[SPARK-52288][PS] Avoid INVALID_ARRAY_INDEX in split/rsplit when ANSI mode is on

[xinrong-meng]

What changes were proposed in this pull request?

Avoid INVALID_ARRAY_INDEX in split/rsplit when ANSI mode is on

Why are the changes needed?

Ensure pandas on Spark works well with ANSI mode on.
Part of https://issues.apache.org/jira/browse/SPARK-52169.

Does this PR introduce any user-facing change?

Yes. INVALID_ARRAY_INDEX no longer fails split/rsplit when ANSI mode is on

>>> spark.conf.get("spark.sql.ansi.enabled")
'true'
>>> import pandas as pd
>>> pser = pd.Series(["hello-world", "short"])
>>> psser = ps.from_pandas(pser)

FROM

>>> psser.str.split("-", n=1, expand=True)
25/05/28 14:52:10 ERROR Executor: Exception in task 10.0 in stage 2.0 (TID 15)  
org.apache.spark.SparkArrayIndexOutOfBoundsException: [INVALID_ARRAY_INDEX] The index 1 is out of bounds. The array has 1 elements. Use the SQL function `get()` to tolerate accessing element at invalid index and return NULL instead. SQLSTATE: 22003
== DataFrame ==
"__getitem__" was called from
<stdin>:1
...

TO

>>> psser.str.split("-", n=1, expand=True)
       0      1                                                                 
0  hello  world
1  short   None

How was this patch tested?

Unit tests

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

No

[abhiips07]

can we try to calculate f.lit(i+1) outside loop? since this might avoid function calls and object creation during loop execution

literals = [F.lit(i + 1) for i in range(n + 1)]

spark_columns = [F.try_element_at(scol, lit).alias(str(i)) for i, lit in enumerate(literals)]

[xinrong-meng]

Thanks for suggestion!
There might not be a significant perf difference between creating F.lit inside the loop or beforehand, it's just wrapping a Python literal into a Spark expression, which aren’t executed immediately(just nodes in the DAG), and will be deduplicated by Catalyst. With that being said I’d like to keep the original for simplicity, but feel free to share if you have other opinions!

[zhengruifeng]

It seems we can always use the new branch, and remove the configuration read here (which needs one Config RPC)

[xinrong-meng]

What did you mean by "using the new branch"?

[xinrong-meng]

Pending config check introduced in #50972