Skip to content

[HUDI-4601] Read error from MOR table after compaction with timestamp partitioning #6365

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
danny0405 merged 6 commits into from
Aug 17, 2022
Merged

[HUDI-4601] Read error from MOR table after compaction with timestamp partitioning #6365

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
86f98b8
read error from mor after compaction
Aug 11, 2022
02cbe72
fix checkstyle
Aug 11, 2022
4ed8bd8
update
Aug 14, 2022
3b63d1c
update
Aug 15, 2022
62209a9
update from review
Aug 16, 2022
800e7ef
rollback unnecessary change
Aug 16, 2022
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
18 changes: 15 additions & 3 deletions ...rce/hudi-flink/src/main/java/org/apache/hudi/table/format/mor/MergeOnReadInputFormat.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -68,6 +68,7 @@
import java.util.Set;
import java.util.stream.IntStream;

import static org.apache.flink.table.types.logical.LogicalTypeRoot.TIMESTAMP_WITHOUT_TIME_ZONE;
import static org.apache.hudi.hadoop.utils.HoodieInputFormatUtils.HOODIE_COMMIT_TIME_COL_POS;
import static org.apache.hudi.hadoop.utils.HoodieInputFormatUtils.HOODIE_RECORD_KEY_COL_POS;
import static org.apache.hudi.table.format.FormatUtils.buildAvroRecordBySchema;
Expand Down Expand Up @@ -298,10 +299,21 @@ private ParquetColumnarRowSplitReader getReader(String path, int[] requiredPos)
new org.apache.hadoop.fs.Path(path).getParent(),
this.conf.getBoolean(FlinkOptions.HIVE_STYLE_PARTITIONING),
FilePathUtils.extractPartitionKeys(this.conf));

Copy link
Contributor

@danny0405 danny0405 Aug 13, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks, why not fix the partObjects directly when generating it?

Copy link
Contributor Author

@wuwenchi wuwenchi Aug 14, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks a good way! Modified.

LinkedHashMap<String, Object> partObjects = new LinkedHashMap<>();
partSpec.forEach((k, v) -> partObjects.put(k, DataTypeUtils.resolvePartition(
defaultPartName.equals(v) ? null : v,
fieldTypes.get(fieldNames.indexOf(k)))));

// can't convert data by partition value if timestamp is used to partition key.
partSpec.entrySet().stream()
.filter(entry -> fieldTypes
.get(fieldNames.indexOf(entry.getKey()))
.getLogicalType().getTypeRoot() != TIMESTAMP_WITHOUT_TIME_ZONE)
Copy link
Contributor

@danny0405 danny0405 Aug 15, 2022 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why ignore the timestamp partition field directly ? The current code may generates a long number string as partition path field, do you mean the DataTypeUtils.resolvePartition can not handle such a field ? Then we should fix DataTypeUtils.resolvePartition firstly, and we should handle different precision for the timestamp type.

.forEach(entry -> {
String k = entry.getKey();
String v = entry.getValue();
partObjects.put(k, DataTypeUtils.resolvePartition(
defaultPartName.equals(v) ? null : v,
fieldTypes.get(fieldNames.indexOf(k))));
});

return ParquetSplitReaderUtil.genPartColumnarRowReader(
this.conf.getBoolean(FlinkOptions.UTC_TIMEZONE),
Expand Down
54 changes: 54 additions & 0 deletions ...ink-datasource/hudi-flink/src/test/java/org/apache/hudi/table/ITTestHoodieDataSource.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,11 +18,16 @@

package org.apache.hudi.table;

import org.apache.flink.table.types.DataType;
import org.apache.flink.table.types.logical.RowType;
import org.apache.hadoop.fs.FileStatus;
import org.apache.hadoop.fs.Path;
import org.apache.hudi.adapter.TestTableEnvs;
import org.apache.hudi.common.model.DefaultHoodieRecordPayload;
import org.apache.hudi.common.model.HoodieTableType;
import org.apache.hudi.common.table.timeline.HoodieTimeline;
import org.apache.hudi.configuration.FlinkOptions;
import org.apache.hudi.source.FileIndex;
import org.apache.hudi.table.catalog.HoodieHiveCatalog;
import org.apache.hudi.table.catalog.HoodieCatalogTestUtils;
import org.apache.hudi.util.StreamerUtil;
Expand Down Expand Up @@ -56,7 +61,9 @@
import org.junit.jupiter.params.provider.ValueSource;

