Fontconfig directives will help you control our font substitution rules. While synthetic font compositions are not a complete substitute to good pan-unicode fonts , they're the next best thing. You should always consider adding fontconfig tuning to your font packages.
Choosing a ruleset numeral prefix
Fontconfig will sort the ruleset files dropped in /etc/fonts/conf.d/ by filename before evaluating them. Therefore the rules of 01-xxx.conf and 02-yyy.conf will have the same effect a single 01-xy.conf file containing first xxx rules then yyy rules.
Usually that means the files with the smaller prefix will have higher precedence. Unfortunately the fontconfig syntax is pretty flexible and allows prepending effects to existing rules. Careful testing is therefore required if your package deploys rules that interact with the existing fontconfig ruleset.
When in doubt ask the Fedora fontconfig maintainer.
You as a packager know your font is a good substitute to another one, because it has the same metrics, has been forked from the same root, etc. fontconfig unfortunately has no such knowledge. You can teach fontconfig to consider your font first when an application requests this other font by using the following rules.
For releases ≥ Fedora 9
<alias binding="same"> <family>Name of the font to substitute for</family> <accept> <family>Name of your font</family> </accept> </alias>
For previous releases
<match> <test name="family"> <string>Name of the font to substitute for</string> </test> <edit name="family" mode="append" binding="same"> <string>Name of your font</string> </edit> </match>
Fontconfig is usually able to detect the font characteristics itself. Sometimes it needs a little help though:
<alias> <family>Name of your font</family> <default> <family>Generic like sans-serif, serif or monospace</family> </default> </alias>
Registering a font in default families
Fontconfig rules can also be used to add fonts in the substitution lists used to define synthetic default fonts like sans-serif. The effects of those rules are highly dependent on the numeral prefixes of the filenames used.
Simple priority lists
The substitution lists for default families are built using the following blocks:
<alias> <family>Generic like sans-serif, serif, monospace, fantasy or cursive</family> <prefer> <family>Name of your font</family> </prefer> </alias>
When matching a family, fontconfig will use the first suitable font in its substitution list. Thus <family> elements within a <prefer> block are ordered by descending priority, and rulesets with low numeral prefixes take precedence over rulesets with high numeral prefixes:
|51||Local system overrides|
|55-59||High priority latin fonts (distribution general-purpose default fonts)|
|60||Main latin font list|
|61-64||Low priority latin fonts|
|65-69||Fonts with less common encodings, ending with fonts that provide coverage of exotic unicode blocks at the expense of drawing quality|
The main default of simple substitution lists is they order whole fonts without taking the language being rendered into account. This would work fine if fonts never overlapped. All but the simplest real-world fonts, however, cover multiple scripts. And the Unicode consortium itself made the dubious choice of regrouping different regional variants of some glyphs at unique codepoints.
- If a font contains a very desirable drawing of script A, and a botched drawing of script B, you can not assign it a high priority for the sake of script A users without making script B users miserable. Since there are only so many good free/libre-open fonts, if the first group is numerous avoiding this font for the sake of the second group may not be possible.
- Sometimes the drawing conventions of the same script are not the same in different locales, so putting any font that contains this script in a default font substitution list will make some users miserable. Yet you do need to support users of this script in default fonts too.
Therefore it's very difficult to create a setup that pleases everyone using fontconfig default substitution lists only. Some locales will be sacrificed, which is not what we want. We need to complement the default lists using another method:
<match> <test name="lang"> <string>Locale code like zh-cn</string> </test> <test name="family"> <string>Generic like sans-serif, serif, monospace, fantasy or cursive</string> </test> <edit name="family" mode="prepend" binding="same"> <string>Name of your font</string> </edit> </match>
This rule will override default substitution lists for a particular locale. Since it uses the prepend mode, it follows an inverted priority: rulesets with high numeral prefixes take precedence over rulesets with low numeral prefixes. If you need to ship such a rule, please use the following prefix range:
- Using them in digital documents, for example, will expose you to nasty surprises when trying to render the result on other systems
- Typically some script + latin, but more complex combinations are common. Users hate switching fonts when typing text so font designers will try to draw every script a target group may use in a single font.
- Unless a smart font that contains every regional variant of this script thanks to the locl Opentype feature is available.