Build Status Gem Version SemVer

Track changes to your models, for auditing or versioning. See how a model looked at any stage in its lifecycle, revert it to any version, or restore it after it has been destroyed.


Version Documentation

Table of Contents

1. Introduction

1.a. Compatibility

paper_trail branch tags ruby activerecord
unreleased master >= 2.3.0 >= 4.2, < 6
9 9-stable v9.x >= 2.3.0 >= 4.2, < 6
8 8-stable v8.x >= 2.2.0 >= 4.2, < 5.2
7 7-stable v7.x >= 2.1.0 >= 4.0, < 5.2
6 6-stable v6.x >= 1.9.3 >= 4.0, < 5.2
5 5-stable v5.x >= 1.9.3 >= 3.0, < 5.1
4 4-stable v4.x >= 1.8.7 >= 3.0, < 5.1
3 3.0-stable v3.x >= 1.8.7 >= 3.0, < 5
2 2.7-stable v2.x >= 1.8.7 >= 3.0, < 4
1 rails2 v1.x >= 1.8.7 >= 2.3, < 3

1.b. Installation

  1. Add PaperTrail to your Gemfile.

    gem 'paper_trail'

  2. Add a versions table to your database and an initializer file for configuration:

    bundle exec rails generate paper_trail:install
    bundle exec rake db:migrate

    If using rails_admin, you must enable the experimental Associations feature. For more information on this generator, see section 5.c. Generators.

  3. Add has_paper_trail to the models you want to track.

    class Widget < ActiveRecord::Base
  4. If your controllers have a current_user method, you can easily track who is responsible for changes by adding a controller callback.

    class ApplicationController
      before_action :set_paper_trail_whodunnit

1.c. Basic Usage

Your models now have a versions method which returns the "paper trail" of changes to your model.

widget = Widget.find 42
# [<PaperTrail::Version>, <PaperTrail::Version>, ...]

Once you have a version, you can find out what happened:

v = widget.versions.last
v.event # 'update', 'create', 'destroy'. See also: Custom Event Names
v.whodunnit # ID of `current_user`. Requires `set_paper_trail_whodunnit` callback.
widget = v.reify # The widget as it was before the update (nil for a create event)

PaperTrail stores the pre-change version of the model, unlike some other auditing/versioning plugins, so you can retrieve the original version. This is useful when you start keeping a paper trail for models that already have records in the database.

widget = Widget.find 153                                 # 'Doobly'

# Add has_paper_trail to Widget model.

widget.versions                             # []
widget.update_attributes name: 'Wotsit'             # 'Doobly'
widget.versions.last.event                  # 'update'

This also means that PaperTrail does not waste space storing a version of the object as it currently stands. The versions method gives you previous versions; to get the current one just call a finder on your Widget model as usual.

Here's a helpful table showing what PaperTrail stores:

Event create update destroy
Model Before nil widget widget
Model After widget widget nil

PaperTrail stores the values in the Model Before row. Most other auditing/versioning plugins store the After row.

1.d. API Summary

An introductory sample of common features.

When you declare has_paper_trail in your model, you get these methods:

class Widget < ActiveRecord::Base

# Returns this widget's versions.  You can customise the name of the
# association, but overriding this method is not supported.

# Return the version this widget was reified from, or nil if it is live.
# You can customise the name of the method.

# Returns true if this widget is the current, live one; or false if it is from
# a previous version.

# Returns who put the widget into its current state.

# Returns the widget (not a version) as it looked at the given timestamp.

# Returns the widget (not a version) as it was most recently.

# Returns the widget (not a version) as it became next.

And a PaperTrail::Version instance (which is just an ordinary ActiveRecord instance, with all the usual methods) has methods such as:

# Returns the item restored from this version.
version.reify(options = {})

# Return a new item from this version
version.reify(dup: true)

# Returns who put the item into the state stored in this version.

# Returns who changed the item from the state it had in this version.

# Returns the next version.

# Returns the previous version.

# Returns the index of this version in all the versions.

# Returns the event that caused this version (create|update|destroy).

This is just a sample of common features. Keep reading for more.

