[FEA] Return null for the row where null exists in starts, or sizes, or steps

[firestarman]

#9839 has implemented the per-list sequence, but it will blow up when null exists in starts, or sizes, or steps.

Asking to return nulls silently instead (or can choose to return null) for this case is to align with the behavior of Spark. Spark will return nulls for rows that contain at least one null. For example,

start	end	step	sequence(start, end, step)
0	5	1	[0, 1, 2, 3, 4, 5]
null	2	2	null
-3	null	1	null
1	3	null	null
null	3	null	null

[jrhemstad]

We specifically chose not to emulate this behavior. This behavior can be emulated in the plugin by stripping nulls in the inputs and then ANDing the output with the original input null mask.

[ttnghia]

Firstly, you do pre-processing in the plugin, computing the AND bitmask to find the null rows, and setting size = 0 for the rows containing null. By doing so, the corresponding output lists from cudf will be empty list. Then, you do post-processing (also in the plugin), setting null for the corresponding output rows.

[firestarman]

Thanks for the info.
Is there any reason not to do this ? I think handling it in plugin is not as easy as in cudf.

[github-actions]

This issue has been labeled inactive-30d due to no recent activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be labeled inactive-90d if there is no activity in the next 60 days.

[github-actions]

This issue has been labeled inactive-90d due to no recent activity in the past 90 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed.

[GregoryKimball]

Closing based on discussion in #9839