use of org.apache.hadoop.hive.ql.metadata.HiveException in project hive by apache.
the class TestVectorSerDeRow method deserializeAndVerify.
void deserializeAndVerify(Output output, DeserializeRead deserializeRead, VectorRandomRowSource source, Object[] expectedRow) throws HiveException, IOException {
deserializeRead.set(output.getData(), 0, output.getLength());
PrimitiveCategory[] primitiveCategories = source.primitiveCategories();
for (int i = 0; i < primitiveCategories.length; i++) {
Object expected = expectedRow[i];
PrimitiveCategory primitiveCategory = primitiveCategories[i];
PrimitiveTypeInfo primitiveTypeInfo = source.primitiveTypeInfos()[i];
if (!deserializeRead.readNextField()) {
throw new HiveException("Unexpected NULL when reading primitiveCategory " + primitiveCategory + " expected (" + expected.getClass().getName() + ", " + expected.toString() + ") " + " deserializeRead " + deserializeRead.getClass().getName());
}
switch(primitiveCategory) {
case BOOLEAN:
{
Boolean value = deserializeRead.currentBoolean;
BooleanWritable expectedWritable = (BooleanWritable) expected;
if (!value.equals(expectedWritable.get())) {
TestCase.fail("Boolean field mismatch (expected " + expected + " found " + value + ")");
}
}
break;
case BYTE:
{
Byte value = deserializeRead.currentByte;
ByteWritable expectedWritable = (ByteWritable) expected;
if (!value.equals(expectedWritable.get())) {
TestCase.fail("Byte field mismatch (expected " + (int) expected + " found " + (int) value + ")");
}
}
break;
case SHORT:
{
Short value = deserializeRead.currentShort;
ShortWritable expectedWritable = (ShortWritable) expected;
if (!value.equals(expectedWritable.get())) {
TestCase.fail("Short field mismatch (expected " + expected + " found " + value + ")");
}
}
break;
case INT:
{
Integer value = deserializeRead.currentInt;
IntWritable expectedWritable = (IntWritable) expected;
if (!value.equals(expectedWritable.get())) {
TestCase.fail("Int field mismatch (expected " + expected + " found " + value + ")");
}
}
break;
case LONG:
{
Long value = deserializeRead.currentLong;
LongWritable expectedWritable = (LongWritable) expected;
if (!value.equals(expectedWritable.get())) {
TestCase.fail("Long field mismatch (expected " + expected + " found " + value + ")");
}
}
break;
case DATE:
{
DateWritable value = deserializeRead.currentDateWritable;
DateWritable expectedWritable = (DateWritable) expected;
if (!value.equals(expectedWritable)) {
TestCase.fail("Date field mismatch (expected " + expected.toString() + " found " + value.toString() + ")");
}
}
break;
case FLOAT:
{
Float value = deserializeRead.currentFloat;
FloatWritable expectedWritable = (FloatWritable) expected;
if (!value.equals(expectedWritable.get())) {
TestCase.fail("Float field mismatch (expected " + expected + " found " + value + ")");
}
}
break;
case DOUBLE:
{
Double value = deserializeRead.currentDouble;
DoubleWritable expectedWritable = (DoubleWritable) expected;
if (!value.equals(expectedWritable.get())) {
TestCase.fail("Double field mismatch (expected " + expected + " found " + value + ")");
}
}
break;
case STRING:
case CHAR:
case VARCHAR:
case BINARY:
{
byte[] stringBytes = Arrays.copyOfRange(deserializeRead.currentBytes, deserializeRead.currentBytesStart, deserializeRead.currentBytesStart + deserializeRead.currentBytesLength);
Text text = new Text(stringBytes);
String string = text.toString();
switch(primitiveCategory) {
case STRING:
{
Text expectedWritable = (Text) expected;
if (!string.equals(expectedWritable.toString())) {
TestCase.fail("String field mismatch (expected '" + expectedWritable.toString() + "' found '" + string + "')");
}
}
break;
case CHAR:
{
HiveChar hiveChar = new HiveChar(string, ((CharTypeInfo) primitiveTypeInfo).getLength());
HiveCharWritable expectedWritable = (HiveCharWritable) expected;
