use of org.apache.drill.exec.vector.ValueVector in project drill by apache.
the class RepeatedMapVector method load.
@Override
public void load(SerializedField metadata, DrillBuf buffer) {
final List<SerializedField> children = metadata.getChildList();
final SerializedField offsetField = children.get(0);
offsets.load(offsetField, buffer);
int bufOffset = offsetField.getBufferLength();
for (int i = 1; i < children.size(); i++) {
final SerializedField child = children.get(i);
final MaterializedField fieldDef = MaterializedField.create(child);
ValueVector vector = getChild(fieldDef.getLastName());
if (vector == null) {
// if we arrive here, we didn't have a matching vector.
vector = BasicTypeHelper.getNewVector(fieldDef, allocator);
putChild(fieldDef.getLastName(), vector);
}
final int vectorLength = child.getBufferLength();
vector.load(child, buffer.slice(bufOffset, vectorLength));
bufOffset += vectorLength;
}
assert bufOffset == buffer.capacity();
}
use of org.apache.drill.exec.vector.ValueVector in project drill by apache.
the class UnionAllRecordBatch method doWork.
@SuppressWarnings("resource")
private IterOutcome doWork() throws ClassTransformationException, IOException, SchemaChangeException {
if (allocationVectors != null) {
for (ValueVector v : allocationVectors) {
v.clear();
}
}
allocationVectors = Lists.newArrayList();
transfers.clear();
// If both sides of Union-All are empty
if (unionAllInput.isBothSideEmpty()) {
for (int i = 0; i < outputFields.size(); ++i) {
final String colName = outputFields.get(i).getPath();
final MajorType majorType = MajorType.newBuilder().setMinorType(MinorType.INT).setMode(DataMode.OPTIONAL).build();
MaterializedField outputField = MaterializedField.create(colName, majorType);
ValueVector vv = container.addOrGet(outputField, callBack);
allocationVectors.add(vv);
}
container.buildSchema(BatchSchema.SelectionVectorMode.NONE);
return IterOutcome.OK_NEW_SCHEMA;
}
final ClassGenerator<UnionAller> cg = CodeGenerator.getRoot(UnionAller.TEMPLATE_DEFINITION, context.getFunctionRegistry(), context.getOptions());
cg.getCodeGenerator().plainJavaCapable(true);
// Uncomment out this line to debug the generated code.
// cg.getCodeGenerator().saveCodeForDebugging(true);
int index = 0;
for (VectorWrapper<?> vw : current) {
ValueVector vvIn = vw.getValueVector();
// get the original input column names
SchemaPath inputPath = SchemaPath.getSimplePath(vvIn.getField().getPath());
// get the renamed column names
SchemaPath outputPath = SchemaPath.getSimplePath(outputFields.get(index).getPath());
final ErrorCollector collector = new ErrorCollectorImpl();
// cast data types (Minortype or DataMode)
if (hasSameTypeAndMode(outputFields.get(index), vw.getValueVector().getField())) {
// Transfer column
MajorType outputFieldType = outputFields.get(index).getType();
MaterializedField outputField = MaterializedField.create(outputPath.getAsUnescapedPath(), outputFieldType);
/*
todo: Fix if condition when DRILL-4824 is merged
If condition should be changed to:
`if (outputFields.get(index).getPath().equals(inputPath.getAsUnescapedPath())) {`
DRILL-5419 has changed condition to correct one but this caused regression (DRILL-5521).
Root cause is missing indication of child column in map types when it is null.
DRILL-4824 is re-working json reader implementation, including map types and will fix this problem.
Reverting condition to previous one to avoid regression till DRILL-4824 is merged.
Unit test - TestJsonReader.testKvgenWithUnionAll().
