Missing warning for jj diff treating positional arguments as paths

[BatmanAoD]

I've seen jj log warn me about treating a positional argument as a path instead of a revision, but it appears that jj diff <foo> doesn't give the same warning, even if <foo> doesn't exist. Instead, jj diff <foo> just has no output if <foo> is a revision, which is very surprising and misleading.

Could jj diff have a warning about positional arguments that aren't paths to existing files?

[ilyagr]

There is an issue and even an unfinished pull request for this. See #505.

[martinvonz]

I think you're right. I may just be too colored by my use of hg, where there can be exactly one value associated with each flag (but flags and values can be repeated). I'm pretty sure I've seen that clap supports -p a b c syntax, so I guess it can't be that uncommon.

I haven't decided what I think about that yet, but is the suggestion then to not have any positional arguments? So even jj new main would have to be written jj new -r main? Maybe I'll get used to it quickly but it sounds a bit annoying.

[necauqua]

you probably missed it, but no, the last (optional) positional arg (before 0 up to tons of optional repeating and interleaving -rs and -ps) is a revision, it's perfect

I think literally grep -r paths* -e 'have to use -e for the actual pattern now' is like what I proposed

[martinvonz]

Ah, I see. I didn't look very carefully, and maybe I'm also too used to the hg CLI where you can put positional arguments anywhere between flags.

[necauqua]

The funny part is that since -p accepts any number of args and -r accepts one arg, it should work to have -p paths* -r rev some-positional-arg

I think clap handles that fine 😄

[arxanas]

I opened #3809 as my preferred solution.