Posts

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/.

  • On my machine this expands to the absolute path /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".

  • On my machine this expands to the absolute path /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").

  • On my machine this expands to the absolute path /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:

• Go to http://localhost:8080/settings
• Or, if you're already in a notebook, click the Livebook icon in the top left and click 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.

Tracing the Default Setting

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.

Finding 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.

Installation

Steps

1. Create a top-level directory to keep our application pair. This is temporary as the project matures.

   1. mkdir beacon_sample

2. Clone GitHub - BeaconCMS/beacon: Beacon CMS to ./beacon.

   1. git clone git@github.com:BeaconCMS/beacon.git

3. Start with our first step from the Readme

   1. Create an umbrella phoenix app
   2. mix phx.new --umbrella --install beacon_sample

4. Go to the umbrella project directory

   1. cd beacon_sample/

5. Initialize git

   1. git init

6. Commit the freshly initialized project

   1. Initial commit of Phoenix v1.6.15 as of the time of this writing.
   2. I prefer to capture the version and everything scaffolded as-is. This allows us to revert back to the pristine state if we ever need to.

7. Add :beacon as a dependency to both apps in your umbrella project

   # Local:
   {:beacon, path: "../../../beacon"},
   # Or from GitHub:
   {:beacon, github: "beaconCMS/beacon"},

   1. Add to apps/beacon_sample/mix.exs and apps/beacon_sample_web/mix.exs under the section defp deps do.
   2. We choose the local version to override commits as needed. When the project solidifies, the GitHub repository will be far more ideal.
   3. I'll want to research the git dependency as I believe we can specify commits? There's possibly no need to have a local revision at all.
8. Run mix deps.get to install the dependencies.

9. Commit the changes.

   1. Add :beacon as a dependency to both apps in your umbrella project seems like a good enough commit message.

10. Configure Beacon Repo

    1. Add the Beacon.Repo under the ecto_repos: section in config/config.exs.

    2. 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

11. Commit the changes.

    1. Configure Beacon Repo subject with Configure the beacon repository in our dev only environment for now. body.

12. Create a BeaconDataSource module that implements Beacon.DataSource.Behaviour

    1. 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

    2. Add that DataSource to your config/config.exs

       config :beacon,
         data_source: BeaconSample.BeaconDataSource

13. Commit the changes.

    1. Configure BeaconDataSource

14. Make router (apps/beacon_sample_web/lib/beacon_sample_web/router.ex) changes to cover Beacon pages.

    1. 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

    2. 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

    3. Comment out existing scope.

       # scope "/", BeaconSampleWeb do
       #   pipe_through :browser
       
       #   get "/", PageController, :index
       # end

15. Commit the changes.

    1. Add routing changes

16. 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"]}!")}
      """
    })

17. Run ecto.reset to create and seed our database(s).

    1. cd apps/beacon_sample.
    2. mix ecto.setup (as our repos haven't been created yet).
    3. mix ecto.reset thereafter.
18. We can skip to Step 22 now that the SafeCode package works as expected.

19. 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

    1. If you remove the line <%= @inner_content %>, seeding seems to complete.

    2. Running mix phx.server throws another error:

       ** (RuntimeError) invalid_node:
       
       assigns . :val

    3. It looks like safe_code is problematic and needs to be surgically removed from Beacon for now.

20. In Beacon's repository, remove SafeCode.Validator.validate_heex! function calls from the loaders

    1. lib/beacon/loader/layout_module_loader.ex
    2. lib/beacon/loader/page_module_loader.ex
    3. lib/beacon/loader/component_module_loader.ex

21. Fix the seeder to work without SafeCode.

    1. Change line 49 in 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 %>.

22. Commit the seeder changes.

    1. Add component seeds

23. 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

24. Commit the Page Management router changes.

    1. Add Page Management routes

25. Navigate to http://localhost:4000/beacon/home to view the main CMS page.

    1. You should see Header, Some Values, and Page Footer with a zoom-in cursor over the page.

26. Navigate to http://localhost:4000/beacon/blog/beacon_is_awesome to view the blog post.

    1. You should see Header, A blog, and Page Footer with a zoom-in cursor over the page.

27. Navigate to http://localhost:4000/page_management/pages to view the Page Management section.

    1. You should see Listing Pages, Reload Modules, a list of pages, and New Page.

Playground

We should put the page management through its paces to determine weak points.

