Contribution instructions should state dataset limitations (CRAN)

[tomjemmett]

Currently the contribution section on Readme.md doesn't state any of the limitations imposed by CRAN, namely that the entire package must be <5MB in size.

[chrismainey]

Agreed. We should probably write guidance for other ways to submit larger datasets. Emphasis on this being for training, not a general data warehouse. Hacktoberfest?

[tomjemmett]

Here are some of my (highly opinionated) thoughts on this. I think we should aim to follow the Tidyverse Style Guide where possible.

• datasets should be designed for teaching how to do things in R, so should be easy to understand and relevant datasets to a general audience
• datasets need to be relatively small in size; CRAN has a limit of 5MB for the entire package, so each dataset should be no more than 500KB in size. You can check with object.size()
• datasets should not contain any sensitive or disclosive information; they are being released publicly. The data ideally should be from a published source, or synthetic/generated data
• datasets that come from other sources must be licensed under a suitable license for reshaping, e.g. MIT, GPL, OGL, CC. Attribution must be included to the source data
• datasets should be saved as a tibble - you can use as_tibble() to convert
• datasets should be named using camel case, as should all columns within the dataset
• datasets should be documented with roxygen2; this documentation should be a high level overview
• datasets should have a vignette that describes what the data is in more detail than the documentation goes into as well as containing a useful example (ideally examples) of how to use the data demonstrating useful R functions
• vignettes should use tidyverse functions and avoid base R and data.table; this is more so because the introductory training NHS-R offers focussed on the tidyverse
• vignettes should not require the use of too many extra packages. Any packages you use must be included in the Suggests section of DESCRIPTION

[Lextuga007]

Adding to this list:

• each function should have an example that is more than the use of glimpse() which is now listed in the Get Started vignette. An example being in ons_mortality the example is how to view the data in wide form with each date as a column.