Closed seabbs closed 2 months ago
Also, we have cover some of this in the EpiAware
README
but opinions on nice Julia set ups e.g. having Revise.jl
TestEnv.jl
etc in the basic environment of your local Julia installation.
Opinions on setting up VS-Code to run nicely, setting up REPL to be a quick interface to running example Julia code. For example, hardly anyone seems to know about (@v1.10) pkg> activate --temp
to make a quick fire-and-forget environment to try out (say) the Turing examples in.
I see this as separate from getting set up to run the code and more about interacting with the code. The reason for this is the getting setup is really package independent and could live in a few places whilst this advice would be tailored to things you want to do with EpiAware
and sometimes would grow into longer form documentation.
things like
EpiAware
struct?
_We recommend Accessors.jl
. As an example you could do this:...._
Quick thoughts:
Arviz.jl
examplesRCall
or conversely JuliaCall
from R e.g. running EpiAware
from R (RStudio editor) and then getting the chain info back to do your favourite tidybayes
based workflow?It sounds like we agree this is a good idea. I think step 1 is making the skeleton and step 2 (likely a new issue or series of issues is content). That should help manage PRs
Assuming we lumping the skeleton in with the getting started assigning @SamuelBrand1
The new getting started with julia (i.e #301) resolved a lot of the Julia level questions here. Now I think wee need a doc which
The skeleton for this is in place and now just needs some content
As Julia is so modular we have no need to reimplement some of our helper functionality as it can be used from other packages. However, this attitude across the Julia ecosystem makes packages very hard to use for everyone but experts.
I think a way to mitigate this might be to have a hints and tips section (or something like useful functionality from other packages) where we collect some of these ideas.
Things that come to mind are:
Accessors.jl
(#90)|
andcondition
MCMCChain
objects and to visualise them for diagnostics etc.Are there any examples of good practice at this in the ecosystem we can be inspired by?