zhijiangW commented on a change in pull request #6911: [FLINK-10656] Refactor
org.apache.flink.runtime.io.network.api.reader…
URL: https://github.com/apache/flink/pull/6911#discussion_r228020917
##########
File path:
flink-runtime/src/main/java/org/apache/flink/runtime/io/network/api/reader/IterationReader.java
##########
@@ -21,34 +21,21 @@
import org.apache.flink.runtime.event.TaskEvent;
import org.apache.flink.runtime.util.event.EventListener;
-import java.io.IOException;
-
/**
- * The basic API for every reader.
+ * Reader for iteration.
*/
-public interface ReaderBase {
-
- /**
- * Returns whether the reader has consumed the input.
- */
- boolean isFinished();
-
- //
------------------------------------------------------------------------
- // Task events
- //
------------------------------------------------------------------------
-
- void sendTaskEvent(TaskEvent event) throws IOException;
-
- void registerTaskEventListener(EventListener<TaskEvent> listener,
Class<? extends TaskEvent> eventType);
-
- //
------------------------------------------------------------------------
- // Iterations
- //
------------------------------------------------------------------------
+public interface IterationReader {
void setIterativeReader();
void startNextSuperstep();
boolean hasReachedEndOfSuperstep();
+ /**
+ * Returns whether the reader has consumed the input.
+ */
+ boolean isFinished();
Review comment:
Yes, I just confirmed it is actually only used in iteration, so all the
current methods in `ReaderBase` are for iteration. If we change it to
`IterationReaderBase`, the corresponding `AbstractReader` should also be
changed to `AbstractIterationReader`, because the current `AbstractReader` is
also only for iteration.
But I think it still seems a little strange to let `IterationReaderBase` in
the upper layer, because the current `Reader` interface which is used for
common read extends the `IterationReaderBase`. This relationship seems not make
sense.
If we have two independent reader base interfaces, and one is for iteration,
the other is for common case. Then the specific record reader can implement
both of interfaces for iteration scenario or only implement common reader
interface for non-iteration scenario.
So it may need reorganize the current relationships if we want to change
this part. Easily change the name to `IterationReaderBase` may bring literal
confusing.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
