Is this a right place to put turing-dependent utility functions? about sr2turingpluto.jl HOT 8 CLOSED

That place is in src/require/turing in StatisticalRethinking.jl. The same is true for StatisticalRethinkingPlots.jl. Both hold subdirs named src/require/[Stan|Turing|...] which will be included by Requires.jl if e.g. Turing or StanSample is included.

from sr2turingpluto.jl.

StatisticalRethinkingStan and StatisticalRethinkingTuring only provide the (package) environment and the (basic) snippets/clips/codes in Pluto notebooks. These are Julia projects, not packages.

from sr2turingpluto.jl.

I am taking a look how your versions of the Jupiter notebooks would work in Pluto. I've completed first stabs for chapters 2, 3 and 7 (that's how I started this) and 4 partially. I think they work well but are in some cases too long for Pluto.

My view on these projects is a user downloads them once and then the downloaded version becomes your own work copy as you continue to explore, extent and add stuff (e.g. exercises?) to the local project repo.

from sr2turingpluto.jl.

Your versions are called Chapter-02.jl, Chapter_03.jl, etc.. If you don't agree with using your versions, just let me know, I'll keep them locally and remove them from the master repo. If we leave them there we should list you as an author (in fact the first author) and I will take that along as I update the README.

from sr2turingpluto.jl.

There are multiple ways of doing this in Julia.

I opted to use projects, which are not registered (but can be tagged) and a user can store these wherever they like. That local directory gives users read & write access. But projects are not really intended to hold source code to be included in other packages, Projects are typically saved in the repository including a Manifest.toml file.

The other way, as suggested in your question, would be to turn these into a package. Packages are stored in .julia/packages and users can not write to that directory. The notebooks would have to be duplicated to a different location if they need to be modified. Or such users could work with the package dev-ed. By using a package, we could add src code and drop the whole Requires.jl approach.

StatisticalRethinking.jl for about 4 years or so indeed was Jupiter notebook based. It became rather unwieldy over time and when I started to think about also supporting Turing.

But nothing is cast in concrete!

from sr2turingpluto.jl.

Hi!

Thanks for the explanation. At the moment the question is not urgent at all, just having a small piece of code (method to sample from ModeResult objects, produced by optimize()), which relies on Turing, but also generic enough to be put in the package, rather than copy-paste in notebooks.

In principle, I can contribute it to the Turing.jl itself, but it might take a while before merging. So, my though was to put it in the StatisticalRethinking somewhere.

from sr2turingpluto.jl.

I would suggest to maybe put it into a new file StatisticalRethinking/src/require/Turing/turing_sample.jl?

I know there is currently no sample method on turing quap objects in Turing (as far as I know), that would be another good method to put in that file if you need it. Or maybe we're talking in fact about the same functionality here.

By adding include("turing_sample.jl") to src/require/Turing.jl it will be loaded if Turing is present.

from sr2turingpluto.jl.

Yes, sure, will do

from sr2turingpluto.jl.