00:00:15.480
hello
00:00:16.439
everybody Welcome to the state of Ruby
00:00:19.960
tooling as mentioned my name is Vinnie
00:00:22.680
I'm a part of the Ruby developer
00:00:24.720
experience team our team created the
00:00:27.279
Ruby LSP we contribute to s aspects of
00:00:30.960
gradual typing for Ruby and we also
00:00:33.600
contribute to projects like IRB uh Ruby
00:00:36.920
debug Ardo essentially anything related
00:00:39.920
to developer experience uh if you want
00:00:42.520
to know a little bit more about me you
00:00:43.719
can find me at Vino on those links over
00:00:48.120
there I want to start by asking a
00:00:50.520
question is having options a good thing
00:00:54.800
now it's tempting to believe that the
00:00:56.640
answer to this question is always
00:00:58.719
positive and that being able to choose
00:01:00.840
is going to lead to flexibility and a
00:01:03.600
sense of freedom and autonomy so it can
00:01:06.960
never be a negative thing to be able to
00:01:09.479
choose but in reality it is a little bit
00:01:12.080
more complicated than that this paper
00:01:14.799
from the year 2000 when choice is
00:01:17.000
demotivating wanted to understand what
00:01:19.240
would happen if they were to increase
00:01:21.840
the complexity and the number of options
00:01:24.759
related to a decision someone was making
00:01:27.439
would they still feel like they got this
00:01:30.200
sense of autonomy and freedom from being
00:01:32.600
able to choose and it turns out that's
00:01:35.040
not always the case you have a risk of
00:01:37.280
inducing something called Choice
00:01:39.240
overload which is essentially when the
00:01:41.439
effort to make the decision has become
00:01:43.680
so high that people are going to start
00:01:45.640
cutting Corners to make the decision
00:01:48.520
they might consider a smaller portion of
00:01:50.920
the information that's available because
00:01:53.040
processing everything requires too much
00:01:55.600
effort they might simplify their
00:01:57.759
decision- Mak process or they might try
00:02:00.280
to eliminate the effort completely by
00:02:03.039
just preferring whatever is the default
00:02:04.920
or making no choice
00:02:07.640
whatsoever today I would like to talk to
00:02:09.599
you about options in the context of
00:02:12.360
developer
00:02:13.519
experience and where are we in the Ruby
00:02:16.200
Community when it comes to developer
00:02:18.120
experience I want to address one point
00:02:20.440
which is that it's common to hear people
00:02:22.040
say that creating tooling for Ruby is
00:02:24.560
difficult because Ruby is a dynamic
00:02:26.519
language and of course there is truth to
00:02:28.920
that STC analysis is not as accurate in
00:02:32.120
Dynamic languages as we just saw Matt's
00:02:34.760
talking about uh
00:02:36.480
upstairs but I don't think that that's
00:02:38.400
the main thing holding us back today and
00:02:40.760
Ruby is not the only Dynamic language in
00:02:43.640
existence JavaScript is also Dynamic
00:02:46.640
python is also Dynamic and the answer
00:02:49.560
for more accurate static analysis is
00:02:52.000
gradual typing this in the case of
00:02:54.319
JavaScript typescript for python it was
00:02:57.000
type hints and we have our own gradual
00:02:59.239
typing solution ions like surve and
00:03:00.920
steep so I don't believe that today that
00:03:03.040
is the main source of difficulty we face
00:03:05.080
with our Tooling in my opinion the main
00:03:07.560
challenge for a modern developer
00:03:09.159
experience is the level of fragmentation
00:03:11.760
that we have in our
00:03:13.840
tooling let's follow the story of a new
00:03:16.280
developer so this is a junior developer
00:03:18.480
just starting out and they want to have
00:03:20.440
a complete developer experience with
00:03:22.040
everything they would expect today while
00:03:24.239
trying out rust and Ruby for the very
00:03:26.959
first
00:03:27.959
time the first thing they have to do is
00:03:30.879
install the language so they need to be
00:03:32.360
able to install rust and manage
00:03:34.319
different rust versions the official way
00:03:36.799
