/inst/extdata folder

[eva0marques]

@Spatiotemporal-Exposures-and-Toxicology @dawranadeep @MAKassien @dzilber @sigmafelix @Sanisha003
I created the folder /inst/extdata (see https://r-pkgs.org/data.html) that is supposed to store all data created and used in our package. For example, I stored a table with NLCD classes names, abbreviation, meaning and associated colors in nlcd_classes.csv. It is used for renaming nlcd ratio in my .R function.

[kyle-messier]

@eva0marques This is an interesting solution. I am not seeing the documentation mention how it deals with data size and storage on GitHub.

[eva0marques]

Yes, I don't know. It is more for small data I guess, like dictionary or small csv files that need to be used within .R functions. It is not for the input or output data. It was @sigmafelix idea.

[sigmafelix]

All packages submitted to CRAN should abide the size limit of <=10 MB in tarball size and <=5 MB in documentation/data (cf. 10th bullet). I think we need to reduce the data size to as small as possible since we aim to submit this package to CRAN. Large data-only packages are usually hosted in non-CRAN locations (e.g., spDataLarge package). Our study has a different situation where dataset sizes often exceed several gigabytes. I supposed this would be a fundamental question about the purpose of NRTAPModel package: do we want to share the full workflow with data for full reproducibility or want to make the pipeline be able to be maintained with CI/TDD? If the former is our ultimate goal, all raw datasets need to be stored elsewhere on the internet; however, if the objective is the latter, I think that sharing our pipeline, fully tested functions, and regularly maintained raw data sources (e.g., URLs) without all raw datasets might be fine. Users will be able to download the entire raw datasets with the functions if they want to reproduce our results.