Matplotlib: tick labels are inconsist with font setting (LaTeX text example)

Question

I have the following simple python code:

import numpy as np
import matplotlib.pyplot as plt 

plt.rc( 'font', size=20, family="Times" )   # use a font with serifs

# the following line triggers the problem
plt.rc( 'text', usetex=True )               # activate LaTeX text rendering

fig = plt.figure( figsize=(8,6) )           # (width,height) in inches
ax1 = fig.add_subplot( 1, 1, 1 )            # rows cols plotnumber

ax1.plot( np.linspace(1,10,10), np.linspace(1,10,10)**2 )

ax1.set_xlabel( r'\textit{x} in a.u.' )
ax1.set_ylabel( r'\textit{y} in a.u.' )

plt.show()

This results in the following figure:

As you can see, the tick-labels have a too thin font compared with the axes-labels (or the axes-labels are too thick). I have found out that this is due to activating the LaTeX text rendering (see comment in the code), but I have no clue how to change this as I do not want to switch the LaTeX text rendering off.

Any idea why the font-thickness (what is the plural of thickness?) is inconsistent and how to change that?

Update 1: Following the suggestion from llap42, a hack would be to do

plt.xticks([2, 4, 6, 8, 10], ['2', '4', '8', '10' ])

But that is only a hack and there has to be a better solution.

Ninfaningal

looks like usetex only applied to the labels. If you have a number in the label, is it thicker? (btw: thicknesses) – Ence

@ElmarPeise numbers in the label are getting thicker, so usetex also applies to numbers – Ninfaningal

Does the ordering of the calls to plt.rc() matter? i.e. have you tried plt.rc('text', usetex=True) first, and then plt.rc('font', size=20, family="Times")? – Kosak

#12921300 – Kosak

@llap42 just tried it: the ordering does not matter, will now look into the suggested link – Ninfaningal

The problem is that the ticklabels do not obay to the font setting. This can better be seen when trying plt.rc('font',family = 'sans-serif', size=20). As can be seen the labels are sans-serif, but the ticklabels do not change. So instead of asking about 'font thickness', you should reword your question to why the ticklabels do not follow the font settings. – Pitching

I put an issue about this on the matplotlib page: github.com/matplotlib/matplotlib/issues/8436 – Pitching

@Pitching indeed, good point. I guess you can put that into the answer, since it seems to be the only "proper" (still a hack, but not as bad as the one I put into the update) solution then at the moment. – Ninfaningal

Answer 1

As said in the comments, this is rather an issue of ticklabels not obeying the font setting when used with latex.

This issue seems to only occur when using a ScalarFormatter (which is the default formatter for axes). I've posted an issue about this on GitHub.

A workaround may be to use a different Formatter. E.g a StrMethodFormatter:

import matplotlib.pyplot as plt 
import matplotlib.ticker

plt.rc( 'text', usetex=True ) 
plt.rc('font',family = 'sans-serif',  size=20)

fig , ax = plt.subplots(figsize=(5,3))

ax.set_xlabel( r'\textit{x} in a.u.' )
ax.set_ylabel( r'\textit{y} in a.u.' )

fmt = matplotlib.ticker.StrMethodFormatter("{x}")
ax.xaxis.set_major_formatter(fmt)
ax.yaxis.set_major_formatter(fmt)

plt.tight_layout()
plt.show()

Pitching

yes, that works, thanks also for opening an issue on GitHub – Ninfaningal

Can this be used somehow for power notation in log axes? – Carsoncarstensz

Thanks! I have trouble with this problem for a whole afternoon! – Walhalla

Answer 2

An alternative solution is to set the font latex uses to a sans-serif font. How to achieve this is explained over on tex.stackexchange.

One solution would be to use the sfmath latex package. To add it to the preamble, use

    plt.rc('text.latex', preamble=r'\usepackage[cm]{sfmath}')

This also works for logarithmic scales, where the other proposed solution fails.

Esdraelon