Add record_batch! macro for easy record batch creation

[timsaucer]

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

In apache/datafusion#12846 we added a macro to easily create a record batch for use in building up unit tests. It would be nice to move this upstream to this repo.

Describe the solution you'd like

Port code and optionally add remaining data types.

Describe alternatives you've considered

None

Additional context

apache/datafusion#12846

[Xuanwo]

Hi, I like this idea. How about renaming this macro to record_batch!() like vec!()?

[timsaucer]

I incorporated your recommendation in the last push of the PR in datafusion.

[alamb]

I incorporated your recommendation in the last push of the PR in datafusion.

I think it looks really nice -- thank you @timsaucer and @Xuanwo

[alamb]

I think this is a good first issue for someone who wants to try their hand out at contributing

[ByteBaker]

take

[ByteBaker]

@alamb this is the basic idea I have. Lemme know if I should change this.

let schema: Arc<Schema> = ...;
let options: RecordBatchOptions = ...;
let array: Arc<Int32Array> = ...; // Can have more arrays

record_batch!(schema); // Creates empty RecordBatch
record_batch!(schema, [array]); // Returns a result
record_batch!(schema, [array], options); // Creates w/ options, returns a result

Or do we want something that allows us to write a template-like, visually intuitive structure for creating an RB? For instance:

record_batch!{
    // Always required
    [
     "id", Int32, false // More fields can go in next lines
    ],
    // Following bit be optional. 
    // Not giving data would create an empty RecordBatch
    [
     Int32Array[1, 2, 3, 4, 5] // More arrays can go in next lines
    ],
    // `RecordBatchOptions` with `match_field_names` = false & `row_count` = Some(10)
    // This bit is optional, and the absence will just use the default options
    {true, 10}

}

Also, how do we reconcile the return types in this case?

• Using just Schema would create a RecordBatch directly
• Any other combination would return Result<RecordBatch>

[alamb]

Or do we want something that allows us to write a template-like, visually intuitive structure for creating an RB? For instance:

Thanks @ByteBaker -- maybe you can check out the implementation that @timsaucer put into DataFusion in apache/datafusion#12846