More clarity about convenience implementations

[travisbrown]

This is a pretty big change, but it fixes the last major problem with the public API that I think should be addressed before 1.0.

The change in #102 (first published in 0.12.0) adds RawFacade as a supertype of Facade that provides index parameters for Facade's methods. This was to maintain source compatibility, so that people working with Facade wouldn't have to change their implementations.

This naming doesn't seem ideal for the long term. Calling the "real" type class (i.e. the one used as a constraint) RawFacade and having Facade be essentially just a convenience class for implementers hurts discoverability and readability—"raw" just doesn't really mean anything in this context.

There aren't all that many implementations of RawFacade out in the world—a few in the support packages here, one in circe-jawn, and maybe several others. Updating this handful of implementations to use an explicit convenience trait seems like a small price to pay to gain some clarity in the API in 1.0, which (hopefully) will be around for a long time.

The change here renames the type class itself from RawFacade to Facade (and the context helper class from RawFContext to FContext), and moves all of the convenience implementation definitions (NoIndexFacade, SimpleFacade, MutableFacade, and NullFacade) into the Facade companion object.

It also removes the empty () parameter lists from jnull, jfalse, and jtrue, since unlike arrayContext(), etc., there's probably never a good reason for them to return mutable values.

I've also improved the API docs in a few places—e.g. where the Facade docs were just copy-pasted for the implementation traits, and I've added type annotations in some places where the inferred type would be more specific than it should be.

[larsrh]

I don't have a clear enough idea of Jawn's internals in order to make an informed review of this, but the small amount of changes in the support modules and the tests suggests this is a conservative-enough change.

[rossabaker]

👍 I find this easier to understand. It is not too burdensome on library integrators, and I don't expect most downstream users to notice.

[travisbrown]

Thanks all! I'll go ahead and publish a 1.0.0-RC3 that I think should be an actual candidate for release.

[travisbrown]

Merged master to resolve conflict after #221.