Emacs author stats

Emacs author stats

[Lars Magne Ingebrigtsen]

I wondered whether switching to git had any effect, developer wise, so I
whipped up some stats:

http://lars.ingebrigtsen.no/2015/04/23/the-effect-of-version-control-systems-on-emacs-developers/

The Emacs Lisp functions I wrote to make the stats are on github (linked
from the article), so feel free to fix the data if my methodology is
unsound.  :-)

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: Emacs author stats

[Eli Zaretskii]

> From: Lars Magne Ingebrigtsen <larsi@gnus.org>
> Date: Thu, 23 Apr 2015 19:41:03 +0200
> 
> I wondered whether switching to git had any effect, developer wise, so I
> whipped up some stats:
> 
> http://lars.ingebrigtsen.no/2015/04/23/the-effect-of-version-control-systems-on-emacs-developers/

Thanks.

This part warrants a comment:

  [...] I would suspect that the numbers before, say, 2007 are
  suspect because maintainers would commit patches with themselves as
  the “author”, I think.

This suspicion is not true: whoever committed the changes even back
then, always mentioned the actual author.

Re: Emacs author stats

[John Wiegley]

> "Uhm… Well, attracting droves of new developers didn’t happen, apparently.
> Except for the spike the month git was introduced, we’re within the normal
> range of developers, I think.
> 
> So… was the changeover worth it? Probably. But the impact was a bit over-sold
> on both sides, I think."

You really need to wait a bit more before making judgments on whether it was
worthwhile or not.  People are still getting used to the change, and many
people haven't even heard of it (I hadn't until a few weeks ago!).

John

Re: Emacs author stats

[Lars Magne Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

> This suspicion is not true: whoever committed the changes even back
> then, always mentioned the actual author.

Ah, right.  I thought that the...  er...  previous vc didn't
differentiate, but I assumed that the previous one was cvs.  But perhaps
there was something else in between cvs and bzr that I'm forgetting?

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: Emacs author stats

[Eli Zaretskii]

> From: Lars Magne Ingebrigtsen <larsi@gnus.org>
> Cc: emacs-devel@gnu.org
> Date: Thu, 23 Apr 2015 20:44:07 +0200
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > This suspicion is not true: whoever committed the changes even back
> > then, always mentioned the actual author.
> 
> Ah, right.  I thought that the...  er...  previous vc didn't
> differentiate, but I assumed that the previous one was cvs.

What I said was true even when Emacs used RCS.

> But perhaps there was something else in between cvs and bzr that I'm
> forgetting?

Not officially, no.  (Miles used arch.)

Re: Emacs author stats

[Artur Malabarba]

2015-04-23 18:49 GMT+01:00 Eli Zaretskii <eliz@gnu.org>:
>> From: Lars Magne Ingebrigtsen <larsi@gnus.org>
>> Date: Thu, 23 Apr 2015 19:41:03 +0200
>>
>> I wondered whether switching to git had any effect, developer wise, so I
>> whipped up some stats:
>>
>> http://lars.ingebrigtsen.no/2015/04/23/the-effect-of-version-control-systems-on-emacs-developers/
>
> Thanks.
>
> This part warrants a comment:
>
>   [...] I would suspect that the numbers before, say, 2007 are
>   suspect because maintainers would commit patches with themselves as
>   the “author”, I think.
>
> This suspicion is not true: whoever committed the changes even back
> then, always mentioned the actual author.

Either way, I think the relevant metric is precisely the number of
*committers*. Commiters are the ones most affected by this change.
Authors who submit patches and bug fixes couldn't care less whether we
use bzr or git (or do I misunderstand something?).

Re: Emacs author stats

[Lars Magne Ingebrigtsen]

Artur Malabarba <bruce.connor.am@gmail.com> writes:

> Either way, I think the relevant metric is precisely the number of
> *committers*. Commiters are the ones most affected by this change.
> Authors who submit patches and bug fixes couldn't care less whether we
> use bzr or git (or do I misunderstand something?).

I've now added the stats for the committers as a chart at the bottom
here:

http://lars.ingebrigtsen.no/2015/04/23/the-effect-of-version-control-systems-on-emacs-developers/

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: Emacs author stats

[Eli Zaretskii]

> Date: Thu, 23 Apr 2015 19:48:10 +0100
> From: Artur Malabarba <bruce.connor.am@gmail.com>
> Cc: Lars Magne Ingebrigtsen <larsi@gnus.org>, emacs-devel <emacs-devel@gnu.org>
> 
> Either way, I think the relevant metric is precisely the number of
> *committers*.

Not for us, no.  Committers don't write code, they just commit.  What
we need is more active developers, i.e. authors.

Re: Emacs author stats

[Dmitry Gutov]

On 04/23/2015 10:55 PM, Eli Zaretskii wrote:

>> Either way, I think the relevant metric is precisely the number of
>> *committers*.
>
> Not for us, no.  Committers don't write code, they just commit.  What
> we need is more active developers, i.e. authors.

Of course we do, but people who just send patches are much less affected 
by the choice of the VCS.

Re: Emacs author stats

[Eli Zaretskii]

> Date: Thu, 23 Apr 2015 22:59:23 +0300
> From: Dmitry Gutov <dgutov@yandex.ru>
> CC: larsi@gnus.org, emacs-devel@gnu.org
> 
> On 04/23/2015 10:55 PM, Eli Zaretskii wrote:
> 
> >> Either way, I think the relevant metric is precisely the number of
> >> *committers*.
> >
> > Not for us, no.  Committers don't write code, they just commit.  What
> > we need is more active developers, i.e. authors.
> 
> Of course we do, but people who just send patches are much less affected 
> by the choice of the VCS.

That's not what we were told when this issue came up in the past.  To
send a patch, you need to clone the repository, develop and test the
patch, send it for review, update it several times according to
comments, rebase it as master gets new commits, etc.  All of which
involves the VCS, and so many people said they will not consider
becoming contributors unless we switched to Git.

Re: Emacs author stats

[Artur Malabarba]

[-- Attachment #1: Type: text/plain, Size: 1008 bytes --]

Fair enough.
On Apr 23, 2015 10:10 PM, "Eli Zaretskii" <eliz@gnu.org> wrote:

> > Date: Thu, 23 Apr 2015 22:59:23 +0300
> > From: Dmitry Gutov <dgutov@yandex.ru>
> > CC: larsi@gnus.org, emacs-devel@gnu.org
> >
> > On 04/23/2015 10:55 PM, Eli Zaretskii wrote:
> >
> > >> Either way, I think the relevant metric is precisely the number of
> > >> *committers*.
> > >
> > > Not for us, no.  Committers don't write code, they just commit.  What
> > > we need is more active developers, i.e. authors.
> >
> > Of course we do, but people who just send patches are much less affected
> > by the choice of the VCS.
>
> That's not what we were told when this issue came up in the past.  To
> send a patch, you need to clone the repository, develop and test the
> patch, send it for review, update it several times according to
> comments, rebase it as master gets new commits, etc.  All of which
> involves the VCS, and so many people said they will not consider
> becoming contributors unless we switched to Git.
>
>

[-- Attachment #2: Type: text/html, Size: 1507 bytes --]

Re: Emacs author stats

[Peder O. Klingenberg]

On Thu, Apr 23 2015 at 23:10, Eli Zaretskii wrote:

> To send a patch, you need to clone the repository, develop and test
> the patch, send it for review, update it several times according to
> comments, rebase it as master gets new commits, etc.  All of which
> involves the VCS, and so many people said they will not consider
> becoming contributors unless we switched to Git.

I am not a committer, but I have contributed a patch or two.  The basis
for my contributions has always been git-controlled.  Even before the
official switchover, I used the git mirror of the bzr tree.  I have no
desire to learn more VC systems than I absolutely have to, and these
days, git is a must for almost everything.  But had the git mirror not
existed, I'm convinced I could have muddled through the necessary
cut'n'paste to download the bzr tree.