import java.io.File;
import java.util.Arrays;
import java.util.Collection;
import java.util.Collections;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
Expand All @@ -66,9 +73,11 @@
import java.util.stream.Collectors;
import java.util.stream.Stream;

import static org.apache.hudi.configuration.FlinkOptions.PARTITION_DEFAULT_NAME;
import static org.apache.hudi.utils.TestConfigurations.catalog;
import static org.apache.hudi.utils.TestConfigurations.sql;
import static org.apache.hudi.utils.TestData.assertRowsEquals;
import static org.junit.jupiter.api.Assertions.assertEquals;
import static org.junit.jupiter.api.Assertions.assertThrows;
import static org.junit.jupiter.api.Assertions.assertTrue;

Expand Down Expand Up @@ -1395,6 +1404,51 @@ void testWriteReadWithComputedColumns() {
assertRowsEquals(result2, "[+I[3]]");
}

@Test
public void testReadMorTableWithCompactionAndTimestampPartition() throws TableNotExistException, InterruptedException {
TableEnvironment tableEnv = batchTableEnv;
String createSql = sql("t1")
.field("id int")
.field("ts timestamp(3)")
.pkField("id")
.partitionField("ts")
.option(FlinkOptions.PATH, tempFile.getAbsolutePath())
.option(FlinkOptions.TABLE_TYPE, FlinkOptions.TABLE_TYPE_MERGE_ON_READ)
.option(FlinkOptions.PARTITION_PATH_FIELD, "ts")
.option("hoodie.compact.inline", true)
.option(FlinkOptions.COMPACTION_DELTA_COMMITS, 2)
.option(FlinkOptions.PARTITION_FORMAT, FlinkOptions.PARTITION_FORMAT_DASHED_DAY)
.end();
tableEnv.executeSql(createSql);

execInsertSql(tableEnv, "insert into t1 values (1, TIMESTAMP '2022-08-11 10:05:59')");
List<Row> result1 = CollectionUtil.iterableToList(() -> tableEnv.sqlQuery("select * from t1").execute().collect());

execInsertSql(tableEnv, "insert into t1 values (2, TIMESTAMP '2022-08-11 11:05:59')");
List<Row> result2 = CollectionUtil.iterableToList(() -> tableEnv.sqlQuery("select * from t1").execute().collect());

ObjectPath objectPath = ObjectPath.fromString(tableEnv.getCurrentDatabase() + ".t1");
String currentCatalog = tableEnv.getCurrentCatalog();
DataType dataType = tableEnv.getCatalog(currentCatalog).get().getTable(objectPath).getSchema().toRowDataType();
RowType rowType = (RowType) dataType.getLogicalType();
FileIndex fileIndex = FileIndex.instance(
new Path(tempFile.getAbsolutePath()), tableEnv.getConfig().getConfiguration(), rowType);
List<Map<String, String>> partitions =
fileIndex.getPartitions(Collections.singletonList("ts"), PARTITION_DEFAULT_NAME.defaultValue(), false);
assertEquals(1, partitions.size());
assertEquals("2022-08-11", partitions.get(0).get("ts"));

FileStatus[] fileStatuses = fileIndex.getFilesInPartitions();
// should have two log files and one parquet file
assertEquals(3, fileStatuses.length);
assertEquals(1,
Arrays.stream(fileStatuses).filter(file -> file.getPath().getName().endsWith("parquet")).count()
);

assertRowsEquals(result1, "[+I[1, 2022-08-11T10:05:59]]");
assertRowsEquals(result2, "[+I[1, 2022-08-11T10:05:59], +I[2, 2022-08-11T11:05:59]]");
}

// -------------------------------------------------------------------------
// Utilities
// -------------------------------------------------------------------------
Expand Down