Optimize `like`/`ilike` kernels for StringView

[alamb]

Is your feature request related to a problem or challenge? Please describe what you are trying to do.

@XiangpengHao added support for StringView to the like/ilike kernels in #5931

#5931 (review)

This PR does not leverage the special 4 bytes inlined prefix for large string views, which might be able to optimize certain cases (specifically for quickly testing if the starts_with variant of like doesn't match without having to consult the actual strings, as @wjones127 discusses in #5931 (review)

does not leverage the inlined 4 bytes to shortcut some comparison. I plan to leave this as a future work as it can significantly complicate the code. (I also doubt that we can gain any benefit)

I agree we should only start adding specialized paths if we determine it is worth it. starts_with might be, especially if the parameter is short enough to always be able to use the prefix. We can decide based on benchmarks which are appropriate.

I added a benchmark in #5931

Describe the solution you'd like
Investigate various ways to make the benchmark faster. Run the benchmark like this:

cargo bench --bench comparison_kernels -- utf8view

Describe alternatives you've considered
Maybe it is fast enough as is

Additional context

[xinlifoobar]

take

[xinlifoobar]

I read through the code and have some initial ideas. There are serval sub-optimizations:

Good to go:

• Startwith/IStartwith when the needle has length < 4

Side effects (might be, this could be removed completed by specialized string view functions):

• Contains: when the needle has length < 4 and the first 4 bytes match.
• Regex: when the regex has match expected < 4 and the first 4 bytes match.

I would start with startwith and istartwith and needs some inputs for the latter 2. Please let me know your thoughts on generating better implementations :).

CC @alamb @XiangpengHao

[XiangpengHao]

Sounds good to me! Here's the current comparsion function that has similar handling for 4/12 bytes:

https://github.com/apache/arrow-rs/blob/master/arrow-array/src/array/byte_view_array.rs#L340-L399

[alamb]

It seems to me like this ticket was completed with #6231

@xinlifoobar has some ideas for improving other kernels (like regexp) here #5951 (comment)

I did file #6370 for regexp_is_match_utf8

Let me know if there are other kernels that we should update