What I'm saying is that, as a non-committer and infrequent contributor,
Emacs' choice of VC system was not a big factor in my decision to
contribute to Emacs.  Had bzr been the only choice, and had it proved
irksome to use, I might not have continued to contribute, but that was
never an issue, as the git mirror was perfectly adequate.

Compared to the hassle of copyright assignment, Emacs' choice of VC was
lost in the noise.  And I did not find the copyright process especially
burdensome, just a bit time consuming.  (What do we want?  Instant
gratification!  When do we want it? NOW!)

I was in favour of the move to git, because it seemed from the outside
like the right thing to do, but the impact on me was very close to zero.
And I didn't voice my opinion at the time because I felt it wasn't my
decision to make, in part exactly because I was not a frequent
contributor, and knew that the outcome would not influence my
contribution frequency at all.

The things preventing me from contributing more have nothing to do with
versioning systems.  It is partly that annoying thing called life which
my family insists I engage in occasionally, but even more, it's a
distinct lack of itches to scratch.  Emacs mostly does what I need
really, really well already.  So thanks, all of you!

...Peder...
-- 
I wish a new life awaited _me_ in some off-world colony.

Re: Emacs author stats

[Stefan Monnier]

> Not for us, no.  Committers don't write code, they just commit.  What
> we need is more active developers, i.e. authors.

I'd be happy to get help from people who do nothing but install those
patches where we said "LGTM".


        Stefan

Re: Emacs author stats

[chad]

> On 23 Apr 2015, at 11:48, Artur Malabarba <bruce.connor.am@gmail.com> wrote:
> 
> Either way, I think the relevant metric is precisely the number of
> *committers*. Commiters are the ones most affected by this change.
> Authors who submit patches and bug fixes couldn't care less whether we
> use bzr or git (or do I misunderstand something?).

A relatively large number of people suggested that it would be
easier for people to follow and submit patches if the repository
used git, because that is a very common contribution mode for other
large projects. The hope was (I believe) that this would let people
slowly step-up into contributor/committers.

I suspect its too early to see a real impact of that (theoretical)
idea yet, but emacs is the only large project I still follow at this
point, so that’s a raw guess.

~Chad

Re: Emacs author stats

[Stephen J. Turnbull]

Artur Malabarba writes:

 > Either way, I think the relevant metric is precisely the number of
 > *committers*. Commiters are the ones most affected by this change.
 > Authors who submit patches and bug fixes couldn't care less whether we
 > use bzr or git (or do I misunderstand something?).

Back when "bzr branch" from Savannah sometimes took several hours or
didn't finish at all, I'm sure that some potential authors found that
plus the request to verify bugs and patch compatibility with trunk
HEAD made it worth their while to remain potential.

And as Eli says, it's authors that matter to the volume of
contributions.

Re: Emacs author stats

[Artur Malabarba]

You're right that attracting huge droves of developers didn’t happen,
and it's always nice to see numbers and data, but 5 months is not a
very relevant sample here for the purpose of finding an
upward/downward pattern.
Note how much the number has fluctuated over the last 4 years, the
range of some flucuations is as large as the average value.
We'd need to wait another year, and probably average over every 2 or 3
months, in order to see any patterns.

Also, I think number of committers would be a more relevant metric,
altough this number would be even smaller and less reliable.

2015-04-23 18:41 GMT+01:00 Lars Magne Ingebrigtsen <larsi@gnus.org>:
> I wondered whether switching to git had any effect, developer wise, so I
> whipped up some stats:
>
> http://lars.ingebrigtsen.no/2015/04/23/the-effect-of-version-control-systems-on-emacs-developers/
>
> The Emacs Lisp functions I wrote to make the stats are on github (linked
> from the article), so feel free to fix the data if my methodology is
> unsound.  :-)
>
> --
> (domestic pets only, the antidote for overdose, milk.)
>    bloggy blog: http://lars.ingebrigtsen.no
>
>
>