to do that is with rust
00:03:39.120
up if they wish to write automated tests
00:03:42.480
for their program they don't have to
00:03:44.480
configure anything the test framework is
00:03:46.840
built into the language and you can just
00:03:48.720
start writing your tests and executing
00:03:50.480
them straight away with no setup step
00:03:53.959
required if they want to adopt
00:03:56.159
formatting they can just use rust format
00:03:59.000
which is the official formatter for rust
00:04:01.319
maintained by the rust
00:04:03.319
team similarly if they want to adopt
00:04:05.799
linting they can use clippy which is the
00:04:08.439
official linter for the rust language
00:04:10.280
also maintained by the rust team and if
00:04:13.000
they want to integrate all of that into
00:04:14.760
the editor and have Rich features like
00:04:17.000
completion and goat definition then they
00:04:19.880
can use rust analyzer the official
00:04:22.280
language server for rust and uh also
00:04:25.199
maintained by the rust team which
00:04:27.199
integrates with all of the other default
00:04:28.960
tools with with zero configuration
00:04:31.720
required and if we keep keep taking a
00:04:33.960
look at any other category of tooling
00:04:35.720
for rust the story here is always the
00:04:38.440
same but this well integrated experience
00:04:41.840
didn't happen by accident rust has a
00:04:45.520
dedicated developer tooling team with
00:04:48.000
several different sub teams that are
00:04:49.600
each responsible for those individual
00:04:51.639
tools and the overarching team is
00:04:54.520
responsible for Designing this unified
00:04:56.960
cohesive experience and also for coordin
00:04:59.960
the efforts of all of the sub teams so
00:05:01.880
that they're all building towards that
00:05:03.840
same
00:05:05.600
direction what we have here is very
00:05:07.960
modern default tooling but also guidance
00:05:10.600
coming from the language itself about
00:05:12.680
what your developer experience should be
00:05:14.360
when you're interacting with rust which
00:05:17.400
leads to Consolidated tooling and in
00:05:19.720
turn has several benefits that the rust
00:05:22.000
Community gets to enjoy integration is
00:05:25.280
easy because the teams are explicitly
00:05:27.800
working towards that making sure that
00:05:29.199
they use invention over configuration so
00:05:31.680
all of the tools will integrate out of
00:05:34.000
the
00:05:34.720
box it minimizes the decisions for new
00:05:37.360
developers in fact you don't have to put
00:05:39.199
in any effort in evaluating options you
00:05:41.160
can just use the set of default tools
00:05:43.600
and have a great developer
00:05:46.120
experience it also minimizes the
00:05:48.400
configuration you need because you
00:05:50.319
rarely ever if if at any time need to
00:05:52.840
configure something in order to make the
00:05:54.759
tools work the tools already work out of
00:05:57.039
the box and you only configure things if
00:05:59.000
you want a custom imiz
00:06:00.960
Behavior you get a reduced learning
00:06:03.319
curve as well because the set of tools
00:06:05.160
that you learn how to use in one rust
00:06:07.360
project is the same sort of tools that
00:06:09.199
you're going to be using on other R
00:06:10.639
projects so your knowledge
00:06:13.000
transfers and finally you also get a
00:06:15.759
very high collaboration environment if
00:06:18.080
one developer contributes a new feature
00:06:19.880
to any rust tool then the entire
00:06:22.319
Community gets to
00:06:24.599
benefit when our Junior developer
00:06:27.039
switches to Ruby they're going to have a
00:06:29.639
a very different Journey at the very
00:06:32.400
first step they're already faced with a
00:06:34.639
decision how do I install Ruby there's
00:06:37.759
no official way on how to install Ruby
00:06:40.280
so they're going to have to evaluate
00:06:41.880
several different
00:06:43.360
options on how they can do that and here
00:06:46.479
on the slide we have 11 different
00:06:48.440
options some of which allow you to
00:06:50.759
install Ruby others allow you to switch
00:06:53.160
your Ruby version and a few of them
