Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Matthias Koeppe]

Here's the link to the ticket for convenience. 
https://trac.sagemath.org/ticket/29141

On Sunday, February 2, 2020 at 9:15:59 AM UTC-5, Matthias Koeppe wrote:
>
> I have created #29141 (Meta-ticket: Upgrades and other changes that 
> require dropping py2 support or separate package versions for py2/py3) to 
> summarize this thread. 
>
> On Wednesday, January 22, 2020 at 5:22:15 AM UTC-5, E. Madison Bray wrote:
>>
>> On Fri, Jan 10, 2020 at 1:03 AM Volker Braun  wrote: 
>> > 
>> > * I think its not too difficult to write code that is Python 2.7 + 3.x 
>> (for high enough x) compatible, so its not a super pressing issue 
>> > * We do have a Python 2 buildbot to test for regressions 
>> > * For semver reasons we should drop Python 2.7 support in Sage 10, not 
>> 9.1 
>> > 
>> > Having said that, I'm fine with an accelerated Sage 9 -> Sage 10 
>> schedule, maybe a month or two instead of the usual 3-4. Though first we 
>> should take the opportunity and see if there are any outstanding Python 3 
>> bugs now that we have more data. For example it would be nice if a build 
>> with SAGE_DEBUG=yes would pass tests. There are a few more regressions, 
>> e.g. #28817 
>>
>> Since this thread had gone on quite long and it has perhaps gotten 
>> lost, I would just like to once again endorse Volker's suggestion 
>> here.  We have already put enormous effort (hundreds of hours) into 
>> making Sage Python 2/3 compatible and making it relatively easy to 
>> write compatible code (and most new code will not even encounter these 
>> compatibility issues). 
>>
>> Let us have a short release cycle for 9.1 and add a prominent 
>> deprecation message in 9.1 for Python 2 builds, then Sage 10.0 can 
>> implicitly start dropping support for Python 2.  If there are some 
>> patches that are needed (e.g. to support ipython7) they will have to 
>> be applied eventually anyways so if some niche packing system must 
>> apply it for Sage 9.1 I see little harm. 
>>
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/511c3883-b3d7-4650-aa3c-0e9ebb5e0973%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Matthias Koeppe]

I have created #29141 (Meta-ticket: Upgrades and other changes that require 
dropping py2 support or separate package versions for py2/py3) to summarize 
this thread. 

On Wednesday, January 22, 2020 at 5:22:15 AM UTC-5, E. Madison Bray wrote:
>
> On Fri, Jan 10, 2020 at 1:03 AM Volker Braun  > wrote: 
> > 
> > * I think its not too difficult to write code that is Python 2.7 + 3.x 
> (for high enough x) compatible, so its not a super pressing issue 
> > * We do have a Python 2 buildbot to test for regressions 
> > * For semver reasons we should drop Python 2.7 support in Sage 10, not 
> 9.1 
> > 
> > Having said that, I'm fine with an accelerated Sage 9 -> Sage 10 
> schedule, maybe a month or two instead of the usual 3-4. Though first we 
> should take the opportunity and see if there are any outstanding Python 3 
> bugs now that we have more data. For example it would be nice if a build 
> with SAGE_DEBUG=yes would pass tests. There are a few more regressions, 
> e.g. #28817 
>
> Since this thread had gone on quite long and it has perhaps gotten 
> lost, I would just like to once again endorse Volker's suggestion 
> here.  We have already put enormous effort (hundreds of hours) into 
> making Sage Python 2/3 compatible and making it relatively easy to 
> write compatible code (and most new code will not even encounter these 
> compatibility issues). 
>
> Let us have a short release cycle for 9.1 and add a prominent 
> deprecation message in 9.1 for Python 2 builds, then Sage 10.0 can 
> implicitly start dropping support for Python 2.  If there are some 
> patches that are needed (e.g. to support ipython7) they will have to 
> be applied eventually anyways so if some niche packing system must 
> apply it for Sage 9.1 I see little harm. 
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/3ab827ab-cf2d-4b67-967d-6b08c06de3b1%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[E. Madison Bray]

On Fri, Jan 10, 2020 at 1:03 AM Volker Braun  wrote:
>
> * I think its not too difficult to write code that is Python 2.7 + 3.x (for 
> high enough x) compatible, so its not a super pressing issue
> * We do have a Python 2 buildbot to test for regressions
> * For semver reasons we should drop Python 2.7 support in Sage 10, not 9.1
>
> Having said that, I'm fine with an accelerated Sage 9 -> Sage 10 schedule, 
> maybe a month or two instead of the usual 3-4. Though first we should take 
> the opportunity and see if there are any outstanding Python 3 bugs now that 
> we have more data. For example it would be nice if a build with 
> SAGE_DEBUG=yes would pass tests. There are a few more regressions, e.g. #28817

Since this thread had gone on quite long and it has perhaps gotten
lost, I would just like to once again endorse Volker's suggestion
here.  We have already put enormous effort (hundreds of hours) into
making Sage Python 2/3 compatible and making it relatively easy to
write compatible code (and most new code will not even encounter these
compatibility issues).

Let us have a short release cycle for 9.1 and add a prominent
deprecation message in 9.1 for Python 2 builds, then Sage 10.0 can
implicitly start dropping support for Python 2.  If there are some
patches that are needed (e.g. to support ipython7) they will have to
be applied eventually anyways so if some niche packing system must
apply it for Sage 9.1 I see little harm.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAOTD34bmEgw_CxV6qzwUi6tVa%3DhpeGO0BV4wUOULRUv4iNamVA%40mail.gmail.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[E. Madison Bray]

On Tue, Jan 21, 2020 at 5:19 PM Timo Kaufmann  wrote:
>
>
>
> Am Freitag, 17. Januar 2020 16:07:08 UTC+1 schrieb E. Madison Bray:
>>
>> On Wed, Jan 15, 2020 at 2:37 AM Matthias Koeppe
>>  wrote:
>> >
>> > On Tuesday, January 14, 2020 at 12:01:42 PM UTC-5, Frédéric Chapoton wrote:
>> >>
>> >> So here is my proposal.
>> >>
>> >> * Starting from now, we allow ourselves to move on, using 9.1 betas and 
>> >> further releases for external python3 updates, including switch to 
>> >> ipython7, which seems to me the most urgent matter. But we also do not 
>> >> introduce python3-only code in our own code base if we can avoid it.
>>
>> I'm not exactly sure what you mean by this, but it might be agreeable
>> to me:  I'm completely okay with adding Python 3 only dependencies and
>> even *code/features* so long as it's done without breaking
>> backwards-compatibility.  When it comes to sage-the-distribution
>> (which I sense is where most of this friction is coming from, and yet
>> another reason to better separate sage from the
>> sage-the-distribution), if there are dependencies you want to update
>> that are Python 3-only I'd say go for it, but make it a separate SPKG
>> so that previous versions of the dependency can still work on Python 2
>> builds.
>
>
> As others have said already, supporting both python2 and python3 dependencies 
> is much more work than only supporting python2 or only supporting python3 
> versions.

I don't think it is.  I think that maintaining Python 2 support for
now (and gradually phasing it out) is the path of least resistance.

> I think dropping pyhon2 support with the next release remains the best 
> option. Even better of course would be to keep the option for bugfix releases 
> for 8.9 / 9.0. That would be a nice thing to have in general, but we could 
> also just make an exception for this one release.

Yes, that would be better.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAOTD34YdA%2B_RioirQLxQ6SZL_weGRBT1BjX96Yshw%2B%3Dk9P4jQw%40mail.gmail.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[William Stein]

On Tue, Jan 21, 2020 at 9:49 AM Emmanuel Charpentier
 wrote:
>
> I second that. Unless someone presents us with a pressing case of a large 
> Python 2-based library unportable to Python 3 (and therefore already in a 
> dead-end), our limited workforce is better employed to maintain our Python 3 
> Sage.

I'm just going to endorse this decision yet again based on (for me)
new evidence:

1. I personally spent a lot of frustrating time last week working on
some Python code to make it worth well with both Python2 and Python3
versions of Sage.  The code involved a lot of string manipulation
(display of output, sending data over networks connections, etc.), and
it was a *very* painful experience with expressions that are valid in
both Python2 and Python3, but behave differently in each, etc.I
would not wish such work on anybody.

2. Also I read the blog linked to from
https://news.ycombinator.com/item?id=22036773 which talks about how
Mercurial's decision to support Python2 so long made their support of
Python3 more difficult, and continues to have a significant negative
impact moving forward.

So I endorse Sage developers using their limited resources to move
Sage forward to be Python 3 only, and let old versions of Sage provide
backward compatibility for people with lots of Python2 Sage code.

 -- William

-- 
William (http://wstein.org)

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CACLE5GAoDc-kxyHJGRKS_hA%2Br_KgLYi2ONZtKdpNRo-VA7UPLA%40mail.gmail.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Emmanuel Charpentier]

I second that. Unless someone presents us with a pressing case of a large 
Python 2-based library unportable to Python 3 (and therefore already in a 
dead-end), our limited workforce is better employed to maintain our Python 
3 Sage.

In other words, "if you need a Python 2-specific fix, speak *now* or 
forever hold your peace..."

Corollary : we should update the places where we published our plans for 
Python 2 compatibility and point to the present thread, which can be 
summarized by "The Sagemath developers have decided not to maintain Python 
2 compatibility beyond Sagemath 9.0." as an acceptable first-order 
approximation.

I agree that a Python 2-compatible maintenance branch *would* be a Good 
Thing (TM) to have. But I doubt that our present workforce is enough to 
maintain it. A french proverb says that "L'enfer est pavé de bonnes 
intentions" (Hell is paved with good intents) ; I think that we have such a 
case...

Now, for the cleanup of Python 2-specific idioms, I suspect that this will 
take a long time. And defining the set of Python 2 expressions that would 
be advantageously rewritten with (different) Pyhon 3 idioms is by no way an 
easy task. I would suggest to keep it as an informal goal, and to treat it 
incrementally.

HTH,

Le mardi 21 janvier 2020 17:19:33 UTC+1, Timo Kaufmann a écrit :
>
>
>
> Am Freitag, 17. Januar 2020 16:07:08 UTC+1 schrieb E. Madison Bray:
>>
>> On Wed, Jan 15, 2020 at 2:37 AM Matthias Koeppe 
>>  wrote: 
>> > 
>> > On Tuesday, January 14, 2020 at 12:01:42 PM UTC-5, Frédéric Chapoton 
>> wrote: 
>> >> 
>> >> So here is my proposal. 
>> >> 
>> >> * Starting from now, we allow ourselves to move on, using 9.1 betas 
>> and further releases for external python3 updates, including switch to 
>> ipython7, which seems to me the most urgent matter. But we also do not 
>> introduce python3-only code in our own code base if we can avoid it. 
>>
>> I'm not exactly sure what you mean by this, but it might be agreeable 
>> to me:  I'm completely okay with adding Python 3 only dependencies and 
>> even *code/features* so long as it's done without breaking 
>> backwards-compatibility.  When it comes to sage-the-distribution 
>> (which I sense is where most of this friction is coming from, and yet 
>> another reason to better separate sage from the 
>> sage-the-distribution), if there are dependencies you want to update 
>> that are Python 3-only I'd say go for it, but make it a separate SPKG 
>> so that previous versions of the dependency can still work on Python 2 
>> builds. 
>>
>
> As others have said already, supporting both python2 and python3 
> dependencies is much more work than only supporting python2 or only 
> supporting python3 versions.
>
> I think dropping pyhon2 support with the next release remains the best 
> option. Even better of course would be to keep the option for bugfix 
> releases for 8.9 / 9.0. That would be a nice thing to have in general, but 
> we could also just make an exception for this one release.
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/459616b9-b12b-4588-b684-7833536e2db9%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Timo Kaufmann]

Am Freitag, 17. Januar 2020 16:07:08 UTC+1 schrieb E. Madison Bray:
>
> On Wed, Jan 15, 2020 at 2:37 AM Matthias Koeppe 
> > wrote: 
> > 
> > On Tuesday, January 14, 2020 at 12:01:42 PM UTC-5, Frédéric Chapoton 
> wrote: 
> >> 
> >> So here is my proposal. 
> >> 
> >> * Starting from now, we allow ourselves to move on, using 9.1 betas and 
> further releases for external python3 updates, including switch to 
> ipython7, which seems to me the most urgent matter. But we also do not 
> introduce python3-only code in our own code base if we can avoid it. 
>
> I'm not exactly sure what you mean by this, but it might be agreeable 
> to me:  I'm completely okay with adding Python 3 only dependencies and 
> even *code/features* so long as it's done without breaking 
> backwards-compatibility.  When it comes to sage-the-distribution 
> (which I sense is where most of this friction is coming from, and yet 
> another reason to better separate sage from the 
> sage-the-distribution), if there are dependencies you want to update 
> that are Python 3-only I'd say go for it, but make it a separate SPKG 
> so that previous versions of the dependency can still work on Python 2 
> builds. 
>

As others have said already, supporting both python2 and python3 
dependencies is much more work than only supporting python2 or only 
supporting python3 versions.

