[
https://issues.apache.org/jira/browse/FLINK-6094?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16180188#comment-16180188
]
ASF GitHub Bot commented on FLINK-6094:
---------------------------------------
Github user shaoxuan-wang commented on a diff in the pull request:
https://github.com/apache/flink/pull/4471#discussion_r140951770
--- Diff:
flink-libraries/flink-table/src/main/scala/org/apache/flink/table/plan/rules/datastream/DataStreamJoinRule.scala
---
@@ -0,0 +1,93 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.flink.table.plan.rules.datastream
+
+import org.apache.calcite.plan.{RelOptRule, RelOptRuleCall, RelTraitSet}
+import org.apache.calcite.rel.RelNode
+import org.apache.calcite.rel.convert.ConverterRule
+import org.apache.flink.table.api.TableConfig
+import org.apache.flink.table.calcite.FlinkTypeFactory
+import org.apache.flink.table.plan.nodes.FlinkConventions
+import org.apache.flink.table.plan.nodes.datastream.DataStreamJoin
+import org.apache.flink.table.plan.nodes.logical.FlinkLogicalJoin
+import org.apache.flink.table.plan.schema.RowSchema
+import org.apache.flink.table.runtime.join.WindowJoinUtil
+import scala.collection.JavaConverters._
+
+class DataStreamJoinRule
+ extends ConverterRule(
+ classOf[FlinkLogicalJoin],
+ FlinkConventions.LOGICAL,
+ FlinkConventions.DATASTREAM,
+ "DataStreamJoinRule") {
+
+ override def matches(call: RelOptRuleCall): Boolean = {
+ val join: FlinkLogicalJoin = call.rel(0).asInstanceOf[FlinkLogicalJoin]
+ val joinInfo = join.analyzeCondition
+
+ val (windowBounds, remainingPreds) =
WindowJoinUtil.extractWindowBoundsFromPredicate(
+ joinInfo.getRemaining(join.getCluster.getRexBuilder),
+ join.getLeft.getRowType.getFieldCount,
+ join.getRowType,
+ join.getCluster.getRexBuilder,
+ TableConfig.DEFAULT)
+
+ // remaining predicate must not access time attributes
+ val remainingPredsAccessTime = remainingPreds.isDefined &&
+ WindowJoinUtil.accessesTimeAttribute(remainingPreds.get,
join.getRowType)
+
+ // Check that no event-time attributes are in the input.
+ val rowTimeAttrInOutput = join.getRowType.getFieldList.asScala
+ .exists(f => FlinkTypeFactory.isRowtimeIndicatorType(f.getType))
+
+ if (!windowBounds.isDefined && !remainingPredsAccessTime &&
!rowTimeAttrInOutput) {
--- End diff --
@fhueske, we actually agree quite a lot on the concern of infinite size you
have raised. The same problem does not only exist in joining, but also in other
cases, for example GROUPBY, where the grouping-key and associated state can be
unlimited in terms of the size that the state of Flink can not hold them all.
IMO, there is not an easy way to completely eliminate this just through the
validation of query planner/optimizer, so I think it is not a good idea to only
allow the unbounded-joining after a certain operators, like non-windowed
aggregation (in fact, as mentioned above, the grouping-key of aggregation may
also be infinite, so this does not ensure the finite state for joining
operator).
On the other hand, I think the finite state can only be ensured by the
users by giving some hints/controls. We need instruct users to properly set
those control knobs, such that their jobs will not run out of space. One hint
we currently have is state ttl. (I think @hequn8128 has already added this for
this unbounded joining). Maybe here we can add a check on state ttl to force
users set a proper value. What do you think?
> Implement stream-stream proctime non-window inner join
> -------------------------------------------------------
>
> Key: FLINK-6094
> URL: https://issues.apache.org/jira/browse/FLINK-6094
> Project: Flink
> Issue Type: New Feature
> Components: Table API & SQL
> Reporter: Shaoxuan Wang
> Assignee: Hequn Cheng
>
> This includes:
> 1.Implement stream-stream proctime non-window inner join
> 2.Implement the retract process logic for join
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
