Closed maurolepore closed 3 years ago
I have comments about this (which I think are understood, but want to make it explicit for posterity):
pactaCore
is exactly to clean up the messy file structure of PACTA, not to hide itinst/
is OK as an interim step in the refactoring process, but the goal should be to migrate the code back to the root of the directory.Rbuildignore
, but this is because Docker and packages aim to solve the same problem. I do think the Dockerfile still has value. With all of that said, #49 is the obvious reason to add all of this scaffolding, and having this snapshot test structure will allow us (hopefully) to quickly develop/ refactor and extract the core code back out of inst
.
All of that to say, I back this, but with many caveats, and I think it is important we deal with it quickly.
This PR provides the new
local_pacta()
-- a test fixutre like many others in the withr package.Also:
The messy file structure of PACTA moved to inst/extdata/context so the root is a clean, small package that only wraps calls to docker.
The main call to docker now happens in the package. We could expose arguments for the user to specify the location of input/ output/ and pacta-data/ on the host, to specify image:tag, command to run, etc.
With this interface docker-compose could be not a system requirement for user but instead a convenience for developers.