Enable type1 fonts in matplotlib

[ppurka]

Essentially what the ticket title says. Some publications require all the fonts to be embedded in the document and require the use of type-1 fonts. It is not very hard to enable these fonts.

---

Apply trac_14664-type1-fonts.patch​ to devel/sage

[vbraun]

This isn't really about type 1 fonts, right? You are switching TeXing labels on/off. How about calling it plot(latex=bool).

It would be nice if we could autodetect the prerequisites and default to the right thing.

[ppurka]

It is about type 1 fonts, as far as I can tell. The first thing publications tell is that type 3 fonts should not be present in the document. Some of them silently fail when you try to upload your document - the upload website just says that the upload didn't succeed.

As for references, I found these links earlier:

• ​http://nerdjusttyped.blogspot.sg/2010/07/type-1-fonts-and-matplotlib-figures.html
• ​http://matplotlib.1069221.n5.nabble.com/Type-1-font-in-figures-needed-td10294.html
• ​http://brainlogging.wordpress.com/2011/11/16/type-1-vs-3-fonts/

About using latex=bool: Unless someone knows that matplotlib has its own math renderer and doesn't use latex by default for rendering math, it will not be clear that something like latex=True will switch fonts to type 1 fonts.

[vbraun]

rcParams['text.usetex'] = False means that you disable TeX entirely, right? Maybe we want separate plot(usetex=<bool>, type1_fonts=<bool>)? Or plot(tex=[None(default)|True|False|"type1"])?

[ppurka]

False is the default. You can try the following in your sage installation

sage: from matplotlib import rcParams
sage: rcParams['text.usetex']
False

matplotlib ​does not require latex to render math fonts.

[vbraun]

What I'm saying is: Whether or not to use LaTeX is completely orthogonal to which fonts to use. But your patch links type1_fonts to both. You should be a able to use LaTeX without Type1 and the built-in renderer with Type1 fonts.

[vbraun]

The following setting might be relevant:

sage: matplotlib.rcParams['pdf.fonttype']
3

[ppurka]

Replying to vbraun:

What I'm saying is: Whether or not to use LaTeX is completely orthogonal to which fonts to use. But your patch links type1_fonts to both. You should be a able to use LaTeX without Type1 and the built-in renderer with Type1 fonts.

I am not sure if the latter is possible.

But you are right about decoupling latex from type1. I will make that change.

[ppurka]

Replying to vbraun:

The following setting might be relevant:

sage: matplotlib.rcParams['pdf.fonttype']
3

From an initial reading, I believe this is something completely different and is only used with the internal matplotlib tex engine.

[ppurka]

I changed the option to font_style. It take three strings now, "type1", "latex" and "default". In my limited testing just turning on "latex" by itself gives me embedded type 1 fonts, but maybe there are cases where this is not true.

[vbraun]

Font style means serif/sans serif. How about typeset="latex"?

[ppurka]

Thanks. I have changed it to typeset.

[vbraun]

I get fuzz 2 with sage-5.10.rc0

[vbraun]

• can we get a ValueError if typeset is not one of the allowed values?
• If you change matplotlib rcParams then you should switch it back when you are finished

[ppurka]

Replying to vbraun:

• can we get a ValueError if typeset is not one of the allowed values?

I will do that in a day or two. I need to compile 5.10rc0 first.

• If you change matplotlib rcParams then you should switch it back when you are finished

Not resetting the rcParams actually affects only those plots which are a mixture of Sage and (direct) matplotlib calls. Currently, there is nothing in Sage that actually does this. I will make the change nevertheless because there might be someone doing this manually.

[ppurka]

Updated the patch. The "default" does not mess with the settings at all now.

[vbraun]

Looks good but unfortunately conflicts with #14640. The latter has already positive review, so you'll have to rebase your patch.

[ppurka]

Done. Thanks.

[vbraun]

You are restoring rc_backup before saving it; The fuzz 2 ended up inserting a hunk at the wrong place.

[ppurka]

So sorry. I should have checked the git diff. Thanks for catching this.

[ppurka]

Apply to devel/sage

[ppurka]

ok. Now it should be done.

[vbraun]

Looks good to me