Re: [Matplotlib-users] legend(loc='best') not so great

[Damon M. <dam...@gm...>]

On Wed, Sep 12, 2012 at 5:57 PM, Aronne Merrelli
<aro...@gm...> wrote:
> On Tue, Sep 11, 2012 at 10:49 AM, Damon McDougall
> <dam...@gm...> wrote:
>> On Tue, Sep 11, 2012 at 3:34 PM, Neal Becker <ndb...@gm...> wrote:
>>> OK, I've attached my sanitized example
>>>
>>
>> ImportError: No module named pandas.
>>
>> Can you provide an example that doesn't depend on pandas?
>>
>
> I was playing with this example to remove the pandas stuff, and It
> looks a lot like the check for the best legend location is just not
> accounting correctly for the multiple legend entries in this case.
> Here is an even more minimal example, that I think reproduces the same
> problem as Neal's original example.
>
> colors = ['b','g','r']
> for n in range(3):
>     plt.scatter([n,],[n,],color=colors[n])
> plt.legend(['foo','foo','foo'],loc='best')
>
>
> Note that if you just put one legend name in there, the location is
> perfectly fine, it is just as it extends downward for the second and
> third names, that it covers over the point.
>
> FYI: Im using v1.1.0 mpl, in EPD 7.2.
>
> HTH,
> Aronne
>

Aronne, thanks for taking the time to produce a nice example. I can
now recreate the problem. I'm not sure how invasive it will be to fix
this, so it possibly won't make it in for version 1.2, but possibly a
bugfix release after the fact.

For anyone who's interested, here's the github issue:
https://github.com/matplotlib/matplotlib/issues/1235

-- 
Damon McDougall
http://www.damon.is-a-geek.com
B2.39
Mathematics Institute
University of Warwick
Coventry
West Midlands
CV4 7AL
United Kingdom