DEV Community: Mark

The .iex.exs file

Mark — Thu, 06 Feb 2020 08:03:11 +0000

The .iex.exs file is a "must have" convenience.

You can use it to set aliases, set imports and add define helper functions for every IEx session you run!

Here's what's in my global one:

import_if_available(Ecto.Query)
import_if_available(Ecto.Changeset)

defmodule AC do
  def update(schema, changes) do
    schema
    |> Ecto.Changeset.change(changes)
    |> Repo.update()
  end

  IEx.configure(
    colors: [
      syntax_colors: [
        number: :light_yellow,
        atom: :light_cyan,
        string: :light_black,
        boolean: [:light_blue],
        nil: [:magenta, :bright]
      ],
      ls_directory: :cyan,
      ls_device: :yellow,
      doc_code: :green,
      doc_inline_code: :magenta,
      doc_headings: [:cyan, :underline],
      doc_title: [:cyan, :bright, :underline]
    ],
    default_prompt:
      [
        # ANSI CHA, move cursor to column 1
        "\e[G",
        :light_magenta,
        # plain string
        "🧪 iex",
        ">",
        :white,
        :reset
      ]
      |> IO.ANSI.format()
      |> IO.chardata_to_string()
  )
end

And in this project, my local .iex.exs has:

alias Campsite.{Accounts, Content, Extract, Game, Repo}
alias Accounts.{User, Credential}
alias Content.{Article, Comment, Episode, Podcast}
alias Game.Request

With these in place, every time IEx starts up, commonly used modules are automatically aliased, Ecto queries are easier and there's a nice 🧪 prompt making it clear the terminal is in IEx. 🎉

Request a free email-based Elixir course from Alchemist.Camp

Querying nested JSONB fields with Ecto

Mark — Mon, 30 Dec 2019 10:37:54 +0000

Ecto has a map type, which is implemented in Postgres as a JSONB field. It's particularly useful for situations involving user-submitted data, where we don't know the exact shape the data will come in.

Consider this schema for user-submitted reviews:

defmodule MyApp.Repo.Migrations.CreateReview do
  use Ecto.Migration

  def change do
    create table(:reviews) do
      add :submitter_id, references(:users)
      add :item_id, references(:items)
      add :rating, :integer, default: 0, null: false
      add :metadata, :map

      timestamps()
    end

    create index(:reviews, [:submitter_id])
    create index(:reviews, [:item_id])
    create(unique_index(:reviews, [:submitter_id, :item_id]))

    # How do we create an index a JSONB sub-field?
  end
end

Querying into JSONB fields with raw SQL

Over time, we might come to find that some reviews include an item field in their metadata. These can be queried in the DB with:

select * from metadata
where metadata->'item' is not null;

Digging in further we could examine only reviews with an an item field in the metadata and another nested field price. To find all reviews with such a structure and a price of 35, we can do the following:

select * from metadata
where metadata @> '{"item": {"price": 35}}';

select * from metadata
where metadata->'item'->>'price' = '35'

Note: The difference between -> and ->> is that -> returns a JSON object and ->> returns the value as text!

Ecto Fragments

Ecto provides a handy escape hatch to use SQL for more specialized queries such as these. Assuming Ecto.Query has been imported, you can wrap the SQL you need in a fragment call. The following returns all the items within review metadata fields that have a price:

Repo.all(
  from r in "reviews",
  where: fragment("metadata->'item'->'price' is not null"),
  select: fragment("metadata->'item'")
)

Note: When using the @> JSON syntax, it's often convenient to use the ~s sigil with either | as delimiters and avoid a lot of backslash escaping. For example, using the JSON syntax to get a count of items with price 35 like this:

Repo.one(
  from e in "events",
  where: fragment(~s|metadata @> '{"item":{"price": 35}}'|),
  select: count(e.id)
)

Creating indexes on JSONB sub fields

In a large table where a particularly embedded field is important enough to query, such as price in this case, it's a good idea to add an index.

