It seems unfortunate to me that the feather file format doc page [1]
appears to have been forgotten when those IPC file format docs were written
[2][3]. I will find or make a JIRA to make this info consistent in the docs.

[1]: https://arrow.apache.org/docs/python/feather.html
[2]: https://arrow.apache.org/docs/format/Columnar.html#ipc-file-format
[3]: https://arrow.apache.org/docs/cpp/ipc.html

Aldrin Montana
Computer Science PhD Student
UC Santa Cruz


On Fri, Jul 22, 2022 at 2:46 PM Will Jones <will.jones...@gmail.com> wrote:

> FYI It looks like there is active work to change the Python [1] and R [2]
> implementations (built on C++) to write .arrow instead of .feather.
>
> [1] https://issues.apache.org/jira/browse/ARROW-17089
> [2] https://issues.apache.org/jira/browse/ARROW-17088
>
> On Fri, Jul 22, 2022 at 11:16 AM Aldrin <akmon...@ucsc.edu.invalid> wrote:
>
> > sorry, I meant "...especially *for* the rust community if they are just
> > using IPC directly for file formats."
> >
> > Aldrin Montana
> > Computer Science PhD Student
> > UC Santa Cruz
> >
> >
> > On Fri, Jul 22, 2022 at 11:14 AM Aldrin <akmon...@ucsc.edu> wrote:
> >
> > > I always assumed IPC was when it was in memory, feather was when it was
> > > persisted as files. That distinction/indirection may be totally moot,
> > > though, especially if the rust community is just using IPC directly for
> > > file formats.
> > >
> > > In my mind, feather V1 came just before IPC and started using IPC since
> > > V2. I'm not sure if a feather V3 would ever diverge from IPC format or
> if
> > > feather adds anything that's more filesystem friendly (versus other
> > storage
> > > system interfaces) or makes filesystem performance more predictable.
> > >
> > >
> > > Aldrin Montana
> > > Computer Science PhD Student
> > > UC Santa Cruz
> > >
> > >
> > > On Fri, Jul 22, 2022 at 10:49 AM Andrew Lamb <al...@influxdata.com>
> > wrote:
> > >
> > >> For what it is worth, in the Rust Arrow community, I have only ever
> > heard
> > >> the format referred to as "IPC Format" , we haven't used the term
> > Feather
> > >> that I know of.
> > >>
> > >> On Fri, Jul 22, 2022 at 12:41 PM David Li <lidav...@apache.org>
> wrote:
> > >>
> > >> > FeatherV2 is the same thing as the IPC file format (
> > >> > https://arrow.apache.org/faq/). This confusion keeps coming up, so:
> > how
> > >> > do we want to brand Arrow IPC? As "Feather", or as "Arrow IPC
> Files"?
> > It
> > >> > seems the latter, but even right now Arrow Dataset still writes
> files
> > >> with
> > >> > .feather extension (though there are PRs to update it). I don't
> think
> > >> I've
> > >> > seen a formal discussion/decision on whether to keep using Feather
> or
> > >> IPC
> > >> > as the name (though I likely missed it).
> > >> >
> > >> > On Fri, Jul 22, 2022, at 12:21, Andrew Lamb wrote:
> > >> > > I am not familiar with feather -- there are some Draft PRs to add
> > >> support
> > >> > > for reading [1] and writing [2] the IPC file format which may be
> > >> related
> > >> > to
> > >> > > your question
> > >> > >
> > >> > > [1] https://github.com/apache/arrow-datafusion/pull/1858
> > >> > > [2] https://github.com/apache/arrow-datafusion/pull/1893
> > >> > >
> > >> > >
> > >> > > On Fri, Jul 22, 2022 at 12:10 PM Ian Joiner <
> iajoiner...@gmail.com>
> > >> > wrote:
> > >> > >
> > >> > >> Hi,
> > >> > >>
> > >> > >> I'd like to ask about the level of Feather support in Datafusion.
> > >> Can we
> > >> > >> read and write Feather files? I don't really find any information
> > >> while
> > >> > >> searching through the repo or checking the docs.
> > >> > >>
> > >> > >> Thanks,
> > >> > >> Ian
> > >> > >>
> > >> >
> > >>
> > >
> >
>

Reply via email to