Add cudf::strings::find_re API

[davidwendt]

Description

Adds the cudf::strings::find_re and str.find_re API to libcudf/pylibcudf/cudf. This function returns the character position where the pattern first matches in each row of the input column. If a match is not found, -1 is returned for that corresponding row.

Closes #16729

Checklist

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

[Matt711]

Could you add a pylibcudf test for find_re?

[ttnghia]

How about stream test for the new API?

[Matt711]

Thanks for adding a test. I'll wait to approve until we decide whether to seperate find_re and find_all or combine them under a common file name as per this comment.

[Matt711]

Approving python changes

[bdice]

One question about the pylibcudf API / code location. Otherwise LGTM.

[bdice]

Should this go in a new file, or should this file be renamed? It seems incorrect to have find_re defined in the findall file, since they're different algorithms.

[davidwendt]

I was told to keep match these with the header files. So find_re and findall are both declared in findall.hpp. They are common only in that both use regex I suppose.

[bdice]

I didn't realize find_re was in findall.hpp. Then yes, this is fine.

[bdice]

(But it does raise the question of why find_re is in findall.hpp -- it doesn't "find all")

[Matt711]

IIUC they will be combined into a common file as per https://github.com/rapidsai/cudf/pull/16742/files#r1781466464

[davidwendt]

The find_re has more in common with findall than just find based on the parameters and how they are used just in not what they return.

Perhaps find_re() would be better in contains.hpp which has contains_re()
Since find.hpp has contains() and find() (among other similar functions)
So contains.hpp would have contains_re() and find_re().
I could do that in a separate PR since this one is already approved.

[davidwendt]

/merge