I think dropping pyhon2 support with the next release remains the best 
option. Even better of course would be to keep the option for bugfix 
releases for 8.9 / 9.0. That would be a nice thing to have in general, but 
we could also just make an exception for this one release.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/81d27008-9330-4f47-ae11-b65fd442cbba%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[E. Madison Bray]

On Wed, Jan 15, 2020 at 2:37 AM Matthias Koeppe
 wrote:
>
> On Tuesday, January 14, 2020 at 12:01:42 PM UTC-5, Frédéric Chapoton wrote:
>>
>> So here is my proposal.
>>
>> * Starting from now, we allow ourselves to move on, using 9.1 betas and 
>> further releases for external python3 updates, including switch to ipython7, 
>> which seems to me the most urgent matter. But we also do not introduce 
>> python3-only code in our own code base if we can avoid it.

I'm not exactly sure what you mean by this, but it might be agreeable
to me:  I'm completely okay with adding Python 3 only dependencies and
even *code/features* so long as it's done without breaking
backwards-compatibility.  When it comes to sage-the-distribution
(which I sense is where most of this friction is coming from, and yet
another reason to better separate sage from the
sage-the-distribution), if there are dependencies you want to update
that are Python 3-only I'd say go for it, but make it a separate SPKG
so that previous versions of the dependency can still work on Python 2
builds.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAOTD34YdumAcpD%3DM_xqvJc5WrycB9vpNqurQyfLOLR3V4LzEPA%40mail.gmail.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[E. Madison Bray]

On Tue, Jan 14, 2020 at 9:42 PM Nils Bruin  wrote:
>
> On Tuesday, January 14, 2020 at 9:01:42 AM UTC-8, Frédéric Chapoton wrote:
>>
>> So here is my proposal.
>>
>> * Starting from now, we allow ourselves to move on, using 9.1 betas and 
>> further releases for external python3 updates, including switch to ipython7, 
>> which seems to me the most urgent matter. But we also do not introduce 
>> python3-only code in our own code base if we can avoid it.
>>
>
> I think for development and stability, we should take heed of the fact that 
> trac has some recent tickets addressing issues with py3 that are only being 
> uncovered now. The reality is that prior to 9.0, sage/py3 got only very 
> limited testing outside doctests and developers working on it. Doctests by no 
> means cover the full functionality of sage, and I would expect that in the 
> next few months a lot more issues will be found. I'd be a little hesitant 
> dropping py2 support under those conditions: being able to drop back on py2 
> in these situations for a while seems like a decent mitigation strategy, and 
> would probably help with debugging.
>
> I think similar thinking led to the original proposed and published strategy 
> on the wiki, and it still makes sense. In practice, sage/py3 is not really a 
> proven product yet; just because people won't start really using it until 
> there's actually an official release with it.
>
>  I appreciate the good intentions of "not introducing py3-only code in our 
> own code-base if we can avoid it", but how much does that help? If there's 
> one such change in a file that gets loaded into sage upon start-up, or which 
> needs to be parsed during build, the possibility of running sage on py2 is 
> gone, isn't it?
>
> I agree it's perhaps frustrating that the process of moving sage from py2 to 
> py3 hasn't completed with py2 falling out of support, but having a first 
> official release based on py3, while being the biggest step (and a big 
> applause for everybody who contributed to this herculean task!) is 
> unfortunately not the completion. The bug reporting and fixing that follows 
> it shouldn't just be ignored.

 ^ This says it better than I could.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAOTD34b-%3DW%2Biah4Cd0-nhNFNu1rs9Qq095cyhRM_g_U02q62Kg%40mail.gmail.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[E. Madison Bray]

On Tue, Jan 14, 2020 at 10:41 AM Timo Kaufmann  wrote:
>
> Am Montag, 13. Januar 2020 17:33:56 UTC+1 schrieb E. Madison Bray:
>>
>> On Sat, Jan 11, 2020 at 9:24 AM Antonio Rojas  wrote:
>> >
>> > El viernes, 10 de enero de 2020, 14:54:24 (UTC+1), E. Madison Bray 
>> > escribió:
>> >>
>> >> That seems like the obvious approach to me.  As it is I've long felt
>> >> that Sage should be more flexible in its dependencies where
>> >> possible/necessary.  With most Python packages it's easy as most have
>> >> a .__version__ and its not so hard to define some variable
>> >> like IS_RPY_2 and just have two separate branches.  I have things like
>> >> that all over the place in other packages to support e.g. different
>> >> Numpy versions or work around version-specific bugs.
>> >
>> >
>> > I've opened https://trac.sagemath.org/ticket/28988 for rpy. But at this 
>> > point the major issues are python 3.8 and ipython 7, and I don't see how 
>> > one could support several versions of them without forking hundreds of 
>> > doctests. Both updates require multi-thousand-lines patches, due to 
>> > changes in dict sorting and hashes.
>>
>> That remains a fault of over-reliance on doctests.
>
>
> What else should we rely on? Also, doctests are not the only things that fail 
> with current python3 libraries.

Normal unit tests.  The annoying thing about Sage's doctest-centric
testing framework (which is otherwise good) is that it tends to lead
people to thinking they can't write unit tests.  But in fact there are
unit tests in Sage, and they're just invoked through the doctest
runner by writing "doctests" that run a unit test suite.

>> I don't think
>> downstream packaging is a good enough reason to push sage to rush
>> things in such a way that is not well-communicated to the user
>> community.  If you need to have a multi-thousand-line patch then so be
>> it.  A patch is a patch.
>
>
> That is unfortunate. I agree that "a patch is a patch", but in my view the 
> conclusion should be the opposite: Upstream should strive for no patches to 
> be necessary (except maybe to work around very distro-specific quirks). No 
> 5000 line patches and no 5 line patches.
>
> For me this decision means that sage on nixos will likely be stuck on python2 
> for a while, and I can only hope that the python2 infrastructure keeps 
> working for long enough.

Well, all the more reason to maintain Python 2 support a little longer
than to rush and break things.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAOTD34YreBJwmQS97R5%2Bn2YmpDxeTR8OJ_42Y7G3nQZCktH6Mw%40mail.gmail.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Sébastien Labbé]

On Tuesday, January 14, 2020 at 6:01:42 PM UTC+1, Frédéric Chapoton wrote:
>
> This has been a long discussion already. Let me try to summarize. My 
> question was :
>
>   *Do you agree that sage release 9.1 (and most of the 9.1.betas) will 
> not be kept compatible with Python 2 ?*
>

here is my vote:

+1
 

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/ec12a435-95d8-4225-b2ab-291c227f2e7b%40googlegroups.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Matthias Koeppe]

On Tuesday, January 14, 2020 at 12:01:42 PM UTC-5, Frédéric Chapoton wrote:
>
> So here is my proposal.
>
> * Starting from now, we allow ourselves to move on, using 9.1 betas and 
> further releases for external python3 updates, including switch to 
> ipython7, which seems to me the most urgent matter. But we also do not 
> introduce python3-only code in our own code base if we can avoid it.
>
> * Concerning semantic versioning, I would advocate to use 10.0 as the 
> first release where all compatibility code will be removed, hence pure py3. 
> This may take some time, given the sparse work force. Once done, we allow 
> ourself to use python3-only syntax in our own codebase, such as "yield 
> from".
>
> Therefore, we can use 9.1, 9.2 and so on until somebody™ does the job of 
> cleaning up sagemath codebase from remains of py2-only code.
>
> I have tried to make a balanced and reasonable proposal. Applause or 
> scream at your will.
>

I agree with all of this.  

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/44fe2a96-99b2-4ebb-8128-4ddae3e76f2a%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Matthias Koeppe]

On Tuesday, January 14, 2020 at 8:01:18 PM UTC-5, Michael Orlitzky wrote:
>
>
>   1. Branch off the 9.0 release 
>   2. Drop compatibility in 9.1 
>   3. Proceed as normal 
>   4. If there are any major bugs found in 9.1, backport the fix to 
>  the 9.0 branch, and release a 9.0.1 that supports python-2.x. 
>
> Mathematical bug fixes that merge clean and pass the tests could also 
> make the cut, but ideally step (4) would require little work. This would 
> buy users a few more months of python2 compatibility, and lets the rest 
> of us move on. 
>

I agree that there needs to be a 9.0 maintenance branch – if only to be the 
designated upstream for distributors that want to keep maintaining a py2 
version of sage. 
 

> The main downside is that this relies solely on the release manager to 
> merge bugfix tickets twice, into both release branches. So if he thinks 
> this is a waste of time (understandable), then it's a non-starter. 
>

The maintenance branch could also be maintained by another developer. 

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/c232c89b-b45a-4465-a9e9-03ca8a240c6d%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Michael Orlitzky]

On 1/14/20 12:01 PM, Frédéric Chapoton wrote:
> 
> [A] It seems to me that the conclusion of the vote is rather clear, and
> that we should not feel obliged to make 9.1 compilable with python2.
> 

Since I'm not on that list yet, one thing is for sure: there are going
to be two consecutive versions of sage, one where python-2.x works, and
one where it doesn't. Users won't care too much which versions those are
unless there's a feature or bugfix that they need in the new version --
they can just use the old one.

Most projects solve this with multiple release branches. When you break
backwards-compatibility in a major way, you keep the previous version
(before the break) in a separate branch, and backport critical fixes to
it for a limited time. It's not how the sage release process works, but
I don't see why we couldn't do it this once, to keep everyone happy:

  1. Branch off the 9.0 release
  2. Drop compatibility in 9.1
  3. Proceed as normal
  4. If there are any major bugs found in 9.1, backport the fix to
 the 9.0 branch, and release a 9.0.1 that supports python-2.x.

Mathematical bug fixes that merge clean and pass the tests could also
make the cut, but ideally step (4) would require little work. This would
buy users a few more months of python2 compatibility, and lets the rest
of us move on.

The main downside is that this relies solely on the release manager to
merge bugfix tickets twice, into both release branches. So if he thinks
this is a waste of time (understandable), then it's a non-starter.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/ab573d83-ab79-4bdc-1e23-36f411e26a44%40orlitzky.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Nils Bruin]

On Tuesday, January 14, 2020 at 9:01:42 AM UTC-8, Frédéric Chapoton wrote:
>
> So here is my proposal.
>
> * Starting from now, we allow ourselves to move on, using 9.1 betas and 
> further releases for external python3 updates, including switch to 
> ipython7, which seems to me the most urgent matter. But we also do not 
> introduce python3-only code in our own code base if we can avoid it.
>
>
I think for development and stability, we should take heed of the fact that 
trac has some recent tickets addressing issues with py3 that are only being 
uncovered now. The reality is that prior to 9.0, sage/py3 got only very 
limited testing outside doctests and developers working on it. Doctests by 
no means cover the full functionality of sage, and I would expect that in 
the next few months a lot more issues will be found. I'd be a little 
hesitant dropping py2 support under those conditions: being able to drop 
back on py2 in these situations for a while seems like a decent mitigation 
strategy, and would probably help with debugging.

