In dd-harp/rampdata: Finds data and places data for the RAMP project

Introduction

I need to make an interface to set a path for RAMP. The first goal of the path handling is to give our files a common structure that we recognize for searching and file management. The second goal is to make that structure have a root directory that we can change on the cluster or can set for local compuation on desktops.

Alternatives

The first version of an interface asked the user to pass in a list.

dest <- local_path(
  list(stage = "output", project = "uganda2020", dataset = "district_age2")
)

That seems like using named arguments, but with extra steps.

The options are the following. Use paths.

dest <- local_path("/uganda2020/output/district_age2/1")

Use arguments.

dest <- local_path(
  stage = "output", project = "uganda2020", dataset = "district_age2", version = "1"
)

Use paths with replacement.

dest <- local_path("/uganda2020/output/district_age2/$version")

Scenarios

After using this path for a while, these are some things that I noted I wanted to do.

Give me that file.

filename <- local_path("/uganda2020/output/district_age2/1/pop_2000.csv")

Give me a root directory to ensure it exists.

dest <- local_path("/uganda2020/output/district_age2/1")
if (!dir.exists(dest)) dir.create(dest)
filename <- file.path(dest, paste0("pop_", year, ".csv"))

The Python FBDPath we used for the Forecasting Burden of Disease group had a nice feature that would look like this:

dest <- local_path(project = "uganda2020")
dest <- add_path(dest, stage = "mic", version = "3")
dest <- add_path(dest, dataset = "density", file = "file_2000.csv")

That was added by people who used the parsed path a lot, so it's likely a helpful feature. The trouble is that there isn't a strong conversion capability in R. We could make a sensible conversion.