Paper Trail

Track and record all the changes in your database. Revert back to anytime in history.

How does it work?

PaperTrail lets you record every change in your database in a separate database table called versions. Library generates a new version record with associated data every time you run PaperTrail.insert/2, PaperTrail.update/2 or PaperTrail.delete/2 functions. Simply these functions wrap your Repo insert, update or destroy actions in a database transaction, so if your database action fails you won't get a new version.

PaperTrail is assailed with hundreds of test assertions for each release. Data integrity is an important aim of this project, please refer to the strict_mode if you want to ensure data correctness and integrity of your versions. For simpler use cases the default mode of PaperTrail should suffice.

Example

changeset = Post.changeset(%Post{}, %{
  title: "Word on the street is Elixir got its own database versioning library",
  content: "You should try it now!"
})

PaperTrail.insert(changeset)
# => on success:
# {:ok,
#  %{model: %Post{__meta__: #Ecto.Schema.Metadata<:loaded, "posts">,
#     title: "Word on the street is Elixir got its own database versioning library",
#     content: "You should try it now!", id: 1, inserted_at: ~N[2016-09-15 21:42:38],
#     updated_at: ~N[2016-09-15 21:42:38]},
#    version: %PaperTrail.Version{__meta__: #Ecto.Schema.Metadata<:loaded, "versions">,
#     event: "insert", id: 1, inserted_at: ~N[2016-09-15 21:42:38],
#     item_changes: %{title: "Word on the street is Elixir got its own database versioning library",
#       content: "You should try it now!", id: 1, inserted_at: ~N[2016-09-15 21:42:38],
#       updated_at: ~N[2016-09-15 21:42:38]},
#     item_id: 1, item_type: "Post", originator_id: nil, originator: nil, meta: nil}}}

# => on error(it matches Repo.insert/2):
# {:error, Ecto.Changeset<action: :insert,
#  changes: %{title: "Word on the street is Elixir got its own database versioning library", content: "You should try it now!"},
#  errors: [content: {"is too short", []}], data: #Post<>,
#  valid?: false>, %{}}

post = Repo.get!(Post, 1)
edit_changeset = Post.changeset(post, %{
  title: "Elixir matures fast",
  content: "Future is already here, Elixir is the next step!"
})

PaperTrail.update(edit_changeset)
# => on success:
# {:ok,
#  %{model: %Post{__meta__: #Ecto.Schema.Metadata<:loaded, "posts">,
#     title: "Elixir matures fast", content: "Future is already here, Elixir is the next step!",
#     id: 1, inserted_at: ~N[2016-09-15 21:42:38],
#     updated_at: ~N[2016-09-15 22:00:59]},
#    version: %PaperTrail.Version{__meta__: #Ecto.Schema.Metadata<:loaded, "versions">,
#     event: "update", id: 2, inserted_at: ~N[2016-09-15 22:00:59],
#     item_changes: %{title: "Elixir matures fast", content: "Future is already here, Elixir is the next step!"},
#     item_id: 1, item_type: "Post", originator_id: nil, originator: nil
#     meta: nil}}}

# => on error(it matches Repo.update/2):
# {:error, Ecto.Changeset<action: :update,
#  changes: %{title: "Elixir matures fast", content: "Future is already here, Elixir is the next step!"},
#  errors: [title: {"is too short", []}], data: #Post<>,
#  valid?: false>, %{}}

PaperTrail.get_version(post)
#  %PaperTrail.Version{__meta__: #Ecto.Schema.Metadata<:loaded, "versions">,
#   event: "update", id: 2, inserted_at: ~N[2016-09-15 22:00:59],
#   item_changes: %{title: "Elixir matures fast", content: "Future is already here, Elixir is the next step!"},
#   item_id: 1, item_type: "Post", originator_id: nil, originator: nil, meta: nil}}}

updated_post = Repo.get!(Post, 1)

PaperTrail.delete(updated_post)
# => on success:
# {:ok,
#  %{model: %Post{__meta__: #Ecto.Schema.Metadata<:deleted, "posts">,
#     title: "Elixir matures fast", content: "Future is already here, Elixir is the next step!",
#     id: 1, inserted_at: ~N[2016-09-15 21:42:38],
#     updated_at: ~N[2016-09-15 22:00:59]},
#    version: %PaperTrail.Version{__meta__: #Ecto.Schema.Metadata<:loaded, "versions">,
#     event: "delete", id: 3, inserted_at: ~N[2016-09-15 22:22:12],
#     item_changes: %{title: "Elixir matures fast", content: "Future is already here, Elixir is the next step!",
#       id: 1, inserted_at: ~N[2016-09-15 21:42:38],
#       updated_at: ~N[2016-09-15 22:00:59]},
#     item_id: 1, item_type: "Post", originator_id: nil, originator: nil, meta: nil}}}

