David Widmann

Ah never mind, I don't think it supports components of different shapes. So the alternative would be a NamedTuple, I guess.

IMO ideally we should just throw better error messages for common errors. I think users shouldn't receive long and undecipherable stack traces if possible in the first place. An example...

Often the information in the stacktrace is useful for developers, so I don't think it should be hidden. It happens quite regularly that I have a, surprisingly often correct, hypothesis...

I guess you could post it as a gist? I've never run into the character limit but I assume it is also one reason why according to the [Julia instructions](https://github.com/JuliaLang/julia/blob/bb5b98e72a151c41471d8cc14cacb495d647fb7f/doc/src/devdocs/backtraces.md)...

Maybe https://github.com/TuringLang/Turing.jl/pull/1841 already fixes your use cases without hiding any information.

Part of the redesign of MCMCChains tries to remove the amount of "automagic" that was present in MCMCChains, which could lead to surprising behaviour and type instabilities. Instead users have...

I think the main issue here (different order of parameters in MCMC chain vs MAP result) could be fixed by using natural sort order for MAP results as well (which...

The error seems to be caused by the evaluation of `logpdf(dist, old_val)` in https://github.com/TuringLang/Turing.jl/blob/0d4c8fe74f7abf943f7fe33e85cd51163839683c/src/inference/mh.jl#L148. I don't see how this could possibly work if the dimension of `dist` changes in between...

> However, for the model above, chain = sample(model, SMC(), 2) causes a MethodError regarding Chains. I'm currently updating Turing to AbstractMCMC 0.4 and MCMCChains 2.0 - maybe that will...

I quickly checked your example with https://github.com/TuringLang/Turing.jl/pull/1116 and discovered a small bug unrelated to the `Chains` issue. Unfortunately, the construction of a chain still fails since `parray` in https://github.com/TuringLang/Turing.jl/blob/9dc8a297aaed4e898beba698491116c28e40c533/src/inference/Inference.jl#L346 can't...