You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by GitBox <gi...@apache.org> on 2022/07/13 13:47:58 UTC
[GitHub] [flink] godfreyhe commented on a diff in pull request #19759: [FLINK-27623][table] Add 'table.exec.async-lookup.output-mode' to ExecutionConfigOptions
godfreyhe commented on code in PR #19759:
URL: https://github.com/apache/flink/pull/19759#discussion_r920094045
##########
flink-table/flink-table-planner/src/test/scala/org/apache/flink/table/planner/runtime/stream/sql/AsyncLookupJoinITCase.scala:
##########
@@ -303,13 +308,15 @@ class AsyncLookupJoinITCase(
}
object AsyncLookupJoinITCase {
- @Parameterized.Parameters(name = "LegacyTableSource={0}, StateBackend={1}, ObjectReuse={2}")
+ @Parameterized.Parameters(
+ name = "LegacyTableSource={0}, StateBackend={1}, ObjectReuse={2}, AsyncOutputMode={3}")
def parameters(): JCollection[Array[Object]] = {
Seq[Array[AnyRef]](
- Array(JBoolean.TRUE, HEAP_BACKEND, JBoolean.TRUE),
- Array(JBoolean.TRUE, ROCKSDB_BACKEND, JBoolean.FALSE),
- Array(JBoolean.FALSE, HEAP_BACKEND, JBoolean.FALSE),
- Array(JBoolean.FALSE, ROCKSDB_BACKEND, JBoolean.TRUE)
+ Array(JBoolean.TRUE, HEAP_BACKEND, JBoolean.TRUE, AsyncOutputMode.ALLOW_UNORDERED),
+ Array(JBoolean.TRUE, ROCKSDB_BACKEND, JBoolean.FALSE, AsyncOutputMode.ORDERED),
+ Array(JBoolean.FALSE, HEAP_BACKEND, JBoolean.FALSE, AsyncOutputMode.ORDERED),
+ Array(JBoolean.FALSE, HEAP_BACKEND, JBoolean.TRUE, AsyncOutputMode.ORDERED),
+ Array(JBoolean.FALSE, ROCKSDB_BACKEND, JBoolean.TRUE, AsyncOutputMode.ALLOW_UNORDERED)
Review Comment:
add case: `ObjectReuse=false` and `AsyncOutputMode=AsyncOutputMode.ALLOW_UNORDERED`
##########
flink-table/flink-table-planner/src/main/java/org/apache/flink/table/planner/plan/nodes/exec/common/CommonExecLookupJoin.java:
##########
@@ -382,10 +391,17 @@ private StreamOperatorFactory<RowData> createAsyncLookupJoin(
asyncBufferCapacity);
}
- // force ORDERED output mode currently, optimize it to UNORDERED
- // when the downstream do not need orderness
return new AsyncWaitOperatorFactory<>(
- asyncFunc, asyncTimeout, asyncBufferCapacity, AsyncDataStream.OutputMode.ORDERED);
+ asyncFunc, asyncTimeout, asyncBufferCapacity, convert(asyncOutputMode));
+ }
+
+ private AsyncDataStream.OutputMode convert(
+ ExecutionConfigOptions.AsyncOutputMode asyncOutputMode) {
+ if (inputInsertOnly
+ && asyncOutputMode == ExecutionConfigOptions.AsyncOutputMode.ALLOW_UNORDERED) {
+ return AsyncDataStream.OutputMode.UNORDERED;
+ }
+ return AsyncDataStream.OutputMode.ORDERED;
Review Comment:
do we have any cases that can cover the case: inputInsertOnly is false and asyncOutputMode == ALLOW_UNORDERED
##########
flink-table/flink-table-planner/src/main/java/org/apache/flink/table/planner/plan/nodes/exec/batch/BatchExecLookupJoin.java:
##########
@@ -46,7 +46,7 @@ public BatchExecLookupJoin(
Map<Integer, LookupJoinUtil.LookupKey> lookupKeys,
@Nullable List<RexNode> projectionOnTemporalTable,
@Nullable RexNode filterOnTemporalTable,
- InputProperty inputProperty,
+ @Nullable InputProperty inputProperty,
Review Comment:
can `inputProperty` be null ?
--
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: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org