path: root/2020/info/08.md

# Building reproducible Emacs
Andrew Tropin

[[!template id=vid src="https://mirror.csclub.uwaterloo.ca/emacsconf/2020/emacsconf-2020--08-building-reproducible-emacs--andrew-tropin.webm" subtitles="/2020/subtitles/emacsconf-2020--08-building-reproducible-emacs--andrew-tropin.vtt"]]  
[Download compressed .webm video (18.4M)](https://mirror.csclub.uwaterloo.ca/emacsconf/2020/smaller/emacsconf-2020--08-building-reproducible-emacs--andrew-tropin--vp9-q56-video-original-audio.webm)

It's not always easy to take part of someone's configuration and make
it work, it's almost never easy to move your configuration to fresh OS
installation or hardware.  Not sure that this snippet is enough to
make package work?  Forgot to install ripgrep in your system for
rg.el?  Got a broken version of package on package-install?

There is a way to make an Emacs configuration reliable, composable and
self-contained.  It's possible to freeze package versions, create
systemd unit for emacs daemon, maintain system dependencies and
package subconfigurations in one place with one tool.

The talk explains how to leverage the power of nix package manager and
use-package to make pretty good emacs configuration.

There is a stream record on the same topic:
<https://youtu.be/2_e3kPJQ93s>.  It lacks few interesting points about
composability of such configuration approach, but already have enough
interesting information.  The talk will be a little more structured
and more Emacs-users oriented.

<!-- from the pad --->

- Actual start and end time (EST): Start: 2020-11-28T11.26.34; Q&A:
  2020-11-28T11.40.48; End 2020-11-28T11.43.33

# Questions

## Do you deal with config files such as emacs-custom.el, some which have sensitive data?
Sensitive data is in other directories that aren't shared, and
emacs-custom.el is completely avoided, as it prevents
reproducible/system independent behaviour.

## How did you learn Nix language basics? Just from the the manual?
He referred to the Nix IRC channel.

## What are the main advantages besides switching computers (which most people rarely do)?
Make parts of config available for projects - sharing with other
people.

## Have you tried Guix in place of Nix? (more parens! :) :)
Currently trying it, and also in-process of switching from Nix to Guix.

# Notes
- Emacs configuration is entangled with the system configuration
  (dired uses ls, grep.el uses grep).
  - Reproducible behaviour is therefore not only dependent of Emacs
    compilation/configuration, but also system configuration.
  - "config.el" files configure Emacs, and accompanying "default.nix"
    files make sure that the correct packages/fonts/libraries/etc are
    installed.
- Reproducible development environment: <https://github.com/abcdw/rde>
- Using Org-roam to demo how to config a Nix layer(?)
  - custom.el conflicts with Nix(?)

# Transcript

00:00:00.399 --> 00:00:36.399
My name is Corwin Brust and I will be
talking about getting started with Emacs
today. I have been an Emacs user for a
long time. First of all, thanks and a
huge welcome to the conference from me
and and on behalf and back to the other
people that have been helping to
organize. It's been amazing just to be
involved with that and just, kind of,
see backstage.

00:00:36.399 --> 00:01:33.759
So I've used a lot of different editors
in my time. That's about 25 years as a
professional software engineer. And most
of that time I've been using Emacs. I'll
talk a little bit in a minute (if I can
ever find my slides) about how I got
into Emacs, but I think if you've used
Emacs and a lot of other editors for a
long time, something that you notice
right away is that you get good with it
in a way that stays meaningful. You
learn new things. Those things stick
with you. You learn how to make it do
new tricks and then keep doing those
tricks.

00:01:33.759 --> 00:02:17.200
I want to mention that this
conference--oops, this talk isn't about
how to adjust your configuration
specifically. I don't have a bunch of
good code samples in here. There are
other great talks at the conference,
particularly Andrew's, that I looked at,
that looked like they might be more
aimed at that "hey, I'm just getting
started with Emacs, what are some things
to try to make it more comfortable for
me starting?" This is about how to think
about the problem space. Hopefully, a
good warm up as we start thinking about
some of the lightning talks a little
later on.