if (!hiveChar.equals(expectedWritable.getHiveChar())) {
TestCase.fail("Char field mismatch (expected '" + expectedWritable.getHiveChar() + "' found '" + hiveChar + "')");
}
}
break;
case VARCHAR:
{
HiveVarchar hiveVarchar = new HiveVarchar(string, ((VarcharTypeInfo) primitiveTypeInfo).getLength());
HiveVarcharWritable expectedWritable = (HiveVarcharWritable) expected;
if (!hiveVarchar.equals(expectedWritable.getHiveVarchar())) {
TestCase.fail("Varchar field mismatch (expected '" + expectedWritable.getHiveVarchar() + "' found '" + hiveVarchar + "')");
}
}
break;
case BINARY:
{
BytesWritable expectedWritable = (BytesWritable) expected;
if (stringBytes.length != expectedWritable.getLength()) {
TestCase.fail("Byte Array field mismatch (expected " + expected + " found " + stringBytes + ")");
}
byte[] expectedBytes = expectedWritable.getBytes();
for (int b = 0; b < stringBytes.length; b++) {
if (stringBytes[b] != expectedBytes[b]) {
TestCase.fail("Byte Array field mismatch (expected " + expected + " found " + stringBytes + ")");
}
}
}
break;
default:
throw new HiveException("Unexpected primitive category " + primitiveCategory);
}
}
break;
case DECIMAL:
{
HiveDecimal value = deserializeRead.currentHiveDecimalWritable.getHiveDecimal();
if (value == null) {
TestCase.fail("Decimal field evaluated to NULL");
}
HiveDecimalWritable expectedWritable = (HiveDecimalWritable) expected;
if (!value.equals(expectedWritable.getHiveDecimal())) {
DecimalTypeInfo decimalTypeInfo = (DecimalTypeInfo) primitiveTypeInfo;
int precision = decimalTypeInfo.getPrecision();
int scale = decimalTypeInfo.getScale();
TestCase.fail("Decimal field mismatch (expected " + expectedWritable.getHiveDecimal() + " found " + value.toString() + ") precision " + precision + ", scale " + scale);
}
}
break;
case TIMESTAMP:
{
Timestamp value = deserializeRead.currentTimestampWritable.getTimestamp();
TimestampWritable expectedWritable = (TimestampWritable) expected;
if (!value.equals(expectedWritable.getTimestamp())) {
TestCase.fail("Timestamp field mismatch (expected " + expectedWritable.getTimestamp() + " found " + value.toString() + ")");
}
}
break;
case INTERVAL_YEAR_MONTH:
{
HiveIntervalYearMonth value = deserializeRead.currentHiveIntervalYearMonthWritable.getHiveIntervalYearMonth();
HiveIntervalYearMonthWritable expectedWritable = (HiveIntervalYearMonthWritable) expected;
HiveIntervalYearMonth expectedValue = expectedWritable.getHiveIntervalYearMonth();
if (!value.equals(expectedValue)) {
TestCase.fail("HiveIntervalYearMonth field mismatch (expected " + expectedValue + " found " + value.toString() + ")");
}
}
break;
case INTERVAL_DAY_TIME:
{
HiveIntervalDayTime value = deserializeRead.currentHiveIntervalDayTimeWritable.getHiveIntervalDayTime();
HiveIntervalDayTimeWritable expectedWritable = (HiveIntervalDayTimeWritable) expected;
HiveIntervalDayTime expectedValue = expectedWritable.getHiveIntervalDayTime();
if (!value.equals(expectedValue)) {
TestCase.fail("HiveIntervalDayTime field mismatch (expected " + expectedValue + " found " + value.toString() + ")");
}
}
break;
default:
throw new HiveException("Unexpected primitive category " + primitiveCategory);
}
}
TestCase.assertTrue(deserializeRead.isEndOfInputReached());
}
use of org.apache.hadoop.hive.ql.metadata.HiveException in project hive by apache.
the class TestVectorGroupByOperator method testMemoryPressureFlush.
@Test
public void testMemoryPressureFlush() throws HiveException {
List<String> mapColumnNames = new ArrayList<String>();
mapColumnNames.add("Key");
mapColumnNames.add("Value");
VectorizationContext ctx = new VectorizationContext("name", mapColumnNames);
GroupByDesc desc = buildKeyGroupByDesc(ctx, "max", "Value", TypeInfoFactory.longTypeInfo, "Key", TypeInfoFactory.longTypeInfo);
// Set the memory treshold so that we get 100Kb before we need to flush.
