There are certain instances where you may want to build ElixirLS to run against the version of Elixir/Erlang that you're using. The recommendation from the package is:
If you're packaging these archives in an IDE plugin, make sure to build using the minimum supported OTP version for the best backward-compatibility If you're like me, you may not care to support older versions of Elixir. How do we configure the plugin to run the latest version?
The output I see in VSCode's Output tab (Shift-Command-U on macOS) for the ElixirLS extension:
[Info - 4:33:53 PM] Started ElixirLS v0.13.0
[Info - 4:33:53 PM] ElixirLS built with elixir "1.12.3" on OTP "22"
[Info - 4:33:53 PM] Running on elixir "1.14.2 (compiled with Erlang/OTP 25)" on OTP "25"
[Info - 4:33:53 PM] Elixir sources not found (checking in /home/build/elixir). Code navigation to Elixir modules disabled.
[Info - 4:33:54 PM] Loaded DETS databases in 32ms
[Info - 4:33:54 PM] Starting build with MIX_ENV: test MIX_TARGET: host
[Info - 4:33:55 PM] Compile took 854 milliseconds
There are numerous articles on building from source. What if we'd prefer to build the extension instead?
Let's unpack that Docker command to perform each step:
git clone --recursive --branch v0.13.0 https://github.com/elixir-lsp/vscode-elixir-ls.git /tmp/vscode-elixir-ls
.cd /tmp/vscode-elixir-ls
.npm install
.elixir-ls
directory: cd elixir-ls
.mix deps.get
.cd ..
.npx vsce package
.extensions
directory in $HOME: mkdir -p $HOME/extensions
.cp /tmp/vscode-elixir-ls/elixir-ls-0.13.0.vsix $HOME/extensions
.rm -rf /tmp/vscode-elixir-ls
.It is crucial to install Elixir v1.14.x and Erlang 25.1.x using your favorite method prior to packaging the new extension. I'm using asdf global
to do this, but you could create a local .tool-versions
inside the tmp folder if you wish.
The extension should now live at /tmp/vscode-elixir-ls/elixir-ls-0.13.0.vsix
.
The remaining steps copy the package to a directory the Docker container knows, and it's okay to stop here.
Because the prepublish.bash
file that executes at step #7 runs mix deps.get
, we can eliminate steps 4, 5, and 6.
These commands also compile the extension using MIX_ENV=dev
, which we may not want.
To change this, we can edit the last line in prepublish.bash
to MIX_ENV=prod mix elixir_ls.release -o ../elixir-ls-release
to compile for production.
Putting all of the (now reduced) commands together:
git clone --recursive --branch v0.13.0 https://github.com/elixir-lsp/vscode-elixir-ls.git /tmp/vscode-elixir-ls
cd /tmp/vscode-elixir-ls
npm install
npx vsce package
mkdir -p $HOME/extensions
cp /tmp/vscode-elixir-ls/elixir-ls-0.13.0.vsix $HOME/extensions
rm -rf /tmp/vscode-elixir-ls
We can install the extension from the VSIX file using the UI or the command code --install-extension $HOME/extensions/elixir-ls-0.13.0.vsix
.
To take advantage of the new extension in our projects, we need to rm -rf .elixir_ls
and navigate to an Elixir file.
ElixirLS won't start compiling until an Elixir file is open in the editor, and it'll usually take a few minutes to rebuild everything.
With the new extension installed we should see the change in VSCode's Output tab:
[Info - 4:35:42 PM] Started ElixirLS v0.13.0
[Info - 4:35:43 PM] ElixirLS built with elixir "1.14.2" on OTP "25"
[Info - 4:35:43 PM] Running on elixir "1.14.2 (compiled with Erlang/OTP 25)" on OTP "25"
[Info - 4:35:43 PM] Elixir sources not found (checking in /home/build/elixir). Code navigation to Elixir modules disabled.
[Info - 4:35:48 PM] Loaded DETS databases in 414ms
[Info - 4:35:48 PM] Starting build with MIX_ENV: test MIX_TARGET: host
[Info - 4:35:49 PM] Compile took 1811 milliseconds
I upgraded the excellent asdf
version manager using Homebrew and ran into a snag when trying to perform mix
commands.
