As you know, Flink comes with its own type system. Sometimes you want to be very sure about the type of an expression or column. In those cases, the built-in TYPEOF function is your friend.
SELECT TYPEOF(click_id) FROM `examples`.`marketplace`.`clicks` LIMIT 1;returns
STRING NOT NULL
In the context of a support ticket, you might ask a customer to run TYPEOF(event_time) to confirm if the type of a column they refer to as event_time is STRING, TIMESTAMP or TIMESTAMP_LTZ, which often is hard to disambiguate by just exchanging example values.
DESCRIBE or SHOW CREATE TABLE also provide this information for a column, but this assumes that you’ve already created a table or view that contains the expression. So, here is another example, where I use this function to make sure that the return type of an expression meets my expectations:
SELECT
TYPEOF(window_time),
TYPEOF(avg_viewtime)
FROM (SELECT
window_time,
AVG(view_time) AS avg_viewtime
FROM TABLE(TUMBLE(TABLE examples.marketplace.clicks, DESCRIPTOR(`$rowtime`), INTERVAL '1' MINUTE))
GROUP BY window_end, window_start, window_time)
LIMIT 1;returns
EXPR$0 EXPRE$1
TIMESTAMP_LTZ(3) NOT NULL INT NOT NULL
Now, I realize I wanted avg_viewtime to actually be a DOUBLE. So:
SELECT
TYPEOF(window_time),
TYPEOF(avg_viewtime)
FROM (SELECT
window_time,
AVG(CAST(view_time AS DOUBLE)) AS avg_viewtime
FROM TABLE(TUMBLE(TABLE examples.marketplace.clicks, DESCRIPTOR(`$rowtime`), INTERVAL '1' MINUTE))
GROUP BY window_end, window_start, window_time)
LIMIT 1;returns
EXPR$0 EXPRE$1
TIMESTAMP_LTZ(3) NOT NULL DOBULE NOT NULL
As always all the examples are runnables as is on Confluent Cloud.