[FEA] Enable .str.find_multiple public API

[galipremsagar]

Is your feature request related to a problem? Please describe.
We currently have cudf._lib.strings.find_multiple which is an internal API. We would like to expose it via a public API similar to what we wanted to do in: #4575, this was previously dropped due to lack of ListColumn support: #4569 (comment), which shouldn't be a blocker now.

[bdice]

Additional context (@galipremsagar beat me to filing an issue! 😉):

The gpu-bdb query 18 code uses this private API, cudf._lib.strings.find_multiple, to search a column of string values for a list of target values:
https://github.com/rapidsai/gpu-bdb/blob/f67ea68d4b422c998d0284cc69fc82ed371c69cc/gpu_bdb/queries/q18/gpu_bdb_query_18.py#L106-L126

Describe the solution you'd like
Query 18 should be able to call a public API Series.str.find_multiple that behaves like Series.str.find, rather than a private API. The implementation in libcudf can be refactored to return a list column of integers (previously proposed by @davidwendt here: #6341 (comment)).

[davidwendt]

Technically you can build the lists column in the Python layer by just creating an offsets column and using it along with column returned by the find_multiple API.

s = cudf.Series([ "abc", "bcd", "c", "de" ])
t = cudf.Series(["c", "d"])
result = s.str.find_multiple(t)
# result is now [ 2, -1,   1, 2   0, -1   -1, 0 ]

offsets = [ 0, 2, 4, 6, 8 ]  # just an even sequence  [0 : len(s)*len(t) : len(t)]
# likely there is a numpy way to create this array

Build a ListColumn by using offsets and results as its child columns.

[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.

[bdice]

This API is also a candidate for a separate NLP-focused submodule (which has been discussed among cudf developers a few times but I am not aware of an issue tracking this idea). The pandas API does not have an equivalent of Series.str.find_multiple.

[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]

Closed by #10134

[bdice]

@GregoryKimball Re-opening this one because it’s only partially complete. The proposed Python API has not yet been added, only the C++ changes have been made. See: #10134 (review)