Repo.aggregate(Post, :count, :id) # => 0
PaperTrail.Version.count() # => 3
# same as Repo.aggregate(PaperTrail.Version, :count, :id)

PaperTrail.Version.last() # returns the last version in the db by inserted_at
#  %PaperTrail.Version{__meta__: #Ecto.Schema.Metadata<:loaded, "versions">,
#   event: "delete", id: 3, inserted_at: ~N[2016-09-15 22:22:12],
#   item_changes: %{"title" => "Elixir matures fast", content: "Future is already here, Elixir is the next step!", "id" => 1,
#     "inserted_at" => "2016-09-15T21:42:38",
#     "updated_at" => "2016-09-15T22:00:59"},
#   item_id: 1, item_type: "Post", originator_id: nil, originator: nil, meta: nil}

PaperTrail is inspired by the ruby gem paper_trail. However, unlike the paper_trail gem this library actually results in less data duplication, faster and more explicit programming model to version your record changes.

The library source code is minimal and well tested. It is suggested to read the source code.

Installation

1. Add paper_trail to your list of dependencies in mix.exs:

   def deps do
     [{:paper_trail, "~> 0.14.3"}]
   end

2. Configure paper_trail to use your application repo in config/config.exs:

   config :paper_trail, repo: YourApplicationName.Repo
   # if you don't specify this PaperTrail will assume your repo name is Repo

3. Install and compile your dependency:

   mix deps.get && mix compile

4. Run this command to generate the migration:

   mix papertrail.install

   You might want to edit the types for :item_id or :originator_id if you're using UUID or other types for your primary keys before you execute mix ecto.migrate.

5. Run the migration:

   mix ecto.migrate

Your application is now ready to collect some history!

Does this work with phoenix?

YES! Make sure you do the steps above.

%PaperTrail.Version{} fields:

Configuring the types

If you are using UUID or another type for your primary keys, you can configure the PaperTrail.Version schema to use it.

Example Config

config :paper_trail, item_type: Ecto.UUID,
                     originator_type: Ecto.UUID,
                     originator_relationship_options: [references: :uuid]

Example User

defmodule Acme.User do
  use Ecto.Schema

  @primary_key {:uuid, :binary_id, autogenerate: true}
  schema "users" do
    field :email, :string

    timestamps()
  end

Remember to edit the types accordingly in the generated migration.

Version origin references:

PaperTrail records have a string field called origin. PaperTrail.insert/2, PaperTrail.update/2, PaperTrail.delete/2 functions accept a second argument to describe the origin of this version:

PaperTrail.update(changeset, origin: "migration")
# or:
PaperTrail.update(changeset, origin: "user:1234")
# or:
PaperTrail.delete(changeset, origin: "worker:delete_inactive_users")
# or:
PaperTrail.insert(new_user_changeset, origin: "password_registration")
# or:
PaperTrail.insert(new_user_changeset, origin: "facebook_registration")

Version originator relationships

You can specify setter/originator relationship to paper_trail versions with originator assignment. This feature is only possible by specifying :originator keyword list for your application configuration:

# In your config/config.exs
config :paper_trail, originator: [name: :user, model: YourApp.User]
# For most applications originator should be the user since models can be updated/created/deleted by several users.

Note: You will need to recompile your deps after you have added the config for originator.

Then originator name could be used for querying and preloading. Originator setting must be done via :originator or originator name that is defined in the paper_trail configuration:

user = create_user()
# all these set originator_id's for the version records
PaperTrail.insert(changeset, originator: user)
{:ok, result} = PaperTrail.update(edit_changeset, originator: user)
# or you can use :user in the params instead of :originator if this is your config:
# config :paper_trail, originator: [name: :user, model: YourApplication.User]
{:ok, result} = PaperTrail.update(edit_changeset, user: user)
result[:version] |> Repo.preload(:user) |> Map.get(:user) # we can access the user who made the change from the version thanks to originator relationships!
PaperTrail.delete(edit_changeset, user: user)

Also make sure you have the foreign-key constraint in the database and in your version migration file.

