Kafka _timestamp queries show different counts

[metalshanked]

Expected behavior

Example:- the below query returns a different count when run from the trino python client (v0.321) and opposed to via Java apps like Datagrip etc

The query works fine when i revert trino to 0.320 SQLAlchemy v1.4.46

SELECT count(_message) as result
FROM mykafka."default"."kafka-topic" 
WHERE   "_timestamp" >= TIMESTAMP '2023-02-22 19:49:54 UTC'
AND   "_timestamp" <= TIMESTAMP '2023-02-22 19:50:58.001000 UTC'
LIMIT 1

df_kafka = pd.read_sql(sql_kafka, engine)


Actual behavior

Counts should remain same

Steps To Reproduce

try the above query via trino + sqlalchemy and via a JAVA IDE to compare

Log output

No response

Operating System

Windows

Trino Python client version

0.321.

Trino Server version

407

Python version

3.11

Are you willing to submit PR?

• [ ] Yes I am willing to submit a PR!

[mdesmet]

This query doesn't contain any Python typed values (aka prepared statements). Can you validate in your Trino cluster if the executed query is the same. In this case the trino-python-client doesn't do anyting more than sending the exact query to the server.

[hashhar]

@metalshanked closing this as I cannot reproduce. Feel free to reopen if you still have the issue.

The only relevant thing I can think of is whether the client you're using has a different session timezone. You can see it in the Trino UI on the query details page.

[metalshanked]

Apologies @hashhar for the delay in response. I see UTC in the timezone. Will check further. Thanks!