1. Add another more robust layout.

   1. Can we bring in JS frameworks like Vue? My guess is no, the layout looks to start under a <main>.
   2. Inject javascript at the bottom, this should load at the bottom of our <body> section.
   3. Try CDN urls first, then localhost.
2. Add another stylesheet. How do we use stylesheet_urls?

3. Add another more robust component.

   1. Can we use LiveView slots here? We're on 0.17.7.

4. 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.

   1. What migrations are possibly included by Phoenix? Only users?
   2. Add a user profile page.

Notes

• The dependency safe_code was a problem during my first two attempts.

  • The third attempt on 11/6/2022 has no issues so far.

• I ran into issues by failing to add a BeaconWeb scope and adding it as BeaconSampleWeb instead.

  • Navigating to http://localhost:4000/page/home throws an 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.

  • Fixed this as I generated a new repository. There really aren't a ton of steps.
• As of 3/16 page management only covers the page. The layout, component, and stylesheet models are not covered yet.
• Stylesheets are injected into the <head> as inline <style> tags.
• Layout sits under <body><div data-phx-main="true">
• Running the server (mix phx.server) immediately boots our Beacon components before it shows the url.

Originally, I wrote up a post trying to give a 2020 - 2021 overview that got hosed with a local git repo of this blog. I'm using the moment to remind myself that backups are important. It's also important to complete ideas for posts or journals quickly, even if something doesn't feel complete. Letting those linger for days without a git commit that hit the server is a genuine problem and I need to at the very least create and push to a new branch often.

One change that happened at the end of 2020, I started the journal section to try to capture bite-sized rough ideas. I had started a journal at work with notes in files like Phoenix Developer Diary.txt and I looked for a solution to merge my different diaries. The excellent Claire Codes has an extremely consistent diary at clairecodes and served as my main source of inspiration.

I've gone all-in learning Elixir by participating in my first Advent of Code in 2020. I tapered off pretty quickly as I had serious problems working through loops and control flow. Seeing other examples on Elixir Forum helped immensely as I had slowly gotten better at reading the code. Later on in the year, I decided to take a TodoMVC sample through to a LiveView version with a little help from other resources on the internet. I had also started a diary where I wanted to capture the approaches I took each day I worked on the example. I have a plan to try to tackle my version from scratch but I'm also looking at other application ideas.

While the Advent of Code and TodoMVC was good to get my feet wet, I learned far more by pushing through Exercism exercises. If you're on Exercism and curious, my solutions can be found here. I highly recommend using Exercism to learn any language it covers as the recently released version 3 makes for a great experience. Exercises feel a bit more "real world" and less like brain teasers that happen to use programming concepts. Even if I happened to look at the HINTS.md file, it never felt like cheating as it would only guide us toward a solution, not implement it.

After attending the excellent ElixirConf 2021 virtually, I've started working with Livebook in a few examples. I wanted to highlight the 3 notebooks that use the excellent spider_man package to crawl 3 websites: Elixir Jobs, Elixir Radar Jobs, and Elixir Companies. Parsing the DOM of each required slowly stretching far outside my comfort zone. It's also worth mentioning that in the Elixir Jobs example, I left a problem I found under the Sorting the Results section. Due to the zero-width space, the section throws the message ** (SyntaxError) nofile:5:1: unexpected token: "​" (column 1, code point U+200B).

Coming to the end of 2021, I'm looking forward to immersing myself deeper in the Elixir ecosystem. Livebook is also a great way to get your feet wet with Elixir concepts, like a powerful language scratchpad. There have been other life changes since January 2020 but those deserve separate posts when I can get to them. Fortunately, the pandemic hasn't been harsh on my family or extended family at all, which I consider an extreme blessing. I can't say we weren't impacted by the last 2 years but things could've been much worse.

In my last post, I mentioned the transition to Gridsome and it has been relatively pain free. I owe a lot of this to the existing community and the great list of starter resources. If a concept isn't explained or clear in the docs, chances are you can gain some insight from the various starters.