Storing version meta data

You might want to add some meta data that doesn't belong to originator and origin fields. Such data could be stored in one object named meta in paper_trail versions. Meta field could be passed as the second optional parameter to PaperTrail.insert/2, PaperTrail.update/2, PaperTrail.delete/2 functions:

company = Company.changeset(%Company{}, %{name: "Acme Inc."})
  |> PaperTrail.insert(meta: %{slug: "acme-llc"})

# You can also combine this with an origin:
edited_company = Company.changeset(company, %{name: "Acme LLC"})
  |> PaperTrail.update(origin: "documentation", meta: %{slug: "acme-llc"})

# Or even with an originator:
user = create_user()
deleted_company = Company.changeset(edited_company, %{})
  |> PaperTrail.delete(origin: "worker:github", originator: user, meta: %{slug: "acme-llc", important: true})

Strict mode

This is a feature more suitable for larger applications. Models can keep their version references via foreign key constraints. Therefore it would be impossible to delete the first and current version of a model if the model exists in the database, it also makes querying easier and the whole design more relational database/SQL friendly. In order to enable strict mode:

# In your config/config.exs
config :paper_trail, strict_mode: true

Strict mode expects tracked models to have foreign-key reference to their first_version and current_version. These columns must be named first_version_id, and current_version_id in their respective model tables. A tracked model example with a migration file:

# In the migration file: priv/repo/migrations/create_company.exs
defmodule Repo.Migrations.CreateCompany do
  def change do
    create table(:companies) do
      add :name,       :string, null: false
      add :founded_in, :date

      # null constraints are highly suggested:
      add :first_version_id, references(:versions), null: false
      add :current_version_id, references(:versions), null: false

      timestamps()
    end

    create unique_index(:companies, [:first_version_id])
    create unique_index(:companies, [:current_version_id])
  end
end

# In the model definition:
defmodule Company do
  use Ecto.Schema

  import Ecto.Changeset

  schema "companies" do
    field :name, :string
    field :founded_in, :date

    belongs_to :first_version, PaperTrail.Version
    belongs_to :current_version, PaperTrail.Version, on_replace: :update # on_replace: is important!

    timestamps()
  end

  def changeset(struct, params \\ %{}) do
    struct
    |> cast(params, [:name, :founded_in])
  end
end

When you run PaperTrail.insert/2 transaction, first_version_id and current_version_id automagically gets assigned for the model. Example:

company = Company.changeset(%Company{}, %{name: "Acme LLC"}) |> PaperTrail.insert
# {:ok,
#  %{model: %Company{__meta__: #Ecto.Schema.Metadata<:loaded, "companies">,
#     name: "Acme LLC", founded_in: nil, id: 1, inserted_at: ~N[2016-09-15 21:42:38],
#     updated_at: ~N[2016-09-15 21:42:38], first_version_id: 1, current_version_id: 1},
#    version: %PaperTrail.Version{__meta__: #Ecto.Schema.Metadata<:loaded, "versions">,
#      event: "insert", id: 1, inserted_at: ~N[2016-09-15 22:22:12],
#      item_changes: %{name: "Acme LLC", founded_in: nil, id: 1, inserted_at: ~N[2016-09-15 21:42:38]},
#      originator_id: nil, origin: "unknown", meta: nil}}}

When you PaperTrail.update/2 a model, current_version_id gets updated during the transaction:

edited_company = Company.changeset(company, %{name: "Acme Inc."}) |> PaperTrail.update(origin: "documentation")
# {:ok,
#  %{model: %Company{__meta__: #Ecto.Schema.Metadata<:loaded, "companies">,
#     name: "Acme Inc.", founded_in: nil, id: 1, inserted_at: ~N[2016-09-15 21:42:38],
#     updated_at: ~N[2016-09-15 23:22:12], first_version_id: 1, current_version_id: 2},
#    version: %PaperTrail.Version{__meta__: #Ecto.Schema.Metadata<:loaded, "versions">,
#      event: "update", id: 2, inserted_at: ~N[2016-09-15 23:22:12],
#      item_changes: %{name: "Acme Inc."}, originator_id: nil, origin: "documentation", meta: nil}}}

Additionally, you can put a null constraint on origin column, you should always put an origin reference to describe who makes the change. This is important for big applications because a model can change from many sources.

