Condo

Condo is a multi-tenant adapter for Ecto. You can use this to create new schemata in PostgreSQL. It's main advantages are:

Compile-time migrations
Prioritize Ecto.Repo functions for modifying and reading data

Installation

If available in Hex, the package can be installed by adding condo to your list of dependencies in mix.exs:

def deps do
  [
    {:condo, "~> 0.1.0"}
  ]
end

Documentation can be generated with ExDoc and published on HexDocs. Once published, the docs can be found at https://hexdocs.pm/condo.

Setup with Mix Tasks

To set Condo up, you'll need to decide on two things.

What prefix would you like for your tenants?
Where would you like to store your migrations?

I recommend setting your migration prefix to a description of your data instead of simply calling it tenant_ as is the default. For some that's store_ (SaaS) and for others that's region_ (sharding by and for geolocation.) You can do this by:

config :condo,
  prefix: "tenant_"

Next up is simply supplying a place where you would like your migrations written to. Condo will figure out the path based on the module name supplied. If you're managing tenants by App.Company then you would ideally want the migration module namespace to be App.Company.Migrations. You can set this as so:

config :condo,
  migration_namespace: "App.Company.Migrations"

And Condo will now store the migrations in the lib/app/company/migrations folder. Magic!

Generating a Migration

First up is the initial migration, you create one as so:

mix condo.gen.migration create_products

This will create a template that should look as so:

defmodule BlitzPG.LeagueMatches.Migrations.CreateProducts do
  use Ecto.Migration

  def version, do: 20191122003859

  def change do

  end
end

You'll notice that unlike Ecto's migration generator, Condo adds a version() function. Since Condo's migrations are compile-time and not run-time, we don't have access to reading the timestamp off of the file name. However, Condo does prefix all of the file names with the timestamp since this does help significantly with telling the order of the migrations and how they're run. The same result is created with exposing the timestamp in the module as a function instead of in the filename.

Running and Rolling Back a Migration

As you would think, Condo's migration commands are not that different from Ecto's and even accept similar argument for feature parity.

mix condo.migrate
mix condo.migrate -r App.ObscureRepo

mix condo.rollback
mix condo.rollback -r App.ObscureRepo

Condo's Syntactic Sugar

Condo doesn't just aim to give you easier migration handling with non-public postgres schemas, but also to make sure your application code better manages it.

Schema Prefix

Although it's quite simple to do it yourself you can get a schema prefix made for you as so:

# DIY
Repo.all(Product, prefix: "store_#{store.id}")

# Condo Method with a struct supplied
Repo.all(Product, prefix: Condo.prefix(store))

# Or pass in a binary, atom, or integer without the struct needed
Repo.all(Product, prefix: Condo.prefix(store.id))
Repo.all(Product, prefix: Condo.prefix(store.uuid))
Repo.all(Product, prefix: Condo.prefix(:north_america))

# Get back to the public schema just-in-case
Condo.prefix(:public)
# => "public"

# Set the prefix on an Ecto.Queryable.t()
import Ecto.Query
User |> where([u], u.email == ^email) |> Condo.prefix(:foo) |> Repo.one()

Condo's Repo

Just like apartmentex, Condo supports CRUD functions in parity to Ecto.Repo. To achieve this, we just do some meta-programming and keep it simple. The main different is that we add a second parameter just before the final options to specify a region. This will also run a function to figure out which Repo to use if you're sharding tenants over multiple instances.

Condo.Repo.all(Game, :na1) == Repo.NA1.all(Game, prefix: Condo.prefix(:na1))
Condo.Repo.all(Game, :sa1) == Repo.SA1.all(Game, prefix: Condo.prefix(:sa1))

To Do

Schema module to cache migration SQL for creating and running new migrations
Setup tests

BenMorganIO / condo