CREATE INDEX ON reviews((metadata->'item'->>'price'));

Inside Ecto migrations, we can accomplish this by wrapping the command in an execute(). Returning to the question at the top of this page, we can simply add the execute to the bottom of the change function in our migration:

create index(:reviews, [:submitter_id])
create index(:reviews, [:item_id])
create(unique_index(:reviews, [:submitter_id, :item_id]))

# How do we create an index on a JSONB sub-field?
execute("CREATE INDEX ON reviews((metadata->'item'->>'price'));")

Request a free email-based Elixir course from Alchemist.Camp

Elixir Structs are maps with checks and default values

Mark — Sat, 28 Dec 2019 11:10:10 +0000

After our crash course, Elixir maps made effortless, the next logical building block is structs.

What are Elixir Structs?

As explained in the official docs, Structs are extensions built on top of maps that provide compile-time checks and default values. Maps, as we covered, are one of Elixir's basic data structures. They're the primary way we store key-value pairs:

> users = %{
  "sam" => %{age: 22},
  "pat" => %{age: 58}
}

> user1 = Map.get(users, "sam")
%{age: 22}

> Map.get(user1, :age)
22

# Or in one step...
> get_in(users, ["pat", :age])
58

Aside from getting values from maps via standard library functions like Map.get/3 or Kernel.get_in/2, there are also two built-in syntaxes:

The dot syntax, which throws errors on missing keys

> user1.weight
** (KeyError) key :weight not found in: %{age: 22}

The [] syntax, which is forgiving:

> users[:nobody]
nil

Note that the dot syntax assumes keys are atoms and if keys are Strings as above, only the bracket syntax can be used. For more on working with maps, see: Elixir maps made effortless

Structs are built on top of maps

Let's try pasting a simple module with a defstruct into iex and then poking around at it:

defmodule User do
  defstruct age: :nil, name: "anonymous"
end

> user1 = %User{age: 32}
%User{age: 32, name: "anonymous"}

> user2 = %User{}
%User{age: nil, name: "anonymous"}

Using a struct makes it possible to define default values. It also lets us be confident that any User we create will at least have the two keys of age and name. It also makes it possible for us to match various kinds of structs in a case statement, like this:

case user1 do
  %User{} -> IO.puts("This is a user")
  %Admin{} -> IO.puts("This is an admin")
  _ -> IO.puts("I don't know what this is")
end

This will do a compile-time check to ensure both the structs are defined and then it will check the hidden __struct__ key of user1 to see which kind of struct it is. If the Admin struct isn't defined, then you'll see this:

Admin.__struct__/0 is undefined, cannot expand struct Admin

Enforcing keys

We can go a step further. By using @enforce_keys at the top of a struct module, we can enforce that a specific set of keys are used when creating a struct

defmodule User do
  @enforce_keys [:age, :name]
  defstruct age: :nil, name: "anonymous", favorite_color: "purple"
end

Then, we'll see the following in iex:

> bob = %User{}
(ArgumentError) the following keys must also be given when building struct User: [:age, :name]
> bob = %User{name: "Bob", age: 48, favorite_food: "steak"}
** (KeyError) key :favorite_food not found
> bob = %User{name: "Bob", age: "48"}
%User{age: 48, favorite_color: "purple", name: "Bob"}

Now, users must have names and ages, but favorite colors are optional. Any other key is invalid. Again, this is enforced at compile-time, so it is possible to patch together a struct that violates the specified behavior by directly setting the __struct__ field rather than using the %User{} syntax. This isn't a good idea to abuse, but it can be useful in some situations when building libraries.

Here's how we could create a "user" who's missing a required key and includes a key that isn't part of the %User{} struct:

> evil_bob = %{__struct__: User, name: "Bob", favorite_food: "steak"}
> case evil_bob do  
>   %User{} -> IO.puts("This is a user")
>   _ -> IO.puts("This isn't a user")
> end
This is a user
:ok

😱

A hands-on example

If you want to dig into a simple, plain Elixir project using Structs, take a look at the Tic-tac-toe game board screencast!

Request a free email-based Elixir course from Alchemist.Camp

Setting up PrismJS with Phoenix

Mark — Mon, 16 Dec 2019 09:47:02 +0000

PrismJS is a great option for code syntax highlighting. It's used by Stripe, along with many other companies that deeply value documentation.

It supports a vast array of plugins and the total script weighs in at just 2k, gziped. Extra programming languages add 0.3-0.5KB each and themes are around 1KB. All in all, it's a lot of functionality for very little cost to your page load times!

