feat: Add stringview support to encode and decode and bit_length

[jonathanc-n]

Which issue does this PR close?

Closes #13330 .

Rationale for this change

Add string view to some functions to prevent regression during @timsaucer df python migration.

What changes are included in this PR?

Added support for utf8view to bit_length, array_to_string, and encode

Are these changes tested?

[jonathanc-n]

@timsaucer Do you think testing is necessary here? Can add if so. Also, are you able to show how to replicate the optimizer error for ConcatFunc.

[timsaucer]

Wonderful. Thank you. I'll try to review tomorrow.

I did dive into the ConcatFunc and I found (and resolved) two issues:

1. when doing concat on pure literals we aren't allowing for utf8view or large utf8view with values. This isn't caught in the unit tests because they don't have these variants
2. simplify_concat is changing the merged types to utf8, which is what caused the schema error.

I'll put up the PR today or tomorrow to fix both of these, with unit tests.

[timsaucer]

ConcatFunc: #13346

[jonathanc-n]

Will add tests tomorrow.

[alamb]

Thank you @jonathanc-n -- I marked this PR as draft as you seem to be planning to add more tests. Please mark it as ready for review with it is ready for another look

Thanks a lot for sorting this out

[jonathanc-n]

@timsaucer Will push some tests, however after going through it, date_bin need to wait on this conversion to work:
Unsupported CAST from Timestamp(Nanosecond, None) to Utf8View (a issue can be opened for this in arrow-rs to do this as a double cast is probably not ideal). named_struct also has this problem, if you want the output to be defaulted to be utf8view this has to be solved in arrow for array_struct. Even if i defined the schema in datafusion to output Utf8View or input Utf8View, there will be a schema mismatch or a conversion error, due to the data originally being Utf8 (please correct me if i'm wrong on this).

I changed this pr to only deal with support stringview for array_to_string, encoding, and bit_length

[Omega359]

I updated array_to_string and string_to_array in a separate PR - it's a bit more extensive than what this PR contains.

[Omega359]

@timsaucer Will push some tests, however after going through it, date_bin need to wait on this conversion to work: Unsupported CAST from Timestamp(Nanosecond, None) to Utf8View (a issue can be opened for this in arrow-rs to do this as a double cast is probably not ideal).

Is this an arrow-rs issue or is this #13363 ?

[Omega359]

@timsaucer Will push some tests, however after going through it, date_bin need to wait on this conversion to work: Unsupported CAST from Timestamp(Nanosecond, None) to Utf8View (a issue can be opened for this in arrow-rs to do this as a double cast is probably not ideal).

Is this an arrow-rs issue or is this #13363 ?

Possibly both. I think arrow-cast is missing the _,Utf8View for primitive & temporal casting here: https://github.com/apache/arrow-rs/blob/1d580ec1996e75bb9be4c1c880871d0dd2718ad4/arrow-cast/src/cast/mod.rs#L1465

This PR in arrow-rs I think may fix the issue

[alamb]

Thank you @jonathanc-n @Omega359 and @timsaucer

I updated the title of this PR to reflect what it currently contains and it now looks good to me

Please let me know if I missed anything

[alamb]

I think this means the output type will be Utf8 but for this type i think that is totally fine (there is no need to preserve the view encoding)

[alamb]

Thanks again @jonathanc-n