Examples with ExprNodeDesc org.apache.hadoop.hive.ql.plan.ExprNodeDesc used on opensource projects

Example 16 with ExprNodeDesc

use of org.apache.hadoop.hive.ql.plan.ExprNodeDesc in project hive by apache.

the class TestVectorGroupByOperator method testKeyTypeAggregate.

private void testKeyTypeAggregate(String aggregateName, FakeVectorRowBatchFromObjectIterables data, Map<Object, Object> expected) throws HiveException {
    List<String> mapColumnNames = new ArrayList<String>();
    mapColumnNames.add("Key");
    mapColumnNames.add("Value");
    VectorizationContext ctx = new VectorizationContext("name", mapColumnNames);
    Set<Object> keys = new HashSet<Object>();
    AggregationDesc agg = buildAggregationDesc(ctx, aggregateName, GenericUDAFEvaluator.Mode.PARTIAL1, "Value", TypeInfoFactory.getPrimitiveTypeInfo(data.getTypes()[1]));
    ArrayList<AggregationDesc> aggs = new ArrayList<AggregationDesc>();
    aggs.add(agg);
    ArrayList<String> outputColumnNames = new ArrayList<String>();
    outputColumnNames.add("_col0");
    outputColumnNames.add("_col1");
    GroupByDesc desc = new GroupByDesc();
    desc.setVectorDesc(new VectorGroupByDesc());
    desc.setOutputColumnNames(outputColumnNames);
    desc.setAggregators(aggs);
    ((VectorGroupByDesc) desc.getVectorDesc()).setProcessingMode(ProcessingMode.HASH);
    ExprNodeDesc keyExp = buildColumnDesc(ctx, "Key", TypeInfoFactory.getPrimitiveTypeInfo(data.getTypes()[0]));
    ArrayList<ExprNodeDesc> keysDesc = new ArrayList<ExprNodeDesc>();
    keysDesc.add(keyExp);
    desc.setKeys(keysDesc);
    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);
    out.setOutputInspector(new FakeCaptureOutputOperator.OutputInspector() {

        private int rowIndex;

        private String aggregateName;

        private Map<Object, Object> expected;

        private Set<Object> keys;

        @Override
        public void inspectRow(Object row, int tag) throws HiveException {
            assertTrue(row instanceof Object[]);
            Object[] fields = (Object[]) row;
            assertEquals(2, fields.length);
            Object key = fields[0];
            Object keyValue = null;
            if (null == key) {
                keyValue = null;
            } else if (key instanceof ByteWritable) {
                ByteWritable bwKey = (ByteWritable) key;
                keyValue = bwKey.get();
            } else if (key instanceof ShortWritable) {
                ShortWritable swKey = (ShortWritable) key;
                keyValue = swKey.get();
            } else if (key instanceof IntWritable) {
                IntWritable iwKey = (IntWritable) key;
                keyValue = iwKey.get();
            } else if (key instanceof LongWritable) {
                LongWritable lwKey = (LongWritable) key;
                keyValue = lwKey.get();
            } else if (key instanceof TimestampWritable) {
                TimestampWritable twKey = (TimestampWritable) key;
                keyValue = twKey.getTimestamp();
            } else if (key instanceof DoubleWritable) {
                DoubleWritable dwKey = (DoubleWritable) key;
                keyValue = dwKey.get();
            } else if (key instanceof FloatWritable) {
                FloatWritable fwKey = (FloatWritable) key;
                keyValue = fwKey.get();
            } else if (key instanceof BooleanWritable) {
                BooleanWritable bwKey = (BooleanWritable) key;
                keyValue = bwKey.get();
            } else if (key instanceof HiveDecimalWritable) {
                HiveDecimalWritable hdwKey = (HiveDecimalWritable) key;
                keyValue = hdwKey.getHiveDecimal();
            } else {
                Assert.fail(String.format("Not implemented key output type %s: %s", key.getClass().getName(), key));
            }
            String keyValueAsString = String.format("%s", keyValue);
            assertTrue(expected.containsKey(keyValue));
            Object expectedValue = expected.get(keyValue);
            Object value = fields[1];
            Validator validator = getValidator(aggregateName);
            validator.validate(keyValueAsString, expectedValue, new Object[] { value });
            keys.add(keyValue);
        }

