matplotlib-devel Mailing List for matplotlib

Jouni K. Seppänen wrote:
> The matplotlib.use() function was always a no-op after importing pylab,
> and there were occasional questions about it on the mailing list. It was
> changed fairly recently to raise an error in this case. Bug #1855454
> points out that this is not backward compatible: an application that
> calls use() too late didn't do anything before the change, but now gets
> stopped with a traceback. I suggest changing the error to a warning,
> which I think would still deliver the message but would be compatible
> with (broken) legacy code. What do you think?
> 
> http://sourceforge.net/tracker/index.php?func=detail&aid=1855454&group_id=80706&atid=560720
> 
Good point. I probably went overboard when making the original change 
in October, so I have now changed it to a warning as you suggest.
Eric

On Dec 25, 2007 5:39 AM, Mark Bakker <ma...@gm...> wrote:
> Tedd, Michael -
>
> Sorry to join this discussion late, but wouldn't it be easier to use a
> representation of an ellipse in complex variables? That's what I have been
> using and it seems pretty quick to me.
The main point we are discussion is not how to represent the ellipse,
but what geometric primitives to use to render the ellipse. Earlier
versions of mpl used a polygon, then we added support for a 4 spline
approximation to reduce the error of the approximation. This too did
not have sufficient accuracy, and so Michael added an 8 spline
approximation (Ellipse) as well as an adaptive approximation (Arc).
JDH

The matplotlib.use() function was always a no-op after importing pylab,
and there were occasional questions about it on the mailing list. It was
changed fairly recently to raise an error in this case. Bug #1855454
points out that this is not backward compatible: an application that
calls use() too late didn't do anything before the change, but now gets
stopped with a traceback. I suggest changing the error to a warning,
which I think would still deliver the message but would be compatible
with (broken) legacy code. What do you think?
http://sourceforge.net/tracker/index.php?func=detail&aid=1855454&group_id=80706&atid=560720
-- 
Jouni