*/
if (outputFields.get(index).getPath().equals(inputPath)) {
ValueVector vvOut = container.addOrGet(outputField);
TransferPair tp = vvIn.makeTransferPair(vvOut);
transfers.add(tp);
// Copy data in order to rename the column
} else {
final LogicalExpression expr = ExpressionTreeMaterializer.materialize(inputPath, current, collector, context.getFunctionRegistry());
if (collector.hasErrors()) {
throw new SchemaChangeException(String.format("Failure while trying to materialize incoming schema. Errors:\n %s.", collector.toErrorString()));
}
ValueVector vv = container.addOrGet(outputField, callBack);
allocationVectors.add(vv);
TypedFieldId fid = container.getValueVectorId(SchemaPath.getSimplePath(outputField.getPath()));
ValueVectorWriteExpression write = new ValueVectorWriteExpression(fid, expr, true);
cg.addExpr(write);
}
// Cast is necessary
} else {
LogicalExpression expr = ExpressionTreeMaterializer.materialize(inputPath, current, collector, context.getFunctionRegistry());
if (collector.hasErrors()) {
throw new SchemaChangeException(String.format("Failure while trying to materialize incoming schema. Errors:\n %s.", collector.toErrorString()));
}
// cast to the one with the least restriction
if (vvIn.getField().getType().getMode() == DataMode.REQUIRED && outputFields.get(index).getType().getMode() != DataMode.REQUIRED) {
expr = ExpressionTreeMaterializer.convertToNullableType(expr, vvIn.getField().getType().getMinorType(), context.getFunctionRegistry(), collector);
if (collector.hasErrors()) {
throw new SchemaChangeException(String.format("Failure while trying to materialize incoming schema. Errors:\n %s.", collector.toErrorString()));
}
}
// Insert a cast before the Union operation
if (vvIn.getField().getType().getMinorType() != outputFields.get(index).getType().getMinorType()) {
expr = ExpressionTreeMaterializer.addCastExpression(expr, outputFields.get(index).getType(), context.getFunctionRegistry(), collector);
if (collector.hasErrors()) {
throw new SchemaChangeException(String.format("Failure while trying to materialize incoming schema. Errors:\n %s.", collector.toErrorString()));
}
}
final MaterializedField outputField = MaterializedField.create(outputPath.getAsUnescapedPath(), expr.getMajorType());
ValueVector vector = container.addOrGet(outputField, callBack);
allocationVectors.add(vector);
TypedFieldId fid = container.getValueVectorId(SchemaPath.getSimplePath(outputField.getPath()));
boolean useSetSafe = !(vector instanceof FixedWidthVector);
ValueVectorWriteExpression write = new ValueVectorWriteExpression(fid, expr, useSetSafe);
cg.addExpr(write);
}
++index;
}
unionall = context.getImplementationClass(cg.getCodeGenerator());
unionall.setup(context, current, this, transfers);
if (!schemaAvailable) {
container.buildSchema(BatchSchema.SelectionVectorMode.NONE);
schemaAvailable = true;
}
if (!doAlloc()) {
return IterOutcome.OUT_OF_MEMORY;
}
recordCount = unionall.unionRecords(0, current.getRecordCount(), 0);
setValueCount(recordCount);
return IterOutcome.OK;
}
use of org.apache.drill.exec.vector.ValueVector in project drill by apache.
the class TestOutputMutator method replace.
private void replace(ValueVector newVector, SchemaPath schemaPath) {
List<ValueVector> vectors = Lists.newArrayList();
for (VectorWrapper w : container) {
ValueVector vector = w.getValueVector();
if (vector.getField().getPath().equals(schemaPath)) {
vectors.add(newVector);
} else {
vectors.add(w.getValueVector());
}
container.remove(vector);
}
container.addCollection(vectors);
}
use of org.apache.drill.exec.vector.ValueVector in project drill by apache.
the class AbstractContainerVector method getLastPathType.
MajorType getLastPathType() {
if ((this.getField().getType().getMinorType() == MinorType.LIST && this.getField().getType().getMode() == DataMode.REPEATED)) {
// Use Repeated scalar type instead of Required List.
VectorWithOrdinal vord = getChildVectorWithOrdinal(null);
ValueVector v = vord.vector;
if (!(v instanceof AbstractContainerVector)) {
return v.getField().getType();
}
} else if (this.getField().getType().getMinorType() == MinorType.MAP && this.getField().getType().getMode() == DataMode.REPEATED) {
// Use Required Map
return this.getField().getType().toBuilder().setMode(DataMode.REQUIRED).build();
}
return this.getField().getType();
}
use of org.apache.drill.exec.vector.ValueVector in project drill by apache.
the class AbstractMapVector method close.
@Override
public void close() {
for (final ValueVector valueVector : vectors.values()) {
valueVector.close();
}
vectors.clear();
super.close();
}
Aggregations