[PATCH] automv: new experimental extension

[Martijn Pieters]

On Fri, Feb 5, 2016 at 10:20 PM, Augie Fackler <raf at durin42.com> wrote:
>> I'm actually thinking switching to 1.0 just like automv might be an idea here, just for consistency.
>
> Perhaps (I assume you meant addremove). We could always start with
> 1.0, and then try and do something smart using data from various repo
> histories later to try and pick a more sensible default.

Sounds like a sensible approach. I'll look into seeing what kind of
data we can glean from ours.

> Yeah, import modules, but not functions. :)

So obvious now.

> Actually, our current style guide allows "from thing import module",
> where you always do the from ... import ... so that you end up with no
> dots in your module names in the
> code. (https://google.github.io/styleguide/pyguide.html#Imports)
>
> I can't remember the last time I saw something like import x.y and
> then referencing something as x.y.z.

It was a while ago; code that was for a project acquired by Google was
reviewed by a G engineer and they asked to use the full names (which I
pushed back on). Looking at the guide now I see no such requirement.
But we digress.

>> The test applies only to core mercurial, not extensions, which are 'external'. The imports *are* absolute.
>
> Right, but you could add from __future__ import absolute_import and
> suppress the warning ;)

Duh. Facepalm.

-- 
Martijn Pieters