Support string concat || for StringViewArray

[alamb]

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

datafusion-cli

Then run

DataFusion CLI v40.0.0
> create table foo as values ('value1', arrow_cast('one', 'Utf8View')), ('value1', arrow_cast('two', 'Utf8View'));
0 row(s) fetched.
> select column2||'ff' from foo;
Internal error: Data type Utf8View not supported for binary operation 'concat_elements' on string arrays.
This was likely caused by a bug in DataFusion's code and we would welcome that you file an bug report in our issue tracker

Describe the solution you'd like
I would like the query to work

Describe alternatives you've considered
We can probably get the query to work by fixing the coercion logic to coerce to Utf8, but that will be less efficient than implementing native StringView support for concat

The coercion is here:

datafusion/datafusion/expr/src/type_coercion/binary.rs

Lines 116 to 122 in 6e2ff29

	StringConcat => {
	string_concat_coercion(lhs, rhs).map(Signature::uniform).ok_or_else(|| {
	plan_datafusion_err!(
	"Cannot infer common string type for string concat operation {lhs} {op} {rhs}"
	)
	})
	}

The physical expression implementation starts here

datafusion/datafusion/physical-expr/src/expressions/binary.rs

Line 661 in 6e2ff29

StringConcat => binary_string_array_op!(left, right, concat_elements),

Additional context

[alamb]

If anyone wants to work on this, I would suggest starting with a PR that has some tests and does coercsion (which will make the query run) and then we can implement optimziation later

[dharanad]

i'm really interested in working on this issue. Feeling little doubtful, but let me try. I'm might seek some help, if not i will unassign myself .

[alamb]

Thanks @dharanad --- I would recommend starting with a test in string_view.slt that shows the problem

Then I would recommend trying to change the coercion logic so the arguments are coerced to Utf8 (not Utf8view) to get the query passing

Then in a second PR (or maybe even ticket) we can figure out how to implement the native opreration of concat (that is likely much tricker than coercing)

[dharanad]

take

[dharanad]

Linked issue : #11881

@alamb I was thinking the right way to solve the issue would to add support for StringViewArray in arrow::compute::kernels::concat_elements by into introducing a function named concat_elements_utf8view.

Do you have any alternative approaches in mind? If so, any pointers would be greatly appreciated.

[alamb]

@alamb I was thinking the right way to solve the issue would to add support for StringViewArray in arrow::compute::kernels::concat_elements by into introducing a function named concat_elements_utf8view.

Yes, I agree this would be ideal. I recommend calling it concate_elements_bytes_view to be consistent with

https://docs.rs/arrow/latest/arrow/compute/kernels/concat_elements/fn.concat_elements_bytes.html

I started looking into how to explain wiring this into binary.rs and it was a mess, so I just made a skeleton PR here:
#11995

Perhaps you can use that a template ?