Skip to content

FontForge gains ability to reuse OpenType rules for different fonts

Thursday, 22 September 2022 | Rajeesh K Nambiar

FontForge is the long standing libre font development tool: it can be used to design glyphs, import glyphs of many formats (svg, ps, pdf, …), write OpenType lookups or integrate Adobe feature files, and produce binary fonts (OTF, TTF, WOFF, …). It has excellent scripting abilities, especially Python library to manipulate fonts; which I extensively use in producing & testing fonts.

When I wrote advanced definitive OpenType shaping rules for Malayalam and build scripts based on FontForge, I also wanted to reuse the comprehensive shaping rules in all the fonts RIT develop. The challenge in reusing the larger set of rules in a ‘limited’ character set font was that FontForge would (rightly) throw errors that such-and-such glyph does not exist in the font and thus the lookup is invalid. For instance, the definitive OTL shaping rules for Malayalam has nearly 950 glyphs and lookup rules; but a limited character set font like ‘Ezhuthu’ has about 740 glyphs.

One fine morning in 2020, I set out to read FontForge’s source code to study if functionality to safely skip lookups that do not apply to a font (because the glyphs specified in the lookup are not present in the font, for instance) can be added. Few days later, I have modified the core functionality and adapted the Python interface (specifically, the Font.mergeFeature method) to do exactly that, preserving backward compatibility.

Next, it was also needed to expose the same functionality in the graphical interface (via FileMerge Feature info menu). FontForge uses its own GUI toolkit (neither GTK nor Qt); but with helpful pointers from Fredrick Brennan, I have developed the GUI to take a flag (default ‘off’ to retain backward compatibility) that allows the users to try skipping lookup rules that do not apply to the current font. In the process, I had to touch the innards of FontForge’s low-level code and learn about it.

Fig. 1: Fontforge now supports skipping non-existent glyphs when merging a comprehensive OpenType feature file.

This worked fine for our use case, typically ignoring the GSUB lookups of type sub glyph1 glyph2 by glyph3 where glyph3 does not exist in the font. But it did not properly handle the cases when glyph1 or glyph2 were non-existent. I’ve tried to fix the issue but then was unable to spend more time to finish it as Real Life™ caught up; c’est la vie. It was later attempted as part of Free Software Camp mentoring program in 2021 but that didn’t bear fruit.

A couple of weeks ago, Fred followed up now that this functionality is found very useful; so I set aside time again to finish the feature. With fresh eyes, I was able to fix remaining issues quickly, rebase the changes to current master and update the pull request.

The merge request has landed in FontForge master branch this morning. There’s a follow up pull request to update the Python scripting documentation as well. I want to thank Fredrick Brennan and Jeremy Tan for the code reviews and suggestions, and KH Hussain and CVR for sharing the excitement.

This functionality added to FontForge helps immensely in reusing the definitive Malayalam OpenType shaping rules without any modification for all the fonts! 🎉