missing imshow() subplots when using tight_layout()

Hello, I've found an unexpected behavior when I make grids of subplots with imshow and tight_layout the second to last plot is missing, for example:

import numpy as np
import matplotlib.pyplot as plt
plt.figure()
for i in range(16):
    ax= plt.subplot(4,4 ,i+1)
    im=ax.imshow(np.random.normal(size=100).reshape([10,10]))
    plt.tight_layout()
    plt.title(i)
plt.savefig("imshow_4x4.png")

generates:

Whereas the same loop with pcolor works:

import numpy as np
import matplotlib.pyplot as plt
plt.figure()
for i in range(16):
    ax= plt.subplot(4,4 ,i+1)
    im=ax.pcolor(np.random.normal(size=100).reshape([10,10]))
    plt.tight_layout()
    plt.title(i)
plt.savefig("pcolor_4x4.png")

gives:

If I omit the tight_layout() call from the first example it works:

I'm using matplotlib 1.4.3 and numpy 1.9.2 in Anaconda 2.2.0.

... what

I can reproduce this on current master.

plt.tight_layout() is removing an axes from the fig.axes list.

There is a simple work around here, just call plt.tight_layout once outside of the loop and it works correctly.

yes, the work around works, thanks. Calling tight_layout inside the loop is admittedly an inefficient approach but not one I expected to delete a figure...

I agree, something very strange is going on here. Removing the call to imshow makes it work correctly. The bug is in some interaction between plt.tight_layout and plt.subplot(...).

I suggest re-writing this loop as

import numpy as np
import matplotlib.pyplot as plt

fig, axes = plt.subplots(4, 4)

for i, ax in enumerate(axes.ravel()):
    im = ax.imshow(np.random.normal(size=100).reshape([10,10]))
    ax.set_title(i)


plt.tight_layout()

That is f-ed up. I wonder if the axes isn't really removed, just misplaced?
On Aug 21, 2015 4:50 PM, "Thomas A Caswell" notifications@github.com
wrote:

plt.tight_layout() is removing an axes from the fig.axes list.

There is a simple work around here, just call plt.tight_layout once
outside of the loop and it works correctly.

—
Reply to this email directly or view it on GitHub
#4976 (comment)
.

This is my attempted fix of the problem

It seems like tight_layout is not removing the subplots but when you tried to call tight_layout on the figure it resized the subplots that were already on it(which made them bigger because it had avaliable space), then when you add another subplot to the figure with the already resized plot they will disappear because subplot is plotting in their region.

I'm not really sure if this is a good fix though.
https://github.com/Hongyu1230/matplotlib/commit/f21fc17117ee67cf39dae218bf88ea1ef2d572fc

@hamgyu1230 Can you open a pull request with that change (with a better commit message!).

Can you explain why clipping these values at 0 fixes the problem?

#6096

by making these at 0, the subplot can only shrink or stay the same even when there is avaliable space, thus the existing subplots shouldn't be overwritten by any new ones when a new subplot is plotted.

Was this fixed? I have the same problem

Not really, but if you just call tight_layout once, at the end, it all works fine.

This is caused by fully_overlaps returning True for the last axes, and hence the old one gets deleted, as pointed out by #7393. What happens is that the "tight" axes is larger and the new axes overlaps the old axes and deletes it. I'm not sure what we can do to fix this, since the stated behaviour is that if you (partially) cover an axes it gets deleted. Better is to create the axes first (fig, axs = plt.subplots(4, 4)) and then call tight_layout, hence avoiding too much magic.

I actually think the "remove overlapping subplots" behavior is pretty bad and should be deprecated and removed; see #11435 and #7377 (comment).

I agree, but I'm not sure how you deprecate that without a huge rigamarole.

I think(?) you just emit a warning if an axes that would have been removed under the old behavior will no longer be removed with the new one? (telling whoever was relying on the old behavior that they need to explicitly call ax.remove() to suppress the warning)
That means no one can actually benefit from the new behavior until the deprecation elapses, but that seems OK?

@jklymak

I have the same problem

MacOS, matplotlib 3.1.3, MacOSX backend

In my case, @jklymak 's workaround does not work. Once in a while some figures disappear.

I do not want to call subplots since it will create a new figure but I want to use the old one.

code:

        plt.figure(1, figsize=(14, 5))
        plt.clf()
        nr, nc = 1, 2

        ax1 = plt.subplot(nr, nc, 1)
        plt.imshow(im_in, cmap='gray', vmin=0, vmax=255)
        for tbl_idx, tbl in enumerate(all_trace_back_lines):
            plt.plot(np.arange(0, tbl.yarr.shape[0], 1.0), tbl.yarr, c=c, linewidth=w)
            plt.scatter([tbl.x_seed], [tbl.y_seed], c=c, s=3)
            plt.text(tbl.x_seed, tbl.y_seed, f'#{tbl_idx}', size=8)

        plt.gca().add_collection(mc.LineCollection([
            [(20, 20), (20, 20 + global_average_text_scale)]
        ], linewidths=2, colors='orange'))
        plt.text(20, 20, f'{global_average_text_scale:.1f}', size=5)

        plt.subplot(nr, nc, 2, sharex=ax1, sharey=ax1)
        plt.imshow(im_vis_r2l_dp)

        plt.tight_layout()
        plt.show()

example result (once in a while)

Here you are only calling tight layout at the end. If this is indeed reproducible then please provide a self-contained example and open a new issue. If it's not reproducible, because you are really calling tight_layout in the middle of the process sometimes, then it is the same bug. Also if you want to add subplots to an existing figure just call fig.subplots(1,2)

@jklymak

@jklymak Thanks I will look at it to see how to reproduce it

@jklymak

plt.tight_layout() is removing an axes from the fig.axes list.

There is a simple work around here, just call plt.tight_layout once outside of the loop and it works correctly.

@jklymak Thanks you saved my day.

Please why is this code not running
plt.figure(figsize=(10, 10))
for image, labels in data_train.take(1):
for i in range(9):
plt.subplot(3,3,i+1)

    plt.imshow(image[i]).numpy.()astype ('uint8'))
    plt.title(data_cat[labels[i]])
    plt.axis('off')

@Godfreyoise

@Godfreyoise for usage questions please go to https://discourse.matplotlib.org/. You will need to provide a complete code example so that others can reproduce the problem.

Navigation Menu

Repository navigation

missing imshow() subplots when using tight_layout() #4976

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

missing imshow() subplots when using tight_layout() #4976

Description

Activity

tacaswell commented on Aug 21, 2015

tacaswell commented on Aug 21, 2015

douglase commented on Aug 21, 2015

tacaswell commented on Aug 21, 2015

WeatherGod commented on Aug 22, 2015

Hongyu1230 commented on Mar 2, 2016

tacaswell commented on Mar 2, 2016

Hongyu1230 commented on Mar 2, 2016

adizhol commented on Sep 14, 2021

jklymak commented on Sep 14, 2021

jklymak commented on Sep 14, 2021

anntzer commented on Sep 14, 2021

jklymak commented on Sep 14, 2021

anntzer commented on Sep 14, 2021

fzyzcjy commented on Sep 24, 2021

jklymak commented on Sep 24, 2021

fzyzcjy commented on Sep 24, 2021

back2yes commented on Nov 30, 2021

Godfreyoise commented on Mar 17, 2024

Godfreyoise commented on Mar 17, 2024

Godfreyoise commented on Mar 17, 2024

rcomer commented on Mar 17, 2024

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions