View Source Ecto.Multi (Ecto v3.12.2)

Ecto.Multi is a data structure for grouping multiple Repo operations.

Ecto.Multi makes it possible to pack operations that should be performed in a single database transaction and gives a way to introspect the queued operations without actually performing them. Each operation is given a name that is unique and will identify its result in case of success or failure.

If a multi is valid (i.e. all the changesets in it are valid), all operations will be executed in the order they were added.

The Ecto.Multi structure should be considered opaque. You can use %Ecto.Multi{} to pattern match the type, but accessing fields or directly modifying them is not advised.

Ecto.Multi.to_list/1 returns a canonical representation of the structure that can be used for introspection.

Changesets

If multi contains operations that accept changesets (like insert/4, update/4 or delete/4) they will be checked before starting the transaction. If any changeset has errors, the transaction won't even be started and the error will be immediately returned.

Note: insert/4, update/4, insert_or_update/4, and delete/4 variants that accept a function do not perform these checks since the functions are executed after the transaction has started.

Run

Multi allows you to run arbitrary functions as part of your transaction via run/3 and run/5. This is especially useful when an operation depends on the value of a previous operation. For this reason, the function given as a callback to run/3 and run/5 will receive the repo as the first argument, and all changes performed by the multi so far as a map for the second argument.

The function given to run must return {:ok, value} or {:error, value} as its result. Returning an error will abort any further operations and make the whole multi fail.

Example

Let's look at an example definition and usage. The use case we'll be looking into is resetting a password. We need to update the account with proper information, log the request and remove all current sessions:

defmodule PasswordManager do
  alias Ecto.Multi

  def reset(account, params) do
    Multi.new()
    |> Multi.update(:account, Account.password_reset_changeset(account, params))
    |> Multi.insert(:log, Log.password_reset_changeset(account, params))
    |> Multi.delete_all(:sessions, Ecto.assoc(account, :sessions))
  end
end

We can later execute it in the integration layer using Repo:

Repo.transaction(PasswordManager.reset(account, params))

By pattern matching on the result we can differentiate different conditions:

case result do
  {:ok, %{account: account, log: log, sessions: sessions}} ->
    # Operation was successful, we can access results (exactly the same
    # we would get from running corresponding Repo functions) under keys
    # we used for naming the operations.
  {:error, failed_operation, failed_value, changes_so_far} ->
    # One of the operations failed. We can access the operation's failure
    # value (like changeset for operations on changesets) to prepare a
    # proper response. We also get access to the results of any operations
    # that succeeded before the indicated operation failed. However, any
    # successful operations would have been rolled back.
end

We can also easily unit test our transaction without actually running it. Since changesets can use in-memory-data, we can use an account that is constructed in memory as well (without persisting it to the database):

test "dry run password reset" do
  account = %Account{password: "letmein"}
  multi = PasswordManager.reset(account, params)

  assert [
    {:account, {:update, account_changeset, []}},
    {:log, {:insert, log_changeset, []}},
    {:sessions, {:delete_all, query, []}}
  ] = Ecto.Multi.to_list(multi)

  # We can introspect changesets and query to see if everything
  # is as expected, for example:
  assert account_changeset.valid?
  assert log_changeset.valid?
  assert inspect(query) == "#Ecto.Query<from a in Session>"
end

The name of each operation does not have to be an atom. This can be particularly useful when you wish to update a collection of changesets at once, and track their errors individually:

accounts = [%Account{id: 1}, %Account{id: 2}]

Enum.reduce(accounts, Multi.new(), fn account, multi ->
  Multi.update(
    multi,
    {:account, account.id},
    Account.password_reset_changeset(account, params)
  )
end)

Summary

Types

changes()

Map of changes made so far during the current transaction. For any Multi which returns {:ok, result}, its name/0 is added as a key and its result as the value.

failure()

Result of a failed transaction using a Multi.

fun(result)

merge()

name()

Name of an operation in the Multi. Can be any term, as long as it is unique within the list of operations; for example, :insert_post or {:delete_post, 5}.

run()

t()

Functions

all(multi, name, queryable_or_fun, opts \\ [])

Runs a query and stores all entries in the multi.

append(lhs, rhs)

Appends the second multi to the first one.

delete(multi, name, changeset_or_struct_fun, opts \\ [])

Adds a delete operation to the multi.

delete_all(multi, name, queryable_or_fun, opts \\ [])

Adds a delete_all operation to the multi.

error(multi, name, value)

Causes the multi to fail with the given value.

exists?(multi, name, queryable_or_fun, opts \\ [])

Checks if there exists an entry matching the given query and stores a boolean in the multi.

insert(multi, name, changeset_or_struct_or_fun, opts \\ [])

Adds an insert operation to the multi.

insert_all(multi, name, schema_or_source, entries_or_query_or_fun, opts \\ [])

Adds an insert_all operation to the multi.

insert_or_update(multi, name, changeset_or_fun, opts \\ [])

Inserts or updates a changeset depending on whether the changeset was persisted or not.

inspect(multi, opts \\ [])

Inspects results from a Multi

merge(multi, merge)

Merges a multi returned dynamically by an anonymous function.

merge(multi, mod, fun, args)

Merges a multi returned dynamically by calling module and function with args.

new()

Returns an empty Ecto.Multi struct.

one(multi, name, queryable_or_fun, opts \\ [])

Runs a query expecting one result and stores it in the multi.

prepend(lhs, rhs)

Prepends the second multi to the first one.

put(multi, name, value)

Adds a value to the changes so far under the given name.

run(multi, name, run)

Adds a function to run as part of the multi.

run(multi, name, mod, fun, args)

Adds a function to run as part of the multi.

to_list(multi)

Returns the list of operations stored in multi.

update(multi, name, changeset_or_fun, opts \\ [])

Adds an update operation to the multi.

update_all(multi, name, queryable_or_fun, updates, opts \\ [])

Adds an update_all operation to the multi.

Types

changes()

@type changes() :: map()

Map of changes made so far during the current transaction. For any Multi which returns {:ok, result}, its name/0 is added as a key and its result as the value.

failure()

@type failure() ::
  {:error, failed_operation :: name(), failed_value :: any(),
   changes_so_far :: %{required(name()) => any()}}

Result of a failed transaction using a Multi.

fun(result)

@type fun(result) :: (changes() -> result)

merge()

@type merge() :: (changes() -> t()) | {module(), atom(), [any()]}

name()

@type name() :: any()

Name of an operation in the Multi. Can be any term, as long as it is unique within the list of operations; for example, :insert_post or {:delete_post, 5}.

run()

@type run() :: (Ecto.Repo.t(), changes() -> {:ok | :error, any()})

t()

@type t() :: %Ecto.Multi{names: names(), operations: operations()}

Functions

all(multi, name, queryable_or_fun, opts \\ [])

@spec all(
  t(),
  name(),
  queryable :: Ecto.Queryable.t() | fun(Ecto.Queryable.t()),
  opts :: Keyword.t()
) :: t()

Runs a query and stores all entries in the multi.

The name must be unique from other statements in the multi.

The remaining arguments and options are the same as in Ecto.Repo.all/2 does.

Example

Ecto.Multi.new()
|> Ecto.Multi.all(:all, Post)
|> MyApp.Repo.transaction()

Ecto.Multi.new()
|> Ecto.Multi.all(:all, fn _changes -> Post end)
|> MyApp.Repo.transaction()

append(lhs, rhs)

@spec append(t(), t()) :: t()

Appends the second multi to the first one.

All names must be unique between both structures.

Example

iex> lhs = Ecto.Multi.new() |> Ecto.Multi.run(:left, fn _, changes -> {:ok, changes} end)
iex> rhs = Ecto.Multi.new() |> Ecto.Multi.run(:right, fn _, changes -> {:error, changes} end)
iex> Ecto.Multi.append(lhs, rhs) |> Ecto.Multi.to_list |> Keyword.keys
[:left, :right]

delete(multi, name, changeset_or_struct_fun, opts \\ [])

@spec delete(
  t(),
  name(),
  Ecto.Changeset.t()
  | Ecto.Schema.t()
  | fun(Ecto.Changeset.t() | Ecto.Schema.t()),
  Keyword.t()
) :: t()

Adds a delete operation to the multi.

The name must be unique from other statements in the multi.

The remaining arguments and options are the same as in Ecto.Repo.delete/2.

Example

post = MyApp.Repo.get!(Post, 1)
Ecto.Multi.new()
|> Ecto.Multi.delete(:delete, post)
|> MyApp.Repo.transaction()

Ecto.Multi.new()
|> Ecto.Multi.run(:post, fn repo, _changes ->
  case repo.get(Post, 1) do
    nil -> {:error, :not_found}
    post -> {:ok, post}
  end
end)
|> Ecto.Multi.delete(:delete, fn %{post: post} ->
  # Others validations
  post
end)
|> MyApp.Repo.transaction()

delete_all(multi, name, queryable_or_fun, opts \\ [])

@spec delete_all(
  t(),
  name(),
  Ecto.Queryable.t() | fun(Ecto.Queryable.t()),
  Keyword.t()
) :: t()

Adds a delete_all operation to the multi.

Accepts the same arguments and options as Ecto.Repo.delete_all/2 does.

Example

queryable = from(p in Post, where: p.id < 5)
Ecto.Multi.new()
|> Ecto.Multi.delete_all(:delete_all, queryable)
|> MyApp.Repo.transaction()

Ecto.Multi.new()
|> Ecto.Multi.run(:post, fn repo, _changes ->
  case repo.get(Post, 1) do
    nil -> {:error, :not_found}
    post -> {:ok, post}
  end
end)
|> Ecto.Multi.delete_all(:delete_all, fn %{post: post} ->
  # Others validations
  from(c in Comment, where: c.post_id == ^post.id)
end)
|> MyApp.Repo.transaction()

error(multi, name, value)

@spec error(t(), name(), error :: term()) :: t()

Causes the multi to fail with the given value.

Running the multi in a transaction will execute no previous steps and returns the value of the first error added.

exists?(multi, name, queryable_or_fun, opts \\ [])

@spec exists?(
  t(),
  name(),
  queryable :: Ecto.Queryable.t() | fun(Ecto.Queryable.t()),
  opts :: Keyword.t()
) :: t()

Checks if there exists an entry matching the given query and stores a boolean in the multi.

The name must be unique from other statements in the multi.

The remaining arguments and options are the same as in Ecto.Repo.exists?/2.

Example

Ecto.Multi.new()
|> Ecto.Multi.exists?(:post, Post)
|> MyApp.Repo.transaction()

Ecto.Multi.new()
|> Ecto.Multi.exists?(:post, fn _changes -> Post end)
|> MyApp.Repo.transaction()

insert(multi, name, changeset_or_struct_or_fun, opts \\ [])

@spec insert(
  t(),
  name(),
  Ecto.Changeset.t()
  | Ecto.Schema.t()
  | fun(Ecto.Changeset.t() | Ecto.Schema.t()),
  Keyword.t()
) :: t()

Adds an insert operation to the multi.

The name must be unique from other statements in the multi.

The remaining arguments and options are the same as in Ecto.Repo.insert/2.

Example

Ecto.Multi.new()
|> Ecto.Multi.insert(:insert, %Post{title: "first"})
|> MyApp.Repo.transaction()

Ecto.Multi.new()
|> Ecto.Multi.insert(:post, %Post{title: "first"})
|> Ecto.Multi.insert(:comment, fn %{post: post} ->
  Ecto.build_assoc(post, :comments)
end)
|> MyApp.Repo.transaction()

insert_all(multi, name, schema_or_source, entries_or_query_or_fun, opts \\ [])

@spec insert_all(
  t(),
  name(),
  schema_or_source(),
  entries_or_query_or_fun ::
    [map() | Keyword.t()] | fun([map() | Keyword.t()]) | Ecto.Query.t(),
  Keyword.t()
) :: t()

Adds an insert_all operation to the multi.

Accepts the same arguments and options as Ecto.Repo.insert_all/3 does.

Example

posts = [%{title: "My first post"}, %{title: "My second post"}]
Ecto.Multi.new()
|> Ecto.Multi.insert_all(:insert_all, Post, posts)
|> MyApp.Repo.transaction()

Ecto.Multi.new()
|> Ecto.Multi.run(:post, fn repo, _changes ->
  case repo.get(Post, 1) do
    nil -> {:error, :not_found}
    post -> {:ok, post}
  end
end)
|> Ecto.Multi.insert_all(:insert_all, Comment, fn %{post: post} ->
  # Others validations

  entries
  |> Enum.map(fn comment ->
    Map.put(comment, :post_id, post.id)
  end)
end)
|> MyApp.Repo.transaction()

insert_or_update(multi, name, changeset_or_fun, opts \\ [])

@spec insert_or_update(
  t(),
  name(),
  Ecto.Changeset.t() | fun(Ecto.Changeset.t()),
  Keyword.t()
) :: t()

Inserts or updates a changeset depending on whether the changeset was persisted or not.

The name must be unique from other statements in the multi.

The remaining arguments and options are the same as in Ecto.Repo.insert_or_update/2.

Example

changeset = Post.changeset(%Post{}, %{title: "New title"})
Ecto.Multi.new()
|> Ecto.Multi.insert_or_update(:insert_or_update, changeset)
|> MyApp.Repo.transaction()

Ecto.Multi.new()
|> Ecto.Multi.run(:post, fn repo, _changes ->
  {:ok, repo.get(Post, 1) || %Post{}}
end)
|> Ecto.Multi.insert_or_update(:update, fn %{post: post} ->
  Ecto.Changeset.change(post, title: "New title")
end)
|> MyApp.Repo.transaction()

inspect(multi, opts \\ [])

@spec inspect(t(), Keyword.t()) :: t()

Inspects results from a Multi

By default, the name is shown as a label to the inspect, custom labels are supported through the IO.inspect/2 label option.

Options

All options for IO.inspect/2 are supported, it also support the following ones:

:only - A field or a list of fields to inspect, will print the entire map by default.

Examples

Ecto.Multi.new()
|> Ecto.Multi.insert(:person_a, changeset)
|> Ecto.Multi.insert(:person_b, changeset)
|> Ecto.Multi.inspect()
|> MyApp.Repo.transaction()

Prints:

%{person_a: %Person{...}, person_b: %Person{...}}

We can use the :only option to limit which fields will be printed:

Ecto.Multi.new()
|> Ecto.Multi.insert(:person_a, changeset)
|> Ecto.Multi.insert(:person_b, changeset)
|> Ecto.Multi.inspect(only: :person_a)
|> MyApp.Repo.transaction()

Prints:

%{person_a: %Person{...}}

merge(multi, merge)

@spec merge(t(), (changes() -> t())) :: t()

Merges a multi returned dynamically by an anonymous function.

This function is useful when the multi to be merged requires information from the original multi. Hence the second argument is an anonymous function that receives the multi changes so far. The anonymous function must return another multi.

If you would prefer to simply merge two multis together, see append/2 or prepend/2.

Duplicated operations are not allowed.

Example

multi =
  Ecto.Multi.new()
  |> Ecto.Multi.insert(:post, %Post{title: "first"})

multi
|> Ecto.Multi.merge(fn %{post: post} ->
  Ecto.Multi.new()
  |> Ecto.Multi.insert(:comment, Ecto.build_assoc(post, :comments))
end)
|> MyApp.Repo.transaction()

merge(multi, mod, fun, args)

@spec merge(t(), module(), function, args) :: t() when function: atom(), args: [any()]

Merges a multi returned dynamically by calling module and function with args.

Similar to merge/2, but allows to pass module name, function and arguments. The function should return an Ecto.Multi, and receives changes so far as the first argument (prepended to those passed in the call to the function).

Duplicated operations are not allowed.

new()

@spec new() :: t()

Returns an empty Ecto.Multi struct.

Example

iex> Ecto.Multi.new() |> Ecto.Multi.to_list()
[]

one(multi, name, queryable_or_fun, opts \\ [])

@spec one(
  t(),
  name(),
  queryable :: Ecto.Queryable.t() | fun(Ecto.Queryable.t()),
  opts :: Keyword.t()
) :: t()

Runs a query expecting one result and stores it in the multi.

The name must be unique from other statements in the multi.

The remaining arguments and options are the same as in Ecto.Repo.one/2.

Example

Ecto.Multi.new()
|> Ecto.Multi.one(:post, Post)
|> Ecto.Multi.one(:author, fn %{post: post} ->
  from(a in Author, where: a.id == ^post.author_id)
end)
|> MyApp.Repo.transaction()

prepend(lhs, rhs)

@spec prepend(t(), t()) :: t()

Prepends the second multi to the first one.

All names must be unique between both structures.

Example

iex> lhs = Ecto.Multi.new() |> Ecto.Multi.run(:left, fn _, changes -> {:ok, changes} end)
iex> rhs = Ecto.Multi.new() |> Ecto.Multi.run(:right, fn _, changes -> {:error, changes} end)
iex> Ecto.Multi.prepend(lhs, rhs) |> Ecto.Multi.to_list |> Keyword.keys
[:right, :left]

put(multi, name, value)

@spec put(t(), name(), any()) :: t()

Adds a value to the changes so far under the given name.

The given value is added to the multi before the transaction starts. If you would like to run arbitrary functions as part of your transaction, see run/3 or run/5.

Example

Imagine there is an existing company schema that you retrieved from the database. You can insert it as a change in the multi using put/3:

Ecto.Multi.new()
|> Ecto.Multi.put(:company, company)
|> Ecto.Multi.insert(:user, fn changes -> User.changeset(changes.company) end)
|> Ecto.Multi.insert(:person, fn changes -> Person.changeset(changes.user, changes.company) end)
|> MyApp.Repo.transaction()

In the example above there isn't a large benefit in putting the company in the multi, because you could also access the company variable directly inside the anonymous function.

However, the benefit of put/3 is when composing Ecto.Multis. If the insert operations above were defined in another module, you could use put(:company, company) to inject changes that will be accessed by other functions down the chain, removing the need to pass both multi and company values around.

run(multi, name, run)

@spec run(t(), name(), run()) :: t()

Adds a function to run as part of the multi.

The function should return either {:ok, value} or {:error, value}, and receives the repo as the first argument, and the changes so far as the second argument.

Example

Ecto.Multi.run(multi, :write, fn _repo, %{image: image} ->
  with :ok <- File.write(image.name, image.contents) do
    {:ok, nil}
  end
end)

run(multi, name, mod, fun, args)

@spec run(t(), name(), module(), function, args) :: t()
when function: atom(), args: [any()]

Adds a function to run as part of the multi.

Similar to run/3, but allows to pass module name, function and arguments. The function should return either {:ok, value} or {:error, value}, and receives the repo as the first argument, and the changes so far as the second argument (prepended to those passed in the call to the function).

to_list(multi)

@spec to_list(t()) :: [{name(), term()}]

Returns the list of operations stored in multi.

Always use this function when you need to access the operations you have defined in Ecto.Multi. Inspecting the Ecto.Multi struct internals directly is discouraged.

update(multi, name, changeset_or_fun, opts \\ [])

@spec update(t(), name(), Ecto.Changeset.t() | fun(Ecto.Changeset.t()), Keyword.t()) ::
  t()

Adds an update operation to the multi.

The name must be unique from other statements in the multi.

The remaining arguments and options are the same as in Ecto.Repo.update/2.

Example

post = MyApp.Repo.get!(Post, 1)
changeset = Ecto.Changeset.change(post, title: "New title")
Ecto.Multi.new()
|> Ecto.Multi.update(:update, changeset)
|> MyApp.Repo.transaction()

Ecto.Multi.new()
|> Ecto.Multi.insert(:post, %Post{title: "first"})
|> Ecto.Multi.update(:fun, fn %{post: post} ->
  Ecto.Changeset.change(post, title: "New title")
end)
|> MyApp.Repo.transaction()

update_all(multi, name, queryable_or_fun, updates, opts \\ [])

@spec update_all(
  t(),
  name(),
  Ecto.Queryable.t() | fun(Ecto.Queryable.t()),
  Keyword.t(),
  Keyword.t()
) :: t()

Adds an update_all operation to the multi.

Accepts the same arguments and options as Ecto.Repo.update_all/3 does.

Example

Ecto.Multi.new()
|> Ecto.Multi.update_all(:update_all, Post, set: [title: "New title"])
|> MyApp.Repo.transaction()

Ecto.Multi.new()
|> Ecto.Multi.run(:post, fn repo, _changes ->
  case repo.get(Post, 1) do
    nil -> {:error, :not_found}
    post -> {:ok, post}
  end
end)
|> Ecto.Multi.update_all(:update_all, fn %{post: post} ->
  # Others validations
  from(c in Comment, where: c.post_id == ^post.id, update: [set: [title: "New title"]])
end, [])
|> MyApp.Repo.transaction()