-
Notifications
You must be signed in to change notification settings - Fork 106
Closed
Labels
Description
The dynamic documentation for parsnip is setup so that there are specific help pages for each model/engine combinations. These have technical details and parsnip only loads the pages for packages that are loaded. For example, the engines for Poisson regression are only shown if the poissonreg package is loaded.
The pros of this:
- More in-depth help
- Modular help does not overwhelm users with information.
The cons:
- There are hidden package dependencies that users need to have installed to build the docs (even if a PR is to fix a typo). This can include engine specific dependencies.
- The documentation can take a while to compile.
- The use of
@includeRmdcan lead to the annihilation of the namespace file if there is an error building the docs. - Extension packages have their own doc files
Proposal:
Create a separate package that contains all of the engine-specific documentation files (parsnip and extension packages). When a model help file is opened, the same (or similar) engine list is shown but the links are to the documentation package pages.
The pros of this:
- Same as above
- All documentation is in one place
- Contributors to parsnip do not have to deal with building the docs
- Better formalization of documentation functions.
- Simpler, more robust code to create links.
- Less code in parsnip, far fewer R and Rd files.
- Users can find out about new extension packages without having them loaded.
Cons:
- Even more packages required to compile docs.
- Docs can get out of synch with parsnip.
- Possible confusion with having some help files in a different package (or pkgdown site).