/**
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
* http://www.apache.org/licenses/LICENSE-2.0
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.hadoop.hive.llap.io.decode;
import java.io.IOException;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.List;
import org.apache.hadoop.hive.common.io.encoded.EncodedColumnBatch;
import org.apache.hadoop.hive.llap.counters.LlapIOCounters;
import org.apache.hadoop.hive.llap.counters.QueryFragmentCounters;
import org.apache.hadoop.hive.llap.io.api.impl.ColumnVectorBatch;
import org.apache.hadoop.hive.llap.io.api.impl.LlapIoImpl;
import org.apache.hadoop.hive.llap.io.metadata.ConsumerFileMetadata;
import org.apache.hadoop.hive.llap.io.metadata.ConsumerStripeMetadata;
import org.apache.hadoop.hive.llap.metrics.LlapDaemonIOMetrics;
import org.apache.hadoop.hive.ql.exec.vector.BytesColumnVector;
import org.apache.hadoop.hive.ql.exec.vector.ColumnVector;
import org.apache.hadoop.hive.ql.exec.vector.DecimalColumnVector;
import org.apache.hadoop.hive.ql.exec.vector.DoubleColumnVector;
import org.apache.hadoop.hive.ql.exec.vector.ListColumnVector;
import org.apache.hadoop.hive.ql.exec.vector.LongColumnVector;
import org.apache.hadoop.hive.ql.exec.vector.MapColumnVector;
import org.apache.hadoop.hive.ql.exec.vector.StructColumnVector;
import org.apache.hadoop.hive.ql.exec.vector.TimestampColumnVector;
import org.apache.hadoop.hive.ql.exec.vector.UnionColumnVector;
import org.apache.hadoop.hive.ql.exec.vector.VectorizedRowBatch;
import org.apache.orc.CompressionCodec;
import org.apache.orc.impl.PositionProvider;
import org.apache.hadoop.hive.ql.io.orc.encoded.Consumer;
import org.apache.hadoop.hive.ql.io.orc.encoded.EncodedTreeReaderFactory;
import org.apache.hadoop.hive.ql.io.orc.encoded.EncodedTreeReaderFactory.SettableTreeReader;
import org.apache.hadoop.hive.ql.io.orc.encoded.OrcBatchKey;
import org.apache.hadoop.hive.ql.io.orc.encoded.Reader.OrcEncodedColumnBatch;
import org.apache.hadoop.hive.ql.io.orc.RecordReaderImpl;
import org.apache.orc.TypeDescription;
import org.apache.orc.impl.SchemaEvolution;
import org.apache.orc.impl.TreeReaderFactory;
import org.apache.orc.impl.TreeReaderFactory.StructTreeReader;
import org.apache.orc.impl.TreeReaderFactory.TreeReader;
import org.apache.orc.impl.WriterImpl;
import org.apache.orc.OrcProto;
public class OrcEncodedDataConsumer
extends EncodedDataConsumer<OrcBatchKey, OrcEncodedColumnBatch> {
private TreeReaderFactory.TreeReader[] columnReaders;
private int[] columnMapping; // Mapping from columnReaders (by index) to columns in file schema.
private int previousStripeIndex = -1;
private ConsumerFileMetadata fileMetadata; // We assume one request is only for one file.