One particular concept I had a problem with right out of the gate was how to use markdown files from multiple directories. I started with the post type to handle /year/month/day/title routes but I wanted to move to an equivalent of the generic page type from Hexo. In doing research to the search terms I could've used months ago, I stumbled on multiple issues that point out how to do it.

In the file gridsome.config.js, I use the following snippet in the plugins section:

{
  use: '@gridsome/source-filesystem',
  options: {
    path: 'blog/articles/**/*.md',
    typeName: 'Article',
    refs: {
      authors: {
        typeName: 'Author',
        create: true
      },
    }
  }
},
{
  use: '@gridsome/source-filesystem',
  options: {
    path: 'blog/posts/**/*.md',
    typeName: 'Post',
    refs: {
      authors: {
        typeName: 'Author',
        create: true
      },
      categories: {
        typeName: 'Category',
        create: true
      },
      tags: {
        typeName: 'Tag',
        create: true
      },
    }
  }
},

Since Gridsome has a concept of pages already, I chose the word article to represent them instead. As an example, the portfolio page is an article type while this page represents a post type. While hindsight makes this seem intuitive now, I somehow had the impression that you were only allowed one plugin type for safety reasons.

To point out something else, the portfolio page highlights a technique I didn't think was possible at the time. The parent portfolio page is an article type but all the subsequent child pages are markdown files in a separate portfolio directory as a portfolio type. In the plugins section of gridsome.config.js:

{
  use: '@gridsome/source-filesystem',
  options: {
    path: 'blog/portfolio/**/*.md',
    typeName: 'Portfolio',
    refs: {
      authors: {
        typeName: 'Author',
        create: true
      },
    }
  }
},

Coming from Hexo, I opt for placing content in markdown files and having unique layouts defined in the various pages and templates files. As much as Gridsome is a generic website framework, I find that it can be extremely flexible to whatever workflow you wish to create. There are some parts of Hexo I miss like scaffolding new page types or steering me into blog concepts but the transition to Gridsome has been rather smooth. While Gridsome may not be for everyone, I can definitely see how JAMstack has gained traction recently. Barring very few gotchas, working on this site is fun again even in the I-can-see-every-blemish state it's currently in.

It's been over a year since my last post and while unfinished drafts don't count, I thought my blog was due for a change. The move from Octopress to Hexo was relatively uneventful but I found keeping up a little difficult. It wasn't completely on Hexo, I had tweaked things to a point that merging in changes over time became cumbersome and slow. In a previous post, I roughly mentioned the transition and a lot has happened to the web in over 3 years.

Static site generators like Hugo and Gatsby have picked up steam and the feature set of Gatsby, particularly the GraphQL component stood out. I wanted to stick to Vue for as many of my personal projects as possible, so I searched for any static site generator using Vue I could find. Fortunately Gridsome has come along as a nice clone of Gatsby using Vue rather than React and even though it's at v0.7.12 at the time of this post, I've run into very few hurdles.

I don't have the best understanding of JAMstack after working with a sample size of one, but learning GraphQL by only dealing with queries has made this one of the best ways to get my feet wet. I'm by no means an expert but this light interaction compels me to use it more often, as it's mostly been a pleasure to work with. Frameworks like Gridsome and I suspect Gatsby let you focus on almost entirely the frontend. Even though the A in JAMstack stands for APIs, as a backend developer I haven't had to write a single REST, GraphQL or what I'd typically associate with an API like I would with Laravel, Phoenix Framework, or Express.

One thing I miss about Hexo is that it had scaffolding to generate new files. Gridsome is a framework for generic sites, not just blogs, so scaffolding doesn't seem to be included. Coming from Hexo I wanted to keep as much of the existing markdown as possible and I think some of the approaches I've taken may be useful to others. A small example I had a problem understanding is that you can use a @gridsome/source-filesystem plugin multiple times, one for each directory or type. It makes sense in hindsight but none of the starters used the technique nor did the docs seem to suggest it was possible. I'm tempted to create a starter based on my usage patterns but worst case, I plan on writing a post outlining some of these approaches in the near future.

One last thing is a small humblebrag. While the theme for this site draws a few cues from the older version, I wanted to flex my design abilities by focusing on techniques I've learned reading Refactoring UI. By the time this post is published it likely won't be perfect but I think it's a decent first pass that should only get better over time.