I think similar thinking led to the original proposed and published 
strategy on the wiki, and it still makes sense. In practice, sage/py3 is 
not really a proven product yet; just because people won't start really 
using it until there's actually an official release with it.

 I appreciate the good intentions of "not introducing py3-only code in our 
own code-base if we can avoid it", but how much does that help? If there's 
one such change in a file that gets loaded into sage upon start-up, or 
which needs to be parsed during build, the possibility of running sage on 
py2 is gone, isn't it?

I agree it's perhaps frustrating that the process of moving sage from py2 
to py3 hasn't completed with py2 falling out of support, but having a first 
official release based on py3, while being the biggest step (and a big 
applause for everybody who contributed to this herculean task!) is 
unfortunately not the completion. The bug reporting and fixing that follows 
it shouldn't just be ignored.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/87ad6da2-851b-4b3a-9bde-a9bd4f51660d%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[William Stein]

Strong *applause* from me.  (This is a very hard problem and I continue to
be amazed and greatly appreciate what everybody has done related to Python3
support in Sage!)

On Tue, Jan 14, 2020, 9:01 AM Frédéric Chapoton 
wrote:

> This has been a long discussion already. Let me try to summarize. My
> question was :
>
>   *Do you agree that sage release 9.1 (and most of the 9.1.betas) will
> not be kept compatible with Python 2 ?*
>
> Some people give a more or less clear *positive* answer :
>
> Chapoton, Bissey, Gourgoulhon, Kaufmann, Novoseltsev, Koeppe, Mantysalo,
> Rojas, Isuru, Stein, Dunfield, Pasechnik (*12*)
>
> Some people give a strong *negative* answer :
>
> Bruin, Bray (*2*)
>
> Some people did not answer the question in a clear way, but still said
> *something* :
>
> Palmieri, Roe, King, Crisman, Delecroix, Scrimshaw, Kliem, Labbé, Braun (
> *9*)
>
> Among them, I would say that maybe 5 people were rather negative on the
> question.
>
> I hope that I have not forgotten anybody or misinterpreted wildly.
>
> [A] It seems to me that the conclusion of the vote is rather clear, and
> that we should not feel obliged to make 9.1 compilable with python2.
>
> ==> I have therefore changed the controversial wiki page, to make only the
> statement that 9.0 is compilable with python2. No more promise for the
> future.
>
> [B] I understand and approve the concerns about our user base. We should
> really try harder to listen more and help people that are simple users.
>
> So here is my proposal.
>
> * Starting from now, we allow ourselves to move on, using 9.1 betas and
> further releases for external python3 updates, including switch to
> ipython7, which seems to me the most urgent matter. But we also do not
> introduce python3-only code in our own code base if we can avoid it.
>
> * Concerning semantic versioning, I would advocate to use 10.0 as the
> first release where all compatibility code will be removed, hence pure py3.
> This may take some time, given the sparse work force. Once done, we allow
> ourself to use python3-only syntax in our own codebase, such as "yield
> from".
>
> Therefore, we can use 9.1, 9.2 and so on until somebody™ does the job of
> cleaning up sagemath codebase from remains of py2-only code.
>
> I have tried to make a balanced and reasonable proposal. Applause or
> scream at your will.
>
> Frederic
>
>
>
>
>
>
> Le dimanche 5 janvier 2020 20:44:10 UTC+1, Frédéric Chapoton a écrit :
>>
>> Hello,
>>
>> I would like to suggest that the sooner we drop Python 2 support the
>> better. We still need to handle the upgrade to ipython7 and the
>> compatibility with python 3.8. All this will be made very difficult if we
>> insist on maintaining a codebase that is both compatible with python 2 and
>> python 3.
>>
>> So, please vote :
>>
>>
>>
>> Frédéric
>>
>>
>>
>> --
> You received this message because you are subscribed to a topic in the
> Google Groups "sage-devel" group.
> To unsubscribe from this topic, visit
> https://groups.google.com/d/topic/sage-devel/vYlbnAwKATM/unsubscribe.
> To unsubscribe from this group and all its topics, send an email to
> sage-devel+unsubscr...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/sage-devel/9db91206-3a7c-434a-8b07-7656e5b5f458%40googlegroups.com
> 
> .
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CACLE5GDNNBsnmQeDSDor16N8CO_v1P-PV1T9ovRcLPskA7b7%2BQ%40mail.gmail.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Frédéric Chapoton]

This has been a long discussion already. Let me try to summarize. My 
question was :

  *Do you agree that sage release 9.1 (and most of the 9.1.betas) will not 
be kept compatible with Python 2 ?*

Some people give a more or less clear *positive* answer :

Chapoton, Bissey, Gourgoulhon, Kaufmann, Novoseltsev, Koeppe, Mantysalo, 
Rojas, Isuru, Stein, Dunfield, Pasechnik (*12*)

Some people give a strong *negative* answer :

Bruin, Bray (*2*)

Some people did not answer the question in a clear way, but still said 
*something* :

Palmieri, Roe, King, Crisman, Delecroix, Scrimshaw, Kliem, Labbé, Braun (*9*
)

Among them, I would say that maybe 5 people were rather negative on the 
question.

I hope that I have not forgotten anybody or misinterpreted wildly.

[A] It seems to me that the conclusion of the vote is rather clear, and 
that we should not feel obliged to make 9.1 compilable with python2.

==> I have therefore changed the controversial wiki page, to make only the 
statement that 9.0 is compilable with python2. No more promise for the 
future.

[B] I understand and approve the concerns about our user base. We should 
really try harder to listen more and help people that are simple users.

So here is my proposal.

* Starting from now, we allow ourselves to move on, using 9.1 betas and 
further releases for external python3 updates, including switch to 
ipython7, which seems to me the most urgent matter. But we also do not 
introduce python3-only code in our own code base if we can avoid it.

* Concerning semantic versioning, I would advocate to use 10.0 as the first 
release where all compatibility code will be removed, hence pure py3. This 
may take some time, given the sparse work force. Once done, we allow 
ourself to use python3-only syntax in our own codebase, such as "yield 
from".

Therefore, we can use 9.1, 9.2 and so on until somebody™ does the job of 
cleaning up sagemath codebase from remains of py2-only code.

I have tried to make a balanced and reasonable proposal. Applause or scream 
at your will.

Frederic






Le dimanche 5 janvier 2020 20:44:10 UTC+1, Frédéric Chapoton a écrit :
>
> Hello,
>
> I would like to suggest that the sooner we drop Python 2 support the 
> better. We still need to handle the upgrade to ipython7 and the 
> compatibility with python 3.8. All this will be made very difficult if we 
> insist on maintaining a codebase that is both compatible with python 2 and 
> python 3.
>
> So, please vote :
>
>
>
> Frédéric
>
>
>
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/9db91206-3a7c-434a-8b07-7656e5b5f458%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Timo Kaufmann]

Am Montag, 13. Januar 2020 17:33:56 UTC+1 schrieb E. Madison Bray:
>
> On Sat, Jan 11, 2020 at 9:24 AM Antonio Rojas  > wrote: 
> > 
> > El viernes, 10 de enero de 2020, 14:54:24 (UTC+1), E. Madison Bray 
> escribió: 
> >> 
> >> That seems like the obvious approach to me.  As it is I've long felt 
> >> that Sage should be more flexible in its dependencies where 
> >> possible/necessary.  With most Python packages it's easy as most have 
> >> a .__version__ and its not so hard to define some variable 
> >> like IS_RPY_2 and just have two separate branches.  I have things like 
> >> that all over the place in other packages to support e.g. different 
> >> Numpy versions or work around version-specific bugs. 
> > 
> > 
> > I've opened https://trac.sagemath.org/ticket/28988 for rpy. But at this 
> point the major issues are python 3.8 and ipython 7, and I don't see how 
> one could support several versions of them without forking hundreds of 
> doctests. Both updates require multi-thousand-lines patches, due to changes 
> in dict sorting and hashes. 
>
> That remains a fault of over-reliance on doctests.


What else should we rely on? Also, doctests are not the only things that 
fail with current python3 libraries.

I don't think 
> downstream packaging is a good enough reason to push sage to rush 
> things in such a way that is not well-communicated to the user 
> community.  If you need to have a multi-thousand-line patch then so be 
> it.  A patch is a patch. 
>

That is unfortunate. I agree that "a patch is a patch", but in my view the 
conclusion should be the opposite: Upstream should strive for no patches to 
be necessary (except maybe to work around very distro-specific quirks). No 
5000 line patches and no 5 line patches.

For me this decision means that sage on nixos will likely be stuck on 
python2 for a while, and I can only hope that the python2 infrastructure 
keeps working for long enough.

I still don't understand the reasoning here. From my point of view (which 
is of course biased), keeping python2 compatibility has a huge downside. At 
the same time, keeping it has very little upside. The main reasoning seems 
to be to give users time to update their code. But the value-addition of 
keeping python2 in newer sage versions vs just telling users to use 8.9 / 
9.0 with python2 seems very small to me. On the contrary, newer features 
being python3 exclusive might give a good incentive for updating. Otherwise 
we just postpone the problem until we finally do drop support.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/2568ce65-4c70-469c-abf3-50e4b6c5ab10%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[E. Madison Bray]

On Sat, Jan 11, 2020 at 9:24 AM Antonio Rojas  wrote:
>
> El viernes, 10 de enero de 2020, 14:54:24 (UTC+1), E. Madison Bray escribió:
>>
>> That seems like the obvious approach to me.  As it is I've long felt
>> that Sage should be more flexible in its dependencies where
>> possible/necessary.  With most Python packages it's easy as most have
>> a .__version__ and its not so hard to define some variable
>> like IS_RPY_2 and just have two separate branches.  I have things like
>> that all over the place in other packages to support e.g. different
>> Numpy versions or work around version-specific bugs.
>
>
> I've opened https://trac.sagemath.org/ticket/28988 for rpy. But at this point 
> the major issues are python 3.8 and ipython 7, and I don't see how one could 
> support several versions of them without forking hundreds of doctests. Both 
> updates require multi-thousand-lines patches, due to changes in dict sorting 
> and hashes.

That remains a fault of over-reliance on doctests.  I don't think
downstream packaging is a good enough reason to push sage to rush
things in such a way that is not well-communicated to the user
community.  If you need to have a multi-thousand-line patch then so be
it.  A patch is a patch.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAOTD34YdyANvYbS-M3cQqM-h0aA4a8us3wu4XDq2wMoX%2BGmECA%40mail.gmail.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Antonio Rojas]

El viernes, 10 de enero de 2020, 14:54:24 (UTC+1), E. Madison Bray escribió:
>
> That seems like the obvious approach to me.  As it is I've long felt 
> that Sage should be more flexible in its dependencies where 
> possible/necessary.  With most Python packages it's easy as most have 
> a .__version__ and its not so hard to define some variable 
> like IS_RPY_2 and just have two separate branches.  I have things like 
> that all over the place in other packages to support e.g. different 
> Numpy versions or work around version-specific bugs. 
>

I've opened https://trac.sagemath.org/ticket/28988 for rpy. But at this 
point the major issues are python 3.8 and ipython 7, and I don't see how 
one could support several versions of them without forking hundreds of 
doctests. Both updates require multi-thousand-lines patches, due to changes 
in dict sorting and hashes.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/cade379b-146b-46c9-b057-c8b66570553b%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Vincent Delecroix]

Le 10/01/2020 à 14:54, E. Madison Bray a écrit :

On Fri, Jan 10, 2020 at 11:49 AM Isuru Fernando  wrote:



On Fri, Jan 10, 2020 at 3:53 PM E. Madison Bray  wrote:


On Fri, Jan 10, 2020 at 10:41 AM Timo Kaufmann  wrote:


I have said this before, but I feel like the point was dropped out of the discussion so 
I'll stress it again. The major issue here is *not* the compatibility of sage's own 
codebase. A few "from __future__ import"'s are not so bad.

