Antony Lee

Probably best would be for you to write a PR first with the functionality (I don't plan to write the first draft for you, at least :-)), we can then...

Actually, I see that napoleon upstream just merged https://github.com/sphinx-contrib/napoleon/pull/36, fixing the 3.6 issue. Let's try to make them release that first, in which case we can go back to having...

That part I'll probably completely get rid of and replace by https://github.com/pypa/setuptools_scm#retrieving-package-version-at-runtime.

Done. I think the best thing you could do now is to prod the sphinxcontrib.napoleon maintainer for a new release :-)

I think this would be better served by #95? The default behavior of later args overwriting earlier ones is very standard for command line tools and I would be loath...

This seems reasonable modulo API bikeshedding, but I will not have the bandwidth to consider all the options at least for the coming month(s). But I'll keep this open for...

One question, though, is how you propose to handle subcommands in this API. Perhaps `foo = main.add_subcommand("foo")` and then an `@foo.cmd` decorator? (perhaps that would be better named `@foo.add_command`, for...

Agreed that `@main.command("foo", "bar")` is likely the most user-friendly. Do you want to try your hand at proposing a PR? :-)

I try reasonably hard to document any backwards-incompatible change (or any change, in fact) in the changelog, and to use a semver-ish versioning, but I explicitly cannot and will not...

This and #96 both seem reasonable to me, but adding a global option to `defopt.run` seems unsatisfying because sooner or later someone will request to be able to specify this...