dianfu commented on a change in pull request #7235: [FLINK-10976] [table] Add
support for aggregate to table API
URL: https://github.com/apache/flink/pull/7235#discussion_r245566320
##########
File path:
flink-libraries/flink-table/src/main/scala/org/apache/flink/table/api/table.scala
##########
@@ -1226,3 +1301,81 @@ class WindowGroupedTable(
select(withResolvedAggFunctionCall: _*)
}
}
+
+class AggregatedTable(
+ private[flink] val table: Table,
+ private[flink] val groupKeys: Seq[Expression],
+ private[flink] val aggregateFunction: Expression) {
+
+ /**
+ * Performs a selection operation after an aggregate operation. The field
expressions
+ * cannot contain table functions and aggregations.
+ *
+ * Example:
+ *
+ * {{{
+ * val aggFunc: AggregateFunction[_, _] = new MyAggregateFunction
+ * tableEnv.registerFunction("aggFunc", aggFunc);
+ * table.groupBy('key).aggregate("aggFunc(a, b) as (f0, f1,
f2)").select('key, 'f0, 'f1)
+ * }}}
+ */
+ def select(fields: Expression*): Table = {
+ val tableEnv = table.tableEnv
+ val (aggNames, propNames) =
extractAggregationsAndProperties(Seq(aggregateFunction), tableEnv)
+ val projectsOnAgg = replaceAggregationsAndProperties(
+ groupKeys ++ Seq(aggregateFunction), tableEnv, aggNames, propNames)
+ val projectFields = extractFieldReferences(groupKeys ++
Seq(aggregateFunction))
+
+ val aggTable = new Table(tableEnv,
+ Project(projectsOnAgg,
+ Aggregate(groupKeys, aggNames.map(a => Alias(a._1, a._2)).toSeq,
+ Project(projectFields, table.logicalPlan).validate(tableEnv)
+ ).validate(tableEnv)
+ ).validate(tableEnv))
+
+ // expand the aggregate results
+ val projectsOnAggTable =
+ aggTable.logicalPlan.output.take(groupKeys.length) ++
+ expandProjectList(
+ Seq(Flattening(aggTable.logicalPlan.output.last)),
+ aggTable.logicalPlan,
+ tableEnv).zip(extractFieldNames(aggregateFunction)).map(a =>
Alias(a._1, a._2))
+
+ val flattenedAggTable = new Table(
+ tableEnv,
+ Project(
+ projectsOnAggTable.map(UnresolvedAlias),
+ aggTable.logicalPlan).validate(tableEnv))
+
+ val expandedFields = expandProjectList(fields,
flattenedAggTable.logicalPlan, tableEnv)
+ // check there are no aggregate functions in the select after aggregate
+ expandedFields.foreach { f =>
+ unwrap(f, tableEnv) match {
+ case _: TableFunctionCall | _: Aggregation =>
Review comment:
Yes, it will report the following exception if user use TableFunction in the
select:
java.lang.UnsupportedOperationException:
org.apache.flink.table.expressions.TableFunctionCall cannot be transformed to
RexNode.
Personally, I think this exception is not obvious for users as users may
think that this exception may be caused by an implementation bug in Flink and
they may not realize that TableFunction is not allowed to be used here. Adding
a more user friendly exception msg will eliminate this kind of confuse. Does it
make sense to you?
----------------------------------------------------------------
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
