Cassandra Granade

I agree entirely, the plan is to add documentation to both the guide and apiref, including descriptions of a few explicit usecases. At the moment, we wanted to start the...

Agreed, this is a major concern moving forward, and one that we should definitely address. At the moment, I think the only place that `n_outcomes` is explicitly used is as...

I'll take a more detailed look at @whitewhim2718's branch soon, then. In the meantime, tagging @jarthurgross into this Issue, as the `n_outcomes` problem is relevant to his continuous-measurement filtering work...

For concreteness, I can think of two main usecases that should be addressed by a new design: - Continuous-valued outcomes - Multinomial (vector-valued but discrete) - Poisson (semi-infinite but discrete)

Very good point, I've added that to the list as well, then.

Somewhat tangential, but what do people think on which version we should target with this? Personally, I think it makes sense to target a new QInfer version 1.1, so that...

There is one other place that the assumption of integer outcomes gets used, and that's in the `perf_testing` module. While not strictly related, much of the generalization discussed here would...

It's pretty critical to have ``qinfer.__version__`` defined in released builds, so I'd like to figure out how to either fix the builds or fail gracefully if ``version.py`` isn't properly created...

By way of another +1 and showing analogous functionality in other toolchains, the .NET SDK allows for overriding version numbers at the command line with commands like `dotnet build /p:Version=1.0.1`...

> Sounds reasonable to me. Great, thank you! If it's alright, then, I'll get an API proposal for the attribute going so that that isn't blocking an implementation. ♥