At 2022-07-17T09:18:28-0500, Dave Kemper wrote: > On 7/16/22, G. Branden Robinson <g.branden.robin...@gmail.com> wrote: > > I think a new, writable troff register would be a better way to > > manipulate this feature--which is basically a style diagnostic--than > > a warning category. > > This suggests that one day there might be other warnings that fall > into the style-diagnostic category, and thus that establishing a > common prefix, maybe "sd*" or "style-", would be helpful to group all > these knobs together.
Yes, but I'm having trouble thinking of single other example that needs to operate at the level of the formatter as this one does. It's why I am similarly loath to add a new category (or "level") of diagnostic message-printing function to groff. We already have fatal/error/warning/debug (the last a post-1.22.4 addition by me, and which really shouldn't have any call sites in a released version). Backtraces reach underneath this layer of groff's error API and use "errprint"; I guess I could do the same for this thing. Using "debug" won't be right because it prefixes output with the string "debug:". At any rate I am finding it hard to generalize from a single specimen. > > "sentence." seems too broad and too vague. > > True, but with a prefix establishing it as a style diagnostic, a vague > following term might be OK. Conceded. If we have a prefix, I don't hate it. > A more specific choice could be "continuedsentence", but with a prefix > this is even longer than your idea that was "too long" (though it is > easier to parse without CamelCase). Right now "style*something" feels like the least bad, but I feel like "something." is more congruent with existing practice, especially if we don't think of another plausible formatter-level style warning. "continuedsentence" doesn't feel intelligible to me because the warning diagnoses a new sentence suprisingly appearing in the midst of an input line, not a single sentence that is rambling on. Maybe "latesentence", as in "this sentence starts later than good style would lead us to expect"? None of this feels ideal. It's enough to make you want to channel your inner Larry Wall and solve the problem with an impenetrable punctuation sequence. Regards, Branden
signature.asc
Description: PGP signature