This is a quick guide to get it set up on Phoenix (or just about any back-end framework that takes an agnostic approach towards JS). The code shown in this post, as well as on Alchemist Camp uses this setup.

This guide starts from a fresh Phoenix app. It should work with just about any standard app.

Install Prism

First, cd into your assets directory and then install the following two libraries. If you're using yarn:

yarn add prismjs
yarn add babel-plugin-prismjs

Or, if you're using npm:

npm install prismjs
npm install babel-plugin-prismjs

Next, to include it in your front-end, add the following to the top of your main app.js file:
import Prism from "prismjs"

Adding support for more languages

Prism supports syntax highlighting for well over a hundred languages, but it doesn't load them all by default. We can configure which languages load and pick a theme by creating a .babelrc file in our assets directory. Here's an example file:

{
  "presets": ["@babel/preset-env"],
  "plugins": [
    [
      "prismjs",
      {
        "languages": [
          "elixir",
          "erlang",
          "css",
          "javascript",
          "markdown",
          "markup",
          "ruby",
          "scss",
          "toml"
        ],
        // themes can be found in this directory:
        // assets/node_modules/prismjs/themes
        "theme": "okaidia",
        "css": true
      }
    ]
  ]
}

Customizing the theme

I wanted the styling from the okaidia theme for blocks of pre-formated code, but wanted my own styles to control inline code. So I commented out the theme line above, copied the theme file from its directory into the /assets/css directory where Phoenix puts stylesheets. Then removed the styles I didn't want and re-wrote it to match my specific needs. All the Prism themes are is CSS, so you can adjust them without concern about breaking anything else.

How Prism works

Prism will add syntax highlighting to inline code tags and to pre blocks with code inside them. The language of highlighting applied is determined by the class of the element. If the class is language-javascript, the highlighting will be for JavaScript, a class of language-elixir will apply Elixir highlighting, and so on. For the full list, refer to Prism's supported language chart.

You can also set a default language by setting a class on a parent element, such as the body tag.

Add code to a page

Phoenix apps come with an index.html.eex template for the / page. Try replacing yours to the following:

<% code = ~s"""
  defmodule MyModule do
    #stuff
  end
  """%>
<section class="row">
  <article class="column">
    <code class="language-elixir">Enum.at([:a, :b, :c], 1) |> IO.puts()</code>
    <pre><code class="language-elixir"><%= code %></code></pre>
  </article>
</section>

(The top line is just saving our block of code in a variable code, using a HEREDOC string so we don't have to worry about new lines or indentation inside the pre block!)

Here's the result:

Integrating with Earmark

If you're using Earmark to convert Markdown to HTML (as this site does), then you'll need to pass in a code_class_prefix option to generate the appropriate classes on your code blocks. E.g:

  Earmark.as_html!(
    %Earmark.Options{
      code_class_prefix: "lang-",
      smartypants: false
    }
  )

Then you can declare the language of your code blocks in Markdown like so:



```ruby
  puts 8 ** 3
```

Originally published on Alchemist Camp

Running a Gatsby blog inside a Phoenix app

Mark — Mon, 07 Oct 2019 02:35:15 +0000

Let's see what it takes to set up a Gatsby blog inside a Phoenix app!

Like Jekyll, VuePress and other popular static site generators, Gatsby is a easy way to create a snappy blog. Since it generates plain HTML, CSS and JS (aka "a static site"), all that's really needed is to drop the files in a directory that serves static contents.

By default, Phoenix includes a directory for serving static assets at /assets/static. Any files and subdirectories named in the only: key of Plug.Static inside your endpoint.ex file will be loaded from this static assets directory instead of being delegated to your router.

Initial Setup

Make sure you've installed Node, Elixir and Phoenix
Install the Gatsby CLI: npm install -g gatsby-cli
Create a new Phoenix app: mix phx.new my_app
Go into the newly created directory: cd my_app
Generate the DB: mix ecto.setup
Go into the assets directory (of your new app): cd assets
Create a Gatsby site inside assets: gatsby new gatsby https://github.com/gatsbyjs/gatsby-starter-hello-world

Verify the installations were successful

At this point, you should be able to start your Phoenix app by running mix phx.server from the base of your project, the my_app directory and opening localhost:4000 in your web browser.

The static site generated by Gatsby can be run in development mode by running gatsby develop from my_app/assets/gatsby and then opening localhost:8000. It's just a "Hello world!" for now.