I encounterd the error /Users/jbrayton/.asdf/shims/mix: line 13: /usr/local/Cellar/asdf/0.10.2/libexec/bin/asdf: No such file or directory
.
The key to notice here is the path /usr/local/Cellar/asdf/0.10.2/
when the newest version is 0.11.0
, as there is clearly a mismatch.
I restarted my terminal and shell, but the problem persisted. I noticed all the files in ~/.asdf/shims
had the line exec /usr/local/Cellar/asdf/0.10.2/libexec/bin/asdf exec "odbcserver" "$@" # asdf_allow: ' asdf '
.
This line is not what we wanted and indicates the problem.
After looking at the pinned https://github.com/asdf-vm/asdf/issues/785 and then following that to https://github.com/asdf-vm/asdf/issues/1393, the solution rm -rf ~/.asdf/shims; asdf reshim
fixes my problem.
Now, whenever I examine one of the shim files, I see the line exec /usr/local/opt/asdf/libexec/bin/asdf exec "mix" "$@" # asdf_allow: ' asdf '
as expected.
The directory /usr/local/opt
is what I see when I run the command brew --prefix asdf
as the prefix is no longer /usr/local/Cellar/asdf/0.10.2/
or the Cellar location.
This corrective measure should be a more permanent solution moving forward as the prefix /usr/local/opt
should no longer change in the future.
This issue was also somewhat of a perfect storm as Phoenix 1.7 rc.1 dropped two days ago and I had just upgraded a bunch of homebrew packages, including asdf
.
Recently, after a brief outage at work, I wondered if it would be possible to replicate the problem locally using Laravel Valet. My Google search landed on this StackOverflow post, where the answers shot down the idea. Not to be dissuaded by something I read on the internet, I started investigating if it was possible and stumbled upon what I think is a viable solution. There aren't very many hoops to jump through or major quirks so I believe it's not only possible but could be supported out of the box.
In my case, I want to proxy the domain scdn-app.thinkorange.com
through my local version of the Laravel application.
~/.config/valet/config.json
on macOS and change the tld
parameter from test
to com
.valet link scdn-app.thinkorange
to set up our valet configuration to point the domain to this directory.valet secure scdn-app.thinkorange
to set up the SSL certificate.cd ~/.config/valet/dnsmasq.d
.cp tld-test.conf tld-com.conf
.address=/.com/127.0.0.1
and save the file.valet isolate --site scdn-app.thinkorange php@8.1
./etc/hosts
file to redirect the domain to 127.0.0.1
for ipv4 and ::1
. I use the excellent Gas Mask to make this step easier.Now we should have a functional production proxy through our local machine. This configuration creates a few problems around keeping the com
TLD.
Fortunately, a few extra steps are necessary for us to switch back to .test
while also keeping this site functional.
~/.config/valet/config.json
again and change the tld
parameter from com
back to test
. This change will immediately break our site.cd ~/.config/valet/Sites
.ls -al
to list the directory, we'll see our site scdn-app.thinkorange
. Let's change that.mv scdn-app.thinkorange scdn-app.thinkorange.com
.Our site should now be working again. We are also able to continue serving our previous local test domains.
Because we can create a permanently functional system using these steps, I believe it should be possible to create a pull request to reduce the number of hoops we have to jump through.
I'd love to be able to run valet link scdn-app.thinkorange.com.
with a period at the end to denote I'm including the full domain with TLD.
That would eliminate the temporary step of editing the config.json
file, and the Sites
directory would just work(TM) as it would include the .com
directory name.
I don't believe we even need the dnsmasq changes as I'm able to navigate to a functional site without them.
I believe Gas Mask is doing the work, but it's better to be safe than sorry.
If you'd prefer a YouTube video where I stumble through recreating these steps from scratch:
Tell me if you've done this before. You write up a nice little prototype of an idea in Livebook. You then get distracted by life situations like eating, writing an email, or taking a nap. You feel the need to close Livebook or prune the multiple sessions you've had running for weeks now. Because you have a million tabs open (with a session manager) and too many in Livebook to individually check, you restart your computer and let it crash(TM). When you open up Livebook again, "Oh. Shiiiiit" you exclaim. Where the hell did that notebook go? I'm 100% sure I clicked the disk icon, what the hell? If you're like me, you may have created this forked Livebook from memory, possibly taking a better approach.
