Aljoscha Krettek created FLINK-1156:
---------------------------------------
Summary: Bug in Constant Fields Annotation (Semantic Properties)
Key: FLINK-1156
URL: https://issues.apache.org/jira/browse/FLINK-1156
Project: Flink
Issue Type: Bug
Reporter: Aljoscha Krettek
When I change the first test in SemanticPropUtilTest.java to this:
{code}
@Test
public void testConstantWithArrowIndividualStrings() {
String[] constantFields = { "0->0,1", "1->3" };
TypeInformation<?> type = new TupleTypeInfo<Tuple3<Integer, Integer,
Integer>>(BasicTypeInfo.INT_TYPE_INFO,
BasicTypeInfo.INT_TYPE_INFO,
BasicTypeInfo.INT_TYPE_INFO);
TypeInformation<?> outType = new TupleTypeInfo<Tuple4<Integer, Integer,
Integer, Integer>>(BasicTypeInfo.INT_TYPE_INFO,
BasicTypeInfo.INT_TYPE_INFO,
BasicTypeInfo.INT_TYPE_INFO, BasicTypeInfo.INT_TYPE_INFO);
SingleInputSemanticProperties sp =
SemanticPropUtil.getSemanticPropsSingleFromString(constantFields, null, null,
type, outType);
FieldSet fs = sp.getForwardedField(0);
Assert.assertTrue(fs.size() == 2);
Assert.assertTrue(fs.contains(0));
Assert.assertTrue(fs.contains(1));
fs = sp.getForwardedField(1);
Assert.assertTrue(fs.size() == 1);
Assert.assertTrue(fs.contains(2));
}
{code}
It fails. It seems the check whether a tuple field is in range is performed on
the input type and the output type is being ignored.
Also, I'm not sure this whole thing works anymore with [~rmetzger]'s recent
POJO rework and nested tuple field selection rework.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