private CompressionCodec codec;
private List<ConsumerStripeMetadata> stripes;
private final boolean skipCorrupt; // TODO: get rid of this
private final QueryFragmentCounters counters;
private boolean[] includedColumns;
private SchemaEvolution evolution;
public OrcEncodedDataConsumer(
Consumer<ColumnVectorBatch> consumer, int colCount, boolean skipCorrupt,
QueryFragmentCounters counters, LlapDaemonIOMetrics ioMetrics) {
super(consumer, colCount, ioMetrics);
// TODO: get rid of this
this.skipCorrupt = skipCorrupt;
this.counters = counters;
}
public void setFileMetadata(ConsumerFileMetadata f) {
assert fileMetadata == null;
fileMetadata = f;
stripes = new ArrayList<>(f.getStripeCount());
codec = WriterImpl.createCodec(fileMetadata.getCompressionKind());
}
public void setStripeMetadata(ConsumerStripeMetadata m) {
assert stripes != null;
int newIx = m.getStripeIx();
for (int i = stripes.size(); i <= newIx; ++i) {
stripes.add(null);
}
assert stripes.get(newIx) == null;
stripes.set(newIx, m);
}
@Override
protected void decodeBatch(OrcEncodedColumnBatch batch,
Consumer<ColumnVectorBatch> downstreamConsumer) {
long startTime = counters.startTimeCounter();
int currentStripeIndex = batch.getBatchKey().stripeIx;
boolean sameStripe = currentStripeIndex == previousStripeIndex;
try {
ConsumerStripeMetadata stripeMetadata = stripes.get(currentStripeIndex);
// Get non null row count from root column, to get max vector batches
int rgIdx = batch.getBatchKey().rgIx;
long nonNullRowCount = -1;
if (rgIdx == OrcEncodedColumnBatch.ALL_RGS) {
nonNullRowCount = stripeMetadata.getRowCount();
} else {
OrcProto.RowIndexEntry rowIndex = stripeMetadata.getRowIndexEntry(0, rgIdx);
nonNullRowCount = getRowCount(rowIndex);
}
int maxBatchesRG = (int) ((nonNullRowCount / VectorizedRowBatch.DEFAULT_SIZE) + 1);
int batchSize = VectorizedRowBatch.DEFAULT_SIZE;
TypeDescription schema = fileMetadata.getSchema();
if (columnReaders == null || !sameStripe) {
int[] columnMapping = new int[schema.getChildren().size()];
TreeReaderFactory.Context context =
new TreeReaderFactory.ReaderContext()
.setSchemaEvolution(evolution)
.writerTimeZone(stripeMetadata.getWriterTimezone())
.skipCorrupt(skipCorrupt);
StructTreeReader treeReader = EncodedTreeReaderFactory.createRootTreeReader(
schema, stripeMetadata.getEncodings(), batch, codec, context, columnMapping);
this.columnReaders = treeReader.getChildReaders();
this.columnMapping = Arrays.copyOf(columnMapping, columnReaders.length);
positionInStreams(columnReaders, batch.getBatchKey(), stripeMetadata);
} else {
repositionInStreams(this.columnReaders, batch, sameStripe, stripeMetadata);
}
previousStripeIndex = currentStripeIndex;
for (int i = 0; i < maxBatchesRG; i++) {
// for last batch in row group, adjust the batch size
if (i == maxBatchesRG - 1) {
batchSize = (int) (nonNullRowCount % VectorizedRowBatch.DEFAULT_SIZE);
if (batchSize == 0) break;
}
ColumnVectorBatch cvb = cvbPool.take();
// assert cvb.cols.length == batch.getColumnIxs().length; // Must be constant per split.
cvb.size = batchSize;
for (int idx = 0; idx < columnReaders.length; ++idx) {
TreeReader reader = columnReaders[idx];
if (cvb.cols[idx] == null) {
// Orc store rows inside a root struct (hive writes it this way).
// When we populate column vectors we skip over the root struct.
cvb.cols[idx] = createColumn(schema.getChildren().get(columnMapping[idx]), batchSize);
}
cvb.cols[idx].ensureSize(batchSize, false);
reader.nextVector(cvb.cols[idx], null, batchSize);
}
// we are done reading a batch, send it to consumer for processing
downstreamConsumer.consumeData(cvb);
counters.incrCounter(LlapIOCounters.ROWS_EMITTED, batchSize);
}
LlapIoImpl.ORC_LOGGER.debug("Done with decode");
counters.incrTimeCounter(LlapIOCounters.DECODE_TIME_NS, startTime);
counters.incrCounter(LlapIOCounters.NUM_VECTOR_BATCHES, maxBatchesRG);
counters.incrCounter(LlapIOCounters.NUM_DECODED_BATCHES);
} catch (IOException e) {
// Caller will return the batch.
