Jun Qin created FLINK-28650:
-------------------------------
Summary: Flink SQL Parsing bug for METADATA
Key: FLINK-28650
URL: https://issues.apache.org/jira/browse/FLINK-28650
Project: Flink
Issue Type: Bug
Affects Versions: 1.14.4
Reporter: Jun Qin
With the following source/sink tables:
{code:sql}
CREATE TABLE sourceTable (
`key` INT,
`time` TIMESTAMP(3),
`value` STRING NOT NULL,
id INT
)
WITH (
'connector' = 'datagen',
'rows-per-second'='10',
'fields.id.kind'='sequence',
'fields.id.start'='1',
'fields.id.end'='100'
);
CREATE TABLE sinkTable1 (
`time` TIMESTAMP(3) METADATA FROM 'timestamp',
`value` STRING NOT NULL
)
WITH (
'connector' = 'kafka',
...
)
CREATE TABLE sinkTable2 (
`time` TIMESTAMP(3), -- without METADATA
`value` STRING NOT NULL
)
WITH (
'connector' = 'kafka',
...
)
{code}
the following three pass the validation:
{code:sql}
INSERT INTO sinkTable1
SELECT
`time`,
`value`
FROM sourceTable;
INSERT INTO sinkTable2
SELECT
`time`,
`value`
FROM sourceTable;
INSERT INTO sinkTable2 (`time`,`value`)
SELECT
`time`,
`value`
FROM sourceTable;
{code}
but this one does not:
{code:sql}
INSERT INTO sinkTable1 (`time`,`value`)
SELECT
`time`,
`value`
FROM sourceTable;
{code}
It failed with
{code:java}
Unknown target column 'time'
{code}
It seems when providing column names in INSERT, the METADATA have an undesired
effect.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