1.e. Configuration

Many aspects of PaperTrail are configurable for individual models; typically this is achieved by passing options to the has_paper_trail method within a given model.

Some aspects of PaperTrail are configured globally for all models. These settings are assigned directly on the PaperTrail.config object. A common place to put these settings is in a Rails initializer file such as config/initializers/paper_trail.rb or in an environment-specific configuration file such as config/environments/test.rb.

2. Limiting What is Versioned, and When

2.a. Choosing Lifecycle Events To Monitor

You can choose which events to track with the on option. For example, if you only want to track update events:

class Article < ActiveRecord::Base
  has_paper_trail on: [:update]

has_paper_trail installs callbacks for the specified lifecycle events.

There are four potential callbacks, and the default is to install all four, ie. on: [:create, :destroy, :touch, :update].

The versions.event Column

Your versions table has an event column with three possible values:

event callback
create create
destroy destroy
update touch, update

You may also have the PaperTrail::Version model save a custom string in its event field instead of the typical create, update, destroy. PaperTrail adds an attr_accessor to your model named paper_trail_event, and will insert it, if present, in the event column.

a = Article.create
a.versions.size                           # 1
a.versions.last.event                     # 'create'
a.paper_trail_event = 'update title'
a.update_attributes title: 'My Title'
a.versions.size                           # 2
a.versions.last.event                     # 'update title'
a.paper_trail_event = nil
a.update_attributes title: 'Alternate'
a.versions.size                           # 3
a.versions.last.event                     # 'update'

Controlling the Order of AR Callbacks

If there are other callbacks in your model, their order relative to those installed by has_paper_trail may matter. If you need to control their order, use the paper_trail_on_* methods.

class Article < ActiveRecord::Base
  # Include PaperTrail, but do not install any callbacks. Passing the
  # empty array to `:on` omits callbacks.
  has_paper_trail on: []

  # Add callbacks in the order you need.
  paper_trail.on_destroy    # add destroy callback
  paper_trail.on_update     # etc.

The paper_trail.on_destroy method can be further configured to happen :before or :after the destroy event. In PaperTrail 4, the default is :after. In PaperTrail 5, the default will be :before, to support ActiveRecord 5. (see

2.b. Choosing When To Save New Versions

You can choose the conditions when to add new versions with the if and unless options. For example, to save versions only for US non-draft translations:

class Translation < ActiveRecord::Base
  has_paper_trail if: { |t| t.language_code == 'US' },
                  unless: { |t| t.type == 'DRAFT'       }

Choosing Based on Changed Attributes

Starting with PaperTrail 4.0, versions are saved during an after-callback. If you decide whether to save a new version based on changed attributes, use attribute_name_was instead of attribute_name.

Saving a New Version Manually

You may want to save a new version regardless of options like :on, :if, or :unless. Or, in rare situations, you may want to save a new version even if the record has not changed.


2.c. Choosing Attributes To Monitor


You can ignore changes to certain attributes:

class Article < ActiveRecord::Base
  has_paper_trail ignore: [:title, :rating]

Changes to just the title or rating will not create a version record. Changes to other attributes will create a version record.

a = Article.create
a.versions.length                         # 1
a.update_attributes title: 'My Title', rating: 3
a.versions.length                         # 1
a.update_attributes title: 'Greeting', content: 'Hello'
a.versions.length                         # 2
a.paper_trail.previous_version.title      # 'My Title'


Or, you can specify a list of the only attributes you care about:

class Article < ActiveRecord::Base
  has_paper_trail only: [:title]

Only changes to the title will create a version record.

a = Article.create
a.versions.length                         # 1
a.update_attributes title: 'My Title'
a.versions.length                         # 2
a.update_attributes content: 'Hello'
a.versions.length                         # 2
a.paper_trail.previous_version.content    # nil

The :ignore and :only options can also accept Hash arguments.

class Article < ActiveRecord::Base
  has_paper_trail only: { title: { |obj| !obj.title.blank? } }

