Python 2.6 support
79 views
Skip to first unread message
Aaron Meurer
Nov 22, 2014, 3:20:37 PM11/22/14
to sy...@googlegroups.com
While we're talking about changes for 1.0, how do people feel about
dropping Python 2.6 support. I think we decided that we need to keep
Python 3.2 support for the time being because of Debian, but what
about 2.6? According to
https://www.python.org/download/releases/2.6.7/, the Python 2.6 series
hasn't been supported for security releases for over a year. But what
matters is if people are using it. If there's a good case to keep it
for the time being, we can.
Some advantages of dropping support is that there is a lot of nice
syntax backported from Python 3 in Python 2.7 which is unavailable in
Python 2.6. Some useful things in Python 2.7:
- Set literals (like {1, 2, 3} instead of set([1, 2, 3]))
- Set and dictionary comprehensions(like {a for a in stuff} or {a: b
for a, b in stuff})
- OrderedDict in the standard library
- Multiple context managers in a single with statement (I don't know
if we ever do this, but we have a lot of experimental context managers
like assuming() and evaluate() and this makes it easier to combine
them).
- argparse (we can use it instead of optparse in our scripts like isympy)
- collections.Counter (could be useful, it's effectively a multiset)
But again, the main question is if people are using it.
Aaron Meurer
dropping Python 2.6 support. I think we decided that we need to keep
Python 3.2 support for the time being because of Debian, but what
about 2.6? According to
https://www.python.org/download/releases/2.6.7/, the Python 2.6 series
hasn't been supported for security releases for over a year. But what
matters is if people are using it. If there's a good case to keep it
for the time being, we can.
Some advantages of dropping support is that there is a lot of nice
syntax backported from Python 3 in Python 2.7 which is unavailable in
Python 2.6. Some useful things in Python 2.7:
- Set literals (like {1, 2, 3} instead of set([1, 2, 3]))
- Set and dictionary comprehensions(like {a for a in stuff} or {a: b
for a, b in stuff})
- OrderedDict in the standard library
- Multiple context managers in a single with statement (I don't know
if we ever do this, but we have a lot of experimental context managers
like assuming() and evaluate() and this makes it easier to combine
them).
- argparse (we can use it instead of optparse in our scripts like isympy)
- collections.Counter (could be useful, it's effectively a multiset)
But again, the main question is if people are using it.
Aaron Meurer
Ondřej Čertík
Nov 22, 2014, 4:12:53 PM11/22/14
to sympy
Ondrej
Joachim Durchholz
Nov 23, 2014, 6:18:20 AM11/23/14
to sy...@googlegroups.com
Am 22.11.2014 um 21:20 schrieb Aaron Meurer:
> But again, the main question is if people are using it.
Exactly.
> But again, the main question is if people are using it.
Do we have any figures about how many people are still using SymPy with 2.6?
There were some Android issues last time dropping 2.6 came up, IIRC.
> Some advantages of dropping support is that there is a lot of nice
> syntax backported from Python 3 in Python 2.7 which is unavailable
> in Python 2.6.
I see two: User adoption and developer attraction.
User adoption would benefit from speed and reliability. I can't judge
whether the newly available features would help with that, or whether
that would make enough of a difference to be user-noticeable.
Developer attraction: Do we have any reactions on the line of "ugh...
2.6... no thanks"?
Ondřej Čertík
Nov 23, 2014, 2:22:23 PM11/23/14
to sympy
On Sun, Nov 23, 2014 at 4:18 AM, Joachim Durchholz <j...@durchholz.org> wrote:
> Am 22.11.2014 um 21:20 schrieb Aaron Meurer:
>>
>> But again, the main question is if people are using it.
>
>
> Exactly.
> Do we have any figures about how many people are still using SymPy with 2.6?
> There were some Android issues last time dropping 2.6 came up, IIRC.