The issue is that python2 compatibility forces us to use outdated versions of a 
lot of libraries, since many libraries have dropped python2 support a while 
ago. This is a big headache especially for packagers. Those outdated libraries 
are generally not available on distros. At the same time sage is usually not 
compatible with the newer versions. Sage is already difficult to package, and 
that makes it a lot more difficult.


Can you be more specific about this?  What is it about Sage's upstream
codebase maintaining backwards-compatibility for Python 2 that
prevents you from packaging it for Python 3 only, given that it does
support Python 3?  No one is saying that just because upstream support
is maintained for Python 2 for one or two (at the most) more releases,
any downstream packagers have to package it for Python 2.



I'll give an example. sage has rpy2 2.8.2.
rpy2 latest is 3.2.4 but 3.x series is python 3 only and sage requires breaking 
changes to support rpy2 3.x. See 
https://trac.sagemath.org/ticket/28494#comment:6
While python2 support for rpy2 is required, sage codebase can't be updated to 
3.x (unless someone adds code to detect rpy2 version and support both 2.8.2 and 
3.2.4)


That seems like the obvious approach to me.  As it is I've long felt
that Sage should be more flexible in its dependencies where
possible/necessary.  With most Python packages it's easy as most have
a .__version__ and its not so hard to define some variable
like IS_RPY_2 and just have two separate branches.  I have things like
that all over the place in other packages to support e.g. different
Numpy versions or work around version-specific bugs.



+1. I also do that with many Python packages that I develop and depend
on Sage. This is the only way I found to support multiple Sage versions.

--
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/5f87ae84-023d-4cbf-e8fc-75f894ee543a%40gmail.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[E. Madison Bray]

On Fri, Jan 10, 2020 at 11:49 AM Isuru Fernando  wrote:
>
>
> On Fri, Jan 10, 2020 at 3:53 PM E. Madison Bray  wrote:
>>
>> On Fri, Jan 10, 2020 at 10:41 AM Timo Kaufmann  wrote:
>> >
>> > I have said this before, but I feel like the point was dropped out of the 
>> > discussion so I'll stress it again. The major issue here is *not* the 
>> > compatibility of sage's own codebase. A few "from __future__ import"'s are 
>> > not so bad.
>> >
>> > The issue is that python2 compatibility forces us to use outdated versions 
>> > of a lot of libraries, since many libraries have dropped python2 support a 
>> > while ago. This is a big headache especially for packagers. Those outdated 
>> > libraries are generally not available on distros. At the same time sage is 
>> > usually not compatible with the newer versions. Sage is already difficult 
>> > to package, and that makes it a lot more difficult.
>>
>> Can you be more specific about this?  What is it about Sage's upstream
>> codebase maintaining backwards-compatibility for Python 2 that
>> prevents you from packaging it for Python 3 only, given that it does
>> support Python 3?  No one is saying that just because upstream support
>> is maintained for Python 2 for one or two (at the most) more releases,
>> any downstream packagers have to package it for Python 2.
>
>
> I'll give an example. sage has rpy2 2.8.2.
> rpy2 latest is 3.2.4 but 3.x series is python 3 only and sage requires 
> breaking changes to support rpy2 3.x. See 
> https://trac.sagemath.org/ticket/28494#comment:6
> While python2 support for rpy2 is required, sage codebase can't be updated to 
> 3.x (unless someone adds code to detect rpy2 version and support both 2.8.2 
> and 3.2.4)

That seems like the obvious approach to me.  As it is I've long felt
that Sage should be more flexible in its dependencies where
possible/necessary.  With most Python packages it's easy as most have
a .__version__ and its not so hard to define some variable
like IS_RPY_2 and just have two separate branches.  I have things like
that all over the place in other packages to support e.g. different
Numpy versions or work around version-specific bugs.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAOTD34a9wVPCu7LPcH_onCvCaViY70dSUH-DJxR1M5PbHYS%2BFQ%40mail.gmail.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Isuru Fernando]

On Fri, Jan 10, 2020 at 3:53 PM E. Madison Bray 
wrote:

> On Fri, Jan 10, 2020 at 10:41 AM Timo Kaufmann 
> wrote:
> >
> > I have said this before, but I feel like the point was dropped out of
> the discussion so I'll stress it again. The major issue here is *not* the
> compatibility of sage's own codebase. A few "from __future__ import"'s are
> not so bad.
> >
> > The issue is that python2 compatibility forces us to use outdated
> versions of a lot of libraries, since many libraries have dropped python2
> support a while ago. This is a big headache especially for packagers. Those
> outdated libraries are generally not available on distros. At the same time
> sage is usually not compatible with the newer versions. Sage is already
> difficult to package, and that makes it a lot more difficult.
>
> Can you be more specific about this?  What is it about Sage's upstream
> codebase maintaining backwards-compatibility for Python 2 that
> prevents you from packaging it for Python 3 only, given that it does
> support Python 3?  No one is saying that just because upstream support
> is maintained for Python 2 for one or two (at the most) more releases,
> any downstream packagers have to package it for Python 2.
>

I'll give an example. sage has rpy2 2.8.2.
rpy2 latest is 3.2.4 but 3.x series is python 3 only and sage requires
breaking changes to support rpy2 3.x. See
https://trac.sagemath.org/ticket/28494#comment:6
While python2 support for rpy2 is required, sage codebase can't be updated
to 3.x (unless someone adds code to detect rpy2 version and support both
2.8.2 and 3.2.4)

Isuru


> --
> You received this message because you are subscribed to the Google Groups
> "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sage-devel+unsubscr...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/sage-devel/CAOTD34ZqHf%2BrzoSD8juhhk%3D4YF1WB6BHZuxAOO0ZQYk6KhhX_g%40mail.gmail.com
> .
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CA%2B01voMi8UybPwJMB_dq9RfeULarnQ_pWrXj09Fxbc%3D2_kiviA%40mail.gmail.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Vincent Delecroix]

Le 06/01/2020 à 00:24, Nils Bruin a écrit :

On Sunday, January 5, 2020 at 2:51:44 PM UTC-8, Nils Bruin wrote:


I don't think there's a doubt that we'll drop py2.7 support. This Wiki,
however, has been referred to in many places describing the py2/py3
transition.



One of these places is the FAQ, which is official documentation:

http://doc.sagemath.org/html/en/faq/faq-usage.html#can-i-use-sagemath-with-python-3-x

The 9.0 FAQ still mentions:

As of August 2019, most of SageMath works fine with Python 3. However, we
still consider Python 3 support to be experimental and no official Python 3
release has been made yet.

That would probably be good to adjust too: there is now an official Py3
release and I don't think we consider it experimental anymore -- unless
sage in its entirety is still considered experimental :-).



See https://trac.sagemath.org/ticket/28980 (even though Frédéric claims
that it might have been done in another ticket).

--
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/126b6e30-8f93-333e-a436-2ae5015a0dcd%40gmail.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[E. Madison Bray]

On Fri, Jan 10, 2020 at 10:41 AM Timo Kaufmann  wrote:
>
> I have said this before, but I feel like the point was dropped out of the 
> discussion so I'll stress it again. The major issue here is *not* the 
> compatibility of sage's own codebase. A few "from __future__ import"'s are 
> not so bad.
>
> The issue is that python2 compatibility forces us to use outdated versions of 
> a lot of libraries, since many libraries have dropped python2 support a while 
> ago. This is a big headache especially for packagers. Those outdated 
> libraries are generally not available on distros. At the same time sage is 
> usually not compatible with the newer versions. Sage is already difficult to 
> package, and that makes it a lot more difficult.

Can you be more specific about this?  What is it about Sage's upstream
codebase maintaining backwards-compatibility for Python 2 that
prevents you from packaging it for Python 3 only, given that it does
support Python 3?  No one is saying that just because upstream support
is maintained for Python 2 for one or two (at the most) more releases,
any downstream packagers have to package it for Python 2.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAOTD34ZqHf%2BrzoSD8juhhk%3D4YF1WB6BHZuxAOO0ZQYk6KhhX_g%40mail.gmail.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Timo Kaufmann]

I have said this before, but I feel like the point was dropped out of the 
discussion so I'll stress it again. The major issue here is *not* the 
compatibility of sage's own codebase. A few "from __future__ import"'s are 
not so bad.

The issue is that python2 compatibility forces us to use outdated versions 
of a lot of libraries, since many libraries have dropped python2 support a 
while ago. This is a big headache especially for packagers. Those outdated 
libraries are generally not available on distros. At the same time sage is 
usually not compatible with the newer versions. Sage is already difficult 
to package, and that makes it a lot more difficult.

Case in point, my own efforts to update sage on nixos to 9.0 and python3 
have been blocked by incompatibilities in a new sphinx version 
(https://trac.sagemath.org/ticket/28856). I have already put some hours 
into this, but I simply don't have time to maintain a sage package if all 
the compatibility efforts fall to packagers. Another example is Antonio's 
multiple thousand line patch he mentioned earlier.

The alternatives are:

(1) create infrastructure in sage that allows us to use newer dependencies 
in python3 (https://trac.sagemath.org/ticket/28190), and make sage 
compatible with both the py2 and py3 version. This is a lot of effort.

(2) drop python2 support ASAP, e.g. in 9.1, and make sage compatible with 
the newer libraries

(3) continue as-is and leave it to packers to "deal with it"

As I said, (1) is a lot of effort. It would be the nicest solution, but I 
don't think there are enough people willing to do the work quickly enough. 
Especially if that effort will be wasted in a couple of months' time once 
py2 support is dropped.

(2) is my favorite solution. It re-synchronizes upstream with downstream 
distributions (which will most likely no longer package the python2 version 
anyways). I see very little downsides to this. Yes, some people will need 
longer to switch over their codebases. But their existing codebases, having 
been written before 9.1, will work with sage 9.0. So why not continue using 
it as long as the porting efforts take?

And again, I think (3) is a very bad solution. Sage already takes many 
times the packaging effort than most other packages. If this effort 
increases, I won't be willing or able to keep it up.

Sorry for the rant, I hope the email didn't get too long for people to 
read. 

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/288c5857-c4ae-43e9-8d0d-bb67319e470f%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[E. Madison Bray]

On Fri, Jan 10, 2020 at 1:03 AM Volker Braun  wrote:
>
> * I think its not too difficult to write code that is Python 2.7 + 3.x (for 
> high enough x) compatible, so its not a super pressing issue
> * We do have a Python 2 buildbot to test for regressions
> * For semver reasons we should drop Python 2.7 support in Sage 10, not 9.1
>
> Having said that, I'm fine with an accelerated Sage 9 -> Sage 10 schedule, 
> maybe a month or two instead of the usual 3-4. Though first we should take 
> the opportunity and see if there are any outstanding Python 3 bugs now that 
> we have more data. For example it would be nice if a build with 
> SAGE_DEBUG=yes would pass tests. There are a few more regressions, e.g. #28817

Accelerated is fine, but I should add that fully *removing* Python 2
support is not a trivial task either (it's mostly mechanical work, but
there's a LOT of it).  https://trac.sagemath.org/ticket/28000 has a
few steps towards it but is far from complete.