00:02:17.200 --> 00:02:29.680
I'm just gonna quickly make sure I can
see my IRC buffer in case I run into
time. I didn't get my stopwatch started
for this one. So all right, let's dive
in.

00:02:29.680 --> 00:02:53.519
We assume that we want to install
packages and maybe configure some
features. This is particularly from the
perspective of where we're working with
a bunch of people on a team and we want
to get something done. Some of us
probably already have mature Emacs
workflows. Others are installing it for
the first time.

00:02:53.519 --> 00:03:35.840
So the first question is, you know, in
that context: what's the value
proposition? Why should I mess with my
machine, my mature Emacs configuration,
and impose my ideas over the way
somebody else is learning Emacs? Well,
it can be.. I'm off my slides here a
little bit. It can be a little bit
tricky to learn Emacs. One thing that
helps us a lot is if people that we're
working with can tell us, kinda,
keystroke for keystroke at times, what
to do and explain what everything is
doing. Using the same packages can
really help us working together on a
project.

00:03:35.840 --> 00:04:54.840
Speaking from my personal experience, it
took me decades to get to the point
where I was excited to program in Emacs
Lisp. I've programmed in a lot of
programming languages, but Lisp wasn't
on my list. I looked at my config that I
was copy-pasting around from generation
after generation of .emacs file or
recrafting it from hand and from
Internet searches, to get the things
that I needed when I would quickly go
install Emacs at some new job or
contract, and be able to to quickly get
through that workflow that caused me to
install the program. You know, just
little simple one-liners that that got
committed to memory over decades
eventually just led me to a sort of "hey
what's going on here." And I credit Jeff
Goff, my good friend who died earlier in
2020, for my lifelong love of Emacs.
Perhaps EriK and I will talk about that
a little bit more in another talk we
have scheduled, but Jeff was a huge
influence on us in a number of ways and
a huge contributor to the Raku
programming language which is very cool.

00:04:54.840 --> 00:06:23.380
So, understanding how to make a good
decision about splitting up
configuration in a way to share it
across people with really different uses
of Emacs... That's actually a
complicated topic and I want to sort of
back off and stare at it for a second. I
think Emacs is about people, so that
means it's about community. And
community means we're going to invite
disagreement. In fact, that disagreement
isn't necessarily a road-block to our
project. In fact, some of the work that
a community project can invite us to do
is to get closer to each other by
inviting those disagreements, by
learning from them--learning from
different people's styles and from how
they argue, and thinking about why they
have that perspective and what technical
benefits that perhaps radical point of
view might carry away. Some people are
really aggressive arguers, and others
are very passive and really couch their
ideas in distancing terms, to say, "well
probably, this is a good idea" or
"please double check me." Those don't
always necessarily indicate how certain
a person is, because we're different. We
have different ways of communicating
ideas like certainty or excitement.

00:06:24.560 --> 00:07:29.039
When we think about a bunch of really
diverse programmers approaching Emacs,
probably one of our our first really big
challenges is just to pick what we're
going to go after. There are a lot of
existing kit installs and things like
this. My argument is that you could
actually get pretty far just trading
files around. Maybe the more valuable
conversation to have is making the hard
decisions about, well, "should we have
vertical completion," should that be the
out of the box, and the people that want
the traditional splayed out over a
single line completion for example in
the mode line, those people are going to
add a line of config to their own setup?

00:07:29.039 --> 00:08:16.240
The way to get there? I mean, how do we
find out what works? We don't want to
slow down the people that are super
productive with Emacs by asking them to
completely break their workflows and
make it easier for new folks. At the
same time, we do want to make sure those
new people are excited by Emacs and not
turned off by having to learn the entire
jungle of Emacs history in the form of
its unique technical stylings for things
like frames, buffers, and other unique
Emacs viewpoints on important interface
concepts, especially.

