Re-introduce docopt?

[emilio]

Seems like the performance issues docopt had (#46, docopt/docopt.rs#195) are fixed.

Should we restore it?

[Yamakaky]

An alternative would be to use clap, which ia also very cool.

[fitzgen]

I like docopt because it forces us to at least minimally document our command line flags. That is a great property to have.

[Yamakaky]

I too love docopt for that. Clap does the reverse: it generates the --help from the options.

[fitzgen]

I was under the impression that calling .help("...") on clap's builder was optional (meaning that the documentation is optional even if the help text contains all flags). Is that incorrect?

[Yamakaky]

If you don't set it, clap auto-generates the --help.

[jdub]

I'm going to send a quick PR to fix some inaccuracies in the current --help, but I can see it needs more love. If you have a preference either way for docopt / clap, I'd be happy to plug one in.

[Yamakaky]

The only thing I have against clap is that the matches are basically a Map<String, String>, while docopt auto-deserialize in a struct.

[jdub]

Looks like we'd be stuck with that style of interface either way for now, as docopt depends on a compiler plugin to deserialise into a struct, doesn't yet support custom derive, and I suspect Macros 1.1 might not land as soon as we'd hoped anyway.

I'll get started with docopt.

[jdub]

This can be resolved, as I reintroduced docopt… and then switched to clap in #228. 😂

[emilio]

Right, thanks!