Repository navigation
tight_layout is incorrect for certain title sizes #8201
Description
I am using Python 3.4.4 |Anaconda 2.3.0 (64-bit) on win64 bits, with latest matplotlib and pyqt updated by conda install xx
for the following code, if I use unicode title in the for-loop, the images are too small. However, if I use English title, the images' size are expected
This can be observed for both TkAgg and Qt5Agg
#coding=utf-8
import matplotlib
matplotlib.rcParams['backend'] = 'TkAgg'
#~ matplotlib.rcParams['backend'] = 'Qt5Agg'
from pylab import *
import PIL.Image as Image
im=Image.open('Mario cosplay.jpg').convert('L')
orgsize=im.size
newsize=[int(i/4) for i in orgsize]
subplot(2,3,1)
imshow(im, cmap='gray', vmin=0, vmax=255)
title('original pic', fontsize=20);
filters=[Image.NEAREST, Image.BILINEAR, Image.BICUBIC, Image.ANTIALIAS]
filtersName=['NEAREST', 'BILINEAR', 'BICUBIC', 'ANTIALIAS']
for k in range(len(filters)):
subplot(2,3,k+2)
im1=im.resize(newsize, filters[k])
imshow(im1, cmap='gray', vmin=0, vmax=255)
#~ title('原图的1/%i，插值法：%s' % (4, filtersName[k]), fontsize=20) # bug here
title('size: 1/%i，interpolation: %s' % (4, filtersName[k]), fontsize=20) # no bug
gcf().tight_layout()
show()
Activity
[-]unicode title does not go well with gcf().tight_layout(), which yields bad image size[/-][+]tight_layout is incorrect for certain title sizes[/+]anntzer commentedon Apr 22, 2020
This actually does have anything to do with unicode. It's just that for certain title lengths, tight_layout shrinks the axes bbox in a vain attempt of decreasing the title size as well, but of course that doesn't work; for even longer titles it just gives up and doesn't shrink the axes.
Minimalish repro:
jklymak commentedon Apr 22, 2020
So, I guess its a valid question if
tight_layoutandconstrained_layoutshould even care about the title or xlabel for the x-direction calculation, or the ylabel for y-direction calculation. Shrinking the axes doesn't make more room. That makes the code more complicated, but maybe not terribly so.anntzer commentedon Apr 22, 2020
Ah, very good point, I guess we could just exclude them from tightbbox computation (but only in the "irrelevant" direction).