downstreamConsumer.setError(e);
}
}
private ColumnVector createColumn(TypeDescription type, int batchSize) {
switch (type.getCategory()) {
case BOOLEAN:
case BYTE:
case SHORT:
case INT:
case LONG:
case DATE:
return new LongColumnVector(batchSize);
case FLOAT:
case DOUBLE:
return new DoubleColumnVector(batchSize);
case BINARY:
case STRING:
case CHAR:
case VARCHAR:
return new BytesColumnVector(batchSize);
case TIMESTAMP:
return new TimestampColumnVector(batchSize);
case DECIMAL:
return new DecimalColumnVector(batchSize, type.getPrecision(),
type.getScale());
case STRUCT: {
List<TypeDescription> subtypeIdxs = type.getChildren();
ColumnVector[] fieldVector = new ColumnVector[subtypeIdxs.size()];
for(int i = 0; i < fieldVector.length; ++i) {
fieldVector[i] = createColumn(subtypeIdxs.get(i), batchSize);
}
return new StructColumnVector(batchSize, fieldVector);
}
case UNION: {
List<TypeDescription> subtypeIdxs = type.getChildren();
ColumnVector[] fieldVector = new ColumnVector[subtypeIdxs.size()];
for(int i=0; i < fieldVector.length; ++i) {
fieldVector[i] = createColumn(subtypeIdxs.get(i), batchSize);
}
return new UnionColumnVector(batchSize, fieldVector);
}
case LIST:
return new ListColumnVector(batchSize, createColumn(type.getChildren().get(0), batchSize));
case MAP:
List<TypeDescription> subtypeIdxs = type.getChildren();
return new MapColumnVector(batchSize, createColumn(subtypeIdxs.get(0), batchSize),
createColumn(subtypeIdxs.get(1), batchSize));
default:
throw new IllegalArgumentException("LLAP does not support " + type.getCategory());
}
}
private void positionInStreams(TreeReaderFactory.TreeReader[] columnReaders,
OrcBatchKey batchKey, ConsumerStripeMetadata stripeMetadata) throws IOException {
PositionProvider[] pps = createPositionProviders(columnReaders, batchKey, stripeMetadata);
if (pps == null) return;
for (int i = 0; i < columnReaders.length; i++) {
columnReaders[i].seek(pps);
}
}
private void repositionInStreams(TreeReaderFactory.TreeReader[] columnReaders,
EncodedColumnBatch<OrcBatchKey> batch, boolean sameStripe,
ConsumerStripeMetadata stripeMetadata) throws IOException {
PositionProvider[] pps = createPositionProviders(
columnReaders, batch.getBatchKey(), stripeMetadata);
if (pps == null) return;
for (int i = 0; i < columnReaders.length; i++) {
TreeReader reader = columnReaders[i];
// Note: we assume this never happens for SerDe reader - the batch would never have vectors.
// That is always true now; but it wasn't some day, the below would throw in getColumnData.
((SettableTreeReader) reader).setBuffers(batch, sameStripe);
// TODO: When hive moves to java8, make updateTimezone() as default method in
// SettableTreeReader so that we can avoid this check.
if (reader instanceof EncodedTreeReaderFactory.TimestampStreamReader && !sameStripe) {
((EncodedTreeReaderFactory.TimestampStreamReader) reader)
.updateTimezone(stripeMetadata.getWriterTimezone());
}
reader.seek(pps);
}
}
/**
* Position provider used in absence of indexes, e.g. for serde-based reader, where each stream
* is in its own physical 'container', always starting at 0, and there are no RGs.
*/
private final static class IndexlessPositionProvider implements PositionProvider {
@Override
public long getNext() {
return 0;
}
@Override
public String toString() {
return "indexes not supported";
}
}
private PositionProvider[] createPositionProviders(
TreeReaderFactory.TreeReader[] columnReaders, OrcBatchKey batchKey,
ConsumerStripeMetadata stripeMetadata) throws IOException {
if (columnReaders.length == 0) return null;
PositionProvider[] pps = null;
if (!stripeMetadata.supportsRowIndexes()) {
PositionProvider singleRgPp = new IndexlessPositionProvider();
pps = new PositionProvider[stripeMetadata.getEncodings().size()];
for (int i = 0; i < pps.length; ++i) {
pps[i] = singleRgPp;
}
} else {
int rowGroupIndex = batchKey.rgIx;
if (rowGroupIndex == OrcEncodedColumnBatch.ALL_RGS) {
throw new IOException("Cannot position readers without RG information");
}
// TODO: this assumes indexes in getRowIndexes would match column IDs
OrcProto.RowIndex[] ris = stripeMetadata.getRowIndexes();
pps = new PositionProvider[ris.length];
for (int i = 0; i < ris.length; ++i) {
OrcProto.RowIndex ri = ris[i];
if (ri == null) continue;
pps[i] = new RecordReaderImpl.PositionProviderImpl(ri.getEntry(rowGroupIndex));
}
}
return pps;
}
private long getRowCount(OrcProto.RowIndexEntry rowIndexEntry) {
return rowIndexEntry.getStatistics().getNumberOfValues();
}
@Override
public boolean[] getIncludedColumns() {
return includedColumns;
}
public void setIncludedColumns(final boolean[] includedColumns) {
this.includedColumns = includedColumns;
}
public void setSchemaEvolution(SchemaEvolution evolution) {
this.evolution = evolution;
}
public SchemaEvolution getSchemaEvolution() {
return evolution;
}
}