[C++] Implement cast to Binary and FixedSizeBinary

[asfimport]

It appears you can cast from Binary to String but not the other way.

Reporter: Neal Richardson / @nealrichardson
Assignee: Krisztian Szucs / @kszucs

Related issues:

• [C++] C++ array kernels framework and execution buildout (umbrella issue) (is a child of)
• [R] Conversion from arrow::BinaryArray to R character vector not implemented (is depended upon by)

PRs and other links:

• GitHub Pull Request #7470
• GitHub Pull Request #7529

_{Note: This issue was originally created as ARROW-8025. Please see the migration documentation for further details.}

[asfimport]

Wes McKinney / @wesm:
You should be able to use ZeroCopyCastExec https://github.com/apache/arrow/blob/master/cpp/src/arrow/compute/kernels/scalar_cast_internal.h#L71

[asfimport]

Krisztian Szucs / @kszucs:
@wesm Cast to and from FixedSizedBinary will involve some more work and I have a couple of questions:

String/Binary to FixedSizedBinary:

• shall we append values one by one and truncate if necessary

• or add as zero copy cast and check that the value offsets indicate same width values

• or both?

  FixedSizedBinary to String/Binary is more straightforward to zero copy cast and generate the offset values, although I'm not how much sense this direction makes.

[asfimport]

Wes McKinney / @wesm:
I think you can split the FSB work into a separate PR/JIRA.

I would say that when casting from String/Binary to FSB that if a value does not match the byte width then an error should be raised. One issue is that if there are nulls and the nulls are length-0 then zero copy is not possible

If something more nuanced is required later we can pursue it then, this seems like enough for now

Going the other direction as you say is easier

[asfimport]

Wes McKinney / @wesm:
Issue resolved by pull request 7470
#7470

[asfimport]

Daniel Nugent / @nugend:
Was there ever a similar issue for the FixedSizeBinary to String cast? It’s nice to have when you want to ensure that single byte records are interpreted as characters.