We named the site "gatsby" in the previous section so that it's clear, from the context of the enclosing Phoenix project, that this is a Gatsby site.

Generate the static site

In order to run the static site in production, we'll want to build a production bundle. The command to do it is gatsby build. By default this will build the site under a directory called "public", which in our case will be at my_app/assets/gatsby/public.

This isn't ideal.

We need to serve the output of the Gatsby build from our Phoenix static assets directory, but don't want to serve our Gatsby project's src or other directories publicly. There is an --output-dir flag that can be passed to the gatsby build command, but at the time of this writing, it's not possible to pass an output directory outside the Gatsby project, only the name of the directory is configurable.

Create a symlink

One way to get around this issue is to create a symlink from Gatsby's output directory to the static assets directory. First cd into the /my_app/assets/static directory, and then...

On Unix or Mac:

ln -s ../gatsby/public blog

On Windows:

mklink ..\gatsby\public blog

Now, the entire contents of whatever is generated in Gatsby's public output directory will be visible as a "blog" directory under the Phoenix static assets directory.

Enable the blog directory

The final step is to open up endpoint.ex and update the Plug.Static portion near the top to allow it to handle the newly created blog directory (which is a symbolic link of the Gatsby public directory). It should look like this:

defmodule MyAppWeb.Endpoint do
  use Phoenix.Endpoint, otp_app: :demo

  socket "/socket", MyAppWeb.UserSocket,
    websocket: true,
    longpoll: false

  plug Plug.Static.IndexHtml, at: "/blog"

  plug Plug.Static,
    at: "/",
    from: :demo,
    gzip: false,
    only: ~w(blog css fonts images js favicon.ico robots.txt)

  # The rest of the file

After saving these changes, and restarting the Phoenix app, the generated Gatsby site is served from within the Phoenix app at localhost:4000/blog/index.html.

Possible next steps

Now that everything is up and working, You could just build out the web app and static blog. For many apps and sites, no further integration is necessary.

Here are some ideas, though...

See if you can write a simple plug that will read the path from conn.request_path and redirect /blog/ to /blog.index.html. Or alternatively, you could look at a library plug_static_index_html and see if you can use it to handle this for you.
Try installing Absinthe GraphQL and serving your static Gatsby blog GraphQL from the Phoenix back-end. This is a fairly large endeavor, but it's a great learning experience!

Request a free email-based Elixir course from Alchemist.Camp

Elixir lists vs Elixir tuples

Mark — Fri, 12 Jul 2019 03:07:16 +0000

Along with maps, Lists and Tuples are both ubiquitous and somewhat surprising for Elixir learners. Here's a quick guide.

Lists are linked lists

Elixir lists look just like arrays from C, Java or JavaScript, but they're not. They're not laid out contiguously in memory. They're a series of nodes where each node holds a piece of data and a pointer to the next node. The final node points to nothing.

This has performance implications. In order to read the final element in a list, you must traverse every element from the head to the tail of the list! Prepending an element to the front of the list is cheap, O(1), but accessing an element at the end is O(n)

Tuples are more like Arrays

Elixir tuples are stored contiguously in memory. Accessing any element is O(1). On the down-side, inserting or deleting an element involves copying the entire tuple, which is O(n). Remember, like everything else in the language, Tuples are immutable!

List elements are usually of the same type

This is more of a convention than a rule. Since lists implement the Enumerable protocol and it's very common to use functions like Enum.map, Enum.reduce and Enum.each that pass each element into a function, life is easier if everything in the list has the same shape.

Tuples, on the other hand are very often mixed types. They're commonly returned from functions and pattern matched against and in this case the first element is usually a status. For example:

case retrieve_stuff do
  {:ok, stuff} -> process(stuff)
  {:err, msg} -> Logger.error("Error with message: %{msg}")
end

Converting lists and tuples is easy

You can convert a list to a tuple with List.to_tuple(some_list) and you can convert a tuple to a list with Tuple.to_list(some_tuple). For example:

my_list = [1, 2, 3]
my_tuple = {:a, :b, :c}

other_tuple = List.to_tuple(my_list)   # equals {1, 2, 3}
other_list = Tuple.to_list(my_tuple)   # equals [:a, :b, :c]

Common questions

How do I get the nth element of a list?

