
if (useBinarySortableCharsNeedingEscape) {
source.addBinarySortableAlphabets();
}
break;
case LAZY_BINARY:
deserializeRead = new LazyBinaryDeserializeRead(source.primitiveTypeInfos(), useExternalBuffer);
serializeWrite = new LazyBinarySerializeWrite(fieldCount);
break;
case LAZY_SIMPLE:
{
StructObjectInspector rowObjectInspector = source.rowStructObjectInspector();
Configuration conf = new Configuration();
Properties tbl = new Properties();
tbl.setProperty(serdeConstants.FIELD_DELIM, "\t");
tbl.setProperty(serdeConstants.LINE_DELIM, "\n");
byte separator = (byte) '\t';
boolean useLazySimpleEscapes = alternate1;
if (useLazySimpleEscapes) {
tbl.setProperty(serdeConstants.QUOTE_CHAR, "'");
String escapeString = "\\";
tbl.setProperty(serdeConstants.ESCAPE_CHAR, escapeString);
}
LazySerDeParameters lazySerDeParams = getSerDeParams(conf, tbl, rowObjectInspector);
if (useLazySimpleEscapes) {
// LazySimple seems to throw away everything but \n and \r.
boolean[] needsEscape = lazySerDeParams.getNeedsEscape();
StringBuilder sb = new StringBuilder();
if (needsEscape['\n']) {
sb.append('\n');
}
if (needsEscape['\r']) {
sb.append('\r');
}
// for (int i = 0; i < needsEscape.length; i++) {
// if (needsEscape[i]) {
// sb.append((char) i);
// }
// }
String needsEscapeStr = sb.toString();
if (needsEscapeStr.length() > 0) {
source.addEscapables(needsEscapeStr);
}
}
deserializeRead = new LazySimpleDeserializeRead(source.primitiveTypeInfos(), useExternalBuffer, separator, lazySerDeParams);
serializeWrite = new LazySimpleSerializeWrite(fieldCount, separator, lazySerDeParams);
}
break;
default:
throw new Error("Unknown serialization type " + serializationType);
}
VectorDeserializeRow vectorDeserializeRow = new VectorDeserializeRow(deserializeRead);
vectorDeserializeRow.init();
// junk the destination for the 1st pass
for (ColumnVector cv : batch.cols) {
Arrays.fill(cv.isNull, true);
cv.noNulls = false;
}
VectorExtractRow vectorExtractRow = new VectorExtractRow();
vectorExtractRow.init(source.typeNames());
Object[][] randomRows = source.randomRows(100000);
int firstRandomRowIndex = 0;
for (int i = 0; i < randomRows.length; i++) {
Object[] row = randomRows[i];
Output output = serializeRow(row, source, serializeWrite);
vectorDeserializeRow.setBytes(output.getData(), 0, output.getLength());
try {
vectorDeserializeRow.deserialize(batch, batch.size);
} catch (Exception e) {
throw new HiveException("\nDeserializeRead details: " + vectorDeserializeRow.getDetailedReadPositionString(), e);
}
batch.size++;
if (batch.size == batch.DEFAULT_SIZE) {
examineBatch(batch, vectorExtractRow, primitiveTypeInfos, randomRows, firstRandomRowIndex);
firstRandomRowIndex = i + 1;
batch.reset();
}
}
if (batch.size > 0) {
examineBatch(batch, vectorExtractRow, primitiveTypeInfos, randomRows, firstRandomRowIndex);
}
}
Aggregations