Ah! Great catch. Thanks for the PR, and apologies for having to dig through the source of my first/only real Haskell project! :P

I think, technically, this fix results in arguments with the same name but differing style, e.g. ARG and <ARG>, clashing. Though since ARG and <arg> would coexist fine, I am good with this. Both cases are probably smells for a bad CLI anyway.

Will make a new release once I can test and verify. A new issue would be appreciated. :)

My pleasure! I don't know if the merit is Haskell's or yours, but the place I needed to tweak was really easy to find - just start with a function from public API and then grep a bit to find where pDocopt is defined.

I think, technically, this fix results in arguments with the same name but differing style, e.g. ARG and <ARG>, clashing.

I don't view that as a problem, because I don't think we need to distinguish between the two at all. To me, it's really a matter of style: you either choose one or another, and stick to it.

Both cases are probably smells for a bad CLI anyway.

I'm just curious: why do you think so?

For the sake of tying off loose ends:

I'm just curious: why do you think so?

By "bad CLI", I really meant "an inconsistent usage page". A usage page which either uses both ARG and <ARG> or both ARG and <arg> style of positional arguments could just as easily choose one of those styles to use throughout that usage page.

I personally prefer the <brackets> because it reads less ambiguously to me, but I know some others are more accustomed to UPCASE. Either way, there's no reason to mix them in these same usage page, so we shouldn't worry about supporting that in docopt.

Oh, I just misunderstood the meaning of "both cases" back then — I read it as if you meant "both arg and <arg> smell of bad CLI". Consider that particular loose end a false one :)