Fully removing Python 2-isms from the codebase is not necessarily a
prerequisite to dropping Python 2 support, but I would be hesitant to
leave a monstrous partially-hybrid codebase around for too long.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAOTD34Z%3DpXaDP5ByS2pAf45HgL-0gOkrTdcGFzSZsAc24rgR%3Dw%40mail.gmail.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Volker Braun]

* I think its not too difficult to write code that is Python 2.7 + 3.x (for 
high enough x) compatible, so its not a super pressing issue
* We do have a Python 2 buildbot to test for regressions
* For semver reasons we should drop Python 2.7 support in Sage 10, not 9.1

Having said that, I'm fine with an accelerated Sage 9 -> Sage 10 schedule, 
maybe a month or two instead of the usual 3-4. Though first we should take 
the opportunity and see if there are any outstanding Python 3 bugs now that 
we have more data. For example it would be nice if a build with 
SAGE_DEBUG=yes would pass tests. There are a few more regressions, e.g. 
#28817


-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/4e7cd06a-2d73-4530-b235-d4f675cd9cbd%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Dima Pasechnik]

On Tue, Jan 7, 2020 at 9:45 PM Sébastien Labbé  wrote:
>
> The question I have is the folllowing. Suppose a user have 10k lines of 
> SageMath code (notebooks, files, etc.) working well in 8.9 which obviously 
> will get broken in the default 9.0 mostly because of:

notebooks as in sagenb?
This would really be not fun at all, as sagenb does not run on py3.


>  - print
>  - comparisons
>  - iterkeys, iteritems
>  - and few other particular changes
>
> 1. What is the best workflow for such users to adapt to the 9.0, 9.1, etc. 
> series?
>
> 2. How can the version 9.1 of SageMath be made so that it is really helpful?
>
> 3. How can a version 9.1 or 9.0 running Python 2 can be more helpful than the 
> version 8.9 running Python 2? (Let's recall that the old code was running in 
> 8.9 so it does not use any of the new features added in 9.0)

We did see one-liners working in 8.9 and broken in 9.0, to do with graphics.
Every new release not only adds features, it fixed bugs (and
introduces new ones, obviously :-))
Thus, the closer you are to the current release, the better.

>
> Sébastien
>
> --
> You received this message because you are subscribed to the Google Groups 
> "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to sage-devel+unsubscr...@googlegroups.com.
> To view this discussion on the web visit 
> https://groups.google.com/d/msgid/sage-devel/bcd007fd-7b22-4d5d-9aee-44005ebd684b%40googlegroups.com.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAAWYfq3FVs4UfWaagUpyojn8xX8pQF5GD0SBT6uHyh%3DHiVH5vg%40mail.gmail.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Sébastien Labbé]

The question I have is the folllowing. Suppose a user have 10k lines of 
SageMath code (notebooks, files, etc.) working well in 8.9 which obviously 
will get broken in the default 9.0 mostly because of:
 - print
 - comparisons
 - iterkeys, iteritems
 - and few other particular changes

1. What is the best workflow for such users to adapt to the 9.0, 9.1, etc. 
series?

2. How can the version 9.1 of SageMath be made so that it is really 
helpful? 

3. How can a version 9.1 or 9.0 running Python 2 can be more helpful than 
the version 8.9 running Python 2? (Let's recall that the old code was 
running in 8.9 so it does not use any of the new features added in 9.0)

Sébastien

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/bcd007fd-7b22-4d5d-9aee-44005ebd684b%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[kcrisman]

On Tuesday, January 7, 2020 at 2:17:54 PM UTC-5, Jonathan Kliem wrote:
>
> Just a silly question. How do we know that the code remains python2 
> compatible in the first place?
>
> As far as I can see all the patchbots run python3 so it is very easy to 
> break something. 9.0 is supposed to be fully python2 compatible, but are 
> the doctests being tested for that?
>
>
I think on sage-release at least a few people are testing for this, or have 
been, at any rate.
 

> If we don't test the upcoming tickets for python2 compatibility, we have 
> de facto stopped supporting python2 already.
>
>
If we have a few people run with the python 2 config for 9.1 I think that 
is reasonable.  We don't claim the betas are py2 compatible. 

But yes, if we were to support both long-term (not that we are), one would 
want that to be the case.

I'll once again ask if someone knows how I might (easily) make a py2 binary 
for Sage 9.0 using binary-pkg.  I don't think I can necessarily use the 
configure command before running that since everything happens "inside" of 
binary-pkg.

Thanks,
- kcrisman

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/6701320a-97a9-4c0b-bfdf-f800798f6b18%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

['Jonathan Kliem' via sage-devel]

Just a silly question. How do we know that the code remains python2 
compatible in the first place?

As far as I can see all the patchbots run python3 so it is very easy to 
break something. 9.0 is supposed to be fully python2 compatible, but are 
the doctests being tested for that?

If we don't test the upcoming tickets for python2 compatibility, we have de 
facto stopped supporting python2 already.

Jonathan

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/e1faa06c-4935-4080-9895-96ae955948bc%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Dima Pasechnik]

On Tue, 7 Jan 2020, 06:04 William,  wrote:

> This is just some clarification and remarks related to what Karl wrote,
> and not really super relevant to this thread...
>
> On Monday, January 6, 2020 at 7:17:16 AM UTC-8, kcrisman wrote:
>>
>> Because a lot of Sage users are also developers (presumably a higher
>> proportion than for Python or Mathematica!), a lot of Sage work has had
>> developers in mind.  However, I think it is pretty important to pay
>> attention to the users, the vast majority of whom we don't ever hear from,
>> since they aren't paying for Sage.
>>
>> As an example, if there were a coordinated strategy that a company would
>> make, I can't imagine that there wouldn't have been two versions of Sage
>> cell server going at once - or even an addition (temporary) to the
>> drop-down menu saying "Sage with Python 2", for transitioning.
>>
>
> For CoCalc (run by a company), we have many different versions of Sage
> installed for projects to use.  For Jupyter  notebooks, users just
> explicitly select whatever version they want (via kernel selection), and it
> stays selected unless they change it.  For the sage command line, we have a
> "sage_select" command to set the default, or the user can type
> "sage-[version]" to get a specific version.
>
> /ext/bin/sage_select 
> Available versions: develop, 8.2, 8.3, 8.4, 8.5, 8.6, 8.7, 8.8, 8.9, 9.0
>

you could add something like 9.0_py2, as 9.0 is still meant to be fully
functional with Python 2.

>
> For Sage worksheets, they just use whatever is "sage" in the PATH, which
> is a problem since there's a TON of python2 content used on cocalc, much
> made by paying customers (with thousands of students), so we can't break
> it.  We've been discussing how to move forward, slowly but surely, here
> https://github.com/sagemathinc/cocalc/issues/4212/. It's a confusing and
> tricky problem because Sage worksheets don't encode the version of Sage
> they use (dumb design decision by me) -- Jupyter is much better in this
> regard.   In any case, we don't have the luxury of break things for our
> paying customers.
>
>
>> I don't know if I want to ask the very capable, but very busy, Andrey N.
>> to do that!  But once again, sage-devel will never know, because the users
>> didn't pay anything,
>>
>
> Users (at least UTMOST) does sometimes pay a little bit to cover some
> hosting (thanks NSF!), but it doesn't cover what Andrey does, and I fully
> support his decision.  I wish I had the money to pay to support maintenance
> of other versions of the Sage cell server, but I don't.
>
> William
>
> --
> You received this message because you are subscribed to the Google Groups
> "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sage-devel+unsubscr...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/sage-devel/99d1c833-d393-46b0-a7d3-72e184ba047b%40googlegroups.com
> 
> .
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAAWYfq3DNNNO_5hdv6KgRwQpk2ZCkP1shaArgkseBw%3DAGG9f_Q%40mail.gmail.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[E. Madison Bray]

On Tue, Jan 7, 2020 at 2:16 PM Eric Gourgoulhon  wrote:
>
> Le mardi 7 janvier 2020 13:25:04 UTC+1, E. Madison Bray a écrit :
>>
>> On Mon, Jan 6, 2020 at 7:30 PM Eric Gourgoulhon  wrote:
>> >
>> > On the other hand, for the end user the major backwards-incompatibility 
>> > change already happened: a Python 2-only piece of code will break 
>> > immediately in any Sage 9.0 binary. Can we really say to the end user: "to 
>> > solve your issue, download SageMath sources and compile them with 
>> > ./configure --with-python=2" ?
>> > As for developers, the Python 3 switch has been discussed for something 
>> > like 2 years, so what would be the point to extend that (effective) 
>> > deprecation period? (maybe I am missing something here)
>>
>> For this very reason I think there ought to be Python 2 binary
>> releases for 9.0 as well.  I'm building both for Windows, but I'm not
>> in control of the others.
>
>
> One may argue that there are already available Python 2 binaries: the 8.9 
> binaries.
> Is it worth to spend time and energy to build new Python 2 binaries?
> Wouldn't a message like "if you insist in running Python 2-only code, please 
> use the 8.9 binaries" be sufficient?

If that were the case, that should have been communicated clearly
during the 8.9 release, and it wasn't (as it is we don't communicate
changes between versions clearly enough, but that would be a major
omission).

> IMHO, most end users should now that Python 2 is dead (the younger ones even 
> do not know that such a thing had existed) and that "print bla" should be 
> changed to "print(bla)".

Most users I've encountered (especially new users) don't even realize
Sage is written in Python, and don't know anything about Python 2 vs
Python 3.  Fortunately, as you say, for Sage the most major
user-visible change is just print statements and we've already been
warning about that for a while.  But existing users' code can break in
other new an exciting ways (e.g. map).

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAOTD34bJCd8TNy%2Bv7M%3DWcbynkpq-cXtcBsn7a%2BGpHgwEroUovw%40mail.gmail.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Eric Gourgoulhon]

Le mardi 7 janvier 2020 14:35:45 UTC+1, Dima Pasechnik a écrit :
>
> Why 8.9? One can build 9.0 with python2 just fine. 
>
> Yes I know, but I was speaking about the time and energy to actually build 
those and distribute them. IMHO, there are more pressing issues, like the 
handling of ipython7 and Python 3.8 mentioned in Frédéric's original 
message. 
That being said, I am not involved in the preparation of Sage binaries and 
your voice or that of Erik is more important than mine. 

Eric. 

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/625eaa0f-0224-4109-a5a5-a20c059a364d%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Dima Pasechnik]

Why 8.9? One can build 9.0 with python2 just fine.

