Wrap custom iterator result

[galipremsagar]

Description

Fixes: #17165
Fixes: #14481

This PR properly wraps the result of custom iterator.

In [2]: import pandas as pd

In [3]: s = pd.Series([10, 1, 2, 3, 4, 5]*1000000)


# Without custom_iter:

In [4]: %timeit for i in s: True
6.34 s ± 25.8 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

# This PR:

In [4]: %timeit for i in s: True
6.16 s ± 17.4 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)


# On `branch-24.12`:
1.53 s ± 6.27 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

I think custom_iter has to exist. Here is why, invoking any sort of iteration on GPU objects will raise errors and thus in the end we fall-back to CPU. Instead of trying to move the objects from host to device memory (if the object is on host memory only), we will avoid a CPU-to-GPU transfer.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[vyasr]

@galipremsagar could you add a test that shows the undesirable fallback behavior of iter? You should be able to do this by monkey-patching os.environ with CUDF_PANDAS_FAIL_ON_FALLBACK and showing that fallback is triggered undesirably. That ought to be possible without a subprocess, but you might have to define a new custom proxy type without custom_iter to show it.

If that turns out to be too much work we can just document it, but I want to make it as obvious as possible to future devs why this is necessary.

[galipremsagar]

@galipremsagar could you add a test that shows the undesirable fallback behavior of iter? You should be able to do this by monkey-patching os.environ with CUDF_PANDAS_FAIL_ON_FALLBACK and showing that fallback is triggered undesirably. That ought to be possible without a subprocess, but you might have to define a new custom proxy type without custom_iter to show it.

If that turns out to be too much work we can just document it, but I want to make it as obvious as possible to future devs why this is necessary.

I agree, I'll add a test and document the reason behind the custom iterator existence.

[vyasr]

This PR also closes #14481 right?

[galipremsagar]

I agree, I'll add a test and document the reason behind the custom iterator existence.

Done.

This PR also closes #14481 right?

Yes, this also closes #14481

[galipremsagar]

@vyasr @Matt711 This is now ready for review.

[Matt711]

A question and tiny suggestions

[galipremsagar]

Going ahead and merging this PR. @vyasr incase you have additional comments let me know I can address in a follow-up.

[galipremsagar]

/merge

[vyasr]

No worries, this looks great thanks!