[
https://issues.apache.org/jira/browse/FLINK-24176?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17414990#comment-17414990
]
Marios Trivizas edited comment on FLINK-24176 at 9/14/21, 3:06 PM:
-------------------------------------------------------------------
* This is issue is still valid for current flink *master* branch.
* It's traced down to the calcite optimizer which doesn't keep the alias
* Seems that the behaviour is a bit flaky, for example the following query
ends up with the correct fields:
{code:java}
SELECT max(a) as max_a, c as alias_c , sum(a) as sum_a FROM my_table group by
c, HOP(ts, INTERVAL '1' SECOND)
{code}
gives the following columns:
{code:java}
max_a, alias_c, sum_a{code}
* The issue doesn't occur if the newer *toDataStream()* and
*toChangelogStream()* are used instead of the deprecated *toAppendStream()*
** This is because the new methods apply a final projection which fixes any
changes and lost aliases that happen during the optimisation.
was (Author: matriv):
* This is issue is still valid for current flink *master* branch.
* It's traced down to the calcite optimizer which doesn't keep the alias
* Seems that the behaviour is a bit flaky, for example the following query
ends up with the correct fields:
{code:java}
SELECT max(a) as max_a, c as alias_c , sum(a) as sum_a FROM my_table group by
c, HOP(ts, INTERVAL '1' SECOND)
{code}
**
gives the following columns:
{code:java}
max_a, alias_c, sum_a{code}
* The issue doesn't occur if the newer *toDataStream()* and
*toChangelogStream()* are used instead of the deprecated *toAppendStream()*
** This is because the new methods apply a final projection which fixes any
changes and lost aliases that happen during the optimisation.
> Column alias in Flink-SQL is unsuccessful after converting Table to
> DataStream<Row>
> -------------------------------------------------------------------------------------
>
> Key: FLINK-24176
> URL: https://issues.apache.org/jira/browse/FLINK-24176
> Project: Flink
> Issue Type: Bug
> Components: API / DataStream, Table SQL / API
> Affects Versions: 1.12.0
> Environment: java version "1.8.0_291"
> Flink version : 1.12-csadh1.3.0.0
> Flink scala version: 2.12
> Reporter: Wang Guanglei
> Priority: Critical
>
> Hi team,
> I found an interesting issue while using Flink-SQL. If the group window
> function and aggregating operation are used in one SQL statement at the same
> time, some column aliases will be unsuccessful after converting the table to
> DataStream<Row>.
> Following is a part of the code :
> {code:java}
> //代码占位符
> tableEnv.createTemporaryView("my_table", in);
> String sqlQuery = "SELECT c as alias_c, sum(a) as sum_a FROM my_table group
> by c, TUMBLE(ts, INTERVAL '1' SECOND)";
> Table result = tableEnv.sqlQuery(sqlQuery);
> System.out.println("Table schema: " + result.getSchema());
> DataStream<Row> resultDataStream = tableEnv.toAppendStream(result, Row.class);
> final RowTypeInfo type1 = (RowTypeInfo)resultDataStream.getType();
> System.out.println("Field names: " + Arrays.toString(type1.getFieldNames()));
> {code}
> Output:
> Table schema: root
> |-- *alias_c*: STRING
> |-- sum_a: INT
> Field names: [*c*, sum_a]
>
>
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