On Tue, Jan 7, 2020 at 1:16 PM Eric Gourgoulhon  wrote:
>
> Le mardi 7 janvier 2020 13:25:04 UTC+1, E. Madison Bray a écrit :
>>
>> On Mon, Jan 6, 2020 at 7:30 PM Eric Gourgoulhon  wrote:
>> >
>> > On the other hand, for the end user the major backwards-incompatibility 
>> > change already happened: a Python 2-only piece of code will break 
>> > immediately in any Sage 9.0 binary. Can we really say to the end user: "to 
>> > solve your issue, download SageMath sources and compile them with 
>> > ./configure --with-python=2" ?
>> > As for developers, the Python 3 switch has been discussed for something 
>> > like 2 years, so what would be the point to extend that (effective) 
>> > deprecation period? (maybe I am missing something here)
>>
>> For this very reason I think there ought to be Python 2 binary
>> releases for 9.0 as well.  I'm building both for Windows, but I'm not
>> in control of the others.
>
>
> One may argue that there are already available Python 2 binaries: the 8.9 
> binaries.
> Is it worth to spend time and energy to build new Python 2 binaries?
> Wouldn't a message like "if you insist in running Python 2-only code, please 
> use the 8.9 binaries" be sufficient?
> IMHO, most end users should now that Python 2 is dead (the younger ones even 
> do not know that such a thing had existed) and that "print bla" should be 
> changed to "print(bla)".
>
> Eric.
>
> --
> You received this message because you are subscribed to the Google Groups 
> "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to sage-devel+unsubscr...@googlegroups.com.
> To view this discussion on the web visit 
> https://groups.google.com/d/msgid/sage-devel/98c43f60-8613-4a09-b244-51b11a85cfd4%40googlegroups.com.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAAWYfq3epLQdgGgbbo2w5D7BkdgiAN47bixfUOCN8CmUU%2By5Og%40mail.gmail.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Eric Gourgoulhon]

Le mardi 7 janvier 2020 13:25:04 UTC+1, E. Madison Bray a écrit :
>
> On Mon, Jan 6, 2020 at 7:30 PM Eric Gourgoulhon  > wrote: 
> > 
> > On the other hand, for the end user the major backwards-incompatibility 
> change already happened: a Python 2-only piece of code will break 
> immediately in any Sage 9.0 binary. Can we really say to the end user: "to 
> solve your issue, download SageMath sources and compile them with 
> ./configure --with-python=2" ? 
> > As for developers, the Python 3 switch has been discussed for something 
> like 2 years, so what would be the point to extend that (effective) 
> deprecation period? (maybe I am missing something here) 
>
> For this very reason I think there ought to be Python 2 binary 
> releases for 9.0 as well.  I'm building both for Windows, but I'm not 
> in control of the others. 
>

One may argue that there are already available Python 2 binaries: the 8.9 
binaries. 
Is it worth to spend time and energy to build new Python 2 binaries?
Wouldn't a message like "if you insist in running Python 2-only code, 
please use the 8.9 binaries" be sufficient?
IMHO, most end users should now that Python 2 is dead (the younger ones 
even do not know that such a thing had existed) and that "print bla" should 
be changed to "print(bla)". 

Eric. 

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/98c43f60-8613-4a09-b244-51b11a85cfd4%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[kcrisman]

William, thanks for this relevant (if not directly, as you say) point of 
view.

For CoCalc (run by a company), we have many different versions of Sage 
> installed for projects to use.  For Jupyter  notebooks, users just 
> explicitly select whatever version they want (via kernel selection), and it 
> stays selected unless they change it.  For the sage command line, we have a 
> "sage_select" command to set the default, or the user can type 
> "sage-[version]" to get a specific version.
>
> /ext/bin/sage_select 
> Available versions: develop, 8.2, 8.3, 8.4, 8.5, 8.6, 8.7, 8.8, 8.9, 9.0
>
> For Sage worksheets, they just use whatever is "sage" in the PATH, which 
> is a problem since there's a TON of python2 content used on cocalc, much 
> made by paying customers (with thousands of students), so we can't break 
> it.  We've been discussing how to move forward, slowly but surely, here 
> https://github.com/sagemathinc/cocalc/issues/4212/. It's a confusing and 
> tricky problem because Sage worksheets don't encode the version of Sage 
> they use (dumb design decision by me) -- Jupyter is much better in this 
> regard.   In any case, we don't have the luxury of break things for our 
> paying customers.
>

Exactly.  By the way, even 8.2 - impressive!
 

>  
>
>> I don't know if I want to ask the very capable, but very busy, Andrey N. 
>> to do that!  But once again, sage-devel will never know, because the users 
>> didn't pay anything, 
>>
>
> Users (at least UTMOST) does sometimes pay a little bit to cover some 
> hosting (thanks NSF!), but it doesn't cover what Andrey does, and I fully 
> support his decision. 
>

Of course, and I hope that was clear in what I said.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/1a5ab514-ae17-4da9-a67e-6e60668fd1d1%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[E. Madison Bray]

On Tue, Jan 7, 2020 at 12:06 AM rjf  wrote:
>
> just curious when this ends.  Python 4 awaits.

It's already ended.  Python 4 is not going to be the major
backwards-compatibility breaker that Python 3 was.  It's just going to
be the next release after 3.9, at which point we can also hopefully
finally stop talking about "Python 3" and just call it "Python" again.
This is similar to how Jinja2 is called Jinja2 because it was
completely incompatible with Jinja 1.x and so needed a "different
name" as it were.  But now that Jinja2 is long-since the de facto
Jinja, Jinja 3.0 will just be referred to as "Jinja" again; no more
Jinja2.

In the more extreme end of things Perl 6 has finally stopped being
"Perl" altogether and is called Raku now.  Fortunately even Python 3
is not as big a "disaster" from a backwards-incompatibility
perspective that it hasn't necessitated a completely new name for the
language; the "Python 3" distinction has only been relevant in the
context of porting from Python 2 and now that that's mostly done in
most of the community we can finally start to drop it soon...

> On Monday, January 6, 2020 at 10:47:29 AM UTC-8, Nils Bruin wrote:
>>
>> On Monday, January 6, 2020 at 10:30:23 AM UTC-8, Eric Gourgoulhon wrote:
>>>
>>>
>>> On the other hand, for the end user the major backwards-incompatibility 
>>> change already happened: a Python 2-only piece of code will break 
>>> immediately in any Sage 9.0 binary. Can we really say to the end user: "to 
>>> solve your issue, download SageMath sources and compile them with 
>>> ./configure --with-python=2" ?
>>
>>
>> That's a different issue. If we can't say that, the appropriate fix is to 
>> host py2 binaries as well (but hide them a little bit).
>
> --
> You received this message because you are subscribed to the Google Groups 
> "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to sage-devel+unsubscr...@googlegroups.com.
> To view this discussion on the web visit 
> https://groups.google.com/d/msgid/sage-devel/c372b39f-6232-48a5-bc98-e452c905ab1c%40googlegroups.com.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAOTD34aFMAwGRF8kZTJRRCi_jfTxcNLYmEtGpMNfXApEHHiiEQ%40mail.gmail.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[E. Madison Bray]

On Mon, Jan 6, 2020 at 7:30 PM Eric Gourgoulhon  wrote:
>
> Le lundi 6 janvier 2020 14:21:56 UTC+1, E. Madison Bray a écrit :
>>
>>
>> I agree with Nils.  There should be at least a one release deprecation
>> period.  Also, while I don't think we use any kind of real semantic
>> versioning, I think we should name a Python 3-only release 10.0 as
>> it's a very major backwards-incompatibility change.
>>
>
> On the other hand, for the end user the major backwards-incompatibility 
> change already happened: a Python 2-only piece of code will break immediately 
> in any Sage 9.0 binary. Can we really say to the end user: "to solve your 
> issue, download SageMath sources and compile them with ./configure 
> --with-python=2" ?
> As for developers, the Python 3 switch has been discussed for something like 
> 2 years, so what would be the point to extend that (effective) deprecation 
> period? (maybe I am missing something here)

For this very reason I think there ought to be Python 2 binary
releases for 9.0 as well.  I'm building both for Windows, but I'm not
in control of the others.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAOTD34YHWWzviONKTikBSh_vqCmxt6pme6DCQb%3D7XXU0iqdELg%40mail.gmail.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[E. Madison Bray]

On Mon, Jan 6, 2020 at 7:23 PM Travis Scrimshaw  wrote:
>
> I agree that the Python3 only version of Sage should be called 10.0 given the 
> large backwards incompatible changes that result from the Python3 change. 
> Furthermore, I also concur that we should release a version 9.1 (on an 
> accelerated schedule) that is our official deprecation version telling people 
> they will need to adapt their code in the next version.

This would be the right thing to do from a software maintenance
standpoint and from a user-facing standpoint.  The vast majority of
Sage users are NOT developers and this community has too much a habit
in general of moving fast and breaking things for the short-term
benefit of power-users/developers (who in general are more able to run
their own builds as needed).

As it is I think we rushed too fast into a Python 3 default release
but in that case I think it was necessary and unavoidable.  Now that
that's out (grace à Frédéric) we can take a minute to reflect on how
that transition goes for users and make the first Python 3-only
release even stronger.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAOTD34b4vSqeaoTaQaFMO5V-RjKfTahYZ%3DSDJS1oE_NET68jBg%40mail.gmail.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[William]

This is just some clarification and remarks related to what Karl wrote, and 
not really super relevant to this thread...

On Monday, January 6, 2020 at 7:17:16 AM UTC-8, kcrisman wrote:
>
> Because a lot of Sage users are also developers (presumably a higher 
> proportion than for Python or Mathematica!), a lot of Sage work has had 
> developers in mind.  However, I think it is pretty important to pay 
> attention to the users, the vast majority of whom we don't ever hear from, 
> since they aren't paying for Sage.
>
> As an example, if there were a coordinated strategy that a company would 
> make, I can't imagine that there wouldn't have been two versions of Sage 
> cell server going at once - or even an addition (temporary) to the 
> drop-down menu saying "Sage with Python 2", for transitioning.
>

For CoCalc (run by a company), we have many different versions of Sage 
installed for projects to use.  For Jupyter  notebooks, users just 
explicitly select whatever version they want (via kernel selection), and it 
stays selected unless they change it.  For the sage command line, we have a 
"sage_select" command to set the default, or the user can type 
"sage-[version]" to get a specific version.

/ext/bin/sage_select 
Available versions: develop, 8.2, 8.3, 8.4, 8.5, 8.6, 8.7, 8.8, 8.9, 9.0

For Sage worksheets, they just use whatever is "sage" in the PATH, which is 
a problem since there's a TON of python2 content used on cocalc, much made 
by paying customers (with thousands of students), so we can't break it.  
We've been discussing how to move forward, slowly but surely, here 
https://github.com/sagemathinc/cocalc/issues/4212/. It's a confusing and 
tricky problem because Sage worksheets don't encode the version of Sage 
they use (dumb design decision by me) -- Jupyter is much better in this 
regard.   In any case, we don't have the luxury of break things for our 
paying customers.
 

> I don't know if I want to ask the very capable, but very busy, Andrey N. 
> to do that!  But once again, sage-devel will never know, because the users 
> didn't pay anything, 
>

Users (at least UTMOST) does sometimes pay a little bit to cover some 
hosting (thanks NSF!), but it doesn't cover what Andrey does, and I fully 
support his decision.  I wish I had the money to pay to support maintenance 
of other versions of the Sage cell server, but I don't.
 
William

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/99d1c833-d393-46b0-a7d3-72e184ba047b%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[rjf]

just curious when this ends.  Python 4 awaits.  
RJF

On Monday, January 6, 2020 at 10:47:29 AM UTC-8, Nils Bruin wrote:
>
> On Monday, January 6, 2020 at 10:30:23 AM UTC-8, Eric Gourgoulhon wrote:
>>
>>
>> On the other hand, for the end user the major backwards-incompatibility 
>> change already happened: a Python 2-only piece of code will break 
>> immediately in any Sage 9.0 binary. Can we really say to the end user: "to 
>> solve your issue, download SageMath sources and compile them with 
>> ./configure --with-python=2" ?
>>
>
> That's a different issue. If we can't say that, the appropriate fix is to 
> host py2 binaries as well (but hide them a little bit).
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/c372b39f-6232-48a5-bc98-e452c905ab1c%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Nils Bruin]

On Monday, January 6, 2020 at 10:30:23 AM UTC-8, Eric Gourgoulhon wrote:
>
>
> On the other hand, for the end user the major backwards-incompatibility 
> change already happened: a Python 2-only piece of code will break 
> immediately in any Sage 9.0 binary. Can we really say to the end user: "to 
> solve your issue, download SageMath sources and compile them with 
> ./configure --with-python=2" ?
>

That's a different issue. If we can't say that, the appropriate fix is to 
host py2 binaries as well (but hide them a little bit).

