Chad Scherrer

Maybe it could be even simpler? A measure "knows" its domain, so the package could just depend on `MeasureTheory` and `Quadrature` and only have one function `Integral -> QuadratureProblem`. The...

Oh, I guess we're starting to split the conversation in different threads. I just made a related comment here: https://github.com/cscherrer/MeasureTheory.jl/pull/130#issuecomment-897637078

Thanks @arbenede , I think this is a great idea. Most of the references I know of are very theoretical, and I think there could be an opportunity for something...

I'm using `v0.4`. Strange, we had been talking for a long time in terms of `Prob` values. I think I'm confusing them with `prob` values in Hakaru. When I switched...

Offhand I see a few considerations worth... considering. For scalar distributions, there are big advantages in MeasureTheory over Distributions for log-density computations. But for random sampling, this is less clear....

> Probably not. Distributions.jl implements samplers for at least some distributions that can benefit from it, so for those distributions at least, we could just call their samplers. See https://github.com/JuliaStats/Distributions.jl/tree/master/src/samplers...

I think the core of the problem here is that despite that 1. it's easy to get composability by creating and destroying arrays as you go, and 2. with some...

@mschauer does this address your concerns? It's not clear to me what "I really dislike the current state of this" refers to, since I think Distributions etc has the same...

Thanks @gdalle . This is looking good, I think we just need a few more functions and some tests. Weirdly, tests here are failing, for code that passed before, passes...

Sorry that was vague, the failing tests had me pretty distracted. Offhand I didn't see a method for `testvalue`. We'll also need some tests. Could you add some similar to...