MemoryMXBean memoryMXBean = ManagementFactory.getMemoryMXBean();
long maxMemory = memoryMXBean.getHeapMemoryUsage().getMax();
float treshold = 100.0f * 1024.0f / maxMemory;
desc.setMemoryThreshold(treshold);
CompilationOpContext cCtx = new CompilationOpContext();
Operator<? extends OperatorDesc> groupByOp = OperatorFactory.get(cCtx, desc);
VectorGroupByOperator vgo = (VectorGroupByOperator) Vectorizer.vectorizeGroupByOperator(groupByOp, ctx);
FakeCaptureOutputOperator out = FakeCaptureOutputOperator.addCaptureOutputChild(cCtx, vgo);
vgo.initialize(hconf, null);
this.outputRowCount = 0;
out.setOutputInspector(new FakeCaptureOutputOperator.OutputInspector() {
@Override
public void inspectRow(Object row, int tag) throws HiveException {
++outputRowCount;
}
});
Iterable<Object> it = new Iterable<Object>() {
@Override
public Iterator<Object> iterator() {
return new Iterator<Object>() {
long value = 0;
@Override
public boolean hasNext() {
return true;
}
@Override
public Object next() {
return ++value;
}
@Override
public void remove() {
}
};
}
};
FakeVectorRowBatchFromObjectIterables data = new FakeVectorRowBatchFromObjectIterables(100, new String[] { "long", "long" }, it, it);
// The 'it' data source will produce data w/o ever ending
// We want to see that memory pressure kicks in and some
// entries in the VGBY are flushed.
long countRowsProduced = 0;
for (VectorizedRowBatch unit : data) {
countRowsProduced += 100;
vgo.process(unit, 0);
if (0 < outputRowCount) {
break;
}
// Set an upper bound how much we're willing to push before it should flush
// we've set the memory treshold at 100kb, each key is distinct
// It should not go beyond 100k/16 (key+data)
assertTrue(countRowsProduced < 100 * 1024 / 16);
}
assertTrue(0 < outputRowCount);
}
use of org.apache.hadoop.hive.ql.metadata.HiveException in project hive by apache.
the class TestVectorUDFAdaptor method testMultiArgumentUDF.
@Test
public void testMultiArgumentUDF() {
// create a syntax tree for a function call "testudf(col0, col1, col2)"
ExprNodeGenericFuncDesc funcDesc;
TypeInfo typeInfoStr = TypeInfoFactory.stringTypeInfo;
TypeInfo typeInfoLong = TypeInfoFactory.longTypeInfo;
TypeInfo typeInfoDbl = TypeInfoFactory.doubleTypeInfo;
GenericUDFBridge genericUDFBridge = new GenericUDFBridge("testudf", false, ConcatTextLongDoubleUDF.class.getName());
List<ExprNodeDesc> children = new ArrayList<ExprNodeDesc>();
children.add(new ExprNodeColumnDesc(typeInfoStr, "col0", "tablename", false));
children.add(new ExprNodeColumnDesc(typeInfoLong, "col1", "tablename", false));
children.add(new ExprNodeColumnDesc(typeInfoDbl, "col2", "tablename", false));
VectorUDFArgDesc[] argDescs = new VectorUDFArgDesc[3];
for (int i = 0; i < 3; i++) {
argDescs[i] = new VectorUDFArgDesc();
argDescs[i].setVariable(i);
}
funcDesc = new ExprNodeGenericFuncDesc(typeInfoStr, genericUDFBridge, genericUDFBridge.getUdfName(), children);
// create the adaptor for this function call to work in vector mode
VectorUDFAdaptor vudf = null;
try {
vudf = new VectorUDFAdaptor(funcDesc, 3, "String", argDescs);
} catch (HiveException e) {
// We should never get here.
assertTrue(false);
throw new RuntimeException(e);
}
// with no nulls
VectorizedRowBatch b = getBatchStrDblLongWithStrOut();
vudf.evaluate(b);
byte[] result = null;
byte[] result2 = null;
try {
result = "red:1:1.0".getBytes("UTF-8");
result2 = "blue:0:0.0".getBytes("UTF-8");
} catch (Exception e) {
;
}
BytesColumnVector out = (BytesColumnVector) b.cols[3];
int cmp = StringExpr.compare(result, 0, result.length, out.vector[1], out.start[1], out.length[1]);
assertEquals(0, cmp);
assertTrue(out.noNulls);
// with nulls
b = getBatchStrDblLongWithStrOut();
b.cols[1].noNulls = false;
vudf.evaluate(b);
out = (BytesColumnVector) b.cols[3];
assertFalse(out.noNulls);
assertTrue(out.isNull[1]);
// with all input columns repeating
b = getBatchStrDblLongWithStrOut();
b.cols[0].isRepeating = true;
b.cols[1].isRepeating = true;
b.cols[2].isRepeating = true;
vudf.evaluate(b);
out = (BytesColumnVector) b.cols[3];
assertTrue(out.isRepeating);
cmp = StringExpr.compare(result2, 0, result2.length, out.vector[0], out.start[0], out.length[0]);
assertEquals(0, cmp);
assertTrue(out.noNulls);
}
use of org.apache.hadoop.hive.ql.metadata.HiveException in project hive by apache.
the class TestVectorUDFAdaptor method testLongUDF.