There is a better way to handle this scenario. Livebook has had autosaves since 0.4:
The feature was added in this PR according to the changelog:
https://github.com/livebook-dev/livebook/pull/736
To find your autosave files:
For the Desktop application and CLI in production: ~/Library/Application Support/livebook/autosaved/
.
/Users/jbrayton/Library/Application Support/livebook/autosaved/
.For the dev environment: in config/dev.exs
, this is set as config :livebook, :data_path, Path.expand("tmp/livebook_data/dev"
.
/Users/Shared/repositories/personal/elixir/livebook/tmp/livebook_data/dev/autosaved/
.For the test environment: in config/test.exs
this is set as Path.expand("tmp/livebook_data/test")
.
/Users/Shared/repositories/personal/elixir/livebook/tmp/livebook_data/test/autosaved/
.Notebooks are saved by day in the autosave directory and the date corresponds to when they were created (when you immediately click the New notebook button).
To view or change your autosave directory in the CLI:
Settings
under the Home
and Learn
links.For the Desktop application, the port will be randomized but you can either change the URL to tack on /settings
after the port or click around to the settings page as described earlier.
If you are curious as to how this setting gets configured, we can start by looking at Livebook.Settings.default_autosave_path()
in https://github.com/livebook-dev/livebook/blob/main/lib/livebook/settings.ex#L32-L34.
We follow Livebook.Config.data_path()
to https://github.com/livebook-dev/livebook/blob/main/lib/livebook/config.ex#L76-L78 then the Erlang function :filename.basedir(:user_data, "livebook")
.
Running this in Livebook we get the output "/Users/jbrayton/Library/Application Support/livebook"
, precisely where the desktop app stores its files.
What lead me to this discovery, after vaguely remembering autosave was a thing, was looking for files on my computer.
I purposefully install and use the locate
command because I find it far easier to use than remembering the find -name
syntax.
Here's the output for checking that the word autosave
is in any directory or file name:
⋊> ~ locate autosaved/
/Users/Shared/repositories/personal/elixir/livebook/tmp/livebook_data/dev/autosaved/2022_10_31/18_25_03_mapset_drills_hedh.livemd
/Users/Shared/repositories/personal/elixir/livebook/tmp/livebook_data/dev/autosaved/2022_11_03/18_12_21_teller_bank_challenge_pv4e.livemd
/Users/Shared/repositories/personal/elixir/livebook/tmp/livebook_data/dev/autosaved/2022_11_03/18_13_39_untitled_notebook_pidb.livemd
/Users/Shared/repositories/personal/elixir/livebook/tmp/livebook_data/dev/autosaved/2022_11_03/19_31_57_dockyard_academy_amas_p75r.livemd
/Users/Shared/repositories/personal/elixir/livebook/tmp/livebook_data/dev/autosaved/2022_11_03/20_02_17_intro_to_timescale_jm7r.livemd
/Users/Shared/repositories/personal/elixir/livebook/tmp/livebook_data/dev/autosaved/2022_11_08/11_10_21_untitled_notebook_ervg.livemd
/Users/Shared/repositories/personal/elixir/livebook/tmp/livebook_data/dev/autosaved/2022_11_22/19_15_12_untitled_notebook_p75e.livemd
What I found interesting was that my files in ~/Library/Application Support/livebook/autosaved/
did not show up.
Had I not realized there could be different locations, I may have overlooked the notebook I was looking for all along.
I have no clue why locate
doesn't scour the directories in ~/Library
it should have access to but that's a problem for another day.
