[
https://issues.apache.org/jira/browse/FLINK-2998?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15241491#comment-15241491
]
ASF GitHub Bot commented on FLINK-2998:
---------------------------------------
Github user fhueske commented on a diff in the pull request:
https://github.com/apache/flink/pull/1838#discussion_r59752067
--- Diff:
flink-tests/src/test/java/org/apache/flink/test/javaApiOperators/CustomDistributionITCase.java
---
@@ -175,6 +175,77 @@ else if (pIdx > 0 && pIdx < dist.getParallelism() - 1)
{
env.execute();
}
+ @Test
+ public void testPartitionKeyLessDistribution() throws Exception{
+ /*
+ * Test the number of keys less than the number of distribution
fields
+ */
+
+ ExecutionEnvironment env =
ExecutionEnvironment.createLocalEnvironment();
+
+ DataSet<Tuple3<Integer, Long, String>> input1 =
CollectionDataSets.get3TupleDataSet(env);
+ final TestDataDist2 dist = new TestDataDist2();
+
+ env.setParallelism(dist.getParallelism());
+
+ DataSet<Boolean> result = DataSetUtils
+ .partitionByRange(input1, dist, 0)
+ .mapPartition(new
RichMapPartitionFunction<Tuple3<Integer, Long, String>, Boolean>() {
+
+ @Override
+ public void
mapPartition(Iterable<Tuple3<Integer, Long, String>> values, Collector<Boolean>
out) throws Exception {
+ int
pIdx = getRuntimeContext().getIndexOfThisSubtask();
+
+ for
(Tuple3<Integer, Long, String> s : values) {
+
boolean correctlyPartitioned = true;
+
if (pIdx == 0) {
+
Integer[] upper = dist.boundaries[0];
+
if (s.f0.compareTo(upper[0]) > 0) {
+
correctlyPartitioned = false;
+
}
+
} else if (pIdx > 0 && pIdx < dist.getParallelism() - 1) {
+
Integer[] lower = dist.boundaries[pIdx - 1];
+
Integer[] upper = dist.boundaries[pIdx];
+
if (s.f0.compareTo(upper[0]) > 0 || (s.f0.compareTo(lower[0]) <= 0)) {
+
correctlyPartitioned = false;
+
}
+
} else {
+
Integer[] lower = dist.boundaries[pIdx - 1];
+
if ((s.f0.compareTo(lower[0]) <= 0)) {
+
correctlyPartitioned = false;
+
}
+
}
+
+
if (!correctlyPartitioned) {
+
fail("Record was not correctly partitioned: " + s.toString());
+
}
+ }
+ }
+ }
+ );
+
+ result.output(new DiscardingOutputFormat<Boolean>());
+ env.execute();
+ }
+
+ @Test(expected = IllegalArgumentException.class)
+ public void testPartitionMoreThanDistribution() throws Exception{
+ /*
+ * Test the number of keys larger than the number of
distribution fields
+ */
+
+ ExecutionEnvironment env =
ExecutionEnvironment.createLocalEnvironment();
+
+ DataSet<Tuple3<Integer, Long, String>> input1 =
CollectionDataSets.get3TupleDataSet(env);
+ final TestDataDist2 dist = new TestDataDist2();
+
+ DataSet<Tuple3<Integer, Long, String>> result = DataSetUtils
+ .partitionByRange(input1, dist, 0, 1, 2);
+
+ result.output(new DiscardingOutputFormat<Tuple3<Integer, Long,
String>>());
--- End diff --
I think you can remove this code. The test should fail at
`partitionByRange(...)`, right?
> Support range partition comparison for multi input nodes.
> ---------------------------------------------------------
>
> Key: FLINK-2998
> URL: https://issues.apache.org/jira/browse/FLINK-2998
> Project: Flink
> Issue Type: New Feature
> Components: Optimizer
> Reporter: Chengxiang Li
> Priority: Minor
>
> The optimizer may have potential opportunity to optimize the DAG while it
> found two input range partition are equivalent, we does not support the
> comparison yet.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