Use Enum.at/2. For example, Enum.at([1, 2, 3, 4], 2) will return 3 since lists are zero-indexed.

How do I get the nth element of a tuple?

Use Kernel.elem/2. For example, elem({:a, :b, :c}, 0) will return :a since tuples are also zero-indexed.

How do I insert an element into a list?

Use List.insert_at/3. For example, List.insert_at([:foo, :bar, :baz], 1, :yolo) returns [:foo, :yolo, :bar, :baz].

You can delete elements in the same way with List.delete_at/2.

If you're prepending to the beginning of a list (which is very performant), then you can also use a cons syntax and insert an element like this:

my_list = [1, 2, 3]
[:hello | my_list]  # returns [:hello, 1, 2, 3]

How do I insert an element into a tuple?

Use Tuple.insert_at/3. For example, Tuple.insert_at({:ok, :stuff}, 1, :yolo) returns {:ok, :yolo, :stuff}.

You can delete elements in the same way with Tuple.delete_at/2.

How do I get the length of a list or a tuple?

Use Kernel.tuple_size/1 for tuples and use Enum.count/1 for lists.

What's a good reference if I need to know more?

It's best to go straight to the official docs!

Request a free email-based Elixir course from Alchemist.Camp

Elixir maps made effortless

Mark — Tue, 09 Jul 2019 07:56:06 +0000

Getting used to working with Elixir maps can be one of the most painful aspects of really getting comfortable with the language. If you're coming from a language like Java or Ruby, the fact that everything is immutable can be frustrating to deal with. If you're coming from JavaScript, you'll have that problem and be spoiled by having native maps (Objects, in JS speak) fit perfectly with JSON.

The good news is, there are really only a couple of points that trip people up!

Variables are immutable in Elixir

// JavaScript
a = {foo: 42}
b = a
b.foo   // equals 42
b.foo = 5
a   // equals {foo: 5}

# Elixir
a = %{foo: 42}
b = a
b.foo   # equals 42
b.foo = 5   # b and b.foo are immutable so we get an error
# ** (CompileError) iex:5: cannot invoke remote function b.foo/0 inside a match

b = %{b| foo: 5}
b   # now b is reassigned and bound to %{foo: 5}
a.foo   # still equals 42

All "updating" of maps involves reassigning a variable to a new map

Here are a few common ways:

put adds a new value to a map: a = Map.put(a, bar: 5) Now a is %{foo: 42, bar: 5}
delete removes a value from a map: a = Map.delete(a, :foo) Now a is %{bar: 5}
put_new works like put, but does nothing if the key already exists: a = Map.put_new(a, :bar, 10) Doesn't replace the existing key, so a is still %{bar: 5}
merge adds/updates multiple values into a map: a = Map.merge(a, %{foo: "stuff", baz: -5}) Now a is %{foo: "stuff", bar: 5, baz: -5}

Note that since Elixir variables are immutable, the map functions above created new maps instead of changing a itself. Without reassigning a to the new maps with the a = at the front of each of the examples above, a would be unchanged.

Map keys can be Strings or Atoms (or anything!)

In Elixir, you'll run into two forms of map keys. Some, like the examples above, are atoms and some are strings. Very rarely, you may find integers, floats or even other types, including nested maps used as keys of maps!

The following are all valid maps:

a = %{:some_atom => :foo}
i = %{1 => 52}
f = %{1.5 => i} - the value of f is now: %{1.5 => %{1 => 52}}
m = %{f => "wat???"} - the value of m is now: %{%{1.5 => %{1 => 52}} => "wat???"}

Strings are the same thing as Erlang binaries.

They're represented with quotes. "foo"
Longer strings or strings with quotation marks in them can be made with sigils ~s{I'm a string made from a sigil and can have "quoted" parts}
In a key-value form, string keys use an arrow syntax. %{"foo" => "stuff", "bar" => 5}
When being used to access values, they use a bracket syntax. a["foo"] is "stuff" and b["foo"] is 5

Atoms are unique symbols. They don't get garbage collected so don't generate them dynamically.

