todo_comment_linter could get an argument for valid TODO patterns

[MichaelChirico]

This can allow TODO only if they fit certain patterns.

This could be used to enforce a rule like "All TODO must be tied to a bug, not a person".

## BAD
# TODO(michaelchirico): fix this
# TODO: add testing

## GOOD
# TODO(#nnn): fix this
# TODO(#mmm): add testing

Maybe an argument like allow_regex or allow_style_regex, in this case allow_regex = "TODO[(][#][0-9]+[)]:" could work.

[AshesITR]

Maybe just expose the regex as an optional argument and call it regex=?

todo_comment_linter(regex = "# TODO\\(#\\d+\\):") seems fine to me.

[MichaelChirico]

The question then is how to interact this with the existing todo argument:

lintr/R/comment_linters.R

Line 164 in 191e3b7

todo_comment_linter <- function(todo = c("todo", "fixme")) {

Should we be looking for matches of paste0(todo, regex)? Then the call would be todo_comment_linter(regex = "\\(#\\d+\\):").

[AshesITR]

I would let regex take precedence with a warning if both arguments are present.

[MichaelChirico]

After #2433, I'm thinking the approach here will be:

function(todo = c("todo", "fixme"), except_regex = NULL)

Where except_regex is used to filter out lints matched by todo. It also makes sense verbally: Don't use todo unless it matches one of except_regex.

I think we should support case-sensitive matching too -- I would want TODO(#1234) to be OK but not todo(#1234). Probably letting todo= be case-insenstive but except_regex= case-sensitive makes sense, otherwise we could add a third argument.

[AshesITR]

I don't think we need a third argument. You can simply turn on ignore-case by prepending except_regex with "(?i)" if necessary.