I have a matplotlib script that starts ...
import matplotlib as mpl
import matplotlib.pyplot as plt
import matplotlib.font_manager as fm
mpl.rcParams['xtick.labelsize']=16
...
I've used the command
fm.findSystemFonts()
to get a list of the fonts on my system. I've discovered the full path to a .ttf file I'd like to use,
'/usr/share/fonts/truetype/anonymous-pro/Anonymous Pro BI.ttf'
I've tried to use this font without success using the following commands
mpl.rcParams['font.family'] = 'anonymous-pro'
and
mpl.rcParams['font.family'] = 'Anonymous Pro BI'
which both return something like
/usr/lib/pymodules/python2.7/matplotlib/font_manager.py:1218: UserWarning: findfont: Font family ['anonymous-pro'] not found. Falling back to Bitstream Vera Sans
Can I use the mpl.rcParams dictionary to set this font in my plots?
EDIT
After reading a bit more, it seems this is a general problem of determining the font family name from a .ttf file. Is this easy to do in linux or python ?
In addition, I've tried adding
mpl.use['agg']
mpl.rcParams['text.usetex'] = False
without any success
Whew I made it in under 100 lines, @nim this also explains in more detail how dangerous it is, some modifications change completely the behaviour of the font property and font size.
prerequisites: Matplotlib and a font folder at the same level of the script containing the ttf font-file calibri.ttf
But this is what I have for you as an easter egg:
Specifying a font family:
If all you know is the path to the ttf, then you can discover the font family name using the
get_name
method:Specifying a font by path:
You can use the fc-query myfile.ttf command to check the metadata information of a font according to the Linux font system (fontconfig). It should print you names matplotlib will accept. However the matplotlib fontconfig integration is rather partial right now, so I'm afraid it's quite possible you'll hit bugs and limitations that do not exist for the same fonts in other Linux applications.
(this sad state is hidden by all the hardcoded font names in matplotlib's default config, as soon as you start trying to change them you're in dangerous land)