Hi,
While I'm generally in favor of accepting this soon, I'm -1 on accepting
it right now because it seems the PR hasn't had enough review attention
on it (I posted some comments).
A spec is an important document that will bind us for years, so let's
make sure we write something that will not bother us in the future.
Regards
Antoine.
Le 05/12/2024 à 03:58, Sutou Kouhei a écrit :
Hi,
I would like to propose standardizing how to pas statistics
through the C data interface.
Motivation:
* We want to pass not only Apache Arrow data but also
statistics of them through the C data interface for query
planning.
Approach:
* Define a standardized schema for statistics.
* Represent statistics as an Apache Arrow array that uses
the schema.
* Pass the statistics Apache Arrow array through the C data
interface like a normal Apache Arrow array.
Note that we don't define a new interface for statistics. We
just use the existing C data interface. A statistics Apache
Arrow array is passed through a separated API call.
See also:
* The discussion of this:
https://lists.apache.org/thread/z0jz2bnv61j7c6lbk7lympdrs49f69cx
* The PR of this proposal that includes the statistics
schema definition:
https://github.com/apache/arrow/pull/43553
* The preview URL of the PR:
http://crossbow.voltrondata.com/pr_docs/43553/format/CDataInterfaceStatistics.html
Note:
* I implemented this proposal only in C++. The
implementation is already merged into apache/arrow. Should
we have one more implementation like format specification
change?
http://crossbow.voltrondata.com/pr_docs/43553/format/Changing.html#at-least-two-reference-implementations
The vote will be open for at least 72 hours.
[ ] +1 Accept this proposal
[ ] +0
[ ] -1 Do not accept this proposal because...
Thanks,