Bang(!) functions:

PaperTrail also supports PaperTrail.insert!, PaperTrail.update!, PaperTrail.delete!. Naming of these functions intentionally match Repo.insert!, Repo.update!, Repo.delete! functions. If PaperTrail is on strict_mode these bang functions will update the version references of the model just like the normal PaperTrail operations.

Bang functions assume the operation will always be successful, otherwise functions will raise Ecto.InvalidChangesetError just like Repo.insert!, Repo.update! and Repo.delete!:

changeset = Post.changeset(%Post{}, %{
  title: "Word on the street is Elixir got its own database versioning library",
  content: "You should try it now!"
})

inserted_post = PaperTrail.insert!(changeset)
# => on success:
# %Post{__meta__: #Ecto.Schema.Metadata<:loaded, "posts">,
#   title: "Word on the street is Elixir got its own database versioning library",
#   content: "You should try it now!", id: 1, inserted_at: ~N[2016-09-15 21:42:38],
#   updated_at: ~N[2016-09-15 21:42:38]
# }
#
# => on error raises: Ecto.InvalidChangesetError !!

inserted_post_version = PaperTrail.get_version(inserted_post)
# %PaperTrail.Version{__meta__: #Ecto.Schema.Metadata<:loaded, "versions">,
#   event: "insert", id: 1, inserted_at: ~N[2016-09-15 21:42:38],
#   item_changes: %{title: "Word on the street is Elixir got its own database versioning library",
#     content: "You should try it now!", id: 1, inserted_at: ~N[2016-09-15 21:42:38],
#     updated_at: ~N[2016-09-15 21:42:38]},
#   item_id: 1, item_type: "Post", originator_id: nil, originator: nil, meta: nil}

edit_changeset = Post.changeset(inserted_post, %{
  title: "Elixir matures fast",
  content: "Future is already here, Elixir is the next step!"
})

updated_post = PaperTrail.update!(edit_changeset)
# => on success:
# %Post{__meta__: #Ecto.Schema.Metadata<:loaded, "posts">,
#   title: "Elixir matures fast", content: "Future is already here, you deserve to be awesome!",
#   id: 1, inserted_at: ~N[2016-09-15 21:42:38],
#   updated_at: ~N[2016-09-15 22:00:59]}
#
# => on error raises: Ecto.InvalidChangesetError !!

updated_post_version = PaperTrail.get_version(updated_post)
# %PaperTrail.Version{__meta__: #Ecto.Schema.Metadata<:loaded, "versions">,
#   event: "update", id: 2, inserted_at: ~N[2016-09-15 22:00:59],
#   item_changes: %{title: "Elixir matures fast", content: "Future is already here, Elixir is the next step!"},
#   item_id: 1, item_type: "Post", originator_id: nil, originator: nil
#   meta: nil}

PaperTrail.delete!(updated_post)
# => on success:
# %Post{__meta__: #Ecto.Schema.Metadata<:deleted, "posts">,
#   title: "Elixir matures fast", content: "Future is already here, Elixir is the next step!",
#   id: 1, inserted_at: ~N[2016-09-15 21:42:38],
#   updated_at: ~N[2016-09-15 22:00:59]}
#
# => on error raises: Ecto.InvalidChangesetError !!

PaperTrail.get_version(updated_post)
# %PaperTrail.Version{__meta__: #Ecto.Schema.Metadata<:loaded, "versions">,
#   event: "delete", id: 3, inserted_at: ~N[2016-09-15 22:22:12],
#   item_changes: %{title: "Elixir matures fast", content: "Future is already here, Elixir is the next step!",
#   id: 1, inserted_at: ~N[2016-09-15 21:42:38],
#   updated_at: ~N[2016-09-15 22:00:59]},
#   item_id: 1, item_type: "Post", originator_id: nil, originator: nil, meta: nil}

Repo.aggregate(Post, :count, :id) # => 0
PaperTrail.Version.count() # => 3
# same as Repo.aggregate(PaperTrail.Version, :count, :id)

PaperTrail.Version.last() # returns the last version in the db by inserted_at
#  %PaperTrail.Version{__meta__: #Ecto.Schema.Metadata<:loaded, "versions">,
#   event: "delete", id: 3, inserted_at: ~N[2016-09-15 22:22:12],
#   item_changes: %{"title" => "Elixir matures fast", content: "Future is already here, Elixir is the next step!", "id" => 1,
#     "inserted_at" => "2016-09-15T21:42:38",
#     "updated_at" => "2016-09-15T22:00:59"},
#   item_id: 1, item_type: "Post", originator_id: nil, originator: nil, meta: nil}

Working with multi tenancy

Sometimes you have to deal with applications where you need multi tenancy capabilities, and you want to keep tracking of the versions of your data on different schemas (PostgreSQL) or databases (MySQL).