When represented alone, they start with a colon. :foo
In a key-value form, they can use the standard arrow syntax or simply have a trailing colon. %{foo: "stuff, bar: 5} is the same as %{:foo => "stuff", :bar => 5}
When being used to access values, they can use a dot syntax. a.foo is the same as a[:foo]

Working with deeply nested maps

The above techniques are enough to do anything with Elixir maps, however immutability makes working with deeply nested maps a bit cumbersome. Intermediate steps would be required to build up the exact structure you want to reassign a given variable to.

Of course you're always free to write your own helpers, but for the 90% case, the built-in get_in, put_in and related functions will do the job. They're part of Kernel, not Map, because they also operate on other types of data, so they don't have a Map. prefix.

users = %{
  "sam" => %{age: 22},
  "pat" => %{age: 58}
}

get_in(users, ["sam", :age])
# returns 22

put_in(users["pat"][:age], 28)
# returns %{"sam" => "{age: 22}, "pat" => %{age: 28}}

Request a free email-based Elixir course from Alchemist.Camp

ASDF, the version manager for all your languages

Mark — Mon, 13 May 2019 16:54:19 +0000

Why use ASDF?

ASDF is a version manager for programming languages. It's somewhat like RVM is for Ruby or NVM is for Node but it also supports Erlang, Elixir, Haskel, Ocaml, PHP, Python, Rust and many other languages.

Prereqs

This guide assumes you have homebrew and Xcode command line tools and nothing else. To see the setup of those from a fresh macOS Mojave install, see this short video.

Clone the repo

The ASDF repo page on will have directions for cloning the newest repo github: https://github.com/asdf-vm/asdf

git clone https://github.com/asdf-vm/asdf.git ~/.asdf --branch v0.7.1

Next, add it to your shell:

echo -e '\n. $HOME/.asdf/asdf.sh' >> ~/.bashrc
echo -e '\n. $HOME/.asdf/completions/asdf.bash' >> ~/.bashrc

Restart your shell (or just open a new tab to work from) and run asdf to verify that it's installed.

Install the plugin dependencies

brew install \
  coreutils automake autoconf openssl \
  libyaml readline libxslt libtool unixodbc \
  unzip curl gpg wxmac

Basic commands

asdf plugin-list-all: shows all the plugins (i.e., languages) available
asdf plugin-add <language>: installs language
asdf list-all <language>: shows all available versions of language
asdf list <language>: shows all installed versions of language
asdf install <language> <version>:
asdf current: shows currently enabled languages
asdf global <language> <version>: enables the chosen version of a language

## Installing Node
Node requires an extra step:

bash ~/.asdf/plugins/nodejs/bin/import-release-team-keyring
asdf plugin-add nodejs
asdf list-all nodejs
asdf install nodejs <version>
asdf global nodejs <version>

## Installing Erlang

asdf plugin-add erlang
asdf list-all erlang

According the readme on the ASDF Erlang plugin repo we need to pass a couple of flags to the install command:

export KERL_CONFIGURE_OPTIONS="--disable-debug --without-javac"
asdf install erlang <version>
asdf global erlang <version>

Installing Elixir

asdf plugin-add elixir
asdf list-all elixir
asdf install elixir <version>
asdf global elixir <version>

Installing Ruby

asdf plugin-add ruby
asdf list-all ruby
asdf install ruby <version>
asdf global ruby <version>

Verify you have what you need

asdf current

Installing the Phoenix Framework

If you're reading this on Alchemist Camp, you're likely using ASDF for Elixir and also want to set up Phoenix:

mix local.hex
mix archive.install hex phx_new 1.4.5

Done!

ASDF is a very handy tool for setting up dev machines and keeping the versions of whichever languages you may need. It's a big improvement to have one unified tool over several language specific ones.

ASDF also supports per-project configuration via .tool-versions files and a number of other things not covered in this setup guide.

Originally published at Alchemist Camp

Deploying a Phoenix app to Heroku

Mark — Sat, 11 May 2019 12:56:44 +0000

In this quick guide, we'll build a new Phoenix app, set up Heroku and deploy! Heroku may not be the ideal choice for an Elixir app but Elixir and Phoenix is an ideal stack to run on Heroku!

If you haven't already installed and set up Elixir on your computer, go through the setup part of lesson 1 and then you'll be ready to follow along here. These are the steps we'll go over:

Sign up for Heroku (here)
Install the Heroku Toolbelt (download here)
Create a new Phoenix app
Create a new app for it on Heroku
Add the Phoenix static buildpack
Create a free tier database on Heroku
Change some settings in our project
Make a Procfile and elixir_buildpack.config for Heroku
Push everything to Heroku to deploy!