00:08:16.240 --> 00:09:04.800
The encouragement here is to keep the
initialization for a project team
together as a crucible. Rather than
necessarily following our defaults of
finding the simplest configuration that
generally work and letting people
customize it, what if we tried to look
for fairly specific configurations that
we'll expect essentially all of our
developers to be using, at least when
they submit bug reports. In particular,
with this, I think that degree of
experimentation can drive back into the
Emacs development process. In the
development mailing list...

00:09:04.800 --> 00:09:15.120
I'm hoping I'll get a timing cue here.

00:09:15.120 --> 00:09:50.720
In the context of Emacs development as a
greater entity, we see some of these
struggles. Should we change this
default? Sometimes we can have the sense
that defaults in Emacs will never
change. The conversation is too
difficult. I think one thing that can
help us get there is evidence that says,
"hey my 30- to 40-person project is
using this set of bindings and here's
what we learned about brand new Emacs
users trying to come in and get work
done with that." (Amin: Yeah you still
have a couple more minutes)

00:09:50.720 --> 00:10:02.000
Oh, beautiful. Okay, great. I will try
to get through my last few slides that i
cut in my last walkthrough, but I think
i'm going quicker today, thank you. Thank
you.

00:10:02.000 --> 00:10:43.040
So let's just recap real quick: in
theory, Emacs works out of the box. That
means we're free to experiment. We can
throw it all away and start over. As an
organizational principle... I don't know
what I was thinking on that slide,
excuse me. Bringing it back around to
the free and open source software
community, our goal is to enable users
to unlock their computers, to do as much
with them as possible.

00:10:43.040 --> 00:10:57.920
That's the context to take with project
initialization, but sometimes it could
make sense to put some gloves on. I've
thrown up on the screen here just a
couple of other ideas, ways to maybe
think outside of the box.

00:10:57.920 --> 00:11:29.120
As you're putting together project nets,
my words of encouragement are to
experiment with it, try different
things, and think really specifically
about how different the development
users might be from each other as you
define standards for configuring the
user environment of Emacs specifically
for developing on a project. That's
pretty much my talk.

00:11:29.120 --> 00:11:32.959
If there's any time, I would take a
couple questions.

00:11:32.959 --> 00:11:53.839
(Amin: Thank you for your awesome talk, Corwin.
I think we have one or two minutes for a
few questions. Do you have the pad open
or would you like me to read the
questions for you?)

00:11:53.839 --> 00:12:03.519
Oh, I managed to close the pad and I am
trying to open it again. All right,
there it opened.

00:12:03.519 --> 00:12:09.360
Bringing it onto a screen where I can
see it. Will you read me the first
question while I drag windows around,
please?

00:12:09.360 --> 00:12:15.600
(Amin: Sure. It says, "do you use Emacs
as a community building tool?")

00:12:15.600 --> 00:12:19.760
Do I use Emacs as a community building
tool, or how do I?

00:12:19.760 --> 00:12:22.182
(Amin: It just says do you.) 

00:12:22.182 --> 00:13:14.639
Yes, absolutely. I think Emacs is an
ambassador to the GNU tool chain. I
think that in the fullness of time, we
will see an Emacs that makes and that
makes iOS and Android and other
closed-source tools dream. That's why
they mock us and call Emacs an operating
system. It's because it could be, if we
cared for it to be. It's quite a
threatening product from the perspective
of how many problem spaces it can
address, how many types of users it can
satisfy, the things that we can do to
make it robust in those environments. I
mean, we're always thinking about the
weak points, but is Emacs a community
building tool? Heck yeah.

00:13:14.639 --> 00:13:26.880
(Amin: There's like one or two more
questions. I think they're more
long-form so it might be better if you
took them off stream so you could keep
the schedule on time.)

00:13:26.880 --> 00:13:33.360
I would love to take those questions
offline. I will respond to you in
writing if we don't get to it in a
breakout room.

00:13:33.360 --> 00:13:38.000
Thanks so much for joining us. I can't
wait to see the rest of the conference.
See you there!

00:13:38.000 --> 00:13:42.800
(Amin: Awesome. Thank you again so much, Corwin.)