BinData can't be instantiated

[jmuhlich]

I don't think dataclasses can usefully subclass a non-dataclass or really any non-cooperative class. For example our BinData ultimately subclasses str and I discovered there's no way to actually instantiate one due to errors from either the str constructor or the pydantic validator. How would the core string value actually get passed in here, anyway? The other subclasses of ConstrainedStr seem OK since they aren't also dataclasses. Maybe BinData should be a top-level class with an attribute value: base64Binary rather than inheriting.

[jmuhlich]

Simply making that proposed change (moving the content to a value attribute) makes 14/32 of the example files at https://github.com/ome/ome-model/tree/master/specification/samples/2016-06 parse without error. I didn't inspect them all to see if the returned object looks correctly structured, but it's a good start. I'll look into how to make that change in OVERRIDE.

[tlambert03]

yeah I recall some of the constrained strings (with additional properties) being hard to model... but I didn't appreciate that it couldn't even be instantiated! That'll never do! :)
how many of those files are parsing without error before that fix? (we should update the tests here to use those)

[jmuhlich]

Zero!

[tlambert03]

fwiw, I'm more than happy to use whatever "ugliness" is necessary for now to just get the output model working. So if using the current OVERRIDE pattern is not working, and you want to just put a conditional somewhere in the body of the code with a "# FIXME:" or something like that. That's fine with me and we can refactor later