lanyuanxiaoyao/hudi
1
0
Fork 0
Code Activity

[HUDI-1104] Adding support for UserDefinedPartitioners and SortModes to BulkInsert with Rows (#3149)

Browse Source
This commit is contained in:
Sivabalan Narayanan
2021-07-07 11:15:25 -04:00
committed by GitHub
parent 55ecbc662e
commit ea9e5d0e8b
31 changed files with 618 additions and 82 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
hudi-spark-datasource/hudi-spark2/src/main/java/org/apache/hudi/internal/DefaultSource.java
View File

@@ -19,6 +19,7 @@
package org.apache.hudi.internal;
import org.apache.hudi.DataSourceUtils;
import org.apache.hudi.config.HoodieInternalConfig;
import org.apache.hudi.config.HoodieWriteConfig;
import org.apache.spark.sql.SaveMode;
@@ -62,7 +63,10 @@ public class DefaultSource extends BaseDefaultSource implements DataSourceV2,
String tblName = options.get(HoodieWriteConfig.TABLE_NAME.key()).get();
// 1st arg to createHooodieConfig is not really reuqired to be set. but passing it anyways.
HoodieWriteConfig config = DataSourceUtils.createHoodieConfig(options.get(HoodieWriteConfig.AVRO_SCHEMA.key()).get(), path, tblName, options.asMap());
boolean arePartitionRecordsSorted = HoodieInternalConfig.getBulkInsertIsPartitionRecordsSorted(
options.get(HoodieInternalConfig.BULKINSERT_ARE_PARTITIONER_RECORDS_SORTED).isPresent()
? options.get(HoodieInternalConfig.BULKINSERT_ARE_PARTITIONER_RECORDS_SORTED).get() : null);
return Optional.of(new HoodieDataSourceInternalWriter(instantTime, config, schema, getSparkSession(),
getConfiguration()));
getConfiguration(), arePartitionRecordsSorted));
}
}

4
hudi-spark-datasource/hudi-spark2/src/main/java/org/apache/hudi/internal/HoodieBulkInsertDataInternalWriter.java
View File

@@ -37,9 +37,9 @@ public class HoodieBulkInsertDataInternalWriter implements DataWriter<InternalRo
public HoodieBulkInsertDataInternalWriter(HoodieTable hoodieTable, HoodieWriteConfig writeConfig,
String instantTime, int taskPartitionId, long taskId, long taskEpochId,
StructType structType) {
StructType structType, boolean arePartitionRecordsSorted) {
this.bulkInsertWriterHelper = new BulkInsertDataInternalWriterHelper(hoodieTable,
writeConfig, instantTime, taskPartitionId, taskId, taskEpochId, structType);
writeConfig, instantTime, taskPartitionId, taskId, taskEpochId, structType, arePartitionRecordsSorted);
}
@Override

6
hudi-spark-datasource/hudi-spark2/src/main/java/org/apache/hudi/internal/HoodieBulkInsertDataInternalWriterFactory.java
View File

@@ -35,18 +35,20 @@ public class HoodieBulkInsertDataInternalWriterFactory implements DataWriterFact
private final HoodieTable hoodieTable;
private final HoodieWriteConfig writeConfig;
private final StructType structType;
private final boolean arePartitionRecordsSorted;
public HoodieBulkInsertDataInternalWriterFactory(HoodieTable hoodieTable, HoodieWriteConfig writeConfig,
String instantTime, StructType structType) {
String instantTime, StructType structType, boolean arePartitionRecordsSorted) {
this.hoodieTable = hoodieTable;
this.writeConfig = writeConfig;
this.instantTime = instantTime;
this.structType = structType;
this.arePartitionRecordsSorted = arePartitionRecordsSorted;
}
@Override
public DataWriter<InternalRow> createDataWriter(int partitionId, long taskId, long epochId) {
return new HoodieBulkInsertDataInternalWriter(hoodieTable, writeConfig, instantTime, partitionId, taskId, epochId,
structType);
structType, arePartitionRecordsSorted);
}
}

6
hudi-spark-datasource/hudi-spark2/src/main/java/org/apache/hudi/internal/HoodieDataSourceInternalWriter.java
View File

