Jark Wu created FLINK-7448:
------------------------------
Summary: Keep the data type unchanged when register an existing
field as rowtime
Key: FLINK-7448
URL: https://issues.apache.org/jira/browse/FLINK-7448
Project: Flink
Issue Type: Improvement
Reporter: Jark Wu
Currently, when we convert a DataStream to Table, we can register an existing
field as rowtime field. But the rowtime field will become
{{TimeIndicator(Timestamp)}} type, not {{TimeIndicator(Long)}} type. It makes
user confused that the result becomes Timestamp type not long when select the
rowtime field.
{code}
// stream schema: [Long, Int, Double, Float], take the first field as rowtime
val table = stream.toTable(tEnv, 'rowtime.rowtime, 'int, 'double, 'float)
table.printSchema()
// - TimeIndicator(Timestamp), Int, Double, Float
table.select('rowtime)
// "1970-01-01 00:00:00.007"
{code}
Such as in the test {{TimeAttributesITCase#testCalcMaterialization2}}, the
original rowtime field is long type, but the selected result becomes timestamp
type. Rowtime attribute shouldn't change
the type, otherwise, the result of batch and stream is not the same.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
