Test groups for equality before doing fontMath kerning math

[benkiel]

This fixes #22, but will be breaking to a lot of code (though it's more correct). Perhaps better to add a strict option?

[benkiel]

@behdad @typesupply

[codecov]

Codecov Report

Merging #245 (4852acf) into master (2da8061) will increase coverage by 0.10%.
The diff coverage is 97.22%.

@@            Coverage Diff             @@
##           master     #245      +/-   ##
==========================================
+ Coverage   88.46%   88.57%   +0.10%     
==========================================
  Files          13       13              
  Lines        2359     2390      +31     
  Branches      305      307       +2     
==========================================
+ Hits         2087     2117      +30     
  Misses        195      195              
- Partials       77       78       +1     

Flag	Coverage Δ
unittests	88.57% <97.22%> (+0.10%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
Lib/fontMath/mathKerning.py	90.69% <85.71%> (-0.30%)	⬇️
Lib/fontMath/test/test_mathKerning.py	99.06% <100.00%> (+0.13%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2da8061...4852acf. Read the comment docs.

[typesupply]

I'm on a deadline now, so this will take me at least a few days to review. The non-equal groups was a feature, not a bug, but I'll have to try to remember why I did it the way that I did. I wrote this particular bit of code around 2003 and all that I remember off the top of my head is that it was very difficult. 😟

[benkiel]

I think that this may be better with a strict option, so one can decide what they want, the default set to False, will re-work. No rush on this at all, the issue has been sitting a long time.

[behdad]

Thanks Ben. I see Tal's comment about this being intentional.

If, say, group1 has a particular glyph A in a group, but other side doesn't have A in any group, one might want to assume that it was an omission in group2 and hence add it anyway? I find that dubious.

The alternative is to split the groups when mismatches occur. We have code to do that in fontTools.misc.classifyTools.