[SPARK-56976][SQL] Support Two-Argument IF(condition, value) Returning NULL

[Richard-Durkee]

What changes were proposed in this pull request?

Add an auxiliary constructor to the If expression that accepts two arguments, defaulting falseValue to Literal(null). This allows IF(condition, value) to return NULL when the condition is false.

Why are the changes needed?

The two-argument IF expression is supported by Trino and Presto. Users migrating queries from these systems currently get an arity error.

Does this PR introduce any user-facing change?

Yes. IF(condition, value) now returns value when true and NULL when false, instead of throwing an error.

How was this patch tested?

New golden file SQL test (if-two-arg.sql). Existing conditional-functions.sql and ifCoercion.sql tests pass with no regressions.

Was this patch authored or co-authored using generative AI tooling?

No

[nnguyen168]

Hi @Richard-Durkee, thanks for adding this compatibility feature! The change is clean and well-tested.

I noticed this aligns with MySQL/Trino/BigQuery behavior. A couple of thoughts:

1. The auxiliary constructor approach is elegant — keeps backward compatibility.
2. Golden file tests are comprehensive.

Minor: Have you considered adding a note to the SQL function documentation about this two-argument form?

Let me know if I can help with anything!

[Richard-Durkee]

Hi @Richard-Durkee, thanks for adding this compatibility feature! The change is clean and well-tested.

I noticed this aligns with MySQL/Trino/BigQuery behavior. A couple of thoughts:

1. The auxiliary constructor approach is elegant — keeps backward compatibility.
2. Golden file tests are comprehensive.

Minor: Have you considered adding a note to the SQL function documentation about this two-argument form?

Let me know if I can help with anything!

Thanks! The @ExpressionDescription annotation was updated with the two-argument form, my understanding is that this auto-generates the SQL function docs. If this feature would be useful to you, consider leaving a comment on the Jira issue.