Thu Jun 7 19:33:18 CDT 2012
On Thu, Jun 7, 2012 at 2:50 PM, Bob McElrath <firstname.lastname@example.org> wrote:
> It turns out to get decent quality SVG figures, you have to set the
> figure.figsize fairly high.
Unfortunately it's much more complicated than that. I experimented a
LOT when I first wrote the inline backend, hoping to make SVG the
default. And unfortunately, while most simple plots do indeed look
much nicer as SVG, in production work I kept running into corner cases
where the SVG generated by MPL was just not quite right (missing
elements, clipping lines wrong, poor transparency handling, etc.). Of
course, if I was a good citizen I'd have taken the time to record
those as test cases so we could fix them in MPL, but I just didn't
have the bandwidth to to both ipython and mpl work (this is when we
were just building the kernel system and the qtconsole, so our task
list was monstrous).
I would *love* to be able to switch to SVG by default, but until we
can trust that the SVG that comes out of MPL is consistently as solid
as the pngs, or have the time to spend fixing those issues, I think
it's best to keep the png defaults.
The problem with fixing this is that I've never been able to see the
issues in easy to reproduce examples. They tend to only happen with
very complex figures that I only can generate pulling full-blown
research datasets out, so they make for poor test cases.
I'll make it a priority from now on to try to at least record these
cases in an isolated place, so that hopefully we can dig into them in
MPL. Because I know it's terrible behavior to put down mpl's
rendering without providing a test case we can use to improve the
More information about the IPython-dev