00:06:55.240
allow you to do both and in addition to
00:06:58.360
that many of these these options are
00:07:00.440
written as pure shell scripts which for
00:07:03.680
some of them make them shell specific
00:07:05.680
and operating system specific so in
00:07:08.680
reality only certain combinations are
00:07:11.080
actually available depending on your
00:07:14.240
setup if they decide to adopt automated
00:07:17.520
testing they have another decision to
00:07:19.360
make but in this case not because we
00:07:21.840
don't have a default test framework is
00:07:24.199
because we have
00:07:25.560
two both mini test and test unit are
00:07:28.759
included as a part of your Ruby
00:07:30.520
installation so it's not possible to
00:07:32.599
provide developers with a completely
00:07:34.960
zero testing zero configuration testing
00:07:38.240
experience at the very minimum you still
00:07:40.400
need to add the gem to your gem file
00:07:42.120
because we don't know which one of the
00:07:43.240
defaults you're going to
00:07:45.759
pick if they decide to adopt typing they
00:07:48.639
have another decision to make they can
00:07:50.599
adopt gradual type Checkers like Steep
00:07:52.800
and Survey but they also have uh at
00:07:55.800
their disposal more novel approaches
00:07:58.080
like typro which is a type profiler and
00:08:00.680
you can get varying levels of type
00:08:02.560
checking with other tools as
00:08:05.680
well for formatting and linting and here
00:08:08.879
I put both in the same category because
00:08:11.159
in the rubby Community they're it's not
00:08:13.120
so clearly separated you also have
00:08:16.159
multiple options that you can choose
00:08:17.720
from and different combinations that you
00:08:19.599
can make with all sorts of
00:08:22.039
linters and if you want to try to
00:08:24.120
integrate all of that into the editor
00:08:26.639
then uh they're going to be evaluating
00:08:28.599
families of Link servers you have the
00:08:31.400
type checking family which focus on
00:08:34.200
features where type checking is
00:08:35.640
particularly important like completion
00:08:37.959
and goats definition you have the
00:08:40.240
general family of language servers like
00:08:42.080
Ruby LSP and solar graph which try to
00:08:44.560
give you the best experience possible
00:08:46.920
even if no type system is present and
00:08:49.560
then you have the formatting family of
00:08:51.040
language servers that provide you with
00:08:54.279
formatting if we continue looking at any
00:08:56.880
other category of tooling for Ruby the
00:08:59.000
story here is the same we have a lot of
00:09:01.720
fragmentation at every single
00:09:04.120
level if we were to consider that our
00:09:06.959
developer wants to have a full
00:09:08.760
experience so if they want tools from
00:09:10.839
all of the all of the categories that we
00:09:12.839
saw and if we limit that they can only
00:09:15.519
pick one per category which is not the
00:09:17.959
case you can combine multiple linters
00:09:19.800
multiple language servers but for the
00:09:21.680
sake of the argument let's say they can
00:09:23.760
only pick one and keeping in mind that
00:09:26.440
the lists are not exhaustive there are
00:09:28.440
many tools that are not on those lists
00:09:30.959
but there are entire categories of tools
00:09:32.680
missing like debugging which I didn't
00:09:34.720
put
00:09:35.760
up with these constraints rust has one
00:09:39.600
possible combination of how you can
00:09:41.560
configure your environment the set of
00:09:44.000
default tools and Ruby has the very
00:09:47.000
conservative number of 11,000 ways which
00:09:50.040
you can configure your
00:09:52.079
environment if we were to visualize that
00:09:54.760
rust's visualization would be a straight
00:09:57.120
line where on each level you just pick
00:09:58.959
the next default tool and for Ruby you
00:10:03.160
can't quite put that on a slide because
00:10:05.000
there are simply way too many
00:10:06.480
combinations to
00:10:08.000
visualize it's no wonder that it's
00:10:10.959
extremely difficult to guarantee that
00:10:12.920
Integrations are going to work properly
