xushiyan commented on code in PR #18040:
URL: https://github.com/apache/hudi/pull/18040#discussion_r2748979018
##########
hudi-flink-datasource/hudi-flink/src/main/java/org/apache/hudi/source/reader/HoodieRecordEmitter.java:
##########
@@ -18,19 +18,75 @@
package org.apache.hudi.source.reader;
+import org.apache.flink.api.common.eventtime.Watermark;
import org.apache.flink.api.connector.source.SourceOutput;
import org.apache.flink.connector.base.source.reader.RecordEmitter;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import org.apache.hudi.common.fs.FSUtils;
import org.apache.hudi.source.split.HoodieSourceSplit;
/**
* Default Hoodie record emitter.
- * @param <T>
+ *
+ * <p>This emitter handles watermark emission based on split information.
+ *
+ * @param <T> The type of records to emit
*/
public class HoodieRecordEmitter<T> implements
RecordEmitter<HoodieRecordWithPosition<T>, T, HoodieSourceSplit> {
+ private static final Logger LOG =
LoggerFactory.getLogger(HoodieRecordEmitter.class);
+ private HoodieSourceSplit lastSplit = null;
+ private long watermark = Long.MIN_VALUE;
@Override
public void emitRecord(HoodieRecordWithPosition<T> record, SourceOutput<T>
output, HoodieSourceSplit split) throws Exception {
+ if (lastSplit == null || !split.splitId().equals(lastSplit.splitId())) {
+ long newWatermark = extractWatermark(split);
+ if (newWatermark < watermark) {
+ LOG.warn(
+ "Received a new split with lower watermark. Previous watermark =
{}, current watermark = {}, previous split = {}, current split = {}",
+ watermark,
+ newWatermark,
+ lastSplit,
+ split);
+ } else {
+ watermark = newWatermark;
+ output.emitWatermark(new Watermark(watermark));
+ LOG.debug("Watermark = {} emitted based on split = {}", watermark,
split);
+ }
+
+ lastSplit = split;
+ }
+
output.collect(record.record());
split.updatePosition(record.fileOffset(), record.recordOffset());
Review Comment:
ok saw the comment:
> emit the watermark of previous finished split
so do we have to couple the record processing with split level emission? is
there a method hook for split processing like `onSplitFinish()` where emitting
watermark would be more appropriate?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
