Re: VC hg backend that use mercurial command server.

[Dmitry Gutov]

Hi Andrii,

On 13.04.2019 23:08, Andrii Kolomoiets wrote:

> I'm finally ready to present you VC hg backend that uses mercurial
> command server. It allows to communicate with single mercurial process
> over pipe rather than execute hg process for every command.
>
> Also there is a possibility to answer to hg questions interactively.
> It's useful during merge when hg is in doubt about right action.
>
> Besides that it has some changes and improvements described in
> commentary section of the package.

Thank you for this proposal.

> Hg process can be controlled via pipe (Linux, MacOS) or pty (TRAMP,
> Windows).
>
> I think this package can make working with hg repositories more
> pleasant :)

I've tried it out briefly, but I hope others on this list who use 
Mercurial would share their impressions with us as well.

FWIW, my main brief impression is that it's prettier (which is a good 
thing, of course). The benefits of the approach alone didn't overcome 
the sluggishness of the monster that is the Mozilla repository (the only 
Hg repo I have).

> I want to contribute to improve built-in hg support.
>
> May I ask which way it can be done better?
>
> I see that all other backends are process based, so can this
> single-process backend be acceptable?

I don't think "process based" is a requirement, so on that front we're 
good. But speaking of conventions, like other backends, it will need a 
list of implemented/unimplemented backend commands in the commentary at 
the top.

> If so then is it better to replace
> hg backend (saving customizable options as much as possible) or provide
> two backends which share same options?

Generally speaking, we'd first add it as an option and then see about 
removing the older one.

But, again, I wonder what others think.

> Source code is here:
> https://github.com/muffinmad/emacs-vc-hgcmd/blob/master/vc-hgcmd.el
>
> Hope you'll find it useful ;)

Thanks again.