Quick start

We'll call our app Hello, so here's what we need to run:

mix phx.new hello
cd hello
git init
heroku create --buildpack "https://github.com/HashNuke/heroku-buildpack-elixir.git"
heroku buildpacks:add https://github.com/gjaldon/heroku-buildpack-phoenix-static.git
heroku addons:create heroku-postgresql:hobby-dev

Application files to change

Heroku stores secrets in system environment variables instead of a secrets file, so we can delete prod.secrets.exs. Then, we can update prod.exs to get the secrets we need from System.get_env. While we're there we can also set the app to use SSL and update the host name to whatever Heroku named our app when we created it. Finally add the database settings to the bottom of the file:

use Mix.Config

config :hello, HelloWeb.Endpoint,
  load_from_system_env: true,
  url: [scheme: "https", host: "NAME_OF_HEROKU_APP.herokuapp.com", port: 443],
  force_ssl: [rewrite_on: [:x_forwarded_proto]],
  cache_static_manifest: "priv/static/cache_manifest.json",
  secret_key_base: Map.fetch!(System.get_env(), "SECRET_KEY_BASE")

# Do not print debug messages in production
config :logger, level: :info

# Configure your database
config :hello, Hello.Repo,
  adapter: Ecto.Adapters.Postgres,
  pool_size: String.to_integer(System.get_env("POOL_SIZE") || "10"),
  ssl: true,
  url: System.get_env("DATABASE_URL")

If your app is going to use channels, then update your user_socket.ex file to add a timeout for sockets. Heroku limits them to 55 seconds, so set the time out to something like 45 seconds to be safe:

transport(:websocket, Phoenix.Transports.WebSocket, timeout: 45_000)

Application settings for Heroku

By default Phoenix uses a pool size of 10. This is very low and ideally we want to use nearly as much as is available. On Heroku we have a maximum pool of 20, so we'll set our application to use 18 and just use 2 for mix tasks. We'll also generate a secret key with Phoenix and set that in Heroku so that the app can get it from the system environment in production:

heroku config:set POOL_SIZE=18
heroku run "POOL_SIZE=2 mix hello.task"
mix phx.gen.secret
heroku config:set SECRET_KEY_BASE="<SECRET PHOENIX JUST GENERATED ABOVE>"

New files to add

In the root directory of our app, we'll add two more files for Heroku—one called Procfile that has just one line:

web: MIX_ENV=prod mix phx.server

and one called elixir_buildpack.config that holds some configuration settings:

erlang_version=20.1
elixir_version=1.7.2
always_rebuild=false
runtime_path=/app

Important: Make sure that both files have LF line endings. Heroku throws errors on CRLF!

Deploy!

git add .
git commit -m "prepares for heroku"
git push heroku master

Originally published at Alchemist Camp

TypeScript libraries with incorrectly typed members

Mark — Sat, 27 Apr 2019 06:44:33 +0000

I recently covered an easy way to silence TypeScript when dealing with 3rd party libries with missing or incorrect types.

An additional question that post brought up was:

How do I get fix TypeScript "namespace has no exported member" errors?

The problem

Let's say you're using the popular JavaScript game framework Phaser, which has multiple complex deeply nested objects. The types provided are virtually always out of date with the library itself since Phaser is complex, quickly moving and mostly used by people writing JS rather than TS.

With an existing TypeScript project, if you've been updating your types as you go, there's no issue. But when you migrate a project from JS to TS, you'll have a lot of errors to work through. Even a new project on the newest version of the framework can be an editor full of red squiggly lines and type errors.

Worse still, parts your code must refer to members of Phaser. Even if you write declare const Phaser: any; at the top of your app to silence all TS errors from importing or using Phaser, you've still got two problems. First, it's overkill and you probably do want to get your app typed, including the Phaser portions at some point. Second, you're probably going to make your own class which has a member of type Phaser.Game. Since Phaser.Game is also very complex and generally ships with incomplete or out of date types, you'll end up with with the "namespace has no exported member error".