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/5063b25b-a348-4b41-8d5f-3a68e3eb8ec5%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[Eric Gourgoulhon]

Le lundi 6 janvier 2020 14:21:56 UTC+1, E. Madison Bray a écrit :
>
>
> I agree with Nils.  There should be at least a one release deprecation 
> period.  Also, while I don't think we use any kind of real semantic 
> versioning, I think we should name a Python 3-only release 10.0 as 
> it's a very major backwards-incompatibility change. 
>
>
On the other hand, for the end user the major backwards-incompatibility 
change already happened: a Python 2-only piece of code will break 
immediately in any Sage 9.0 binary. Can we really say to the end user: "to 
solve your issue, download SageMath sources and compile them with 
./configure --with-python=2" ?
As for developers, the Python 3 switch has been discussed for something 
like 2 years, so what would be the point to extend that (effective) 
deprecation period? (maybe I am missing something here)

Eric.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/e88ab0aa-63f7-4635-b927-2b04741789c1%40googlegroups.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Travis Scrimshaw]

I agree that the Python3 only version of Sage should be called 10.0 given 
the large backwards incompatible changes that result from the Python3 
change. Furthermore, I also concur that we should release a version 9.1 (on 
an accelerated schedule) that is our official deprecation version telling 
people they will need to adapt their code in the next version.

Best,
Travis


-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/04cbe9fc-0d8f-4e9c-869a-282f3a37961b%40googlegroups.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Andrey Novoseltsev]

On Monday, 6 January 2020 03:13:02 UTC-7, Frédéric Chapoton wrote:
>
> the wiki promise was written by myself alone, and I think it should be 
> changed. I think there will be very few (or  no) people that will desire to 
> compile sage 9.0 with python 2. And 9.1 even more. People that want to be 
> up to date do not care about py2.
> I therefore strongly vote for the immediate drop of python2, and for 
> removing the promise that 9.1 will be py2-compatible
>

I think that is the most important voice so far. When a single person on 
his own accord made a statement and posted it on a wiki page it should not 
be considered an official and unbreakable promise. When that same person 
wants to pull it back, and for a good reason, why should not we allow this?

We can talk about what would be nice, like running multiple cell servers, 
but let's pay attention to a) what is really happening and b) what are 
people actually willing to do in the future. I am personally not willing to 
work on multiple cell servers - those who really want and have machines 
available can quite easily set up their own server, I have tried hard to 
make it happen a long time ago. Others HAVE to switch to Python 3 sooner or 
later no matter what. Well, that time is now. I've timed the switch as best 
as I could releasing Python 3 version on Christmas and giving those 
preparing for the new term two weeks to adjust. Even with running another 
server some adjustments would be necessary.

Having a deprecation for print statement a year ago would be very useful 
and it was discussed and agreed on - as far as usefulness goes. But nobody 
actually worked on that change, which, I suspect, would be much easier than 
maintaining compatibility with two Pythons now.

I want to reiterate again that I am personally not affected by the decision 
made here in any way. But after years of adjusting SageMathCell to new 
versions of Sage, IPython, and now Python I think I have some idea of 
effort involved for those who do the porting. And I strongly believe that 
efforts of these people are too valuable to be spent on this support.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/1c27defa-db0d-4e0e-826a-7b82d620c5b5%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[kcrisman]

On Monday, January 6, 2020 at 8:21:56 AM UTC-5, E. Madison Bray wrote:
>
> On Sun, Jan 5, 2020 at 10:06 PM Nils Bruin > 
> wrote: 
> > 
> > I think our wiki vetoes that idea. See 
> https://wiki.sagemath.org/Python3-Switch : 
> > 
> > Compiling with Python 2 
> > 
> > After version 9.0, if you really want so, you can still build and use 
> SageMath with Python 2, as follows. 
> > 
> > make configure 
> > ./configure --with-python=2 
> > make build 
> > 
> > Beware that you will need to call the second line again if you ever call 
> "make distclean". 
> > 
> > This will work until version 9.1 at least. Then the backward 
> compatibility with Python 2 will no longer be ensured. 
> > 
> > If you want to drop py2 compatibility very soon, the only option is to 
> release 9.1 basically right now, identical to 9.0, and then get on with 
> developing 9.2. That's a nasty thing to do. Based on previous release 
> schedules, people would be justified in expecting that <=9.1 is the 
> "current" release until at least June 2020 or so. So we're stuck with py2 
> compatibility until that time. 
>
> I agree with Nils.  There should be at least a one release deprecation 
> period.  Also, while I don't think we use any kind of real semantic 
> versioning, I think we should name a Python 3-only release 10.0 as 
> it's a very major backwards-incompatibility change. 
>
> Instead for 9.1 let us display a prominent deprecation message in any 
> Python 2 builds when running Sage, including a link to a guide for 
> porting existing code (for Sage there is not all that much to 
> do--mostly any generic Python 2 to 3 porting guide will do, plus some 
> Sage-specific caveats of which I can't think of many). 
>


Because a lot of Sage users are also developers (presumably a higher 
proportion than for Python or Mathematica!), a lot of Sage work has had 
developers in mind.  However, I think it is pretty important to pay 
attention to the users, the vast majority of whom we don't ever hear from, 
since they aren't paying for Sage.

As an example, if there were a coordinated strategy that a company would 
make, I can't imagine that there wouldn't have been two versions of Sage 
cell server going at once - or even an addition (temporary) to the 
drop-down menu saying "Sage with Python 2", for transitioning. I don't know 
if I want to ask the very capable, but very busy, Andrey N. to do that! 
 But once again, sage-devel will never know, because the users didn't pay 
anything, so they may just decide to drop it.  I know it's a headache for 
any authors using PreTeXt who used "print" statements, because it's not 
just changing your own code, but releasing a new version of a book. 
 Presumably many many worksheets (sagenb or Jupyter) also are in the same 
boat - and one can say "oh, there was time" but we all know that many Sage 
users do not actually upgrade very frequently, because Sage is so usable 
and stable in the core functionality, so they may not have been at all 
aware of the necessity.  They just know their code is breaking.

This doesn't mean we shouldn't have switched to Py3 or Jupyter or whatever 
the backwards-incompatible flavor of the month is - that much is clear. 
 But the only way the community grows is by attracting people long enough 
to have a certain percentage become developers, and that does mean some 
boring "customer service" for our volunteer force.  I would call the work 
done at ask.sagemath.org and pretextbook.org actually very valuable in 
these terms, even if little of it goes directly toward developing new 
capabilities in Sage.  (CoCalc also performs a lot of this customer base 
development/service but since it is at least partly non-volunteer I put it 
in a different category.)

I think the easiest way out of this whole business actually is to have a 
relatively small bugfix 9.1 changeset, with OFFICIAL BINARIES in Python 2 
posted, and those left up indefinitely, clearly marked as "Python 2".   
(Having a choice of Sage cell server as well for some indeterminate period 
would be good, but again it is a lot to presume on Andrey alone for that.) 
 Does anyone know whether it is easy to use binary-pkg to create Python 2 
binaries - it is as simple as adding a line in the sage.yaml file, or would 
one have to do something on the command line e.g. an env var first?

- kcrisman

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/4293103a-5f9b-434c-a67a-8cf4aa257797%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[E. Madison Bray]

On Sun, Jan 5, 2020 at 10:06 PM Nils Bruin  wrote:
>
> I think our wiki vetoes that idea. See 
> https://wiki.sagemath.org/Python3-Switch :
>
> Compiling with Python 2
>
> After version 9.0, if you really want so, you can still build and use 
> SageMath with Python 2, as follows.
>
> make configure
> ./configure --with-python=2
> make build
>
> Beware that you will need to call the second line again if you ever call 
> "make distclean".
>
> This will work until version 9.1 at least. Then the backward compatibility 
> with Python 2 will no longer be ensured.
>
> If you want to drop py2 compatibility very soon, the only option is to 
> release 9.1 basically right now, identical to 9.0, and then get on with 
> developing 9.2. That's a nasty thing to do. Based on previous release 
> schedules, people would be justified in expecting that <=9.1 is the "current" 
> release until at least June 2020 or so. So we're stuck with py2 compatibility 
> until that time.

I agree with Nils.  There should be at least a one release deprecation
period.  Also, while I don't think we use any kind of real semantic
versioning, I think we should name a Python 3-only release 10.0 as
it's a very major backwards-incompatibility change.

Instead for 9.1 let us display a prominent deprecation message in any
Python 2 builds when running Sage, including a link to a guide for
porting existing code (for Sage there is not all that much to
do--mostly any generic Python 2 to 3 porting guide will do, plus some
Sage-specific caveats of which I can't think of many).

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAOTD34aWri4C-U4XqGHcO77UB6%2BNgAvvbofBtxfuWkgU4UGR4w%40mail.gmail.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Simon King]

Hi Nils,

On 2020-01-06, Nils Bruin  wrote:
> In fact, I think it would make a lot of sense to make 9.1 the last release 
> to support py2 and therefore make 9.1 more a bugfix release: concentrate on 
> fixes of issues that arise as a result of 9.0 and fix them, so that there's 
> a relatively reliable point in the history of sage to run on py2. That 
> would automatically target significant new features for 9.2.

Having a bugfix release as the last py-2-supporting release sounds like
a good plan to me. So, +1 to this.

Also I think that the additional burden to average developers will be
small: If (on top of bug fixes) there would be new code in v9.1, then
making it work with py-2 would basically involve some "from __future__
import", isn't it?

Best regards,
Simon

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/quv67d%2478qv%241%40blaine.gmane.org.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Frédéric Chapoton]

the wiki promise was written by myself alone, and I think it should be 
changed. I think there will be very few (or  no) people that will desire to 
compile sage 9.0 with python 2. And 9.1 even more. People that want to be 
up to date do not care about py2.
I therefore strongly vote for the immediate drop of python2, and for 
removing the promise that 9.1 will be py2-compatible

F

Le dimanche 5 janvier 2020 22:06:21 UTC+1, Nils Bruin a écrit :
>
> I think our wiki vetoes that idea. See 
> https://wiki.sagemath.org/Python3-Switch :
>
> Compiling with Python 2 
>
> After version 9.0, if you really want so, you can still build and use 
> SageMath with *Python 2*, as follows. 
>
> make configure./configure --with-python=2make build
>
> Beware that you will need to call the second line again if you ever call 
> "make distclean". 
> This will work until version 9.1 at least. Then the backward compatibility 
> with Python 2 will no longer be ensured. 
>
> If you want to drop py2 compatibility very soon, the only option is to 
> release 9.1 basically right now, identical to 9.0, and then get on with 
> developing 9.2. That's a nasty thing to do. Based on previous release 
> schedules, people would be justified in expecting that <=9.1 is the 
> "current" release until at least June 2020 or so. So we're stuck with py2 
> compatibility until that time.
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/11341de1-238b-4abe-b617-e328af87b639%40googlegroups.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Antonio Rojas]

El lunes, 6 de enero de 2020, 8:43:14 (UTC+1), Nils Bruin escribió:
>
>
> You could say: there is such a way: just check out 9.0 and build that with 
> py2. Perhaps that's enough. However, I would err at the side of caution and 
> make a slightly bigger gesture here and still include 9.1 as well -- that 
> was published as a definite promise already. Once again, for any piece of 
> code where this turns out to be a bother, just wait with the merge until 
> 9.2. I don't think it needs to put a very large load on developers. It will 
> mainly be a pain for the release manager, so he/she should actually get to 
> have a significant voice in this
>

It's not as simple as waiting for 9.2 to merge them. Some of the pending 
changes (ipython 7 and python 3.8 support) require huge patches that touch 
lots of files. This means that they will probably need rebasing for every 
single beta release (this has been the case until now), which makes it not 
really practical to start working on them on trac until the 9.2 cycle 
starts. Until then the burden of maintaining the patches falls on the 
downstream packagers - and the situation is only going to get worse as new 
python3-only versions of dependencies are released and updated on distros. 

