Statistics mode

It's hard to tell whether a rule will actually be useful. It should be possible to run rules over a large corpus of existing code and gather statistics about how many times they'd be applied, possibly with a sample of application sites. Ideally this would be easy for people other than me to use, in order to help others come up with good rules. Possible corpus candidates include:

• Racket itself. Already installed, so no need to download anything (assuming the source code is still around for macro expansion purposes).
• Scribble. It's big, old, it's a single package (scribble-lib), and it hasn't seen much maintenance love over the years.
• The entire package catalog.