Python 2.7 support: dropping it after Arrow 0.16 ?

[Joris Van den Bossche <jo...@gmail.com>]

Hi all,

The support for Python 2.7 has been discussed before. At some point, it was
decided to support it until the end of 2019 (ARROW-331
<https://issues.apache.org/jira/browse/ARROW-331>, the docs state that
pyarrow "pledges to maintain compatibility with Python 2.7 until the end of
2019"). Last year, it was then decided to drop support after the 1.0
release (ARROW-5757 <https://issues.apache.org/jira/browse/ARROW-5757>).
However, now that the 1.0 release is postponed a bit further, we are
thinking to drop support after the 0.16.0 release coming shortly.

Python 2.7 has officially reached end of life since January 1, 2020 (so it
is no longer maintained), and also many other Python packages (such as
numpy and pandas) already dropped support last year.

So the proposal would be that 0.16 is the last release to support Python
2.7, and the 1.0 release later this year will thus be Python 3 only.

Feedback on this timeline is welcome, certainly from communities where
Python 2.7 is still used a lot.

Best,
Joris

Re: Python 2.7 support: dropping it after Arrow 0.16 ?

[Wes McKinney <we...@gmail.com>]

hi Joris,

Personally I am OK with dropping 2.7 after the next major release
(0.16.0). According to pypistats.org there are still a lot of 2.7
installs but we aren't seeing much incoming maintenance support for
this, so it doesn't seem reasonable to burden ourselves with this.

Rather than hacking away 2.7 compatibility code immediately, I might
suggest a period where we stop testing and building packages in our
CI/release process so that,some counterparty comes out of the woodwork
and wants to submit 2.7 patches they can do that. If we remove all the
2.7 compatibility code right away then we're burning the bridge. After
maybe 6 months I would say we should go ahead and remove all the
compatibility C++ and Python code.

- Wes

On Fri, Jan 17, 2020 at 8:39 AM Joris Van den Bossche
<jo...@gmail.com> wrote:
>
> Hi all,
>
> The support for Python 2.7 has been discussed before. At some point, it was
> decided to support it until the end of 2019 (ARROW-331
> <https://issues.apache.org/jira/browse/ARROW-331>, the docs state that
> pyarrow "pledges to maintain compatibility with Python 2.7 until the end of
> 2019"). Last year, it was then decided to drop support after the 1.0
> release (ARROW-5757 <https://issues.apache.org/jira/browse/ARROW-5757>).
> However, now that the 1.0 release is postponed a bit further, we are
> thinking to drop support after the 0.16.0 release coming shortly.
>
> Python 2.7 has officially reached end of life since January 1, 2020 (so it
> is no longer maintained), and also many other Python packages (such as
> numpy and pandas) already dropped support last year.
>
> So the proposal would be that 0.16 is the last release to support Python
> 2.7, and the 1.0 release later this year will thus be Python 3 only.
>
> Feedback on this timeline is welcome, certainly from communities where
> Python 2.7 is still used a lot.
>
> Best,
> Joris