Btw, IPython also doesn't support Python 2.6, and I bet IPython is
> Am 22.11.2014 um 21:20 schrieb Aaron Meurer:
>>
>> But again, the main question is if people are using it.
>
>
> Exactly.
> Do we have any figures about how many people are still using SymPy with 2.6?
> There were some Android issues last time dropping 2.6 came up, IIRC.
used by more people than SymPy, so if it is good for IPython, I think
it will be good for us as well.
Ondrej
>
>> Some advantages of dropping support is that there is a lot of nice
>> syntax backported from Python 3 in Python 2.7 which is unavailable
>> in Python 2.6.
>
> Question is: Does that improve any of the success factors for SymPy?
> I see two: User adoption and developer attraction.
>
> User adoption would benefit from speed and reliability. I can't judge
> whether the newly available features would help with that, or whether that
> would make enough of a difference to be user-noticeable.
>
> Developer attraction: Do we have any reactions on the line of "ugh... 2.6...
> no thanks"?
>
> You received this message because you are subscribed to the Google Groups
> "sympy" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sympy+un...@googlegroups.com.
> To post to this group, send email to sy...@googlegroups.com.
> Visit this group at http://groups.google.com/group/sympy.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/sympy/5471C279.2040302%40durchholz.org.
>
> For more options, visit https://groups.google.com/d/optout.
Roberto Colistete Jr.
Nov 23, 2014, 2:55:42 PM11/23/14
to sy...@googlegroups.com
On 23-11-2014 09:18, Joachim Durchholz wrote:
> Do we have any figures about how many people are still using SymPy
> with 2.6?
> There were some Android issues last time dropping 2.6 came up, IIRC.
AFAIK, only :
> Do we have any figures about how many people are still using SymPy
> with 2.6?
> There were some Android issues last time dropping 2.6 came up, IIRC.
- Android SL4A uses Python 2.6, but there is now QPython which is easier
to install & use, with Python 2.7;
- MeeGo Harmattan (for Nokia N9/N950) has Python 2.6, but Nokia N9/N950
is already 3 years old.
So, basically there are some Android users with old Python SL4A and
MeeGo Harmattan users who depends on Python 2.6 compatibility.
My opinion : announce with some months in advance that SymPy will
drop Python 2.6 compatibility.
Best regards,
Roberto
Joachim Durchholz
Nov 23, 2014, 3:48:08 PM11/23/14
to sy...@googlegroups.com
Am 23.11.2014 um 20:22 schrieb Ondřej Čertík:
> On Sun, Nov 23, 2014 at 4:18 AM, Joachim Durchholz <j...@durchholz.org> wrote:
>> Do we have any figures about how many people are still using SymPy with 2.6?
>> There were some Android issues last time dropping 2.6 came up, IIRC.
>
> Btw, IPython also doesn't support Python 2.6, and I bet IPython is
> used by more people than SymPy, so if it is good for IPython, I think
> it will be good for us as well.
That's a rather weak argument, too:
> On Sun, Nov 23, 2014 at 4:18 AM, Joachim Durchholz <j...@durchholz.org> wrote:
>> Do we have any figures about how many people are still using SymPy with 2.6?
>> There were some Android issues last time dropping 2.6 came up, IIRC.
>
> Btw, IPython also doesn't support Python 2.6, and I bet IPython is
> used by more people than SymPy, so if it is good for IPython, I think
> it will be good for us as well.
- You compare just a specialized niche of the Python users.
- The SymPy 2.6 percentage might be higher than the Python 2.6
percentage (some specialiced niche).
There's a reason why I prefer hard numbers.
BTW weak arguments are all too often a disguise of wishful thinking -
we'd all *like* to drop 2.6, so there's a tendency to name all the
arguments in favor however weak they are, and ignore those against. I'm
currently wearing my Devil's Advocate hat to counter that effect,
actually I'd be fine with dropping 2.6 ;-)
Regards,
Jo
Joachim Durchholz
Nov 23, 2014, 3:49:09 PM11/23/14
to sy...@googlegroups.com
Am 23.11.2014 um 20:55 schrieb Roberto Colistete Jr.:
> My opinion : announce with some months in advance that SymPy will
> drop Python 2.6 compatibility.