You can use the Ecto.Query prefix in order to switch between different schemas/databases for your own data, so you can specify in your changeset where to store your record. Example:

tenant = "tenant_id"
changeset = User.changeset(%User{}, %{first_name: "Izel", last_name: "Nakri"})

changeset
|> Ecto.Queryable.to_query()
|> Map.put(:prefix, tenant)
|> Repo.insert()

PaperTrail also allows you to store the Version entries generated by your activity in different schemas/databases by using the value of the element :prefix on the options of the functions. Example:

tenant = "tenant_id"

changeset =
  User.changeset(%User{}, %{first_name: "Izel", last_name: "Nakri"})
  |> Ecto.Queryable.to_query()
  |> Map.put(:prefix, tenant)

PaperTrail.insert(changeset, [prefix: tenant])

By doing this, you're storing the new User entry into the schema/database specified by the :prefix value (tenant_id).

Note that the User's changeset it's sent with the :prefix, so PaperTrail will take care of the storage of the generated Version entry in the desired schema/database. Make sure to add this prefix to your changeset before the execution of the PaperTrail function if you want to do versioning on a separate schema.

PaperTrail can also get versions of records or models from different schemas/databases as well by using the :prefix option. Example:

tenant = "tenant_id"
id = 1

PaperTrail.get_versions(User, id, [prefix: tenant])

Version timestamps

PaperTrail can be configured to use utc_datetime or utc_datetime_usec for Version timestamps.

# In your config/config.exs
config :paper_trail, timestamps_type: :utc_datetime

Note: You will need to recompile your deps after you have added the config for timestamps.

Postgres datatype support

PaperTrail serializes the version data in JSON and not all native Postgres data types are supported directly. Composite types and range types are two examples which have no native JSON representation.

Developers may derive their own Jason encoder for such types. It should be noted that an encoder can only be defined for a native Elixir base type or struct once in an application and therefore there is a small risk of conflicting encoders.

Suggestions

• PaperTrail.Version(s) order matter,
• Don't delete your paper_trail versions, instead you can merge them
• If you have a question or a problem, do not hesitate to create an issue or submit a pull request

Contributing

set -a
source .env
mix test --trace

Credits

Many thanks to:

• Jose Pablo Castro - Built the repo configuration for paper_trail
• Harold Tafur - Built the :ecto_options option for PaperTrail inserts
• Florian Gerhardt - Fixed rare compile errors for PaperTrail repos
• Alex Antonov - Original inventor of the originator feature
• Moritz Schmale - UUID primary keys feature
• Jason Draper - UUID primary keys feature
• Jonatan Männchen - Added non-regular :binary_id UUID support for originator
• Josh Taylor - Maintenance and new feature suggestions
• Mitchell Henke - Fixed weird elixir compiler warnings
• Iván González - Multi tenancy feature and some minor refactors
• Teo Choong Ping - Fixed paper_trail references for newer Elixir versions
• devvit - Added non-regular primary key tracking support
• rustamtolipov - Added support for Ecto v3
• gabrielpra1 - Added enhanced support for Ecto.Changeset
• Darren Thompson - Added PaperTrail.Multi which makes paper trail transactions more usable
• Harold Tafur - Made PaperTrail.insert accept :ecto_options params(ie. upsert options)
• Attila Szabo - Made %Version[:inserted_at] accept different ecto datetime options
• Rafael Scheffer - Built PaperTrail.Serializer that unifies %Version{} serialization
• Kian Meng Ang - Improved documentation
• Francisco Correia - Made PaperTrail transaction keys and ecto transactions more customizable
• Don Barlow - Made :initial_version_key configurable for strict_mode inserts
• Christoph Schmatzler - Built PaperTrail.insert_or_update feature
• Izel Nakri - The Originator of this library. See what I did there ;)

Additional thanks to:

• Ruby paper_trail gem - Initial inspiration of this project.
• Ecto - For the great API.

License

This source code is licensed under the MIT license. Copyright (c) 2016-present Izel Nakri.