Skip to content

Avoid underestimating size of partitioned hive tables #17677

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
raunaqmorarka merged 2 commits into from
Jun 6, 2023
Merged

Avoid underestimating size of partitioned hive tables #17677

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
14fd6be
Rename in AbstractTestHive from empty to zero table stats
Dith3r May 31, 2023
e62319c
Avoid underestimating size of partitioned hive tables
Dith3r Jun 1, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
75 changes: 75 additions & 0 deletions plugin/trino-hive-hadoop2/src/test/java/io/trino/plugin/hive/TestHive.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,9 @@
*/
package io.trino.plugin.hive;

import com.google.common.collect.ImmutableList;
import com.google.common.net.HostAndPort;
import io.trino.spi.connector.SchemaTableName;
import org.apache.hadoop.net.NetUtils;
import org.testng.SkipException;
import org.testng.annotations.BeforeClass;
Expand Down Expand Up @@ -75,4 +77,77 @@ public void testHiveViewTranslationError()
// TODO: combine this with tests for successful translation (currently in TestHiveViews product test)
}
}

@Override
public void testUpdateBasicPartitionStatistics()
throws Exception
{
SchemaTableName tableName = temporaryTable("update_basic_partition_statistics");
try {
createDummyPartitionedTable(tableName, STATISTICS_PARTITIONED_TABLE_COLUMNS);
// When the table has partitions, but row count statistics are set to zero, we treat this case as empty
// statistics to avoid underestimation in the CBO. This scenario may be caused when other engines are
// used to ingest data into partitioned hive tables.
testUpdatePartitionStatistics(
tableName,
PartitionStatistics.empty(),
ImmutableList.of(BASIC_STATISTICS_1, BASIC_STATISTICS_2),
ImmutableList.of(BASIC_STATISTICS_2, BASIC_STATISTICS_1));
}
finally {
dropTable(tableName);
}
}

@Override
public void testUpdatePartitionColumnStatistics()
throws Exception
{
SchemaTableName tableName = temporaryTable("update_partition_column_statistics");
try {
createDummyPartitionedTable(tableName, STATISTICS_PARTITIONED_TABLE_COLUMNS);
// When the table has partitions, but row count statistics are set to zero, we treat this case as empty
// statistics to avoid underestimation in the CBO. This scenario may be caused when other engines are
// used to ingest data into partitioned hive tables.
testUpdatePartitionStatistics(
tableName,
PartitionStatistics.empty(),
ImmutableList.of(STATISTICS_1_1, STATISTICS_1_2, STATISTICS_2),
ImmutableList.of(STATISTICS_1_2, STATISTICS_1_1, STATISTICS_2));
}
finally {
dropTable(tableName);
}
}

@Override
public void testUpdatePartitionColumnStatisticsEmptyOptionalFields()
throws Exception
{
SchemaTableName tableName = temporaryTable("update_partition_column_statistics");
try {
createDummyPartitionedTable(tableName, STATISTICS_PARTITIONED_TABLE_COLUMNS);
// When the table has partitions, but row count statistics are set to zero, we treat this case as empty
// statistics to avoid underestimation in the CBO. This scenario may be caused when other engines are
// used to ingest data into partitioned hive tables.
testUpdatePartitionStatistics(
tableName,
PartitionStatistics.empty(),
ImmutableList.of(STATISTICS_EMPTY_OPTIONAL_FIELDS),
ImmutableList.of(STATISTICS_EMPTY_OPTIONAL_FIELDS));
}
finally {
dropTable(tableName);
}
}

@Override
public void testStorePartitionWithStatistics()
throws Exception
{
// When the table has partitions, but row count statistics are set to zero, we treat this case as empty
// statistics to avoid underestimation in the CBO. This scenario may be caused when other engines are
// used to ingest data into partitioned hive tables.
testStorePartitionWithStatistics(STATISTICS_PARTITIONED_TABLE_COLUMNS, STATISTICS_1, STATISTICS_2, STATISTICS_1_1, PartitionStatistics.empty());
}
}
13 changes: 13 additions & 0 deletions ...trino-hive/src/main/java/io/trino/plugin/hive/metastore/alluxio/AlluxioHiveMetastore.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,6 +63,7 @@
import static io.trino.plugin.hive.metastore.thrift.ThriftMetastoreUtil.getHiveBasicStatistics;
import static io.trino.plugin.hive.util.HiveUtil.makePartName;
import static io.trino.spi.StandardErrorCode.NOT_SUPPORTED;
import static java.util.AbstractMap.SimpleEntry;
import static java.util.Objects.requireNonNull;

