Ecto.multi with insert and insert_all?

Questions / Help
,
1

Is it possible to use an Ecto.Multi to do an insert and an insert_all where the insert_all needs the primary key resulting from the first insert? E.g. creating a parent record with numerous child records in one transaction.

The insert option can use a 3rd argument of changeset_or_struct_or_fun, but the insert_all function doesn’t allow for a function… so there doesn’t appear to be a good way to capture the primary key that gets generated.

I can wrap the steps inside of a Repo.transaction callback function like this:

Repo.transaction fn ->
  with {:ok, parent} <- ParentContext.create(params),
       {:ok, _children} <- create_children(parent, params)
    do

  else
    {:error, e} -> Repo.rollback(e)
  end
end

but I was curious if the multi supported it. Thanks!

1 Like
2

Yeah you can using Multi.merge/2 (or Multi.run). It looks something like this:

Multi.new()
|> Multi.insert(:user, User.changeset(%User{}, params))
|> Multi.merge(fn %{user: user} ->
  # That is the inserted user from the first part of the multi
  Multi.new()
  |> Multi.insert_all(:posts, build_insert_all_posts_changeset(user))
end)
|> Repo.transaction()
12 Likes
3

What is this function??

1 Like
5

It is a function that you would need to define that returns a list of post changesets (and is suitable for insert_all so it might have to set inserted_at and updated_at).

6

How can you use insert_all with a list of changeset?

7

You can’t directly. Repo.*_all are lower level APIs, which don’t support many schema or changeset features. You can map things from changeset to values for insert_all manually if you don’t need any of the not really supported parts.

8

Thanks for the clarification.

So, is there any example on how to insert a bunch of “children” in a Multi with validation?

Will open a new thread though. Don’t want to necrobump.