On 08/23/2010 11:08 AM, Daniel Hyams wrote:
Thanks for your response Michael!

Deleting the fontList.cache indeed solved the problem. The first sample that I gave below now works fine under both Windows and Linux; whew...I would have never known about that cache file without your help.

So, should the fontList.cache file be deleted each time my app runs, in case new fonts have been installed since the last invocation? How do we keep that cache up to date?
The cache exists because it can take a long time to read through all of the font files on the system each time matplotlib is imported. It might be fast enough to simply stat the dates on all the files in the font directories (without reading their contents)... I don't know if experiments were done on this when the cache was originally added (that pre-dates me).

Mike


On Mon, Aug 23, 2010 at 10:12 AM, Michael Droettboom <md...@stsci.edu <mailto:md...@stsci.edu>> wrote:

    On 08/22/2010 10:00 PM, Daniel Hyams wrote:
    I am searching for advice on how to handle selecting a specific
    font, and using that in a matplotlib figure.  As a background,
    the font will be picked through the wx.FontDialog (common font
    dialog) provided by wxPython.  So, what I will have is the font
    face (Arial, Times New Roman, Algerian, etc. etc.), the weight,
    the style (italic, normal) and the point size.  All I want to do
    is create a matplotlib font that matches this, and use it in the
    plot.  My first try was this:

    import matplotlib.pyplot as plt
    import matplotlib.font_manager as fm

    # this does not work.  The title font is wrong.
    the_font = 'Century Schoolbook'
    fp = fm.FontProperties()
    fp.set_name(the_font)
    fp.set_size(24)
    plt.title('The Title that should be in Century
    Schoolbook',fontproperties=fp)
    print fp
    plt.show()

    But that didn't work.  I know that "Century Schoolbook" is not
    really a font family, but in the docs it says that you can list a
    font there.
    On my RHEL5 system (at least), the full name of that font is
    "Century Schoolbook L".  You do, of course, have to match the font
    name exactly.

    You may want to experiment with using wx.Font.GetFaceName vs.
    wx.Font.GetFamilyString from the font chooser dialog and see if
    either returns the correct result.

    You can see the list of fonts that matplotlib found on your system
    by doing:

    from matplotlib import font_manager
    font_manager.fontManager.ttflist

    (This is not a public API -- but may help with debugging the issue
    here).

    Is the font you are hoping to match with in that list?

    Another thing you may want to try is deleting matplotlib's font
    cache, in case this font was installed after matplotlib was first
    run on your system.  It is is in a file "fontList.cache" and lives
    in the user data area (can't remember the exact location of this
    on Windows off hand).


    The following does work, if I manually set the TrueType file
    explicitly:

    fp = fm.FontProperties()
    fp.set_file('c:\\Windows\Fonts\CENSCBK.TTF')
    fp.set_size(24)
    print fp
    plt.title('The Title that is in Century
    Schoolbook',fontproperties=fp)
    plt.show()

    So I guess the question is....how does one accomplish this,
    portably?  I don't quite understand the ins and outs of fonts.....


    p.s.
    I did take a stab at creating a mapping between the font names /
    weights / styles like this:

    all_fontfiles = fm.win32InstalledFonts()
    allfonts = fm.createFontList(all_fontfiles)
    fontdict = {}
    for f in allfonts:
        fontdict[(f.name <http://f.name>,f.style,f.weight)] = f.fname

    And I think I can get this to work, because this maps me to a TTF
    file for any name, style, and weight combination.  But this
    seemed awfully hacky, and I don't know what problems I'll run
    into on other platforms (obviously, I would have get
    all_fontfiles above differently on each platform).  If this is
    the only way to do it, I guess that's OK, but I thought that
    surely there was a better way.
    That's basically a stricter way of doing what the matplotlib code
    already does.  The matplotlib lookup is actually smarter because
    it handles approximate "nearest neighbor" matches.  I think the
    problem is more likely one of the above (an inexact name match or
    an out-of-date font cache).

    Mike

-- Michael Droettboom
    Science Software Branch
    Space Telescope Science Institute
    Baltimore, Maryland, USA


    
------------------------------------------------------------------------------
    This SF.net email is sponsored by

    Make an app they can't live without
    Enter the BlackBerry Developer Challenge
    http://p.sf.net/sfu/RIM-dev2dev
    _______________________________________________
    Matplotlib-users mailing list
    Matplotlib-users@lists.sourceforge.net
    <mailto:Matplotlib-users@lists.sourceforge.net>
    https://lists.sourceforge.net/lists/listinfo/matplotlib-users




--
Daniel Hyams
dhy...@gmail.com <mailto:dhy...@gmail.com>


------------------------------------------------------------------------------
This SF.net email is sponsored by

Make an app they can't live without
Enter the BlackBerry Developer Challenge
http://p.sf.net/sfu/RIM-dev2dev


_______________________________________________
Matplotlib-users mailing list
Matplotlib-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/matplotlib-users


--
Michael Droettboom
Science Software Branch
Space Telescope Science Institute
Baltimore, Maryland, USA

------------------------------------------------------------------------------
This SF.net email is sponsored by 

Make an app they can't live without
Enter the BlackBerry Developer Challenge
http://p.sf.net/sfu/RIM-dev2dev 
_______________________________________________
Matplotlib-users mailing list
Matplotlib-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/matplotlib-users

Reply via email to