Colorbar not aligning with custom layout

[ybmy001]

cmap='managua' not exists, which listed in https://ultraplot.readthedocs.io/en/latest/colormaps.html

# bug 1
import ultraplot as uplt
import numpy as np

data = np.arange(9).reshape(3, 3)
fig, ax = uplt.subplots()
img = ax.pcolormesh(data, cmap='managua')  # ValueError: Invalid colormap, ...
ax.colorbar(img, loc='r')
fig.savefig(r"D:\bug_managua.png", dpi=500)

[ybmy001]

Bug2, I just want to keep all axis are equal length, but the default behavior is not equal.

import ultraplot as uplt
import numpy as np

array = [
    [1, 1, 2, 2],
    [3, 3, 4, 4],
    [None, 5, 5, None],
]

data = np.arange(9).reshape(3, 3)

fig, axs = uplt.subplots(array, share=0)
for i, ax in enumerate(axs):
    img = ax.pcolormesh(data, cmap='Blues')
    ax.colorbar(img, loc='r')
# axs.set_aspect(1)
fig.savefig(r"D:\bug2.png", dpi=500)

[ybmy001]

When I set axs.set_aspect(1), all axes are equal, but the spaces of all colorbar are not equal. I will adjust space of colorbar next.

import ultraplot as uplt
import numpy as np

array = [
    [1, 1, 2, 2],
    [3, 3, 4, 4],
    [None, 5, 5, None],
]

data = np.arange(9).reshape(3, 3)

fig, axs = uplt.subplots(array, share=0)
for i, ax in enumerate(axs):
    img = ax.pcolormesh(data, cmap='Blues')
    ax.colorbar(img, loc='r')
axs.set_aspect(1)
fig.savefig(r"D:\bug2.png", dpi=500)

[ybmy001]

When add space=0.1 in ax.colorbar(), the result is strange more.

import ultraplot as uplt
import numpy as np

array = [
    [1, 1, 2, 2],
    [3, 3, 4, 4],
    [None, 5, 5, None],
]

data = np.arange(9).reshape(3, 3)

fig, axs = uplt.subplots(array, share=0)
for i, ax in enumerate(axs):
    img = ax.pcolormesh(data, cmap='Blues')
    ax.colorbar(img, loc='r', space=0.1)
axs.set_aspect(1)
fig.savefig(r"D:\bug2.png", dpi=500)

[cvanelteren]

These are not bugs but the tight layout algorithm working behind the scenes.
I cannot reproduce the not finding the colormap you listed, for me it can be loaded just fine:

fig, ax = uplt.subplots()
ax.pcolormesh(np.random.rand(20, 20), cmap="managua")
uplt.show(block=1)

[cvanelteren]

If you want the axis to be truly square you should perhaps consider using imshow rather than pcolormesh.

[cvanelteren]

Closing this. Feel free to re-open if there are issue I missed.

[syrte]

The issue of spacing seems unresolved.
The spaces of all colorbars are not equal, even when the spacing is set explicitly.
@cvanelteren mentioned this is the expected behavior of the tight layout algorithm. Is there any way to get equal spacing?

import ultraplot as uplt
import numpy as np

array = [
    [1, 1, 2, 2],
    [3, 3, 4, 4],
    [None, 5, 5, None],
]

data = np.arange(9).reshape(3, 3)

fig, axs = uplt.subplots(array, share=0, aspect=1)
for i, ax in enumerate(axs):
    img = ax.imshow(data, cmap='Blues')
    ax.colorbar(img, loc='r', space='0em', pad='0em')

[cvanelteren]

The issue is that grispec is compensating for the last plot (number 5). I haven't had time to look at how UltraPlot is implementing this exactly but you can see the layout mechanisms working by changing the array to for example:

array = [
    [1, 1, 1, 2, 2, 2],
    [3, 3, 3, 4, 4, 4],
    [None, None, 5, 5, None, None],
]

Yielding:

this is closer to what you want but not perfect as can be seen by the misalignment of the colorbar on plot number 5. I think the issue is that you are splitting the axes across plots which somehow interferes with how gridspec is modifying the locations.

We are missing a lot of tests for that part of the codebase -- and I would need more time to properly look how the gridspec and layout algorithm is performed, any help would be of course appreciated. I can leave the issue open for now, but I don't expect to have time to working on this before closing the other PRs and then there is summer ;-).

Swapping out to pcolormesh above would work, i.e.:

import ultraplot as uplt
import numpy as np

array = [
    [1, 1, 1, 2, 2, 2],
    [3, 3, 3, 4, 4, 4],
    [None, None, 5, 5, None, None],
]

data = np.arange(9).reshape(3, 3)

fig, axs = uplt.subplots(array, share=0)
for i, ax in enumerate(axs):
    img = ax.pcolormesh(data, cmap="Blues", colorbar="r", colorbar_kw=dict(pad=0))

looks pretty good I think.

[cvanelteren]

Apologies for not being able to answer this with more detail -- there are thousands of lines of code to get through. I will do my best but can only do so much in the time I am given. I do know that ultraplot does the following

• modifies gridspec to be flat, i.e. no nesting allowed
• adds space next to each plot in each cardinal direction for panels and or colorbars (they are hidden by default)

I suspect that the layout algorithm goes wrong in the last step as the colorbar will take some "space" that is variable and the alignment would then move the position in a particular way (many question marks on how this is exactly done, but this is my understanding so far).

[syrte]

Sorry that I am not an expert and don't have time to dig into it in detail at the moment.
Just for future reference: using pcolormesh does not solve the problem, as shown in @ybmy001 's early example.
Indeed, you will get equal spacing between axes and colorbars, but now the axes in the same row have different widths.

As a temporary imperfect workaround, maybe ybmy001 can use aligned grids for now...

[cvanelteren]

Another alternative would be to use a fig colorbar -- but that would only work if the data is in the same range.