-
Notifications
You must be signed in to change notification settings - Fork 341
Description
Hello,
I am writing to request two potentially separate, but related features from testthat.
My understanding is that testthat infrastructure currently only works easily within the context of R package development (e.g., using devtools::test() or Ctrl + Shift + T to automatically run tests). What I'm wondering is if it'd be possible to also easily use testthat in a non-package data analysis R Studio project?
Say for example, in some analysis project associated with tidymodels. I think easy access to tests should both broaden the scope of the testthat package and also help improve the correctness of analytical projects out in the field.
The second part of this feature request is to allow access to an installed package's testthat tests via some user-facing function. Although this may seem like overkill, I've recently encountered an issue where I stumbled upon computational irreproducibility between machines associated with an R package, even though the cause of the computational irreproducibility is unapparent, even after substantial detective work.
My concern is that even if the cause of the irreproducibility were eventually identified in this case (as outlined here), if I were to move to a new machine and if I were without easy access to some validation suite, then I don't think I should be very confident in the computational reproducibility of my code. Making it easy for me to run a validation suite of my installed package on the new machine would help improve my confidence and the creation of said validation suite could be facilitated by testthat.