In December of 2021, Brian Cardarella introduced DockYard Beacon CMS in this series of tweets:
Over the course of the past year, I've created a sample project a total of 3 times to get a better understanding for how it operates. I haven't seen a ton of content on Beacon beyond announcement tweets, the mention in the ElixirConf 2022 keynote, and https://beaconcms.org/. This post covers the complete instructions in the readme with some notes on where to go from here. I had run into a few snags at first but a lot of those initial pain points have been hammered out so far. While a basic "Hello World" sample project is great, I plan on expanding on the sample with deeper dives into how Beacon serves up content. It takes a few novel approaches I haven't seen before to create either a CMS that runs along your application or it can be centralized with multi-tenancy. One CMS can service all of your ancillary marketing sites, blogs, or wherever you need the content.
The following instructions are also listed on the sample application readme so you're welcome to skip them if you want to look at the code.
Create a top-level directory to keep our application pair. This is temporary as the project matures.
mkdir beacon_sample
Clone GitHub - BeaconCMS/beacon: Beacon CMS to ./beacon
.
git clone git@github.com:BeaconCMS/beacon.git
Start with our first step from the Readme
mix phx.new --umbrella --install beacon_sample
Go to the umbrella project directory
cd beacon_sample/
Initialize git
git init
Commit the freshly initialized project
Initial commit of Phoenix v1.6.15
as of the time of this writing.Add :beacon as a dependency to both apps in your umbrella project
# Local:
{:beacon, path: "../../../beacon"},
# Or from GitHub:
{:beacon, github: "beaconCMS/beacon"},
apps/beacon_sample/mix.exs
and apps/beacon_sample_web/mix.exs
under the section defp deps do
.mix deps.get
to install the dependencies.Commit the changes.
Add :beacon as a dependency to both apps in your umbrella project
seems like a good enough commit message.Configure Beacon Repo
Beacon.Repo
under the ecto_repos:
section in config/config.exs
.Configure the database in dev.exs
. We'll do production later.
# Configure beacon database
config :beacon, Beacon.Repo,
username: "postgres",
password: "postgres",
database: "beacon_sample_beacon",
hostname: "localhost",
show_sensitive_data_on_connection_error: true,
pool_size: 10
Commit the changes.
Configure Beacon Repo
subject with Configure the beacon repository in our dev only environment for now.
body.Create a BeaconDataSource module that implements Beacon.DataSource.Behaviour
Create apps/beacon_sample/lib/beacon_sample/datasource.ex
defmodule BeaconSample.BeaconDataSource do
@behaviour Beacon.DataSource.Behaviour
def live_data("my_site", ["home"], _params), do: %{vals: ["first", "second", "third"]}
def live_data("my_site", ["blog", blog_slug], _params), do: %{blog_slug_uppercase: String.upcase(blog_slug)}
def live_data(_, _, _), do: %{}
end
Add that DataSource to your config/config.exs
config :beacon,
data_source: BeaconSample.BeaconDataSource
Commit the changes.
Configure BeaconDataSource
Make router (apps/beacon_sample_web/lib/beacon_sample_web/router.ex
) changes to cover Beacon pages.
Add a :beacon
pipeline. I typically do this towards the pipeline sections at the top, starting at line 17.
pipeline :beacon do
plug BeaconWeb.Plug
end
Add a BeaconWeb
scope.
scope "/", BeaconWeb do
pipe_through :browser
pipe_through :beacon
live_session :beacon, session: %{"beacon_site" => "my_site"} do
live "/beacon/*path", PageLive, :path
end
end
Comment out existing scope.
# scope "/", BeaconSampleWeb do
# pipe_through :browser
# get "/", PageController, :index
# end
Commit the changes.
Add routing changes
Add some components to your apps/beacon_sample/priv/repo/seeds.exs
.