@@ -45,12 +45,14 @@ public class HoodieDataSourceInternalWriter implements DataSourceWriter {
private final HoodieWriteConfig writeConfig;
private final StructType structType;
private final DataSourceInternalWriterHelper dataSourceInternalWriterHelper;
private final Boolean arePartitionRecordsSorted;
public HoodieDataSourceInternalWriter(String instantTime, HoodieWriteConfig writeConfig, StructType structType,
SparkSession sparkSession, Configuration configuration) {
SparkSession sparkSession, Configuration configuration, boolean arePartitionRecordsSorted) {
this.instantTime = instantTime;
this.writeConfig = writeConfig;
this.structType = structType;
this.arePartitionRecordsSorted = arePartitionRecordsSorted;
this.dataSourceInternalWriterHelper = new DataSourceInternalWriterHelper(instantTime, writeConfig, structType,
sparkSession, configuration);
}
@@ -60,7 +62,7 @@ public class HoodieDataSourceInternalWriter implements DataSourceWriter {
dataSourceInternalWriterHelper.createInflightCommit();
if (WriteOperationType.BULK_INSERT == dataSourceInternalWriterHelper.getWriteOperationType()) {
return new HoodieBulkInsertDataInternalWriterFactory(dataSourceInternalWriterHelper.getHoodieTable(),
writeConfig, instantTime, structType);
writeConfig, instantTime, structType, arePartitionRecordsSorted);
} else {
throw new IllegalArgumentException("Write Operation Type + " + dataSourceInternalWriterHelper.getWriteOperationType() + " not supported ");
}

29
hudi-spark-datasource/hudi-spark2/src/test/java/org/apache/hudi/internal/TestHoodieBulkInsertDataInternalWriter.java
View File

@@ -28,9 +28,13 @@ import org.apache.spark.sql.Dataset;
import org.apache.spark.sql.Row;
import org.apache.spark.sql.catalyst.InternalRow;
import org.junit.jupiter.api.Test;
import org.junit.jupiter.params.ParameterizedTest;
import org.junit.jupiter.params.provider.Arguments;
import org.junit.jupiter.params.provider.MethodSource;
import java.util.ArrayList;
import java.util.List;
import java.util.stream.Stream;
import static org.apache.hudi.testutils.SparkDatasetTestUtils.ENCODER;
import static org.apache.hudi.testutils.SparkDatasetTestUtils.STRUCT_TYPE;
@@ -46,16 +50,26 @@ import static org.junit.jupiter.api.Assertions.fail;
public class TestHoodieBulkInsertDataInternalWriter extends
HoodieBulkInsertInternalWriterTestBase {
@Test
public void testDataInternalWriter() throws Exception {
private static Stream<Arguments> configParams() {
Object[][] data = new Object[][] {
{true},
{false}
};
return Stream.of(data).map(Arguments::of);
}
@ParameterizedTest
@MethodSource("configParams")
public void testDataInternalWriter(boolean sorted) throws Exception {
// init config and table
HoodieWriteConfig cfg = getConfigBuilder(basePath).build();
HoodieTable table = HoodieSparkTable.create(cfg, context, metaClient);
// execute N rounds
for (int i = 0; i < 5; i++) {
for (int i = 0; i < 3; i++) {
String instantTime = "00" + i;
// init writer
HoodieBulkInsertDataInternalWriter writer = new HoodieBulkInsertDataInternalWriter(table, cfg, instantTime, RANDOM.nextInt(100000), RANDOM.nextLong(), RANDOM.nextLong(), STRUCT_TYPE);
HoodieBulkInsertDataInternalWriter writer = new HoodieBulkInsertDataInternalWriter(table, cfg, instantTime, RANDOM.nextInt(100000), RANDOM.nextLong(), RANDOM.nextLong(), STRUCT_TYPE,
sorted);
int size = 10 + RANDOM.nextInt(1000);
// write N rows to partition1, N rows to partition2 and N rows to partition3 ... Each batch should create a new RowCreateHandle and a new file
@@ -78,7 +92,7 @@ public class TestHoodieBulkInsertDataInternalWriter extends
Option<List<String>> fileNames = Option.of(new ArrayList<>());
// verify write statuses
assertWriteStatuses(commitMetadata.getWriteStatuses(), batches, size, fileAbsPaths, fileNames);
assertWriteStatuses(commitMetadata.getWriteStatuses(), batches, size, sorted, fileAbsPaths, fileNames);
// verify rows
Dataset<Row> result = sqlContext.read().parquet(fileAbsPaths.get().toArray(new String[0]));
@@ -99,7 +113,8 @@ public class TestHoodieBulkInsertDataInternalWriter extends
String partitionPath = HoodieTestDataGenerator.DEFAULT_PARTITION_PATHS[0];
String instantTime = "001";
HoodieBulkInsertDataInternalWriter writer = new HoodieBulkInsertDataInternalWriter(table, cfg, instantTime, RANDOM.nextInt(100000), RANDOM.nextLong(), RANDOM.nextLong(), STRUCT_TYPE);
HoodieBulkInsertDataInternalWriter writer = new HoodieBulkInsertDataInternalWriter(table, cfg, instantTime, RANDOM.nextInt(100000), RANDOM.nextLong(), RANDOM.nextLong(), STRUCT_TYPE,
false);
int size = 10 + RANDOM.nextInt(100);
int totalFailures = 5;
@@ -131,7 +146,7 @@ public class TestHoodieBulkInsertDataInternalWriter extends
Option<List<String>> fileAbsPaths = Option.of(new ArrayList<>());
Option<List<String>> fileNames = Option.of(new ArrayList<>());
// verify write statuses
assertWriteStatuses(commitMetadata.getWriteStatuses(), 1, size / 2, fileAbsPaths, fileNames);
assertWriteStatuses(commitMetadata.getWriteStatuses(), 1, size / 2, false, fileAbsPaths, fileNames);
// verify rows
Dataset<Row> result = sqlContext.read().parquet(fileAbsPaths.get().toArray(new String[0]));