... and keep it if protests start rolling in I guess :-)
> My opinion : announce with some months in advance that SymPy will
> drop Python 2.6 compatibility.
I'd be +1 for that route.
Sergey Kirpichev
Nov 24, 2014, 3:28:40 AM11/24/14
to sy...@googlegroups.com
On Sunday, November 23, 2014 10:22:23 PM UTC+3, Ondřej Čertík wrote:
Btw, IPython also doesn't support Python 2.6, and I bet IPython is
used by more people than SymPy, so if it is good for IPython, I think
it will be good for us as well.
IPython doesn't support 3.2 too. Should we drop everything, except 2.7 and >= 3.3?
Francesco Bonazzi
Nov 24, 2014, 12:46:30 PM11/24/14
to sy...@googlegroups.com
Found this one:
http://astrofrog.github.io/blog/2013/01/13/what-python-installations-are-scientists-using/
Apparently at the end of 2012 the adoption of Python 2.6 was still around 15%. Python 2.5 was slightly above 1%.
Python 2.5 was released in October 2006, Python 2.6 in October 2008. So, as of now (November 2014), I would expect Python 2.6's adoption to be similar to Python 2.5's adoption in November 2012, which means around 1%. Does this make sense?
http://astrofrog.github.io/blog/2013/01/13/what-python-installations-are-scientists-using/
Apparently at the end of 2012 the adoption of Python 2.6 was still around 15%. Python 2.5 was slightly above 1%.
Python 2.5 was released in October 2006, Python 2.6 in October 2008. So, as of now (November 2014), I would expect Python 2.6's adoption to be similar to Python 2.5's adoption in November 2012, which means around 1%. Does this make sense?
Joachim Durchholz
Nov 24, 2014, 2:15:23 PM11/24/14
to sy...@googlegroups.com
The figures are a bit old. The question, now is how the figures have
shifted in the two years since that survey.
Maybe we should add a survey to SymPy.
Say, start sympy.survey() from the command line, let it collect the
pertinent information and mail it to us. Preferrably after showing the
information to the user and asking him for confirmation whether to send it.
Aaron Meurer
Nov 24, 2014, 2:42:55 PM11/24/14
to sy...@googlegroups.com
IPython seems to be ahead of the curve here (most other libraries I
know of still support 2.6), whereas SymPy has traditionally been
behind the curve (we supported 2.5 for longer than most libraries). At
least that's what I remember. We could probably look up the release
details to be sure, say, for the other libraries in the scipy stack.
Aaron Meurer
> To view this discussion on the web visit https://groups.google.com/d/msgid/sympy/CADDwiVBz9L8XjNkfNmwt5ZSMSKR9qRojKjf1m2XjEmzQxERUJQ%40mail.gmail.com.
know of still support 2.6), whereas SymPy has traditionally been
behind the curve (we supported 2.5 for longer than most libraries). At
least that's what I remember. We could probably look up the release
details to be sure, say, for the other libraries in the scipy stack.
Aaron Meurer
Roberto Colistete Jr.
Nov 24, 2014, 3:05:17 PM11/24/14
to sy...@googlegroups.com
On 24-11-2014 17:41, Aaron Meurer wrote:
> IPython seems to be ahead of the curve here (most other libraries I
> know of still support 2.6), whereas SymPy has traditionally been
> behind the curve (we supported 2.5 for longer than most libraries). At
> least that's what I remember. We could probably look up the release
> details to be sure, say, for the other libraries in the scipy stack.
