NumPy EOL Versions?

classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

NumPy EOL Versions?

Symphoni Bush - NOAA Affiliate
I am trying to find out if there are any end-of-life versions for NumPy, and if so, when do these versions typically become EOL/unsupported? Thanks.


_______________________________________________
NumPy-Discussion mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: NumPy EOL Versions?

mattip
Administrator
This message ended up in my spam box for some reason, sorry for the late
reply.


We typically release a new version twice a year, and support two
versions at a time. So right now we are releasing version 1.20 and will
support 1.19 and 1.20 for the next period. When 1.21 is released, we
will support 1.20 and 1.21. Support for the released version is limited
to bug fixes.


Matti


On 12/16/20 6:24 PM, Symphoni Bush - NOAA Affiliate wrote:
> I am trying to find out if there are any end-of-life versions for
> NumPy, and if so, when do these versions typically become
> EOL/unsupported? Thanks.
>
_______________________________________________
NumPy-Discussion mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: NumPy EOL Versions?

ralfgommers


On Wed, Jan 13, 2021 at 4:41 PM Matti Picus <[hidden email]> wrote:
This message ended up in my spam box for some reason, sorry for the late
reply.


We typically release a new version twice a year, and support two
versions at a time. So right now we are releasing version 1.20 and will
support 1.19 and 1.20 for the next period. When 1.21 is released, we
will support 1.20 and 1.21. Support for the released version is limited
to bug fixes.

Is this two-version thing actually something we're committed to doing? Chuck started doing it at some point, which as release manager is his decision - but I don't think we ever really decided or announced this as a policy.

If we're committed to it, we should probably mention it somewhere in the docs or release how-to. And I should go update our Tidelift data.

Cheers,
Ralf



_______________________________________________
NumPy-Discussion mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: NumPy EOL Versions?

Charles R Harris


On Wed, Jan 13, 2021 at 9:04 AM Ralf Gommers <[hidden email]> wrote:


On Wed, Jan 13, 2021 at 4:41 PM Matti Picus <[hidden email]> wrote:
This message ended up in my spam box for some reason, sorry for the late
reply.


We typically release a new version twice a year, and support two
versions at a time. So right now we are releasing version 1.20 and will
support 1.19 and 1.20 for the next period. When 1.21 is released, we
will support 1.20 and 1.21. Support for the released version is limited
to bug fixes.

Is this two-version thing actually something we're committed to doing? Chuck started doing it at some point, which as release manager is his decision - but I don't think we ever really decided or announced this as a policy.

If we're committed to it, we should probably mention it somewhere in the docs or release how-to. And I should go update our Tidelift data.

Cheers,
Ralf

We only support one version at a time. Supporting two is hard, backports become difficult. The only time we have supported two was the 1.16.x LTS release. At the moment things are a bit different because we are waiting on pyarrow before releasing 1.20.0, it is a temporary measure.

The binannual release cycle is because 300-500 merges per release felt about right. If the merge rate goes way up and stays there we may release more often. But that is unlikely unless we get a lot more support.

Chuck

_______________________________________________
NumPy-Discussion mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: NumPy EOL Versions?

Charles R Harris


On Wed, Jan 13, 2021 at 10:47 AM Charles R Harris <[hidden email]> wrote:


On Wed, Jan 13, 2021 at 9:04 AM Ralf Gommers <[hidden email]> wrote:


On Wed, Jan 13, 2021 at 4:41 PM Matti Picus <[hidden email]> wrote:
This message ended up in my spam box for some reason, sorry for the late
reply.


We typically release a new version twice a year, and support two
versions at a time. So right now we are releasing version 1.20 and will
support 1.19 and 1.20 for the next period. When 1.21 is released, we
will support 1.20 and 1.21. Support for the released version is limited
to bug fixes.

Is this two-version thing actually something we're committed to doing? Chuck started doing it at some point, which as release manager is his decision - but I don't think we ever really decided or announced this as a policy.

If we're committed to it, we should probably mention it somewhere in the docs or release how-to. And I should go update our Tidelift data.

Cheers,
Ralf

We only support one version at a time. Supporting two is hard, backports become difficult. The only time we have supported two was the 1.16.x LTS release. At the moment things are a bit different because we are waiting on pyarrow before releasing 1.20.0, it is a temporary measure.

The binannual release cycle is because 300-500 merges per release felt about right. If the merge rate goes way up and stays there we may release more often. But that is unlikely unless we get a lot more support.

The fmod problem also played a role in making a 1.19.5 release. I hope that is also a one off.

Chuck

_______________________________________________
NumPy-Discussion mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: NumPy EOL Versions?

ralfgommers


On Wed, Jan 13, 2021 at 6:50 PM Charles R Harris <[hidden email]> wrote:


On Wed, Jan 13, 2021 at 10:47 AM Charles R Harris <[hidden email]> wrote:


On Wed, Jan 13, 2021 at 9:04 AM Ralf Gommers <[hidden email]> wrote:


On Wed, Jan 13, 2021 at 4:41 PM Matti Picus <[hidden email]> wrote:
This message ended up in my spam box for some reason, sorry for the late
reply.


We typically release a new version twice a year, and support two
versions at a time. So right now we are releasing version 1.20 and will
support 1.19 and 1.20 for the next period. When 1.21 is released, we
will support 1.20 and 1.21. Support for the released version is limited
to bug fixes.

Is this two-version thing actually something we're committed to doing? Chuck started doing it at some point, which as release manager is his decision - but I don't think we ever really decided or announced this as a policy.

If we're committed to it, we should probably mention it somewhere in the docs or release how-to. And I should go update our Tidelift data.

Cheers,
Ralf

We only support one version at a time. Supporting two is hard, backports become difficult. The only time we have supported two was the 1.16.x LTS release. At the moment things are a bit different because we are waiting on pyarrow before releasing 1.20.0, it is a temporary measure.

The binannual release cycle is because 300-500 merges per release felt about right. If the merge rate goes way up and stays there we may release more often. But that is unlikely unless we get a lot more support.

Thanks for clarifying Chuck, that all makes perfect sense.

Cheers,
Ralf


_______________________________________________
NumPy-Discussion mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/numpy-discussion