Contourf Colorbar

[lkilcher]

The contourf colorbar doesn't work as expected.

The following code:

import matplotlib.pyplot as plt
import numpy as np
import matplotlib.colors as mplc

x, y = np.mgrid[0:1:0.01, 0:1:0.01]
r = np.sqrt(x ** 2 + y ** 2)
z = np.sin(6 * np.pi * r)

norm = mplc.Normalize(0, 0.5)
fig0, ax0 = plt.subplots(1, 1, )
cf0 = ax0.contourf(x, y, z, np.arange(-1., 1.001, .01), norm=norm)
cbar0 = plt.colorbar(cf0,)

Produces the following figure:

Compare this with the colorbar of, e.g. a pcolormesh:

fig2, ax2 = plt.subplots(1, 1)
pc2 = ax2.pcolormesh(x, y, z, norm=norm)
cbar2 = plt.colorbar(pc2)

That's a nice colorbar, and it's what I expect the default behavior to be for contourf also. Is there a reason the contourf colorbar behaves this way? I've looked at using the vmin,vmax args, and the result is the same. Also, the result is the same for tricontourf colorbars as well.

[lkilcher]

I've already worked a little bit on creating a fix for this here. The issue seems to be related to trying to handle creation of colorbar contour lines (not contourf), but I have no idea if this fix causes other issues.

This fix above gives the colorbar I want:

[tacaswell]

That color bar does look "correct". You asked countourf to make contours from [-1, 1] so the color bar covers that range. Outside of [0, .5] the color maps to the limits of the colormap as you asked it to.

attn @efiring

[tacaswell]

If you do

import matplotlib.pyplot as plt
import numpy as np
import matplotlib.colors as mplc

x, y = np.mgrid[0:1:0.01, 0:1:0.01]
r = np.sqrt(x ** 2 + y ** 2)
z = np.sin(6 * np.pi * r)

norm = mplc.Normalize(0, 0.5)
fig0, ax0 = plt.subplots(1, 1, )
cf0 = ax0.contourf(x, y, z, np.arange(0, .5, .01), norm=norm,
                   extend='both', cmap='viridis')
cbar0 = plt.colorbar(cf0,)

you get

which is what I think your intent was.

[lkilcher]

Got it. I was unaware of the extend kwarg. Thanks and sorry. I was starting to think there is a need for a cbar.set_lim method, but I suppose this kind of functionality removes any need for it (i.e. one should always control the color limits via the mappable, rather than the cbar itself). Is that correct?

[tacaswell]

I could have sworn there was a cbar.set_lim method as I remember using it in the past, but I could not find it (which either means my memory is bad or we have that functionality someplace really strange). I suspect that what you say is true (that you should control the color bar limits through the mappable object), but defer that question to @efiring .

This is something where there are multiple places where it might make sense to be able to control it, we may not always be consistent on where it should be controlled, there is a high degree of 'do the right thing' required, and we can't break back compatibility.

I am only aware of the extend kwarg because there was another bug report related to it last week 😉 . I have a patch on the way that will make the extend on the color bar a bit more controllable.

Where would you have looked in the docs to try and find out this sort of stuff?

[lkilcher]

Looking more closely I see that it's in the contourf_demo.py, but it wasn't highlighted prominently enough for me to find it. Is it worth creating a Q+A on SO about how to do this (I don't see one that uses extend), or is this issue good enough?

[tacaswell]

I am indifferent to creating on SO post on this.

[WeatherGod]

Does the problem still happen if you remove the comma in the colorbar() call? Also, next time, please file a new issue rather than resurrecting an old issue, unless the problem is identical.

…

On Dec 26, 2016 3:54 AM, "Ankit Barik" ***@***.***> wrote: The 'extend' keyword gives a line instead of a colorbar. Matplotlib version: 2.0.0rc2+2827.gab98852 Numpy version: 1.13.0.dev0+2498b74 Compiler: gcc Here's the same piece of code generating the following `import matplotlib.pyplot as plt import numpy as np import matplotlib.colors as mplc x, y = np.mgrid[0:1:0.01, 0:1:0.01] r = np.sqrt(x ** 2 + y ** 2) z = np.sin(6 * np.pi * r) norm = mplc.Normalize(0, 0.5) fig0, ax0 = plt.subplots(1, 1, ) cf0 = ax0.contourf(x, y, z, np.arange(0, .5, .01), norm=norm, extend='both', cmap='viridis') cbar0 = plt.colorbar(cf0,)` [image: figure_1] <https://cloud.githubusercontent.com/assets/13730345/21478977/36db3784-cb51-11e6-9ee3-446fd3e19e9a.png> — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#5055 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AARy-BXatTQ8xPtfZxbO7xfV63hyZO3Aks5rL4DigaJpZM4F8eOy> .

[efiring]

The comment above refers to what has now been opened as #7686.