cboettig / build.eml Goto Github PK

@amoeba @mbjones Here's a first proof-of-concept on what an R package that defines a function for approximately each complex type would look like. Curious what you think. Fastest way to browse this is probably check out the pkgdown version of the docs here: https://cboettig.github.io/build.eml/reference/index.html, though I think the real test is in seeing if this is a more useful, more beginner-friendly way to construct EML than we can currently offer in EML. I'm not sure.

As commented in ropensci/EML#252, this is based on a crude parsing of EML XSD files which could definitely be improved upon to pull out the rest of the documentation and examples found there (though I really wish it were easier to do so! parsing the complex referential logic of XSD classes across multiple files really is rather messy to me). Certainly this could be much improved (e.g., to start, adding cross-linking between the documentation files, and including the rest of the documentation strings indicating which value is required or not, etc).

These functions are generated programmatically in a very simple way from this hack-y table I constructed by parsing the XSD files. It would be really useful to me to have a more complete and polished version of the XSD schema+documentation in a format like this or JSON that I could then work with, but not sure when I will get enough time to dive back into XSD extraction.

As also anticipated, this approach creates a handful of namespace collisions with base packages, as noted in #1. (with more collisions possible with other additional common packages). Not sure how best to deal with that, but I'm not really sure these collisions are acceptable.