advanced_config.md

Advanced Config Options

Scenario:

We are building a Phoenix app which shares a database with a legacy Rails appliction. At present, the Rails application manages database migrations. This leaves us several challenges:

1. We need to have a Test DB to run our Phoenix app's tests.
2. Rails uses created_at instead of inserted_at.
3. Our legacy database uses UUID's (Binary ID's) for the primary keys.

Solutions:

Test DB

In our text.exs file, when we configure the db's, add the priv key to both the main and shard db's and begin the paths in each with test. So the main db gets priv: "test" and the shard db's get priv: "test/shards". Now when we run MIX_ENV=test mix ecto.create it will create the main test db and MIX_ENV=test mix shards.create it creates the test shard db's. Now we build migrations using Ecto, specifying the test environment and Ecto will place the migration files in our test directory instead of priv. We create Ecto migrations to build the structure of the existing databases.

DB Field Names

This is relatively straight-forward since we already have a custom schema file for our app:

defmodule MyApp.Schema do
  @moduledoc false

  defmacro __using__(opts \\ []) do
    sharded = Keyword.get(opts, :sharded, false)
    quote do
      use EctoSharding.Schema, sharded: unquote(sharded)
      # set primary keys to UUIDs
      @primary_key {:id, :binary_id, autogenerate: true}
      @foreign_key_type :binary_id
      # use Rails-compatible timestamps
      @timestamps_opts [ inserted_at: :created_at, usecs: true]
    end
  end
end

Since all of our schemas already use this schema as a base, we have one central place to add customizations to the common fields as needed. We've also defaulted sharded to false so that we only need to specify that in the sharded schema. Those for the main DB will not need to include the schema option (this makes sharding a little less obtrusive)