Comment thread
Dith3r marked this conversation as resolved.
Outdated
/**
Expand Down Expand Up @@ -161,6 +162,18 @@ public Map<String, PartitionStatistics> getPartitionStatistics(Table table, List
Map<String, OptionalLong> partitionRowCounts = partitionBasicStatistics.entrySet().stream()
.collect(toImmutableMap(Map.Entry::getKey, entry -> entry.getValue().getRowCount()));

long tableRowCount = partitionRowCounts.values().stream()
.mapToLong(count -> count.orElse(0))
.sum();
if (!partitionRowCounts.isEmpty() && tableRowCount == 0) {
// When the table has partitions, but row count statistics are set to zero, we treat this case as empty
// statistics to avoid underestimation in the CBO. This scenario may be caused when other engines are
// used to ingest data into partitioned hive tables.
partitionBasicStatistics = partitionBasicStatistics.keySet().stream()
.map(key -> new SimpleEntry<>(key, HiveBasicStatistics.createEmptyStatistics()))
.collect(toImmutableMap(SimpleEntry::getKey, SimpleEntry::getValue));
}

Map<String, List<ColumnStatisticsInfo>> colStatsMap = client.getPartitionColumnStatistics(table.getDatabaseName(), table.getTableName(),
ImmutableList.copyOf(partitionBasicStatistics.keySet()), dataColumns);
Map<String, Map<String, HiveColumnStatistics>> partitionColumnStatistics = colStatsMap.entrySet().stream()
Expand Down
17 changes: 16 additions & 1 deletion plugin/trino-hive/src/main/java/io/trino/plugin/hive/metastore/glue/GlueHiveMetastore.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -116,6 +116,7 @@
import javax.inject.Inject;

import java.time.Duration;
import java.util.AbstractMap.SimpleEntry;
import java.util.ArrayList;
import java.util.Comparator;
import java.util.List;
Expand Down Expand Up @@ -318,10 +319,24 @@ public PartitionStatistics getTableStatistics(Table table)
@Override
public Map<String, PartitionStatistics> getPartitionStatistics(Table table, List<Partition> partitions)
{
return columnStatisticsProvider.getPartitionColumnStatistics(partitions).entrySet().stream()
Map<String, PartitionStatistics> partitionBasicStatistics = columnStatisticsProvider.getPartitionColumnStatistics(partitions).entrySet().stream()
.collect(toImmutableMap(
entry -> makePartitionName(table, entry.getKey()),
entry -> new PartitionStatistics(getHiveBasicStatistics(entry.getKey().getParameters()), entry.getValue())));

long tableRowCount = partitionBasicStatistics.values().stream()
.mapToLong(partitionStatistics -> partitionStatistics.getBasicStatistics().getRowCount().orElse(0))
.sum();
if (!partitionBasicStatistics.isEmpty() && tableRowCount == 0) {
// When the table has partitions, but row count statistics are set to zero, we treat this case as empty
// statistics to avoid underestimation in the CBO. This scenario may be caused when other engines are
// used to ingest data into partitioned hive tables.
partitionBasicStatistics = partitionBasicStatistics.keySet().stream()
.map(key -> new SimpleEntry<>(key, PartitionStatistics.empty()))
.collect(toImmutableMap(SimpleEntry::getKey, SimpleEntry::getValue));
}

return partitionBasicStatistics;
}

@Override
Expand Down
14 changes: 14 additions & 0 deletions ...n/trino-hive/src/main/java/io/trino/plugin/hive/metastore/thrift/ThriftHiveMetastore.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -87,6 +87,7 @@
import java.io.IOException;
import java.net.InetAddress;
import java.net.UnknownHostException;
import java.util.AbstractMap.SimpleEntry;
import java.util.ArrayList;
import java.util.Collections;
import java.util.HashMap;
Expand Down Expand Up @@ -387,6 +388,19 @@ public Map<String, PartitionStatistics> getPartitionStatistics(Table table, List
}));
Map<String, OptionalLong> partitionRowCounts = partitionBasicStatistics.entrySet().stream()
.collect(toImmutableMap(Map.Entry::getKey, entry -> entry.getValue().getRowCount()));

long tableRowCount = partitionRowCounts.values().stream()
.mapToLong(count -> count.orElse(0))
.sum();
if (!partitionRowCounts.isEmpty() && tableRowCount == 0) {
Comment thread
Dith3r marked this conversation as resolved.
Outdated
// When the table has partitions, but row count statistics are set to zero, we treat this case as empty
// statistics to avoid underestimation in the CBO. This scenario may be caused when other engines are
// used to ingest data into partitioned hive tables.
partitionBasicStatistics = partitionBasicStatistics.keySet().stream()
.map(partitionName -> new SimpleEntry<>(partitionName, HiveBasicStatistics.createEmptyStatistics()))
.collect(toImmutableMap(SimpleEntry::getKey, SimpleEntry::getValue));
}

Map<String, Map<String, HiveColumnStatistics>> partitionColumnStatistics = getPartitionColumnStatistics(
table.getDbName(),
table.getTableName(),
Expand Down
Loading