alias Beacon.Components
alias Beacon.Pages
alias Beacon.Layouts
alias Beacon.Stylesheets
Stylesheets.create_stylesheet!(%{
site: "my_site",
name: "sample_stylesheet",
content: "body {cursor: zoom-in;}"
})
Components.create_component!(%{
site: "my_site",
name: "sample_component",
body: """
<li>
<%= @val %>
</li>
"""
})
%{id: layout_id} =
Layouts.create_layout!(%{
site: "my_site",
title: "Sample Home Page",
meta_tags: %{"foo" => "bar"},
stylesheet_urls: [],
body: """
<header>
Header
</header>
<%= @inner_content %>
<footer>
Page Footer
</footer>
"""
})
%{id: page_id} =
Pages.create_page!(%{
path: "home",
site: "my_site",
layout_id: layout_id,
template: """
<main>
<h2>Some Values:</h2>
<ul>
<%= for val <- @beacon_live_data[:vals] do %>
<%= my_component("sample_component", val: val) %>
<% end %>
</ul>
<.form let={f} for={:greeting} phx-submit="hello">
Name: <%= text_input f, :name %> <%= submit "Hello" %>
</.form>
<%= if assigns[:message], do: assigns.message %>
</main>
"""
})
Pages.create_page!(%{
path: "blog/:blog_slug",
site: "my_site",
layout_id: layout_id,
template: """
<main>
<h2>A blog</h2>
<ul>
<li>Path Params Blog Slug: <%= @beacon_path_params.blog_slug %></li>
<li>Live Data blog_slug_uppercase: <%= @beacon_live_data.blog_slug_uppercase %></li>
</ul>
</main>
"""
})
Pages.create_page_event!(%{
page_id: page_id,
event_name: "hello",
code: """
{:noreply, Phoenix.LiveView.assign(socket, :message, "Hello \#{event_params["greeting"]["name"]}!")}
"""
})
Run ecto.reset
to create and seed our database(s).
cd apps/beacon_sample
.mix ecto.setup
(as our repos haven't been created yet).mix ecto.reset
thereafter.SafeCode
package works as expected.This is typically where we run into issues with safe_code
on the inner content of the layout seed, specifically:
** (RuntimeError) invalid_node:
assigns . :inner_content
<%= @inner_content %>
, seeding seems to complete.Running mix phx.server
throws another error:
** (RuntimeError) invalid_node:
assigns . :val
safe_code
is problematic and needs to be surgically removed from Beacon for now.In Beacon's repository, remove SafeCode.Validator.validate_heex!
function calls from the loaders
lib/beacon/loader/layout_module_loader.ex
lib/beacon/loader/page_module_loader.ex
lib/beacon/loader/component_module_loader.ex
Fix the seeder to work without SafeCode.
apps/beacon_sample/priv/repo/seeds.exs
under Pages.create_page!
from <%= for val <- live_data[:vals] do %>
to <%= for val <- live_data.vals do %>
.Commit the seeder changes.
Add component seeds
Enable Page Management and the Page Management API in router (apps/beacon_sample_web/lib/beacon_sample_web/router.ex
).
require BeaconWeb.PageManagement
require BeaconWeb.PageManagementApi
scope "/page_management", BeaconWeb.PageManagement do
pipe_through :browser
BeaconWeb.PageManagement.routes()
end
scope "/page_management_api", BeaconWeb.PageManagementApi do
pipe_through :api
BeaconWeb.PageManagementApi.routes()
end
Commit the Page Management router changes.
Add Page Management routes
Navigate to http://localhost:4000/beacon/home to view the main CMS page.
Header
, Some Values
, and Page Footer
with a zoom-in cursor over the page.Navigate to http://localhost:4000/beacon/blog/beacon_is_awesome to view the blog post.
Header
, A blog
, and Page Footer
with a zoom-in cursor over the page.Navigate to http://localhost:4000/page_management/pages to view the Page Management
section.
Listing Pages
, Reload Modules
, a list of pages, and New Page
.We should put the page management through its paces to determine weak points.
Add another more robust layout.
<main>
.<body>
section.stylesheet_urls
?Add another more robust component.
0.17.7
.A replica of Laravel Nova panel of pages. Welcome and Home are Laravel defaults. Users would be useful as we could integrate with phx gen auth
.
The dependency safe_code
was a problem during my first two attempts.
I ran into issues by failing to add a BeaconWeb
scope and adding it as BeaconSampleWeb
instead.
UndefinedFunctionError
as function BeaconSampleWeb.PageLive.__live__/0 is undefined (module BeaconSampleWeb.PageLive is not available)
.The sample isn't as "pristine" as I'd like due to the bug fix but it really shouldn't be a showstopper.
<head>
as inline <style>
tags.<body><div data-phx-main="true">
mix phx.server
) immediately boots our Beacon components before it shows the url.