>  
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/e1ad7cf6-e2c6-482e-83d2-c2992d725efd%40googlegroups.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Nils Bruin]

On Sunday, January 5, 2020 at 7:53:35 PM UTC-8, Andrey Novoseltsev wrote:
>
> Those who do want to keep that promise intact - can you please explain who 
> will be affected and how if we drop Python 2 support? "Regular" users do 
> not have a choice - precompiled binaries are with Python 3 only. Those who 
> can build from source and add configure parameters may just as well 
> checkout 9.0 and compile that.
>
> You can see the problem with the py2/py3 transition already in this 
asksage question about sagecell: 
https://ask.sagemath.org/question/49335/deprecationwarning-on-sagecell-python-3/
 
:

The transition to py3 instantly invalidated a whole bunch of sage-cell 
contents, because there are sufficiently many differences between py2 and 
py3 that almost any code needs some (fairly routine) changes. Just a with 
deprecation in the user interface, we must have a transition period that 
gives people a chance to change over: where there's a relatively 
straightforward way to still run the py2 code unchanged, so that the user 
can plan for an opportune time to make the required changes.

You could say: there is such a way: just check out 9.0 and build that with 
py2. Perhaps that's enough. However, I would err at the side of caution and 
make a slightly bigger gesture here and still include 9.1 as well -- that 
was published as a definite promise already. Once again, for any piece of 
code where this turns out to be a bother, just wait with the merge until 
9.2. I don't think it needs to put a very large load on developers. It will 
mainly be a pain for the release manager, so he/she should actually get to 
have a significant voice in this.

We're late to the game already with making the transition only at the point 
where py2 has fallen out of support.

In fact, I think it would make a lot of sense to make 9.1 the last release 
to support py2 and therefore make 9.1 more a bugfix release: concentrate on 
fixes of issues that arise as a result of 9.0 and fix them, so that there's 
a relatively reliable point in the history of sage to run on py2. That 
would automatically target significant new features for 9.2.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/e4826aac-697f-4001-911b-91b92f77031e%40googlegroups.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Matthias Koeppe]

+1 for dropping python 2 support in 9.1 betas or 9.2 betas if the 9.1 cycle 
is not too long.

Matthias

On Sunday, January 5, 2020 at 2:44:10 PM UTC-5, Frédéric Chapoton wrote:
>
> Hello,
>
> I would like to suggest that the sooner we drop Python 2 support the 
> better. We still need to handle the upgrade to ipython7 and the 
> compatibility with python 3.8. All this will be made very difficult if we 
> insist on maintaining a codebase that is both compatible with python 2 and 
> python 3.
>
> So, please vote :
>
> Do you agree that sage release 9.1 (and most of the 9.1.betas) will not be 
> kept compatible with Python 2 ?
>
> Frédéric
>
>
>
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/497181b0-943c-4985-822b-8ae0a1be1d25%40googlegroups.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Andrey Novoseltsev]

On Sunday, 5 January 2020 16:24:38 UTC-7, Nils Bruin wrote:
>
> On Sunday, January 5, 2020 at 2:51:44 PM UTC-8, Nils Bruin wrote:
>>
>> I don't think there's a doubt that we'll drop py2.7 support. This Wiki, 
>> however, has been referred to in many places describing the py2/py3 
>> transition.
>>
>
> One of these places is the FAQ, which is official documentation:
>
>
> http://doc.sagemath.org/html/en/faq/faq-usage.html#can-i-use-sagemath-with-python-3-x
>
> The 9.0 FAQ still mentions:
>
> As of August 2019, most of SageMath works fine with Python 3. However, we 
> still consider Python 3 support to be experimental and no official Python 3 
> release has been made yet.
>
> That would probably be good to adjust too: there is now an official Py3 
> release and I don't think we consider it experimental anymore -- unless 
> sage in its entirety is still considered experimental :-).
>
>
I fail to see the difference between "officialness" of this documentation 
statement that Python 3 support is experimental and the wiki promise that 
9.1 will work with Python 2. If the first one is wrong and it is OK to 
correct it and move forward, why can't the second one also be corrected?

Those who do want to keep that promise intact - can you please explain who 
will be affected and how if we drop Python 2 support? "Regular" users do 
not have a choice - precompiled binaries are with Python 3 only. Those who 
can build from source and add configure parameters may just as well 
checkout 9.0 and compile that.

Note that I am not a particularly active Sage developer anymore - I 
maintain SageMathCell, which is switched to Python 3 already. So it does 
not affect me and my workload in any way whether 9.1 is Python 2 compatible 
or not. But for the sake of those who do work on transition - please don't 
force them to spend their valuable energy on a pretty much pointless 
compatibility with unsupported languages.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/48277f25-e4e8-4ff5-a353-d27d6c0a618c%40googlegroups.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Nils Bruin]

On Sunday, January 5, 2020 at 2:51:44 PM UTC-8, Nils Bruin wrote:
>
> I don't think there's a doubt that we'll drop py2.7 support. This Wiki, 
> however, has been referred to in many places describing the py2/py3 
> transition.
>

One of these places is the FAQ, which is official documentation:

http://doc.sagemath.org/html/en/faq/faq-usage.html#can-i-use-sagemath-with-python-3-x

The 9.0 FAQ still mentions:

As of August 2019, most of SageMath works fine with Python 3. However, we 
still consider Python 3 support to be experimental and no official Python 3 
release has been made yet.

That would probably be good to adjust too: there is now an official Py3 
release and I don't think we consider it experimental anymore -- unless 
sage in its entirety is still considered experimental :-).

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/b68f89dc-6d9e-4ca4-ad92-46c61ee614cb%40googlegroups.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Nils Bruin]

On Sunday, January 5, 2020 at 2:33:02 PM UTC-8, Timo Kaufmann wrote:
>
> +1 for dropping python 2.7 compatibility
>
> It is unfortunate that the wiki makes that promise. But it is a community 
> wiki, and as such I wouldn't consider it an authorative source. Keeping 
> python2 support will be a *lot* of effort, either for the sage project or 
> (if it just refuses to deal with this problem and keeps using outdated 
> dependencies) for packagers. At the same time, I see little to no benefit 
> in keeping it. People who insist on python2 can keep using 9.0 as long as 
> they need to. It will continue to work for all the purposes it works for 
> now.
>

I don't think there's a doubt that we'll drop py2.7 support. This Wiki, 
however, has been referred to in many places describing the py2/py3 
transition. As far as I'm aware this is the only announcement at the moment 
that comes close to an official py2/py3 transition plan for sage. If you're 
aware of a message elsewhere, we can see how we can unify the two. The 
message on the wiki makes a very definite mention that 9.1 will still 
support building with python2. I don't consider reneging on that an option. 
Breaking definite promises in support makes for *very* bad PR. If there is 
a more "official" place to mention the transition strategy, we can put an 
announcement there, but given the fact that we've already published the 
message on the wiki and pointed at it means that that announcement must be 
as least as lenient wrt. py2 as the wiki message.

It's not such a burden for developers anyway: you can just develop for Py3 
only. Your work will just be merged in 9.2 at the earliest. If we end up 
with loads of 9.2 milestone tickets and no 9.1 ones, releasing 9.1 will be 
quite straightforward.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/84107680-3147-4c9d-932a-61142d96a1d9%40googlegroups.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Timo Kaufmann]

+1 for dropping python 2.7 compatibility

It is unfortunate that the wiki makes that promise. But it is a community 
wiki, and as such I wouldn't consider it an authorative source. Keeping 
python2 support will be a *lot* of effort, either for the sage project or 
(if it just refuses to deal with this problem and keeps using outdated 
dependencies) for packagers. At the same time, I see little to no benefit 
in keeping it. People who insist on python2 can keep using 9.0 as long as 
they need to. It will continue to work for all the purposes it works for 
now.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/1728e934-a4cf-49e1-9976-a7c20c521568%40googlegroups.com.

Re: [sage-devel] Re: drop python2 compatibility in 9.1 ?

[David Roe]

I just added a 9.2 milestone on trac, which isn't the default.
David

On Sun, Jan 5, 2020 at 4:37 PM John H Palmieri 
wrote:

> Can someone with trac admin access and know-how add a 9.2 milestone, so
> people can develop for 9.2 if they want to focus on dropping Python 2
> support?
>
> Meanwhile, 9.1 should include a deprecation warning when people use
> './configure --with-python=2'.
>
>
>
>
> On Sunday, January 5, 2020 at 1:06:21 PM UTC-8, Nils Bruin wrote:
>>
>> I think our wiki vetoes that idea. See
>> https://wiki.sagemath.org/Python3-Switch :
>>
>> Compiling with Python 2
>>
>> After version 9.0, if you really want so, you can still build and use
>> SageMath with *Python 2*, as follows.
>>
>> make configure./configure --with-python=2make build
>>
>> Beware that you will need to call the second line again if you ever call
>> "make distclean".
>> This will work until version 9.1 at least. Then the backward
>> compatibility with Python 2 will no longer be ensured.
>>
>> If you want to drop py2 compatibility very soon, the only option is to
>> release 9.1 basically right now, identical to 9.0, and then get on with
>> developing 9.2. That's a nasty thing to do. Based on previous release
>> schedules, people would be justified in expecting that <=9.1 is the
>> "current" release until at least June 2020 or so. So we're stuck with py2
>> compatibility until that time.
>>
> --
> You received this message because you are subscribed to the Google Groups
> "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sage-devel+unsubscr...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/sage-devel/2a6cdc41-1ef1-4f95-b44c-c062342dc82b%40googlegroups.com
> 
> .
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/CAChs6_k%2BVZ0vNfxhyYP%2B-db1t__bevtk-J3Q8%3DXnRqOHgU3CdA%40mail.gmail.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[John H Palmieri]

Can someone with trac admin access and know-how add a 9.2 milestone, so 
people can develop for 9.2 if they want to focus on dropping Python 2 
support?

Meanwhile, 9.1 should include a deprecation warning when people use 
'./configure --with-python=2'.




On Sunday, January 5, 2020 at 1:06:21 PM UTC-8, Nils Bruin wrote:
>
> I think our wiki vetoes that idea. See 
> https://wiki.sagemath.org/Python3-Switch :
>
> Compiling with Python 2 
>
> After version 9.0, if you really want so, you can still build and use 
> SageMath with *Python 2*, as follows. 
>
> make configure./configure --with-python=2make build
>
> Beware that you will need to call the second line again if you ever call 
> "make distclean". 
> This will work until version 9.1 at least. Then the backward compatibility 
> with Python 2 will no longer be ensured. 
>
> If you want to drop py2 compatibility very soon, the only option is to 
> release 9.1 basically right now, identical to 9.0, and then get on with 
> developing 9.2. That's a nasty thing to do. Based on previous release 
> schedules, people would be justified in expecting that <=9.1 is the 
> "current" release until at least June 2020 or so. So we're stuck with py2 
> compatibility until that time.
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/2a6cdc41-1ef1-4f95-b44c-c062342dc82b%40googlegroups.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Nils Bruin]

I think our wiki vetoes that idea. See 
https://wiki.sagemath.org/Python3-Switch :

Compiling with Python 2 

After version 9.0, if you really want so, you can still build and use 
SageMath with *Python 2*, as follows. 

make configure./configure --with-python=2make build

Beware that you will need to call the second line again if you ever call 
"make distclean". 
This will work until version 9.1 at least. Then the backward compatibility 
with Python 2 will no longer be ensured. 

If you want to drop py2 compatibility very soon, the only option is to 
release 9.1 basically right now, identical to 9.0, and then get on with 
developing 9.2. That's a nasty thing to do. Based on previous release 
schedules, people would be justified in expecting that <=9.1 is the 
"current" release until at least June 2020 or so. So we're stuck with py2 
compatibility until that time.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/d15b09a4-e604-4b9d-bdd4-26801630315e%40googlegroups.com.

[sage-devel] Re: drop python2 compatibility in 9.1 ?

[Eric Gourgoulhon]

+1 for dropping Python 2 compatibility in Sage 9.1

Eric.


-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/76393149-5a47-411d-aa5a-e10edbb57525%40googlegroups.com.