        private FakeCaptureOutputOperator.OutputInspector init(String aggregateName, Map<Object, Object> expected, Set<Object> keys) {
            this.aggregateName = aggregateName;
            this.expected = expected;
            this.keys = keys;
            return this;
        }
    }.init(aggregateName, expected, keys));
    for (VectorizedRowBatch unit : data) {
        vgo.process(unit, 0);
    }
    vgo.close(false);
    List<Object> outBatchList = out.getCapturedRows();
    assertNotNull(outBatchList);
    assertEquals(expected.size(), outBatchList.size());
    assertEquals(expected.size(), keys.size());
}

Example 17 with ExprNodeDesc

use of org.apache.hadoop.hive.ql.plan.ExprNodeDesc 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);
}

Example 18 with ExprNodeDesc

use of org.apache.hadoop.hive.ql.plan.ExprNodeDesc 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);
}

Example 19 with ExprNodeDesc

use of org.apache.hadoop.hive.ql.plan.ExprNodeDesc in project hive by apache.

the class TestUtilities method testSerializeTimestamp.

@Test
public void testSerializeTimestamp() {
    Timestamp ts = new Timestamp(1374554702000L);
    ts.setNanos(123456);
    ExprNodeConstantDesc constant = new ExprNodeConstantDesc(ts);
    List<ExprNodeDesc> children = new ArrayList<ExprNodeDesc>(1);
    children.add(constant);
    ExprNodeGenericFuncDesc desc = new ExprNodeGenericFuncDesc(TypeInfoFactory.timestampTypeInfo, new GenericUDFFromUtcTimestamp(), children);
    assertEquals(desc.getExprString(), SerializationUtilities.deserializeExpression(SerializationUtilities.serializeExpression(desc)).getExprString());
}

Example 20 with ExprNodeDesc

use of org.apache.hadoop.hive.ql.plan.ExprNodeDesc in project hive by apache.

the class TestExecDriver method populateMapRedPlan1.

@SuppressWarnings("unchecked")
private void populateMapRedPlan1(Table src) throws SemanticException {
    ArrayList<String> outputColumns = new ArrayList<String>();
    for (int i = 0; i < 2; i++) {
        outputColumns.add("_col" + i);
    }
    // map-side work
    Operator<ReduceSinkDesc> op1 = OperatorFactory.get(ctx, PlanUtils.getReduceSinkDesc(Utilities.makeList(getStringColumn("key")), Utilities.makeList(getStringColumn("value")), outputColumns, true, -1, 1, -1, AcidUtils.Operation.NOT_ACID));
    addMapWork(mr, src, "a", op1);
    ReduceWork rWork = new ReduceWork();
    rWork.setNumReduceTasks(Integer.valueOf(1));
    rWork.setKeyDesc(op1.getConf().getKeySerializeInfo());
    rWork.getTagToValueDesc().add(op1.getConf().getValueSerializeInfo());
    mr.setReduceWork(rWork);
    // reduce side work
    Operator<FileSinkDesc> op3 = OperatorFactory.get(ctx, new FileSinkDesc(new Path(tmpdir + File.separator + "mapredplan1.out"), Utilities.defaultTd, false));
    List<ExprNodeDesc> cols = new ArrayList<ExprNodeDesc>();
    cols.add(getStringColumn(Utilities.ReduceField.VALUE.toString() + "." + outputColumns.get(1)));
    List<String> colNames = new ArrayList<String>();
    colNames.add(HiveConf.getColumnInternalName(2));
    Operator<SelectDesc> op2 = OperatorFactory.get(new SelectDesc(cols, colNames), op3);
    rWork.setReducer(op2);
}