If the title is not blank, then only changes to the title will create a version record.

a = Article.create
a.versions.length                         # 1
a.update_attributes content: 'Hello'
a.versions.length                         # 2
a.update_attributes title: 'Title One'
a.versions.length                         # 3
a.update_attributes content: 'Hai'
a.versions.length                         # 3
a.paper_trail.previous_version.content    # "Hello"
a.update_attributes title: 'Title Two'
a.versions.length                         # 4
a.paper_trail.previous_version.content    # "Hai"

Configuring both :ignore and :only is not recommended, but it should work as expected. Passing both :ignore and :only options will result in the article being saved if a changed attribute is included in :only but not in :ignore.


You can skip attributes completely with the :skip option. As with :ignore, updates to these attributes will not create a version record. In addition, if a version record is created for some other reason, these attributes will not be persisted.

class Article < ActiveRecord::Base
  has_paper_trail skip: [:file_upload]

2.d. Turning PaperTrail Off

PaperTrail is on by default, but sometimes you don't want to record versions.

Per Process

Turn PaperTrail off for all threads in a ruby process.

PaperTrail.enabled = false

Do not use this in production unless you have a good understanding of threads vs. processes.

A legitimate use case is to speed up tests. See Testing below.

There is also a rails config option that does the same thing.

# in config/environments/test.rb
config.paper_trail.enabled = false

Per HTTP Request

PaperTrail.request(enabled: false) do
  # no versions created

This should satisfy most needs. In the rare case that you need to disable versioning for one model while keeping versioning enabled for other models, use:

# ...
PaperTrail.request.enabled_for_model?(Banana) # => true

For this rare use case, there is no convenient way to pass a block.

PaperTrail installs a callback in your rails controllers. The installed callback will call paper_trail_enabled_for_controller, which you can override.

class ApplicationController < ActionController::Base
  def paper_trail_enabled_for_controller
    # Don't omit `super` without a good reason.
    super && request.user_agent != 'Disable User-Agent'

Because you are unable to controll the order of callback execution, this technique is not recommended, but is preserved for backwards compatability.

It would be better to install your own callback and use PaperTrail.request.enabled= as you see fit.

Per Class


This setting, as with all PaperTrail.request settings, affects only the current request, not all threads.

Per Method

You can call a method without creating a new version using without_versioning. It takes either a method name as a symbol:

@widget.paper_trail.without_versioning :destroy

Or a block:

@widget.paper_trail.without_versioning do
  @widget.update_attributes name: 'Ford'

During without_versioning, PaperTrail is disabled for the whole model (e.g. Widget), not just for the instance (e.g. @widget).

2.e. Limiting the Number of Versions Created

Configure version_limit to cap the number of versions saved per record. This does not apply to create events.

# Limit: 4 versions per record (3 most recent, plus a `create` event)
PaperTrail.config.version_limit = 3
# Remove the limit
PaperTrail.config.version_limit = nil

3. Working With Versions

3.a. Reverting And Undeleting A Model

PaperTrail makes reverting to a previous version easy:

widget = Widget.find 42
widget.update_attributes name: 'Blah blah'
# Time passes....
widget = widget.paper_trail.previous_version  # the widget as it was before the update                                   # reverted

Alternatively you can find the version at a given time:

widget = widget.paper_trail.version_at(  # the widget as it was one day ago                                        # reverted

Note version_at gives you the object, not a version, so you don't need to call reify.

Undeleting is just as simple:

widget = Widget.find(42)
# Time passes....
versions = widget.versions    # versions ordered by versions.created_at, ascending
widget = versions.last.reify  # the widget as it was before destruction                   # the widget lives!

You could even use PaperTrail to implement an undo system; Ryan Bates has!

If your model uses optimistic locking don't forget to increment your lock_version before saving or you'll get a StaleObjectError.

