Template Haskell determinism

[Richard Eisenberg]

On May 31, 2016, at 9:54 AM, Bartosz Nitka <niteria at gmail.com> wrote:

> I'd like to get feedback if this is an acceptable solution and if the problem
> is worth solving.

I don't have an opinion about "worth solving". While I understand your description of the problem and believe you that it crops up in practice, I don't have a grasp on the net effect of this all. Bottom line on this front: I trust your judgment.

As to the breaking change: go for it. Template Haskell churns a good deal between releases (though not within a single major release) and so breaking changes are common. And NameFlavour really should be abstract (it can't be due to the linkage between the template-haskell and ghc packages) and anyone who uses it (including me) is doing something fishy. Accordingly, I'd personally be OK with a breaking change in a minor release around NameFlavour, as long as nothing exported from Language.Haskell.TH is changed. (We do have a way of using CPP to detect minor version bumps, right?)

I hope this helps,
Richard