I created

https://github.com/apache/arrow-experiments

to provide a testing ground for this idea. Might need to add a README
indicating the protocol for adding new branches (and documenting their
existence) and making PRs.

To Micah's questions

> What is the expectation around code reviews for these PRs?

I think from a code review standpoint the main focus should be IP
lineage and licensing. While we aren't going to make Apache releases
out of this repository, if code gets prototyped there and then moved
at a later date into the main repository, we don't want to have to go
spelunking through experiments PRs to make sure that the IP lineage is
clean (3rd party licenses are documented, code is not copy-pasted from
StackOverflow, etc.).

> What does minimal look like here?  Would minimal look like here?  Programmed 
> against the C-ABI (introduce a C only API?).

This is a good question. One possibility for C++ experiments is a
pared down version of the existing C++ project containing only some
portions of arrow/array and pieces like MemoryPool and Buffer, along
with certain toolchain components like googletest and gbenchmark, so
you can write some prototype code with accompanying microbenchmarks
and/or unit tests as part of exploring something new.

Other "small" APIs for experiments could be created. For example one
of my colleagues recently made a small R package containing a
miniature C Arrow implementation that implements the C-ABI with
interfaces with the R C API:

https://github.com/paleolimbot/carrow/tree/master/src

On Fri, Nov 26, 2021 at 1:37 PM Micah Kornfield <emkornfi...@gmail.com> wrote:
>
> >
> > People would make pull requests like any other repository, but it
> > would be a bit more free form than our other repositories. The goal is
> > to get this kind of collaboration (code and the discussions) happening
> > on Arrow community channels.
>
>
> What is the expectation around code reviews for these PRs?   What does
> minimal look like here?  Would minimal look like here?  Programmed against
> the C-ABI (introduce a C only API?).
>
> On Thu, Nov 18, 2021 at 9:33 AM Benson Muite <benson_mu...@emailplus.org>
> wrote:
>
> > On 11/18/21 6:29 PM, Wes McKinney wrote:
> > > On Thu, Nov 18, 2021 at 2:25 AM Antoine Pitrou <anto...@python.org>
> > wrote:
> > >>
> > >>
> > >> Le 18/11/2021 à 02:54, Wes McKinney a écrit :
> > >>>
> > >>> In short I wanted to propose creating a separate git repository under
> > >>> apache/arrow-* for this purpose, to invite these kinds of
> > >>> contributions to our project and to help more R&D work happen inside
> > >>> the Arrow umbrella so we have clean IP lineage. I can't imagine we
> > >>> would ever make releases from this repository but it could serve as a
> > >>> flexible place to put stuff (even in branches that are independent
> > >>> from each other) that may or may not be ready to make its home in one
> > >>> of our production repositories.
> > >>
> > >> What would be the rules for contributing? Is it just a place where
> > >> people store source code?
> > >
> > > People would make pull requests like any other repository, but it
> > > would be a bit more free form than our other repositories. The goal is
> > > to get this kind of collaboration (code and the discussions) happening
> > > on Arrow community channels.
> > This may be helpful. Some of it might also lead to developer/interested
> > user documentation, perhaps similar to the R-Journal
> > https://journal.r-project.org/ but with less formality
> > >
> >
> >

Reply via email to