3.b. Navigating Versions

You can call previous_version and next_version on an item to get it as it was/became. Note that these methods reify the item for you.

live_widget = Widget.find 42
live_widget.versions.length                       # 4, for example
widget = live_widget.paper_trail.previous_version # => widget == live_widget.versions.last.reify
widget = widget.paper_trail.previous_version      # => widget == live_widget.versions[-2].reify
widget = widget.paper_trail.next_version          # => widget == live_widget.versions.last.reify
widget.paper_trail.next_version                   # live_widget

If instead you have a particular version of an item you can navigate to the previous and next versions.

widget = Widget.find 42
version = widget.versions[-2]    # assuming widget has several versions
previous = version.previous
next =

You can find out which of an item's versions yours is:

current_version_number = version.index    # 0-based

If you got an item by reifying one of its versions, you can navigate back to the version it came from:

latest_version = Widget.find(42).versions.last
widget = latest_version.reify
widget.version == latest_version    # true

You can find out whether a model instance is the current, live one -- or whether it came instead from a previous version -- with live?:

widget = Widget.find 42                        # true
widget = widget.paper_trail.previous_version                        # false

And you can perform WHERE queries for object versions based on attributes:

# Find versions that meet these criteria.
PaperTrail::Version.where_object(content: 'Hello', title: 'Article')

# Find versions before and after attribute `atr` had value `v`:
PaperTrail::Version.where_object_changes(atr: 'v')

Using where_object_changes to read YAML from a text column was deprecated in 8.1.0, and will now raise an error.

3.c. Diffing Versions

There are two scenarios: diffing adjacent versions and diffing non-adjacent versions.

The best way to diff adjacent versions is to get PaperTrail to do it for you. If you add an object_changes text column to your versions table, either at installation time with the rails generate paper_trail:install --with-changes option or manually, PaperTrail will store the changes diff (excluding any attributes PaperTrail is ignoring) in each update version. You can use the version.changeset method to retrieve it. For example:

widget = Widget.create name: 'Bob'
# {
#   "name"=>[nil, "Bob"],
#   "created_at"=>[nil, 2015-08-10 04:10:40 UTC],
#   "updated_at"=>[nil, 2015-08-10 04:10:40 UTC],
#   "id"=>[nil, 1]
# }
widget.update_attributes name: 'Robert'
# {
#   "name"=>["Bob", "Robert"],
#   "updated_at"=>[2015-08-10 04:13:19 UTC, 2015-08-10 04:13:19 UTC]
# }
# {}

The object_changes are only stored for creation and updates, not when an object is destroyed.

Please be aware that PaperTrail doesn't use diffs internally. When I designed PaperTrail I wanted simplicity and robustness so I decided to make each version of an object self-contained. A version stores all of its object's data, not a diff from the previous version. This means you can delete any version without affecting any other.

To diff non-adjacent versions you'll have to write your own code. These libraries may help:

For diffing two strings:

  • htmldiff: expects but doesn't require HTML input and produces HTML output. Works very well but slows down significantly on large (e.g. 5,000 word) inputs.
  • differ: expects plain text input and produces plain text/coloured/HTML/any output. Can do character-wise, word-wise, line-wise, or arbitrary-boundary-string-wise diffs. Works very well on non-HTML input.
  • diff-lcs: old-school, line-wise diffs.

For diffing two ActiveRecord objects:

3.d. Deleting Old Versions

Over time your versions table will grow to an unwieldy size. Because each version is self-contained (see the Diffing section above for more) you can simply delete any records you don't want any more. For example:

sql> delete from versions where created_at < 2010-06-01;
PaperTrail::Version.delete_all ['created_at < ?', 1.week.ago]

4. Saving More Information About Versions

4.a. Finding Out Who Was Responsible For A Change

Set PaperTrail.request.whodunnit=, and that value will be stored in the version's whodunnit column.

PaperTrail.request.whodunnit = 'Andy Stewart'
widget.update_attributes name: 'Wibble'
widget.versions.last.whodunnit # Andy Stewart

Setting whodunnit to a Proc

whodunnit= also accepts a Proc, in the rare case that lazy evaluation is required.