00:10:15.040
for every possible combination of way
00:10:17.040
that you can configure your
00:10:19.959
environment if we take a look at
00:10:22.440
different projects that all use
00:10:24.720
different sets of tools you also
00:10:26.839
conclude that learning Ruby itself is is
00:10:29.440
not enough to be able to contribute to
00:10:31.000
all of those for example if you
00:10:33.000
contribute to project a and you're not
00:10:35.279
used to gradual typing but then you have
00:10:37.279
to work on Project B then you're going
00:10:39.600
to have to learn the specifics of steep
00:10:41.480
and RBS in order to be able to make a
00:10:43.600
successful contribution but if you then
00:10:46.079
want to contribute to open source
00:10:47.440
project C now you need to learn the
00:10:49.480
specifics of another gradual typing
00:10:51.279
system surve in order to be able to
00:10:53.480
contribute and this is not exclusive to
00:10:55.959
type Checkers it's the same thing for
00:10:57.320
test Frameworks linters or even
00:10:59.279
debuggers if they use a different
00:11:00.680
debugger you're going to have to learn
00:11:01.760
how to use it to debug your
00:11:04.160
code the essence here is that the set of
00:11:06.760
tools you learn how to use in one
00:11:08.200
project may not fully transfer to other
00:11:10.680
Ruby
00:11:12.519
projects what we have here is missing
00:11:14.800
defaults but also guidance from the
00:11:17.000
language Ruby is not prescriptive about
00:11:19.839
what your developer experience should be
00:11:22.680
and that in practice is delegating the
00:11:25.360
design of the overall developer
00:11:26.880
experience to the community which leads
00:11:29.079
to to fragmentation and in turn we see
00:11:32.120
all of the Opposites of what the rust
00:11:34.279
Community
00:11:35.360
enjoys integration is extremely
00:11:38.160
difficult because we have several
00:11:39.440
implementations of the same tools all of
00:11:42.120
which don't conform to any standardized
00:11:44.040
API so you need custom integration for
00:11:46.720
all of those tools you get an excessive
00:11:49.639
number of decisions for new developers
00:11:52.160
can we really expect that a junior
00:11:54.120
developer who's just starting out is
00:11:56.079
going to understand all of the
00:11:57.200
trade-offs they're making by picking
00:11:59.279
tools from all of those
00:12:01.480
categories because integration is
00:12:04.200
difficult it also increases the amount
00:12:06.120
of configuration required because the
00:12:08.240
easy way out is to ask users for more
00:12:10.320
and more configuration so that they are
00:12:12.360
responsible for ensuring that the tools
00:12:14.440
are going to integrade properly so it
00:12:16.360
degrades our ability to provide a nice
00:12:18.079
out-of-the-box
00:12:20.240
experience it increases the learning
00:12:22.360
curve because the set of tools you learn
00:12:24.040
how to use in one project will not
00:12:26.279
necessarily transfer to other Ruby
00:12:28.120
projects you may need to re learn the
00:12:29.959
same tools you already knew but just a
00:12:31.760
different
00:12:33.440
implementation and it spreads the
00:12:35.399
efforts of our
00:12:37.440
community if we take a look at
00:12:39.320
JavaScript which is another highly
00:12:41.160
fragmented ecosystem they have similar
00:12:44.360
issues they have dependency management
00:12:46.279
fragmentation runtime fragmentation in
00:12:49.160
that sense they're very similar to the
00:12:50.600
Ruby ecosystem but they do have an
00:12:53.959
advantage they have a much larger
00:12:56.360
Community if we take a look at the most
00:12:58.519
recent stack overflow results the number
00:13:01.279
of respondents for JavaScript was at 62%
00:13:04.160
it was the number one programming
00:13:05.480
language and Ruby was just a little over
00:13:09.240
5% so maybe JavaScript has enough people
00:13:12.880
in their community that they can afford
00:13:15.440
to have all of this richness and
00:13:17.199
diversity in tooling they can sustain
00:13:19.399
all of these different options but I
00:13:21.600
