[SPARK-41977][SPARK-41978][CONNECT] SparkSession.range to take float as arguments

[HyukjinKwon]

What changes were proposed in this pull request?

This PR proposes to Spark Connect's SparkSession.range to accept floats. e.g., spark.range(10e10).

Why are the changes needed?

For feature parity.

Does this PR introduce any user-facing change?

No to end users since Spark Connect has not been released yet.
SparkSession.range allows floats.

How was this patch tested?

Unittests enabled back.

[HyukjinKwon]

Merged to master.

[dongjoon-hyun]

May I ask a question?

• The function range's signature is not changed. Does this introduce any behavior change?

[zhengruifeng]

@dongjoon-hyun

good question. range in Connect has the same signature as the PySpark's one, which should only accept intergers.

But PySpark's implementation doesn't check the input type

spark/python/pyspark/sql/session.py

Lines 788 to 796 in 56c7cf3

	if numPartitions is None:
	numPartitions = self._sc.defaultParallelism
	if end is None:
	jdf = self._jsparkSession.range(0, int(start), int(step), int(numPartitions))
	else:
	jdf = self._jsparkSession.range(int(start), int(end), int(step), int(numPartitions))
	return DataFrame(jdf, self)

and in some cases / tests, the floats are used, like range(10e10).

We might need to update the typing for both Connect and PySpark in the future.

[HyukjinKwon]

Yeah .. so technically it should only take ints as that's what the method wants. There are a lot of cases like that in PySpark (e.g., DataFrameReader.jdbc), and a lot of discussions happened.

The take is: type desired signature, and do not type the signature supported by casting (and keep it as a legacy behaviour).
Here I fixed Spark Connect's range to match with the legacy behaviour of the regular PySpark's range.