PaperTrail.request.whodunnit = proc do
  caller.first{ |c| c.starts_with? Rails.root.to_s }

Because lazy evaluation can be hard to troubleshoot, this is not recommended for common use.

Setting whodunnit Temporarily

To set whodunnit temporarily, for the duration of a block, use PaperTrail.request:

PaperTrail.request(whodunnit: 'Dorian MariƩ') do
  widget.update_attributes name: 'Wibble'

Setting whodunnit with a controller callback

If your controller has a current_user method, PaperTrail provides a callback that will assign to whodunnit.

class ApplicationController
  before_action :set_paper_trail_whodunnit

You may want set_paper_trail_whodunnit to call a different method to find out who is responsible. To do so, override the user_for_paper_trail method in your controller like this:

class ApplicationController
  def user_for_paper_trail
    logged_in? ? : 'Public user'  # or whatever

See also: Setting whodunnit in the rails console

Terminator and Originator

A version's whodunnit column tells us who changed the object, causing the version to be stored. Because a version stores the object as it looked before the change (see the table above), whodunnit tells us who stopped the object looking like this -- not who made it look like this. Hence whodunnit is aliased as terminator.

To find out who made a version's object look that way, use version.paper_trail_originator. And to find out who made a "live" object look like it does, call paper_trail_originator on the object.

widget = Widget.find 153                    # assume widget has 0 versions
PaperTrail.request.whodunnit = 'Alice'
widget.update_attributes name: 'Yankee'
widget.paper_trail.originator               # 'Alice'
PaperTrail.request.whodunnit = 'Bob'
widget.update_attributes name: 'Zulu'
widget.paper_trail.originator               # 'Bob'
first_version, last_version = widget.versions.first, widget.versions.last
first_version.whodunnit                     # 'Alice'
first_version.paper_trail_originator        # nil
first_version.terminator                    # 'Alice'
last_version.whodunnit                      # 'Bob'
last_version.paper_trail_originator         # 'Alice'
last_version.terminator                     # 'Bob'

Storing an ActiveRecord globalid in whodunnit

If you would like whodunnit to return an ActiveRecord object instead of a string, please try the paper_trail-globalid gem.

4.b. Associations

To track and reify associations, use paper_trail-association_tracking (PT-AT).

From 2014 to 2018, association tracking was part of PT core as an experimental feature, but many issues were discovered. To attract new volunteers to address these issues, PT-AT was extracted (see

Even though this has always been an experimental feature, we don't want the extraction of PT-AT to be a breaking change. In PT 9, PT-AT was kept as a runtime dependency. In PT 10, it is a development dependency, so if you use it you must add it to your own Gemfile. We will keep PT-AT as a development dependency and continue running the existing tests related to association tracking for as long as is practical.

4.c. Storing Metadata

You can add your own custom columns to your versions table. Values can be given using Model Metadata or Controller Metadata.

Model Metadata

You can specify metadata in the model using has_paper_trail(meta:).

class Article < ActiveRecord::Base
  belongs_to :author
    meta: {
      author_id: :author_id, # model attribute
      word_count: :count_words, # arbitrary model method
      answer: 42, # scalar value
      editor: proc { |article| article.editor.full_name } # a Proc
  def count_words

Metadata from Controllers

You can also store any information you like from your controller. Override the info_for_paper_trail method in your controller to return a hash whose keys correspond to columns in your versions table.

class ApplicationController
  def info_for_paper_trail
    { ip: request.remote_ip, user_agent: request.user_agent }

Advantages of Metadata

Why would you do this? In this example, author_id is an attribute of Article and PaperTrail will store it anyway in a serialized form in the object column of the version record. But let's say you wanted to pull out all versions for a particular author; without the metadata you would have to deserialize (reify) each version object to see if belonged to the author in question. Clearly this is inefficient. Using the metadata you can find just those versions you want:

PaperTrail::Version.where(author_id: author_id)

Metadata can Override PaperTrail Columns

Experts only. Metadata will override the normal values that PT would have inserted into its own columns.