want to argue that a community our size
00:13:24.279
cannot afford to divide our efforts
00:13:26.240
maintaining several different
00:13:28.120
implementations of the same highly
00:13:30.160
complex
00:13:31.639
tools as another statistic if we take a
00:13:34.120
look at the number of non-unique
00:13:35.519
contributors for language servers as of
00:13:38.240
the time when I wrote the slide rust
00:13:40.199
analyzer had
00:13:42.040
874
00:13:43.600
contributors and if we sum up the number
00:13:45.959
of five actively maintained language
00:13:48.160
servers in our community serbet steep
00:13:50.959
solar graph Ruby LSP and typr and this
00:13:54.120
is non-unique so the pandabot is being
00:13:56.079
counted five times and I'm being counted
00:13:58.639
more than one too it's an inflated
00:14:00.759
number and even though it's inflated
00:14:03.199
we're not even close to the number of
00:14:04.800
contributors that rust analyzer has
00:14:07.480
which is a lot more impactful to the
00:14:09.120
quality of our tools than the fact that
00:14:10.680
Ruby is a dynamic
00:14:13.800
language and I want to say that building
00:14:16.279
new tools might make sense I am not
00:14:19.000
arguing that we should stop making new
00:14:21.000
tools the point that I'm trying to make
00:14:23.480
is that you don't automatically get a
00:14:25.000
better developer experience simply by
00:14:26.880
creating more tools not with without
00:14:29.320
considering how they're going to fit
00:14:30.720
into this General picture of developer
00:14:33.160
experience how they're going to
00:14:34.279
integrate with the rest of the
00:14:36.079
ecosystem because developer experience
00:14:38.040
is not about the tools in isolation it's
00:14:39.680
about how they combine to allow you to
00:14:41.959
interact with the
00:14:43.839
language if we go back to the article
00:14:46.360
that we were discussing at the very
00:14:47.639
beginning what we want to avoid is being
00:14:50.440
in the last option of making no choice
00:14:54.320
whatsoever that would mean that
00:14:56.000
evaluating all of the different options
00:14:57.800
requires so much effort it that in the
00:15:00.040
best possible case the developer might
00:15:01.600
be neglecting an entire category of
00:15:03.399
tooling by choice and in the worst
00:15:06.519
possible case they might decide to not
00:15:08.120
try out Ruby and go try some other
00:15:10.160
language that is easier to get
00:15:12.480
started we want to be where rust is
00:15:14.759
where the default option is so good that
00:15:17.680
they can just use the defaults and have
00:15:19.759
a great developer experience without
00:15:21.199
investing anything in the
00:15:23.480
decision after all using the tools is
00:15:26.399
not the developer final goal the
00:15:29.360
developer doesn't have it as it final
00:15:31.600
goal to use rails or to use the Ruby LSP
00:15:34.480
or to use
00:15:35.519
sorbet those are just stepping stones
00:15:38.199
and the goal is to build something so
00:15:40.720
setting up your development environment
00:15:42.240
should be absolutely effortless and it
00:15:44.399
should set you up for Success so that
00:15:46.279
you can focus on what it is that you're
00:15:48.399
trying to
00:15:50.720
build and perception matters right the
00:15:53.240
way that Ruby is perceived is important
00:15:55.560
for
00:15:57.000
us if you search online there's a lot of
00:15:59.560
research into how certain elements of
00:16:01.480
one thing can impact the perception of
00:16:03.759
that thing overall like certain elements
00:16:06.120
of Brands can impact how consumer feel
00:16:09.120
consumers feel about their products or
00:16:11.639
how certain aspects of a location can
00:16:13.880
impact tourism or how certain elements
00:16:16.480
of a restaurant can impact how people
00:16:18.120
feel about the
00:16:20.000
food this effect is called the halo
00:16:22.440
effect it is when one trait impacts the
00:16:25.560
overall perception of that thing I I do
00:16:29.120
not believe that Ruby is uh immune to
00:16:31.519
this effect I don't think that Ruby the
00:16:33.279
language is evaluated independently from
