Ticklabel alignment issue with Arial

[asteppke]

While changing the styling of my plots I also tried to change the font used for the tick labels. Upon closer inspection you can notice that the alignment is somehow off for the y tick label "1" in this example:

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

rcParams = matplotlib.rcParams

rcParams['font.serif'] = ['Times New Roman']
rcParams['font.sans-serif'] = ['Arial']
rcParams['font.family'] = 'sans-serif'


fig = plt.figure(figsize=(6,3))
ax = fig.add_subplot(1,1,1)

t = np.arange(0.0, 1.0, 0.01)
s = 5*t**2+1+np.sin(2*np.pi*t)

line, = ax.plot(t, s, color='blue', lw=2)

ax.tick_params(axis='y', pad=14)        

plt.plot([0,1],[0,2])
plt.savefig('test.png', dpi=300, bbox_inches='tight')
plt.show()

Which results in the following plot:

For some reason the alignment of the tick labels is not consistent, the "1" is clearly closer to the axis lines than the other labels. This does not happen with the default font, so I assume this is due to some interaction between both. Unfortunately Arial or Helvetica is the standard in many publications so I would like to stick to this font.

This happens with matplotlib 1.4.3 on Windows 7 using Python 3.4.1.

[tacaswell]

It looks like the numbers are correctly right alligned. Does

[t.set_ha('center') for t in ax.yaxis.get_ticklabels()]

fix the alignment? It will probably also move the tick labels which I don't remember off the top of my head how to move (but I am sure you can).

[asteppke]

The issue is that the numbers are not correctly right aligned. It might be more obvious in the following picture where I zoomed in and added addtionally digits before the labels:

It seems that matplotlib is not using the correct bounding box for the character but something like a visual bounding box where the glyph "1" has a right margin of almost zero.

This is how right alignment should look like in other plotting packages:

Centering the labels only helps as a workaround in few cases if not more than one digit ever appears in the labels.

[EdwinChan]

I'm having a similar issue here. The code

import matplotlib

# slightly noticeable
matplotlib.rcParams['font.family'] = 'DejaVu Sans'
# slightly noticeable
matplotlib.rcParams['font.family'] = 'Liberation Serif'
# extremely noticeable
matplotlib.rcParams['font.family'] = 'Ubuntu'

matplotlib.rcParams['font.weight'] = 200
matplotlib.rcParams['font.size'] = 48
from matplotlib import pyplot
pyplot.plot([0, 5], [0, 5])
for i in range(3):
  pyplot.annotate(i+1, (0, 0), (1, i+2), 'data', 'data', ha='right')
pyplot.savefig('test.png')

generates the plot

The figure shows that both axis labels and annotations are affected by this problem. It appears that only the Agg backend exhibits this incorrect behavior, not the pgf backend, but I haven't tried other backends yet…

[tacaswell]

Does the same happen on master? @mdboom did some work on this recently.

On Wed, Jul 29, 2015, 9:12 PM Edwin Chan notifications@github.com wrote:

I'm having a similar issue. The misalignment is extremely noticeable using
the Ubuntu font:

[image: test]
https://cloud.githubusercontent.com/assets/906137/8973703/093e4dae-3636-11e5-999a-4401f58e9d0e.png

It appears the Agg backend suffers from this problem but not the pgf
backend. I haven't tried the other backends though…

—
Reply to this email directly or view it on GitHub
#4796 (comment)
.

[EdwinChan]

I've performed two little tests. The first test

import matplotlib
matplotlib.rcParams['font.family'] = 'Ubuntu'
matplotlib.rcParams['font.weight'] = 200
matplotlib.rcParams['font.size'] = 48
from matplotlib import pyplot
pyplot.plot([0, 5], [0, 5])
for i in range(3):
  pyplot.annotate(i+1, (0, 0), (1, i+2), 'data', 'data', ha='right')
pyplot.annotate(matplotlib.__version__, (0, 0), (2, 1), 'data', 'data',
  ha='left')
pyplot.savefig('test.png')

produces

The second test

import matplotlib
matplotlib.rcParams['font.family'] = 'Ubuntu'
matplotlib.rcParams['font.weight'] = 200
matplotlib.rcParams['font.size'] = 48
from matplotlib import pyplot
for i in range(10):
  pyplot.plot([0, 5], [0, 5])
  pyplot.annotate('{:.2f}'.format(i/100), (0, 0), (4, 1), 'data', 'data',
    ha='right')
  pyplot.savefig('test{}.png'.format(i))
  pyplot.clf()

produces a chain of images. When the images are flipped through with your favorite image viewer, the 'timer' in the lower right doesn't jitter around like it did in matplotlib 1.4.3.

[mdboom]

So, it appears this is fixed in git, then?

[EdwinChan]

It seems so.

[tacaswell]

Great, closing as fixed on master.

@asteppke Please ping to have this re-opened if you can reproduce this on the current master branch.