The solution

Work from the outside of your 3rd party library in. Disable warnings on member types by exporting their types in the 3rd party library's namespace.

namespace Phaser {
  export type Game = any;
}

export class MyGame {
    private actors: Array<MyActorType>; 
    private actor: <MyActorType>;
    protected game: Phaser.Game;  // TS errors are silenced here now!
    protected manageAssets(): void { };
    // ... the rest of the MyGame class
}

Now, you'll no longer get the error about Phaser.Game. In other modules, where you use members of Game, you can dig deeper with the same process if there are further nested type issues. Eventually you'll get down to simple objects that are trivial to correctly type.

Fixing these issues as you go, you'll get to a point where all the parts in of the 3rd party library that are used in your app will be correctly typed and the others will be silenced. This is a reasonable trade-off with a large 3rd party library in flux, such as Phaser or Materialize.

Subscribe to more content from logicmason.com

setTimeout vs setImmediate vs process.nextTick

Mark — Mon, 22 Apr 2019 03:23:45 +0000

What's the difference between setTimeout(callback, 0) and process.nextTick(callback)? How about Node's setImmediate(callback)?

On the surface it appears that all three functions do the same thing—they execute the callback after the current event loop, but before anything else. The natural question to ask is, why are there three different functions? Let's run an experiment:

let racer = function() {
  setTimeout(() => console.log("timeout"), 0);
  setImmediate(() => console.log("immediate"));
  process.nextTick(() => console.log("nextTick"));
  console.log("current event loop");
}

racer()

We can see from the output that these callbacks aren't executed in the same order they were written in the source code.

[Running] node "/Users/logicmason/timeouts.js"
current event loop
nextTick
timeout
immediate

[Done] exited with code=0 in 0.203 seconds

Explanation

The first one executed was process.nextTick, which puts its callback at the front of the event queue. It will execute after the code currently being executed but before any I/O events or timers.

Next is "timeout". Since we passed setTimeout a timeout of 0, there's no additional enforced delay before its execution, and it is placed on into the timer queue during the next loop.

Finally, we have setImmediate, which is clearly not as immediate as its name suggests! Its callback is placed in the check queue of the next cycle of the event loop. Since the check queue occurs later than the timer queue, setImmediate will be slower than setTimeout 0.

All in all, the event loop looks like this:

timers -> IO -> poll -> check ->close -> timers -> ...

Timers: callbacks from setInterval or setTimeout
IO callbacks: callbacks from I/O events
Idle: used internally by Node between IO and Poll phases
Poll: retrieve new I/O events
Check: callbacks from setImmediate execute here
Close: handle closed connections like sockets

Challenge time!

What do you expect the output of the following code in Node to be?

let racer1 = function() {
  setTimeout(() => console.log("timeout"), 0);
  setImmediate(() => console.log("immediate"));
  process.nextTick(() => console.log("nextTick"));
}

let racer2 = function() {
  process.nextTick(() => console.log("nextTick"));
  setTimeout(() => console.log("timeout"), 0);
  setImmediate(() => console.log("immediate"));
}

let racer3 = function() {
  setImmediate(() => console.log("immediate"));
  process.nextTick(() => console.log("nextTick"));
  setTimeout(() => console.log("timeout"), 0);
}

racer1()
racer2()
racer3()

Was it what you expected?

Subscribe to more content from logicmason.com

How to make TypeScript shut up about that library

Mark — Fri, 19 Apr 2019 15:56:06 +0000

April 14, 2019

TypeScript is a tradeoff.

“I use typescript when I feel like having hours of migraines using 3rd party libraries and making sure I don’t have a single space or comma out of place! Works like a charm! Get a headache 10/10 times.”

-somebody on Reddit

TypeScript can do wonders for your code editor’s ability to help you with auto-completions, it can enforce contracts between different parts of your code base in the way documentation never could, and it can help a large team move faster.

It can also drive you to the point of madness when dealing with its increasingly pedantic compiler, especially when dealing with third party libraries that have incomplete, incorrect or conflicting types.

In general, it’s good to type as much as possible so you get the benefits TS offers. But how do you silence compiler warnings when it’s not your code and the upstream library just isn’t typed correctly?

You can add this to your codebase:

declare const annoyingLibrary: any;

And then the TypeScript compiler will stop complaining about annoyingLibrary types.

Originally published at logicmason.com on April 14, 2019.