00:16:36.319
its developer experience so a new
00:16:38.440
developer wouldn't know where the
00:16:40.560
language ends and the tooling starts or
00:16:42.759
who maintains what so the developer
00:16:45.000
experience is important for a new
00:16:46.720
developer to evaluate how they feel
00:16:48.639
about ruby and what their first uh
00:16:51.680
contact with the language is how they
00:16:53.519
feel about that the question I want to
00:16:56.560
raise and the concern that I have is
00:16:58.639
could we be losing developers who are
00:17:00.440
genuinely interested in Ruby because
00:17:02.720
getting started may be a little bit more
00:17:04.600
difficult than it is in other languages
00:17:07.280
today let's take a look at another
00:17:10.160
situation if we take a look at Windows
00:17:12.880
setting up a modern development
00:17:14.520
environment on Windows is a little bit
00:17:16.520
more difficult than it is on Mac OS and
00:17:19.480
Linux and the reason for that is that
00:17:21.600
many of the tools that we use don't
00:17:23.400
support Windows in particular many of
00:17:25.919
the version managers don't support
00:17:27.640
Windows so doesn't support Windows but
00:17:30.320
also core methods of the Ruby API are
00:17:32.640
sometimes not implemented for Windows
00:17:36.640
let's take a look at some more
00:17:39.039
statistics in the latest stack Overflow
00:17:41.400
survey professional usage of Windows was
00:17:44.440
at
00:17:45.799
47% and the numbers here don't add up to
00:17:48.760
100 because you could select all of the
00:17:50.919
operating systems that you worked
00:17:53.440
with jetbrains which is a company that
00:17:56.120
makes uh developer tooling and
00:17:58.120
Commercial ID these they're very good if
00:18:00.240
you don't know them check them out they
00:18:02.240
have their own uh survey called the
00:18:04.559
stateof the developer ecosystem and in
00:18:07.200
their survey Windows usage was even
00:18:09.039
higher at
00:18:10.960
64% the rust Community survey reported
00:18:14.360
uh windows in the third place but still
00:18:17.600
with a healthy number of developers
00:18:19.600
about 32% which is roughly about a third
00:18:23.080
of the community it's a decent
00:18:25.320
number and the Ruby on Rails Community
00:18:27.760
survey reported reported about 4% of
00:18:30.600
rubyists using Windows and this number
00:18:34.440
is not actually just Windows it includes
00:18:37.200
WSL which is the Linux subsystem so the
00:18:40.840
number of people using just Windows
00:18:43.159
purely is even smaller than
00:18:45.120
that I think that when we see a number
00:18:47.679
this discrepant from the rest of the
00:18:49.360
programming Community we have to ask
00:18:51.760
ourselves what is happening exactly here
00:18:55.120
if the answer is that rubies just have a
00:18:57.440
predisposition to using EOS and maybe
00:19:00.320
there's no problem but the question I
00:19:02.440
want to raise is could we be losing
00:19:04.120
developers who are genuinely interested
00:19:06.440
in our community and maybe have a hard
00:19:08.240
time getting
00:19:11.000
started what can the future look like
00:19:13.880
for our tooling I want to be super Fair
00:19:16.960
here and mention that rust case is the
00:19:19.280
absolute exception if we take a look at
00:19:21.919
many other popular programming languages
00:19:24.000
like python JavaScript uh PHP most of
00:19:27.520
them are also frag M mented and don't
00:19:29.600
have this official cohesive story of
00:19:32.559
developer experience in fact the only
00:19:35.000
other language that I found to have a
00:19:37.240
somewhat similar experience as rust is
00:19:39.280
go which also has official tooling
00:19:41.320
that's quite
00:19:42.720
solid and there's a relationship here
00:19:45.000
between these lists the fragmented
00:19:47.679
languages are all much much older than
00:19:50.679
go and rust are which makes perfect
00:19:53.919
sense many of the techniques that we use
00:19:56.360
today didn't exist back when Ruby and
00:19:59.120
python were created so it's only natural
00:20:01.159
that the ecosystem of tooling evolved
