On Thu, Oct 24, 2019 at 10:01 AM Micah Kornfield <[email protected]>
wrote:
> As part a PR Ji Liu has made to help populate data for test cases [1], the
> question came up on whether we should provide a more builder classes in
> java for ValueVectors. The proposed implementation would wrap the existing
> Writer classes.
>
> Do people think this would be a valuable addition to the java library? I
> imagine it would be a builder per ValueVectorType. The main benefit I see
> to this is making the library potentially slightly easier to use for
> new-comers, but might not be the most efficient. A straw-man interface is
> listed below.
>
> Thoughts?
>
I can see that it makes writing tests easier, and ease-of-use (esp.
handling the setSafe/setValueCount).
In dremio, we mostly populate value vectors either :
- from arrow buffers (eg. read from parquet)
- from other value vectors (eg. selection vector removal or transfers)
- directly populate the constituent arrow buffers (eg. gandiva)
so, we haven't had a need for explicit builders.
>
> Thanks,
> Micah
>
> class IntVectorBuilder {
> public IntVectorBuilder(BufferAllocator allocator);
>
> IntVectorBuilder add(int value);
> IntVectorBuilder addAll(int[] values);
> IntVectorBuilder addNull();
> // handles null values in array
> IntVectorBuilder addAll(Integer... values);
> IntVectorBuilder addAll(List<Integer> values);
> IntVector build(String name);
> }
>
--
Thanks and regards,
Ravindra.
