Hi all,

This proposal already had reviews from four different committers  on GitHub
and I think it's in good shape to call for a vote.
As was noted, it closely matches the Parquet file logical type and I think
it makes sense to add it to canonical extension types in Arrow.

Best,
Rok

On Wed, Nov 4, 2026 at 7:10 PM Mandukhai Alimaa <[email protected]>
wrote:

> Hi everyone,
>
> Bumping this thread to see if anyone has any additional feedback or
> thoughts on format PR #51248 [1].
>
> If there are no further comments or objections, we will start a vote soon.
>
> Thanks!
>
> [1] https://github.com/apache/arrow/pull/51248
>
> On Thu, Sep 17, 2026 at 2:22 PM Mandukhai Alimaa
> <[email protected]> wrote:
> >
> > Hi everyone,
> >
> > Thanks for the feedback. Following the precedent set by
> > "arrow.parquet.variant", the extension name will be
> > "arrow.parquet.file". I have opened a format change PR [1], please
> > take a look. Thanks.
> >
> > [1] https://github.com/apache/arrow/pull/51248
> >
> > On Wed, Sep 2, 2026 at 9:43 PM Neelesh Salian <[email protected]>
> wrote:
> > >
> > > +1 to use "parquet.file" as well.
> > >
> > > On Wed, Sep 2, 2026 at 7:29 PM Matt Topol <[email protected]>
> wrote:
> > >
> > > > We ended up doing "arrow.parquet.variant" so should we do
> > > > "arrow.parquet.file" for symmetry?
> > > >
> > > > On Wed, Sep 2, 2026, 10:20 PM Gang Wu <[email protected]> wrote:
> > > >
> > > > > +1 for "parquet.file"
> > > > >
> > > > > I would expect it to be very similar to how we define
> "parquet.variant".
> > > > >
> > > > > On Thu, Sep 3, 2026 at 5:22 AM Rok Mihevc <[email protected]>
> wrote:
> > > > >
> > > > > > > Shouldn't the extension name be `parquet.file`? It's really a
> direct
> > > > > > > mapping of the Parquet FILE type, not something we decided to
> > > > > > > standardize independently.
> > > > > >
> > > > > > parquet.file sounds good in that it makes it clear where the
> > > > > specification
> > > > > > comes from.
> > > > > >
> > > > > > > Also I wonder if it's the Arrow community's role to
> standardize such
> > > > > > > extension types, or if we should let the Parquet community do
> it?
> > > > > >
> > > > > > We have non-arrow namespaced types in geoarrow [1], but they are
> > > > > specified
> > > > > > in geoarrow repo. Are you proposing that?
> > > > > >
> > > > > > Parquet already specified the file type, we'd just be adding a
> mirror
> > > > > type
> > > > > > in Arrow to make handling it practical.
> > > > > >
> > > > > > [1]
> https://github.com/geoarrow/geoarrow/blob/main/extension-types.md
> > > > > >
> > > > > > Rok
> > > > > >
> > > > > > On Wed, Sep 2, 2026 at 11:05 PM Antoine Pitrou <
> [email protected]>
> > > > > wrote:
> > > > > >
> > > > > > >
> > > > > > > Shouldn't the extension name be `parquet.file`? It's really a
> direct
> > > > > > > mapping of the Parquet FILE type, not something we decided to
> > > > > > > standardize independently.
> > > > > > >
> > > > > > > Also I wonder if it's the Arrow community's role to
> standardize such
> > > > > > > extension types, or if we should let the Parquet community do
> it?
> > > > > > >
> > > > > > > Regards
> > > > > > >
> > > > > > > Antoine.
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > Le 02/09/2026 à 22:51, Mandukhai Alimaa a écrit :
> > > > > > > > Hi everyone,
> > > > > > > >
> > > > > > > > Following the recent addition and subsequent refinement of
> the FILE
> > > > > > > > LogicalType to the Parquet format specification [1, 2] and
> the
> > > > > > > > corresponding Parquet C++
> > > > > > > > reader/writer support currently underway [3], we (@rok and
> I) would
> > > > > > > > like to propose defining a matching Canonical Extension Type
> for
> > > > the
> > > > > > > > Apache Arrow specification: arrow.file.
> > > > > > > >
> > > > > > > > Motivation
> > > > > > > > As storage formats standardize file-level references, Arrow
> needs a
> > > > > > > > canonical in-memory representation to preserve these
> semantics
> > > > across
> > > > > > > > IPC streams, the C Data Interface, Arrow Flight, and language
> > > > > > > > bindings. Without a canonical extension type, file fields
> lose
> > > > their
> > > > > > > > semantic metadata during serialization.
> > > > > > > >
> > > > > > > > Proposed High-Level Design:
> > > > > > > > The intended design is to directly align with the finalized
> Parquet
> > > > > > spec
> > > > > > > [4].
> > > > > > > >
> > > > > > > > * Extension name: ``arrow.file``.
> > > > > > > > * Storage Type: StructType containing standard optional
> fields:
> > > > > > > > * ``uri`` (Utf8)
> > > > > > > > * ``offset`` (Int64)
> > > > > > > > * `` size`` (Int64)
> > > > > > > > * ``content_type`` (Utf8)
> > > > > > > > * `` checksum`` (Utf8)
> > > > > > > > * ``inline`` (Binary)
> > > > > > > >
> > > > > > > > I have created a GitHub issue to track this [5]. Please
> share your
> > > > > > > > thoughts! Thanks!
> > > > > > > >
> > > > > > > > [1] https://github.com/apache/parquet-format/pull/585
> > > > > > > > [2] https://github.com/apache/parquet-format/pull/603
> > > > > > > > [3] https://github.com/apache/arrow/issues/50891
> > > > > > > > [4]
> > > > > > >
> > > > > >
> > > > >
> > > >
> https://github.com/apache/parquet-format/blob/master/LogicalTypes.md#file
> > > > > > > > [5] https://github.com/apache/arrow/issues/51143
> > > > > > >
> > > > > > >
> > > > > >
> > > > >
> > > >
>

Reply via email to