00:20:03.559
organically and we ended up in a place
00:20:05.760
with more fragmentation whereas rust and
00:20:08.320
go had the opportunity to plan the path
00:20:11.159
ahead from the very beginning and were
00:20:13.080
able to create this compelling story for
00:20:15.039
developer
00:20:16.679
experience but there's nothing stopping
00:20:18.799
us from learning from all of the success
00:20:21.280
that rust has had and bringing that
00:20:23.400
learning to our community and in fact
00:20:25.760
there are other programming languages uh
00:20:27.760
that are catching up to this Elixir
00:20:30.559
recently announced their official
00:20:32.360
language server team which is going to
00:20:33.919
build uh Rich Integrations for the
00:20:36.400
editor for The Elixir
00:20:38.440
language and in the python land uh the
00:20:42.240
problem of having modern integrated by
00:20:44.679
default highly performing tooling was
00:20:47.039
deemed so valuable that there is a
00:20:48.520
company behind it Astro is a company for
00:20:51.440
which the mission is to create the full
00:20:53.520
Suite of next generation of python
00:20:57.159
tooling our tool buing is also becoming
00:20:59.520
more integrated but I do think that
00:21:01.000
there's still a ton of work to be done
00:21:02.600
if we want to match what we're seeing in
00:21:04.960
those communities especially in in
00:21:09.000
Rust however I do also believe that we
00:21:11.360
can have a more integrated experience
00:21:13.520
even
00:21:14.720
today and the answer for me comes from
00:21:17.159
outside of our community it comes from
00:21:19.320
the language server protocol language
00:21:22.080
servers are uniquely tailored to
00:21:24.279
integrate and aggregate with other tools
00:21:27.080
for example they can integrate with your
00:21:29.120
test framework your linter your type
00:21:31.000
Checker your formatter but today in our
00:21:33.720
community we are seeing language server
00:21:36.760
fragmentation and that of course has
00:21:38.760
consequences for the developer
00:21:40.039
experience as well if we take a look at
00:21:42.919
editors many of them have some sort of
00:21:45.679
extension hosts like a separate process
00:21:48.240
that will execute the code in your
00:21:50.039
extensions and is normally shared across
00:21:52.440
all of them if you have multiple
00:21:55.279
language servers connected in addition
00:21:57.640
to having to configure each one of them
00:21:59.520
and ensuring that they don't step on
00:22:00.960
each other's toes they're all going to
00:22:03.039
do duplicate work to index your codebase
00:22:05.679
figure out what is available and then
00:22:07.799
maintain an inmemory representation of
00:22:10.080
your entire code base which could be
00:22:12.039
several gigabytes in memory depending on
00:22:14.400
how large of a project you're working
00:22:16.600
on and then for every single operation
00:22:19.200
you do in the editor like opening
00:22:20.600
documents those operations are also
00:22:22.760
going to be replicated for each language
00:22:25.000
server which are also going to maintain
00:22:27.200
duplicate representations of all of the
00:22:29.520
documents you're interacting
00:22:31.720
with it's a lot of duplicate work for
00:22:34.760
maintainers who end up having to solve
00:22:36.679
the same problems over and over again
00:22:39.200
but it's also increased memory usage and
00:22:42.760
uh under certain conditions one language
00:22:44.919
server could potentially make the others
00:22:46.960
the other ones
00:22:49.400
slower if we could consolidate language
00:22:52.320
servers I think we could do a better job
00:22:54.559
at delivering a more integrated
00:22:56.279
out-of-the-box experience for Ruby
00:22:59.480
and of course I am incredibly biased but
00:23:02.919
the Ruby LSB may help
00:23:05.120
here we've been experimenting for a long
00:23:08.240
time now with what we're calling Ruby
00:23:10.279
LSP add-ons and this is a way for other
00:23:13.360
gems to deliver Integrations for the
00:23:15.559
editor through the Ruby LSP it's
00:23:17.919
essentially uh an API layer for the
00:23:20.320
