legend marker 'o' is incorrectly printed

[ppurka]

Just look at the output of this command

list_plot(range(3), plotjoined=True, marker='o', legend_label='square here, circle there')

The marker in the legend should be a o but it is actually printed as a square. This seems to be a bug in Sage because I couldn't reproduce it in matplotlib. I tried the following two variants:

import matplotlib.pyplot as plt
plt.plot(range(3), range(3), '-o')
plt.legend('square here')
plt.savefig('a.png')

import matplotlib.lines as lines
from matplotlib.figure import Figure

options = {'marker': 'o'}
p = lines.Line2D(range(3), range(3), **options)
p.set_label('square here, circle there')
figure=Figure()
subplot = figure.add_subplot(111)
subplot.add_line(p)
subplot.legend()
from matplotlib.backends.backend_agg import FigureCanvasAgg
figure.set_canvas(FigureCanvasAgg(figure))
figure.savefig('a.png')

The bug dates back to at least sage-4.7. That's the oldest precompiled version of sage I had around here.

---

Attached patch contains a test to the line() function. Apply trac_13690-legend_test_for_line.patch​ to devel/sage.

[jason]

And it's not just list_plot. Here is an example with line:

line(enumerate([1,2,1]), marker = 'o', legend_label='a')

[jason]

I just tried this on my local computer with #13693 applied and the legend showed up correctly. Can someone double-check that the upgrade to matplotlib 1.2.0 fixes this problem?

[ppurka]

Replying to jason:

And it's not just list_plot. Here is an example with line:

line(enumerate([1,2,1]), marker = 'o', legend_label='a')

My second matplotlib example essentially does whatever the line command does. Also plotjoined=True in list_plot calls line directly.

[jason]

Right. Thanks.

[ppurka]

Replying to jason:

I just tried this on my local computer with #13693 applied and the legend showed up correctly. Can someone double-check that the upgrade to matplotlib 1.2.0 fixes this problem?

Yes. Upgrade to mpl-1.2.0 fixes this. I am flummoxed.

Setting dependency to #13693 since fixing that will fix this automatically.

[jason]

And I am relieved! We won't have to track this down.

[ppurka]

I am relieved too! This is a nasty bug; very unreproducible.

[kcrisman]

I assume we should put something in a TESTS block somewhere then for this ticket? In the sense that we could ask someone to visually confirm the output, not that we are actually comparing images.

[ppurka]

Apply to devel/sage

[ppurka]

Replying to kcrisman:

I assume we should put something in a TESTS block somewhere then for this ticket? In the sense that we could ask someone to visually confirm the output, not that we are actually comparing images.

You are right. I attached a patch.

[ppurka]

#13693 is in. I just kicked the patchbot. Let's see if the patch still applies.

[ncohen]

Just a doctest, it cannot hurt :-)

Nathann

[ppurka]

Thanks :)