Using Postgres UPDATE ... FROM in Ecto without raw SQL - postgresql

Question

Based on an Elixir thread from last year, I was able to write a raw SQL query to bulk update records with the values from an unrelated table. However, I would like to be able to generate this query using Ecto.
In the example below, assume there are two tables, cats and dogs, and the cats table has a foreign key (dog_id). I want to link a dog to a cat.
The code below is how I'm doing this manually with Elixir and raw SQL:
cat_ids = [1,2,3] # pretend these are uuids
dog_ids = [4,5,6] # ... uuids
values =
cat_ids
|> Enum.zip(dog_ids)
|> Enum.map(fn {cat_id, dog_id} ->
"('#{cat_id}'::uuid, '#{dog_id}'::uuid)"
end)
|> Enum.join(", ")
sql = """
UPDATE cats as a
SET dog_id = c.dog_id
from (values #{values}) as c(cat_id, dog_id)
where c.cat_id = a.id;
"""
Repo.query(sql)
Is there a way to move this to Repo.update_all or some use of fragments so I'm not manually building the query?

Answer 1

Sure, you can use Ecto syntax, but it ain't much different in my opinion, tho you have to use a Schema, for example in my application I have an user authentication and that's how we update the token:
def update_token(user_id, token) do
Repo.transaction(fn ->
from(t in UserAuthentication, where: t.user_id == ^to_string(user_id))
|> Repo.update_all(set: [token: token])
end
and the UserAuthentication schema looks more or less like:
defmodule MyApp.UserAuthentication do
use Ecto.Schema
import Ecto.Changeset
schema "user_authentication" do
field(:user_id, :integer)
field(:token, :string)
timestamps()
end
def changeset(%__MODULE__{} = user, attrs) do
user
|> cast(attrs, [:user_id, :token])
|> validate_required([:user_id, :token])
end
end
This is good for data validation and works in any DB you attach.