Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.python > #88353 > unrolled thread

Python 3 success story

Back to article view | Back to comp.lang.python

Contents

  Python 3 success story Terry Reedy <tjreedy@udel.edu> - 2015-03-30 19:28 -0400
    Re: Python 3 success story Mario Figueiredo <marfig@gmail.com> - 2015-03-31 01:24 +0100
      Re: Python 3 success story Terry Reedy <tjreedy@udel.edu> - 2015-03-31 02:06 -0400

#88353 — Python 3 success story

Last summer, a college student, who prefers Python to Java and Racket 
(her other CS course languages), wanted to do a year-long AI research 
project using PyBrain (for the ai part), numpy (required for PyBrain), 
and pygame (for animated displays).  We both preferred 3.x.  That was 
not an issue for pygame and numpy as compiled 3.4 Windows binaries  were 
easily available.

PyBrain, however, was '2.7 only'.  So I downloaded it, read the doc for 
2to3, and ran it to do the conversion.  Running the PyBrain test suite 
did not work because it uses fragile doctests. Looking at the non-test 
code, I could see that it mostly used things that did not change in 3.x 
and a few things that should have been converted correctly.  This all 
took maybe an hour, certainly less than 2.

So I suggested going ahead and testing PyBrain by using it.  This 
appears to have worked out well. I believe the only 2-3 issue she ran 
into was a '/' that needed to become '//', that either 2to3 or I missed 
in the initial conversion. She had more problems with exception messages 
that she could not understand.  If anything, those have been improved in 
3.x (and such improvements continue).

-- 
Terry Jan Reedy

[toc] | [next] | [standalone]

#88356

On Mon, 30 Mar 2015 19:28:16 -0400, Terry Reedy <tjreedy@udel.edu>
wrote:

>
>So I suggested going ahead and testing PyBrain by using it.  This 
>appears to have worked out well. I believe the only 2-3 issue she ran 
>into was a '/' that needed to become '//', that either 2to3 or I missed 
>in the initial conversion. She had more problems with exception messages 
>that she could not understand.  If anything, those have been improved in 
>3.x (and such improvements continue).

This is good news, Terry. Have you considered requesting a fork from
the authors? And if you don't get the required attention, just fork it
yourself? Even if you don't plan to maintain it, it would be nice to
have it available for someone else to do it.

A quick look at the GitHub pages reveals the project has pretty much
stalled. Although it could be it achieved a good enough stable status,
this is a machine learning project and those tend to always require
constant work.

[toc] | [prev] | [next] | [standalone]

#88363

On 3/30/2015 8:24 PM, Mario Figueiredo wrote:
> On Mon, 30 Mar 2015 19:28:16 -0400, Terry Reedy <tjreedy@udel.edu>
> wrote:

>> So I suggested going ahead and testing PyBrain by using it.  This
>> appears to have worked out well. I believe the only 2-3 issue she ran
>> into was a '/' that needed to become '//', that either 2to3 or I missed
>> in the initial conversion.
>
> This is good news, Terry. Have you considered requesting a fork from
> the authors?

I considered and rejected the idea at the time because a) I was more 
than busy supervising a GSOC project and b) to work on it seriously, I 
would want to ditch doctest (which was not meant for serious unit 
testing) and convert to using unittest for both 2.7 and 3.x.

 > And if you don't get the required attention, just fork it
> yourself? Even if you don't plan to maintain it, it would be nice to
> have it available for someone else to do it.

Now that the port has been used for some months, I a) can see that 
making it available might be helpful to others, and b) would feel more 
comfortable doing so on an as is, use at your own risk, basis (and 
possibly even accept and apply patches).

> A quick look at the GitHub pages reveals the project has pretty much
> stalled. Although it could be it achieved a good enough stable status,
> this is a machine learning project and those tend to always require
> constant work.

Not knowing git and github adds a barrier, but thanks for the suggestion.

-- 
Terry Jan Reedy

[toc] | [prev] | [standalone]

Back to top | Article view | comp.lang.python

csiph-web