1
0

[HUDI-3516] Implement record iterator for HoodieDataBlock (#4909)

*  Use iterator to void eager materialization to be memory friendly
This commit is contained in:
Bo Cui
2022-03-02 10:19:36 +08:00
committed by GitHub
parent a81a6326d5
commit 3fdc9332e5
14 changed files with 452 additions and 217 deletions

View File

@@ -62,6 +62,7 @@ import org.apache.hudi.common.testutils.FileCreateUtils;
import org.apache.hudi.common.testutils.HoodieMetadataTestTable;
import org.apache.hudi.common.testutils.HoodieTestDataGenerator;
import org.apache.hudi.common.testutils.HoodieTestTable;
import org.apache.hudi.common.util.ClosableIterator;
import org.apache.hudi.common.util.HoodieTimer;
import org.apache.hudi.common.util.Option;
import org.apache.hudi.common.util.collection.ExternalSpillableMap;
@@ -694,23 +695,25 @@ public class TestHoodieBackedMetadata extends TestHoodieMetadataBase {
while (logFileReader.hasNext()) {
HoodieLogBlock logBlock = logFileReader.next();
if (logBlock instanceof HoodieDataBlock) {
for (IndexedRecord indexRecord : ((HoodieDataBlock) logBlock).getRecords()) {
final GenericRecord record = (GenericRecord) indexRecord;
if (enableMetaFields) {
// Metadata table records should have meta fields!
assertNotNull(record.get(HoodieRecord.RECORD_KEY_METADATA_FIELD));
assertNotNull(record.get(HoodieRecord.COMMIT_TIME_METADATA_FIELD));
} else {
// Metadata table records should not have meta fields!
assertNull(record.get(HoodieRecord.RECORD_KEY_METADATA_FIELD));
assertNull(record.get(HoodieRecord.COMMIT_TIME_METADATA_FIELD));
}
try (ClosableIterator<IndexedRecord> recordItr = ((HoodieDataBlock) logBlock).getRecordItr()) {
recordItr.forEachRemaining(indexRecord -> {
final GenericRecord record = (GenericRecord) indexRecord;
if (enableMetaFields) {
// Metadata table records should have meta fields!
assertNotNull(record.get(HoodieRecord.RECORD_KEY_METADATA_FIELD));
assertNotNull(record.get(HoodieRecord.COMMIT_TIME_METADATA_FIELD));
} else {
// Metadata table records should not have meta fields!
assertNull(record.get(HoodieRecord.RECORD_KEY_METADATA_FIELD));
assertNull(record.get(HoodieRecord.COMMIT_TIME_METADATA_FIELD));
}
final String key = String.valueOf(record.get(HoodieMetadataPayload.KEY_FIELD_NAME));
assertFalse(key.isEmpty());
if (enableMetaFields) {
assertTrue(key.equals(String.valueOf(record.get(HoodieRecord.RECORD_KEY_METADATA_FIELD))));
}
final String key = String.valueOf(record.get(HoodieMetadataPayload.KEY_FIELD_NAME));
assertFalse(key.isEmpty());
if (enableMetaFields) {
assertTrue(key.equals(String.valueOf(record.get(HoodieRecord.RECORD_KEY_METADATA_FIELD))));
}
});
}
}
}

View File

@@ -40,6 +40,7 @@ import org.apache.hudi.common.table.log.block.HoodieDataBlock;
import org.apache.hudi.common.table.log.block.HoodieLogBlock;
import org.apache.hudi.common.table.view.TableFileSystemView;
import org.apache.hudi.common.testutils.HoodieTestTable;
import org.apache.hudi.common.util.ClosableIterator;
import org.apache.hudi.common.util.collection.ExternalSpillableMap;
import org.apache.hudi.config.HoodieWriteConfig;
import org.apache.hudi.io.storage.HoodieHFileReader;
@@ -292,14 +293,14 @@ public class TestHoodieBackedTableMetadata extends TestHoodieMetadataBase {
while (logFileReader.hasNext()) {
HoodieLogBlock logBlock = logFileReader.next();
if (logBlock instanceof HoodieDataBlock) {
for (IndexedRecord indexRecord : ((HoodieDataBlock) logBlock).getRecords()) {
final GenericRecord record = (GenericRecord) indexRecord;
// Metadata table records should not have meta fields!
assertNull(record.get(HoodieRecord.RECORD_KEY_METADATA_FIELD));
assertNull(record.get(HoodieRecord.COMMIT_TIME_METADATA_FIELD));
final String key = String.valueOf(record.get(HoodieMetadataPayload.KEY_FIELD_NAME));
assertFalse(key.isEmpty());
try (ClosableIterator<IndexedRecord> recordItr = ((HoodieDataBlock) logBlock).getRecordItr()) {
recordItr.forEachRemaining(indexRecord -> {
final GenericRecord record = (GenericRecord) indexRecord;
assertNull(record.get(HoodieRecord.RECORD_KEY_METADATA_FIELD));
assertNull(record.get(HoodieRecord.COMMIT_TIME_METADATA_FIELD));
final String key = String.valueOf(record.get(HoodieMetadataPayload.KEY_FIELD_NAME));
assertFalse(key.isEmpty());
});
}
}
}