More flexible infrastructure with something like load_urls()?

[mrcaseb]

Thinking about a function that loads a small file in which we maintain the actual download urls so we can change them without having to change nflreadr code.

[tanho63]

Hmm, thinking about this on second pass and I think I’d want to have the documentation and function args the way we have it now? It wouldn’t simplify our file structure, in that case. Maybe we’d be able to programmatically access file names but piggyback/nflverse_download allows us do that too?

[mrcaseb]

I think I wasn't clear enough. I was thinking about an internal function that returns the download url for the loading function. Nothing user facing.
So we would run something like load_url("pbp") and it returns "https://github.com/nflverse/nflverse-data/releases/download/pbp/play_by_play_" which we use in load_pbp

It would allow us to redirect loaders without having to change and/or release nflreadr.

[tanho63]

It would allow us to redirect loaders without having to change and/or release nflreadr.

I like the version-controlled aspect of the current setup and hesitate to get rid of that but I think a csv of the URL fragments might be suitable if you want something programmatic

[mrcaseb]

Yeah we could version control the csv inside the nflreadr repo. It's not a strong opinion. Just thinking about more flexibility, especially given the recent problems with CRAN incoming checks. Lower number of releases means less pain for us.

[tanho63]

Closing as not-planned