Data is not showing on matplotlib 1.4.2

[mgalardini]

Hi,

thanks for this wonderful library; I noticed a weird behaviour when using seaborn with the latest matplotlib (1.4.2, with python 2.7 on linux64). A simple plot like this:

plot(range(10), range(10), 'bo')

which works fine using just matplotlib, appears completely blank after importing seaborn. I did some tests and it seems that the 'dots' marker don't work anymore. From what I've seen it used to work with a previous version of matplotlib (1.3.1).

Thanks for your help,
Marco

[mwaskom]

Thanks for noting this. I think this is probably a matplotlib bug, though it will take a bit of work to understand exactly what is happening. Changing the markeredgewidth (or mew) parameter seems to make the fill show back up when it's > 0.01.

[mwaskom]

I think it may be related to this: matplotlib/matplotlib#3711

[mgalardini]

Thanks for looking into this: for the moment I just downgraded matplotlib to 1.4.0. Let's hope it will get fixed for the next release.

[mdboom]

There is a good chance this is fixed by #3687/#3699.

[mwaskom]

Just wanted to make a seaborn workaround (other than downgrading matplotlib) more explicit. If you do

sns.set_context(rc={'lines.markeredgewidth': 0.1})

the matplotlib markers should show up, and hopefully not look too different from the default.

[twiecki]

Just had the same problem and the set_context fix worked.

[xiaoweiz]

The problem can be replicated on Ubuntu but not OSX. I guess it's related to the Agg backend.

[mwaskom]

The problem can be replicated on Ubuntu but not OSX. I guess it's related to the Agg backend.

Yep that's what I gather from the matplotlib issue.

[QCaudron]

I was able to reproduce this bug under OSX when using %matplotlib inline in IPython Notebooks, which ( I believe ) drops back to the Agg backend. Any chance this is fixable on seaborn's end, perhaps in a not-too-distant update, or does it require communication with the matplotlib guys ?

[mwaskom]

I believe it's fixed upstream in matplotlib and I would expect them do do a point release soon.

[QCaudron]

Excellent ! Thanks. We've just discovered this bug over here, glad to see it's already been taken care of.

[mwaskom]

This workaround should now be in master and I'm going to release a 0.5.1 soon so that this is less of a hassle.

[mgalardini]

Much appreciated, thanks!

[tritemio]

I was just bitten by this bug doing in mpl 1.4.2 and seaborn 0.5.1

plt.plot(range(10), range(10), '+')

The workaround of setting lines.markeredgewidth works.

[histed]

I still see this bug with mpl 1.4.3 and seaborn 0.5.1.

I had to set lines.markeredgewidth to 1.0 to make 'x' and '+' visible.

[mwaskom]

That is a different issue than the one reported here. See matplotlib/matplotlib#4679

[histed]

OK. Just to make sure we're on the same page:
If I do not explicitly set lines.mew, using mpl 1.4.3 and seaborn 0.5.1, nothing is visible when markers are set to 'x', '+', or '.'. Setting lines.mew to 0.1 makes them barely visible. Setting it to 1.0 makes them look good.

[mwaskom]

yes but you can do it in the function call too

[patricksurry]

I ran into this same problem with mpl 1.4.2 and the 'mew' fix works - thanks! I also found that switching to the svg backend in the notebook makes the markers appear.

%config InlineBackend.figure_format = 'svg'

[mwaskom]

FWIW, matplotlib 1.4.2 is pretty old now and there's been both a bugfix release (1.4.3) and a new feature release (1.5) since then.

[mizzao]

I'm still setting this issue with seaborn 0.6.0. Is setting markeredgewidth still the right workaround or is it a different problem now?

[mmagnuski]

@mizzao As stated above - use matplotlib 1.4.3 or later.

[mwaskom]

I'm still setting this issue with seaborn 0.6.0. Is setting markeredgewidth still the right workaround or is it a different problem now?

Be more specific about what "this issue" is. People have discussed multiple issues in this thread, and besides, who knows what you're actually seeing.

Also as @mmagnuski notes, the most relevant piece of information would be the version of matplotlib that you are using.

[mizzao]

Sorry I wasn't being clear. In my case I wasn't seeing markers for x and +, etc, and I was using matplotlib 1.4.3 and seaborn 0.6.0.

It was worked around with sns.set_context(rc={'lines.markeredgewidth': 1}). Weird that it was still necessary since I was using the right version.