language server so that other gems can
00:23:22.679
enhance the features that we
00:23:25.279
provide if one of your dependencies in
00:23:27.720
your project ports an add-on it will be
00:23:31.039
automatically detected and loaded so
00:23:33.360
that your experience gets more tailored
00:23:35.240
to the project that you're working on
00:23:37.240
for example uh standard the linter and
00:23:39.840
formatter recently shipped with an
00:23:42.760
add-on inside the gem if you have
00:23:45.080
standard in your gem file the Ruby LSB
00:23:47.320
will automatically pick that up and
00:23:49.480
integrate with it for
00:23:51.559
linting and alternatively addons can
00:23:54.760
also be distributed as an independent
00:23:57.039
gem which we are also have some examples
00:24:00.000
of
00:24:01.960
those there is no need to write a vs
00:24:04.559
code extension or editor plugin with
00:24:06.640
this approach there's no need to write a
00:24:09.320
brand new language server and decide how
00:24:11.640
you're going to represent documents or
00:24:14.200
how you're going to handle different
00:24:15.520
encodings coming from the editor and
00:24:18.080
there is no need to write your own
00:24:20.880
toolkit of static analysis tools to be
00:24:23.400
able to understand Ruby code all of that
00:24:25.640
is provided by the add-on API
00:24:30.159
in this scenario the Ruby LSP is a
00:24:31.840
little bit more than just a language
00:24:33.159
server but also a platform for editor
00:24:35.919
Integrations and through the add-on API
00:24:38.120
we can make the experience richer based
00:24:40.360
on what type of project the user is
00:24:42.960
currently working on we already have the
00:24:45.360
rails add-on which adds rails specific
00:24:47.960
features to the Ruby LSP we have the
00:24:50.960
rspec add-on adding support for rspec
00:24:53.559
tests as I mentioned we also have the
00:24:56.120
standard add-on which adds a link and uh
00:24:59.640
formatting for for standard if you're
00:25:02.559
using that but we can have add-ons for
00:25:05.480
any kind of tool or framework that could
00:25:08.240
benefit from providing their users with
00:25:10.679
a richer editor
00:25:12.559
experience and because everything's
00:25:14.640
going through the language
00:25:16.159
server and and the language server
00:25:18.240
connects to any editor that supports the
00:25:20.320
protocol then everything that gets
00:25:22.279
contributed by add-ons is also available
00:25:24.440
to any editor it doesn't have to be vs
00:25:27.080
code specific
00:25:29.440
and in the age of artificial
00:25:30.840
intelligence if your editor has an llm
00:25:34.240
and if it provides an API that the Ruby
00:25:36.679
LSP could use to provide extra
00:25:39.000
information to the llm then we could
00:25:41.399
make it so that the Ruby LSP will
00:25:43.640
provide extra information and extra
00:25:45.679
context based on where you're currently
00:25:47.760
editing but also based on everything
00:25:50.120
that was contributed by add-ons this
00:25:53.279
doesn't exist at the moment these apis
00:25:56.039
are not available but I'm hoping they
00:25:57.480
will be available
00:25:59.760
soon I think developer experience for
00:26:02.120
Ruby can be a lot more than just good in
00:26:05.320
fact I believe that we can lead the way
00:26:06.960
when it comes to developer experience in
00:26:08.840
highly Dynamic languages even when
00:26:11.200
there's no type system present and to me
00:26:14.320
there are two important things for us to
00:26:15.960
get there number one is considering a
00:26:18.880
unified vision for developer experience
00:26:21.760
to achieve the level of sophistication
00:26:23.760
that we're seeing in the rust Community
00:26:26.480
we need to be able to have the tools
00:26:28.559
integrate automatically and have the
00:26:30.360
tools create apis to talk to each other
00:26:33.000
building the tools in isolation will not
00:26:35.120
cut it and so to get there we also need
00:26:38.240
to coordinate our efforts join forces
00:26:41.640
and ensure that uh Ruby has a great
00:26:44.159
developer
00:26:45.559
experience thank you very much
