Trino split_statements is default to False

[willshen99]

Apache Airflow Provider(s)

trino

Versions of Apache Airflow Providers

6.0.1

Apache Airflow version

2.9.0

Operating System

Red Hat Enterprise Linux 9.5

Deployment

Virtualenv installation

Deployment details

No response

What happened

I'm following this document to use SQLExecuteQueryOperator to issue multiple queries separated with semicolon to Trino.
I got a syntax error complaining mismatched input ';'. I have to add an additional split_statements=True parameter to get it work.

What you think should happen instead

SQLExecuteQueryOperator defers split_statements to the default value in the run method of the configured hook. Such is not defined in TrinoHook, so it uses default False from DbApiHook and throws the error.

How to reproduce

SQLExecuteQueryOperator(
    task_id='t1',
    sql='SELECT 1; SELECT 2;',
    conn_id='loading_user_trino'     # Adding split_statements=True will work
)

Anything else

No response

Are you willing to submit PR?

• Yes I am willing to submit a PR!

Code of Conduct

• I agree to follow this project's Code of Conduct

[boring-cyborg]

Thanks for opening your first issue here! Be sure to follow the issue template! If you are willing to raise PR to address this issue please do so, no need to wait for approval.

[potiuk]

Feel free to add PR

[HowlyBlood]

Hi, I did not notice I left this message on the PR and not the issue, but I think it's here the best place to discuss about the issue. Sorry.

Hi, this is my first "contribution", even if it is just a suggestion. I'm clearly new to this, but I was wondering if some users has some statement like

CREATE TABLE example (id INT); -- added ; in a comment. If the split is not "comment zone safe", it may raise an error
INSERT INTO example (id) VALUES (1);
SELECT * FROM
or "BEGIN; ... COMMIT;" which makes no sense if executed in multiple statements .
As split_statement was initially false, it was behaving correctly, but now it may not behave as expected

I'm not sure if it's a good solution, but providing split_statements = hasattr(sql,'iter') would prevent any unexpected behaviour: if it's not a iterable, it would not split. No changes from before the update.

I'm still looking for the split_statement's definition, but I think it is splitting str with semicolon and iterables by their elements, so it might be an intuitive way to code : you use an iterable if you want it to be split, otherwise, you use a single string (for a single statement)

["DESCRIBE ... ;","SELECT ...;","BEGIN; ... ; COMMIT;"] --> split_statement = True
"BEGIN; ... ; COMMIT;" --> split_statement = False

I'm not a native English speaker, so tell me if i'm not easy to understand, and i'll try to explain my point of view.

Good job anyway, you pointed out something interesting.

Thanks to this issue, I learnt some new things like @overload, which I had already encountered in Java but never explored. Java wasn't in my career path, but seeing it in Python made me look for explanations. I'm still learning, and this was a great opportunity to expand my beginner knowledge.

See you !