>
> Aaron Meurer
Agree about IPython. While NumPy, MatPlotLib and SciPy still
> IPython seems to be ahead of the curve here (most other libraries I
> know of still support 2.6), whereas SymPy has traditionally been
> behind the curve (we supported 2.5 for longer than most libraries). At
> least that's what I remember. We could probably look up the release
> details to be sure, say, for the other libraries in the scipy stack.
>
> Aaron Meurer
support Python 2.6.
NumPy dropped Python 2.5 support in v1.8.0 (30/10/2013), MatPlotLib
in v1.2.0 (08/01/2013), Scipy in v0.12 (07/04/2013).
As Python 2.5 was released 2 years before Python 2.6, the logic
gives 2015 to be the year when Python 2.6 support will be dropped by
NumPy, MatPlotLib and SciPy.
Francesco Bonazzi
Nov 24, 2014, 4:09:16 PM11/24/14
to sy...@googlegroups.com
On Monday, November 24, 2014 9:05:17 PM UTC+1, Roberto Colistete Jr. wrote:
As Python 2.5 was released 2 years before Python 2.6, the logic
gives 2015 to be the year when Python 2.6 support will be dropped by
NumPy, MatPlotLib and SciPy.
That's a good idea, waiting until support is dropped by other scientific python projects.
Aaron Meurer
Nov 24, 2014, 5:58:13 PM11/24/14
to sy...@googlegroups.com
On Mon, Nov 24, 2014 at 1:05 PM, Roberto Colistete Jr.
<roberto....@gmail.com> wrote:
> On 24-11-2014 17:41, Aaron Meurer wrote:
>>
>> IPython seems to be ahead of the curve here (most other libraries I
>> know of still support 2.6), whereas SymPy has traditionally been
>> behind the curve (we supported 2.5 for longer than most libraries). At
>> least that's what I remember. We could probably look up the release
>> details to be sure, say, for the other libraries in the scipy stack.
>>
>> Aaron Meurer
>
>
> Agree about IPython. While NumPy, MatPlotLib and SciPy still support
> Python 2.6.
>
> NumPy dropped Python 2.5 support in v1.8.0 (30/10/2013), MatPlotLib in
> v1.2.0 (08/01/2013), Scipy in v0.12 (07/04/2013).
It seems we're not behind the curve, then. SymPy dropped support in
<roberto....@gmail.com> wrote:
> On 24-11-2014 17:41, Aaron Meurer wrote:
>>
>> IPython seems to be ahead of the curve here (most other libraries I
>> know of still support 2.6), whereas SymPy has traditionally been
>> behind the curve (we supported 2.5 for longer than most libraries). At
>> least that's what I remember. We could probably look up the release
>> details to be sure, say, for the other libraries in the scipy stack.
>>
>> Aaron Meurer
>
>
> Agree about IPython. While NumPy, MatPlotLib and SciPy still support
> Python 2.6.
>
> NumPy dropped Python 2.5 support in v1.8.0 (30/10/2013), MatPlotLib in
> v1.2.0 (08/01/2013), Scipy in v0.12 (07/04/2013).
0.7.3, released July 13, 2013.
Maybe we should talk with the other communities to see when they plan
to drop support.
Aaron Meurer
>
> As Python 2.5 was released 2 years before Python 2.6, the logic gives
> 2015 to be the year when Python 2.6 support will be dropped by NumPy,
> MatPlotLib and SciPy.
>
> --
> You received this message because you are subscribed to the Google Groups
> "sympy" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sympy+un...@googlegroups.com.
> To post to this group, send email to sy...@googlegroups.com.
> Visit this group at http://groups.google.com/group/sympy.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/sympy/54738F77.7070903%40gmail.com.
> You received this message because you are subscribed to the Google Groups
> "sympy" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sympy+un...@googlegroups.com.
> To post to this group, send email to sy...@googlegroups.com.
> Visit this group at http://groups.google.com/group/sympy.
> To view this discussion on the web visit
Reply all
Reply to author
Forward
0 new messages