bug version 0.23.1? about chainconsumer

[Samreay]

Hi Matteo,

That is an interesting issue, I have no idea why that is happening, especially if the other plot method is working. If you are able to zip up the example and chains I can test on my machine and see if there are any issues that jump out.

The posx and poxy issues Ive seen before from having NaNs or infs in the chains, but those issues are easy to find in the stacktrace because its normally the trace points to the histogram. Here I am actually very much confused, because from what I can see chainconsumer doesn't appear in the stacktrace at all!

[MCostanzi]

Hi,
thanks a lot for the quick reply.
Acatully, since I wanted to be sure that it was not a problem of my chains I used your example "plot_introduction.ipynb" to test it. So the code I'm using for the test is:

np.random.seed(0)
data = np.random.multivariate_normal([0.0, 4.0], [[1.0, 0.7], [0.7, 1.5]], size=10000)

c = ChainConsumer()
c.add_chain(data, parameters=["$x_1$", "$x_2$"])
fig = c.plotter.plot(figsize="column", truth=[0.0, 4.0])

(I've explicitly checked that there are no Nans in the data. Moreover, other library such as "corner" works with the same data vector)

[Samreay]

How odd, that code definitely works, so Im wondering if it is an issue with the environment (ah yes, now I see your original sentence highlighting it was an intro example!)

Can you confirm that you get the same error if you download the pure python code and run it, or if it only happens with the ipython environment?

[MCostanzi]

Mmm, interesting.
So I tried to use the pure python code.
If I just ask to display the plot, i.e.:

fig = c.plotter.plot(display=True, figsize="column", truth=[0.0, 4.0])

I have no problem

If instead I try to save it:

fig = c.plotter.plot(filename="/home/costanzi/Desktop/test.png", figsize="column", truth=[0.0, 4.0])

I got the same error.

[Samreay]

I have no idea, however in this case a generic conda update conda (along with an update to matplotlib, numpy, jupyter and ipython) might work, but Im not sure if youre already on the latest version or not. The other option might be changing the backend graphics ipython is using, but Im not quite sure how to do that in whatever environment you've set up. Happy to help debug if you can provide some environment details, provided running all those updates doesnt fix it. Cheers

…

On Mon, Dec 11, 2017 at 10:35 PM, MCostanzi ***@***.***> wrote: Argh, you are right. If I run the pure python code I have no such problem. Do you have any idea how to solve this environment issue? (sorry to bother you :P ) — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#47 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AGb5WUxixUQkqFaNis829arjKC7fkI9aks5s_ROdgaJpZM4Q9Jql> .

[MCostanzi]

I see, I'll try to do these updates.

Thanks a lot!

[Samreay]

Seems like there is a bug in the latest version of matplotlib which is affecting many users. Ive updated CC to force a good matplotlib version, so please update cc and hopefully all the posx posy issues will disappear and youll be able to save figures again.

[MCostanzi]

Great, it works again now! Thanks a lot :)

[Samreay]

Glad it worked!