Determine usefulness of attr validators as validation is done by fontTools.ufoLib already

[madig]

In [1]: u = ufoLib2.Font()

In [14]: u.info.openTypeOS2WeightClass = -200                                                    

In [15]: u.info.openTypeOS2WeightClass                                                           
Out[15]: -200

[anthrotype]

Yeah, attrs validators only work in the constructor

[madig]

Hm... ufoLib will prevent me from saving the invalid data to disk. It will probably also prevent loading of such data. So.. maybe having validators that only fire on instantiation aren't that useful in ufoLib2?

[anthrotype]

There is attrs.validate function
http://www.attrs.org/en/stable/api.html

[madig]

Which runs all validators, not desirable if you just set one property, no?

[anthrotype]

You want to run that just before serialisation. Maybe Font.save method should call that automatically.

[anthrotype]

I don’t want to wrap every attribute in a @property just to ensure validators are run. Consenting adults

[madig]

Well, ufoLib already validates on (de)serialization, so maybe we can do away with validators entirely? Setters would not be affected. Dunno, just a thought. Although there would be no checks for when you do everything in memory, but I suppose that isn't covered currently anyway.

[madig]

To do:

1. Determine which attributes (should) have validators and cross-check with fontTools.ufoLib to see if they are checked there.
2. Decide if or which of these attributes should be turned into properties so that validators are always run; alternative: drop (all?) validators and let fontTools.ufoLib validate on serialization.

[anthrotype]

i'd say, drop all unnecessary validators that checks data integrity, which are already done in ufoLib.
and let static typechecker deal with the rest

[anthrotype]

running attrs.validate() on attrs objects before serialization is still a good idea, it may catch things before it's too late.