10
hudi-spark-datasource/hudi-spark2/src/test/java/org/apache/hudi/internal/TestHoodieDataSourceInternalWriter.java
View File

@@ -52,7 +52,7 @@ public class TestHoodieDataSourceInternalWriter extends
String instantTime = "001";
// init writer
HoodieDataSourceInternalWriter dataSourceInternalWriter =
new HoodieDataSourceInternalWriter(instantTime, cfg, STRUCT_TYPE, sqlContext.sparkSession(), hadoopConf);
new HoodieDataSourceInternalWriter(instantTime, cfg, STRUCT_TYPE, sqlContext.sparkSession(), hadoopConf, false);
DataWriter<InternalRow> writer = dataSourceInternalWriter.createWriterFactory().createDataWriter(0, RANDOM.nextLong(), RANDOM.nextLong());
String[] partitionPaths = HoodieTestDataGenerator.DEFAULT_PARTITION_PATHS;
@@ -98,7 +98,7 @@ public class TestHoodieDataSourceInternalWriter extends
String instantTime = "00" + i;
// init writer
HoodieDataSourceInternalWriter dataSourceInternalWriter =
new HoodieDataSourceInternalWriter(instantTime, cfg, STRUCT_TYPE, sqlContext.sparkSession(), hadoopConf);
new HoodieDataSourceInternalWriter(instantTime, cfg, STRUCT_TYPE, sqlContext.sparkSession(), hadoopConf, false);
List<HoodieWriterCommitMessage> commitMessages = new ArrayList<>();
Dataset<Row> totalInputRows = null;
@@ -142,7 +142,7 @@ public class TestHoodieDataSourceInternalWriter extends
String instantTime = "00" + i;
// init writer
HoodieDataSourceInternalWriter dataSourceInternalWriter =
new HoodieDataSourceInternalWriter(instantTime, cfg, STRUCT_TYPE, sqlContext.sparkSession(), hadoopConf);
new HoodieDataSourceInternalWriter(instantTime, cfg, STRUCT_TYPE, sqlContext.sparkSession(), hadoopConf, false);
List<HoodieWriterCommitMessage> commitMessages = new ArrayList<>();
Dataset<Row> totalInputRows = null;
@@ -189,7 +189,7 @@ public class TestHoodieDataSourceInternalWriter extends
String instantTime0 = "00" + 0;
// init writer
HoodieDataSourceInternalWriter dataSourceInternalWriter =
new HoodieDataSourceInternalWriter(instantTime0, cfg, STRUCT_TYPE, sqlContext.sparkSession(), hadoopConf);
new HoodieDataSourceInternalWriter(instantTime0, cfg, STRUCT_TYPE, sqlContext.sparkSession(), hadoopConf, false);
DataWriter<InternalRow> writer = dataSourceInternalWriter.createWriterFactory().createDataWriter(0, RANDOM.nextLong(), RANDOM.nextLong());
List<String> partitionPaths = Arrays.asList(HoodieTestDataGenerator.DEFAULT_PARTITION_PATHS);
@@ -227,7 +227,7 @@ public class TestHoodieDataSourceInternalWriter extends
// 2nd batch. abort in the end
String instantTime1 = "00" + 1;
dataSourceInternalWriter =
new HoodieDataSourceInternalWriter(instantTime1, cfg, STRUCT_TYPE, sqlContext.sparkSession(), hadoopConf);
new HoodieDataSourceInternalWriter(instantTime1, cfg, STRUCT_TYPE, sqlContext.sparkSession(), hadoopConf, false);
writer = dataSourceInternalWriter.createWriterFactory().createDataWriter(1, RANDOM.nextLong(), RANDOM.nextLong());
for (int j = 0; j < batches; j++) {