@Test
public void testLongUDF() {
// create a syntax tree for a simple function call "longudf(col0)"
ExprNodeGenericFuncDesc funcDesc;
TypeInfo typeInfo = TypeInfoFactory.longTypeInfo;
GenericUDFBridge genericUDFBridge = new GenericUDFBridge("longudf", false, LongUDF.class.getName());
List<ExprNodeDesc> children = new ArrayList<ExprNodeDesc>();
ExprNodeColumnDesc colDesc = new ExprNodeColumnDesc(typeInfo, "col0", "tablename", false);
children.add(colDesc);
VectorUDFArgDesc[] argDescs = new VectorUDFArgDesc[1];
argDescs[0] = new VectorUDFArgDesc();
argDescs[0].setVariable(0);
funcDesc = new ExprNodeGenericFuncDesc(typeInfo, genericUDFBridge, genericUDFBridge.getUdfName(), children);
// create the adaptor for this function call to work in vector mode
VectorUDFAdaptor vudf = null;
try {
vudf = new VectorUDFAdaptor(funcDesc, 1, "Long", argDescs);
} catch (HiveException e) {
// We should never get here.
assertTrue(false);
}
VectorizedRowBatch b = getBatchLongInLongOut();
vudf.evaluate(b);
// verify output
LongColumnVector out = (LongColumnVector) b.cols[1];
assertEquals(1000, out.vector[0]);
assertEquals(1001, out.vector[1]);
assertEquals(1002, out.vector[2]);
assertTrue(out.noNulls);
assertFalse(out.isRepeating);
// with nulls
b = getBatchLongInLongOut();
out = (LongColumnVector) b.cols[1];
b.cols[0].noNulls = false;
vudf.evaluate(b);
assertFalse(out.noNulls);
assertEquals(1000, out.vector[0]);
assertEquals(1001, out.vector[1]);
assertTrue(out.isNull[2]);
assertFalse(out.isRepeating);
// with repeating
b = getBatchLongInLongOut();
out = (LongColumnVector) b.cols[1];
b.cols[0].isRepeating = true;
vudf.evaluate(b);
// The implementation may or may not set output it isRepeting.
// That is implementation-defined.
assertTrue(b.cols[1].isRepeating && out.vector[0] == 1000 || !b.cols[1].isRepeating && out.vector[2] == 1000);
assertEquals(3, b.size);
}
use of org.apache.hadoop.hive.ql.metadata.HiveException in project hive by apache.
the class GenericUDFReflect method evaluate.
@Override
public Object evaluate(DeferredObject[] arguments) throws HiveException {
// Try to find the class
// Skip class loading if the class name didn't change
boolean classNameChanged = false;
Object newClassName = arguments[0].get();
// name/method name is constant Java String, or constant Text (StringWritable).
if (className == null || ObjectInspectorUtils.compare(className, classNameOI, newClassName, inputClassNameOI) != 0) {
className = ObjectInspectorUtils.copyToStandardObject(newClassName, inputClassNameOI);
String classNameString = classNameOI.getPrimitiveJavaObject(className);
try {
c = JavaUtils.loadClass(classNameString);
} catch (ClassNotFoundException ex) {
throw new HiveException("UDFReflect evaluate ", ex);
}
try {
o = null;
o = ReflectionUtils.newInstance(c, null);
} catch (Exception e) {
// ignored
}
classNameChanged = true;
}
// Try to find the method
// Skip method finding if the method name didn't change, and class name didn't change.
Object newMethodName = arguments[1].get();
if (methodName == null || ObjectInspectorUtils.compare(methodName, methodNameOI, newMethodName, inputMethodNameOI) != 0 || classNameChanged) {
methodName = ObjectInspectorUtils.copyToStandardObject(newMethodName, inputMethodNameOI);
String methodNameString = methodNameOI.getPrimitiveJavaObject(methodName);
try {
m = findMethod(c, methodNameString, String.class, false);
} catch (Exception e) {
throw new HiveException("UDFReflect getMethod ", e);
}
}
Object[] parameterJavaValues = setupParameters(arguments, 2);
try {
return String.valueOf(m.invoke(o, parameterJavaValues));
} catch (Exception e1) {
System.err.println("UDFReflect evaluate " + e1 + " method = " + m + " args = " + Arrays.asList(parameterJavaValues));
}
return null;
}
Aggregations