That sounds fair to me as well.
Le 27/05/2022 à 13:01, Andrew Lamb a écrit :
+1 to the idea on making it a stable interface
On Thu, May 26, 2022 at 6:57 PM Jonathan Keane <jke...@gmail.com> wrote:
I too am +1 (nonbinding) to marking it as stable
-Jon
On Thu, May 26, 2022 at 1:05 PM Neal Richardson <
neal.p.richard...@gmail.com>
wrote:
+1 from me too to mark it as stable. De facto it is stable: there have
been
no modifications to
https://github.com/apache/arrow/blob/master/cpp/src/arrow/c/abi.h since
the
stream was added in 2020.
Neal
On Thu, May 26, 2022 at 12:32 PM Dewey Dunnington <de...@voltrondata.com
wrote:
I'm fairly new to this but have worked on the DuckDB--R bindings
integration and used it in geospatial prototyping for a few things. I
would
love to see the ArrowArrayStream declared as stable to promote its
adoption
(or start the process of finalizing its definition if there is pending
feedback that hasn't yet been incorporated).
On Wed, May 25, 2022 at 6:59 PM Will Jones <will.jones...@gmail.com>
wrote:
The Arrow C Stream Interface is still listed as experimental [1],
though
it
was introduced about 20 months ago [2]. It's being used in the
well-advertised integration between PyArrow/R arrow and DuckDB [3].
Support
was added to both Rust implementations [4][5]. It was discussed in
today's
sync meeting that additional systems have been experimenting with it
[6].
Should we stabilize the API now? Or are there any changes we need to
contemplate based on the experience of our early adopters?
[1] https://arrow.apache.org/docs/7.0/format/CStreamInterface.html
[2] https://github.com/apache/arrow/pull/8052
[3] https://arrow.apache.org/blog/2021/12/03/arrow-duckdb/
[4] https://github.com/apache/arrow-rs/pull/1384
[5] https://github.com/jorgecarleitao/arrow2/pull/857
[6] https://lists.apache.org/thread/x03nck8rpmyd8td6vpz1ctqgno1cbf10