PT Column How bad of an idea? Alternative
item_type terrible idea
item_id terrible idea
event meh paper_trail_event
whodunnit meh PaperTrail.request.whodunnit=
object a little dangerous
object_changes a little dangerous

5. ActiveRecord

5.a. Single Table Inheritance (STI)

PaperTrail supports Single Table Inheritance, and even supports an un-versioned base model, as of 23ffbdc7e1.

class Fruit < ActiveRecord::Base
  # un-versioned base model
class Banana < Fruit

A change in what item_type stores for subclassed models was introduced in PR#1108, recording the subclass name instead of the base class. This simplifies reifying through associations, and allows for a change in PT-AT that fixes issue 594.

For those that have existing version data from STI models, item_type can be updated by using a generator, rails generate paper_trail:update_sti. More information is found in section 5.c. Generators.

5.b. Configuring the versions Association

You may configure the name of the versions association by passing a different name to has_paper_trail.

class Post < ActiveRecord::Base
  has_paper_trail class_name: 'Version', versions: :drafts
end # => NoMethodError

Overriding (instead of configuring) the versions method is not supported. Overriding associations is not recommended in general.

5.c. Generators


Used to set up PT for the first time. Writes, but does not run, a migration file. Creates an initializer for configuration. The migration adds the versions table.

  rails generate paper_trail:install --help
  rails generate paper_trail:install [options]

  # Includes the `object_changes` column, for storing changeset (diff) with each version
  [--with-changes], [--no-with-changes]

Runtime options:
  -f, [--force]                    # Overwrite files that already exist
  -p, [--pretend], [--no-pretend]  # Run but do not make any changes
  -q, [--quiet], [--no-quiet]      # Suppress status output
  -s, [--skip], [--no-skip]        # Skip files that already exist


Updates versions.item_type from base class name to subclass name. If you use STI, and you have records in versions created prior to #1108, you must run this migration.

Writes, but does not run, a migration. The migration scans existing data in the versions table to identify models which use STI. Upon finding each entry which refers to an object from a subclassed model, versions.item_type is updated to reflect the subclass name, providing full compatibility with how the versions association works after PR#1108. In order to more quickly update a large volume of version records, updates are batched to affect 100 records at a time.

Hints can be used in rare cases when the STI structure is modified over time such as when establishing additional intermediary inheritance structure, pointing to a new base class, abandoning STI entirely, or changing the name of the inheritance_column. The vast majority of users will probably never need to supply hints when using this generator.

Let's give an example where the inheritance_column is changed, say originally there is an Animal class that uses the default type column, but after generating 500 Bird and Cat objects is modified to instead use the species column with a line like this:

Animal.inheritance_column = "species"

With this change a hint can be used to build the migration in a way that indicates how to treat the older records. For those first 500 Animal objects, the item_type should be derived from type, and for the rest, from species. We would need to research the ID numbers for versions that utilise type in the object and object_changes data. In this example let's say that those first 500 Animals had version IDs between 249 and 1124. These objects would have been created having an item_type of Animal prior to PR#1108. Of course versions representing all manner of other changes would also be intermingled along with creates and updates for Animal objects. Perhaps another STI model exists for Car that inherits from Vehicle, and 50 Car objects were also built around the same timeframe as Bird and Cat objects, with various creates and updates for those being referened in versions with IDs from 382 to 516. For Vehicle let's say that initially the inheritance_column was set to kind, but then it changed to something else after ID 516 in the versions table. In this situation, to properly use the generator then these hints can be supplied:

rails generate update_sti Animal(type):249..1124 Vehicle(kind):382..516

The resulting migration will include these lines near the top:

# Versions of item_type "Animal" with IDs between 249 and 1124 will be updated based on `type`
# Versions of item_type "Vehicle" with IDs between 382 and 516 will be updated based on `kind`
hints = {"Animal"=>{249..1124=>"type"}, "Vehicle"=>{382..516=>"kind"}}

It is important to note that the IDs are not those of the Bird, Cat, or Car objects, but rather the IDs from the create, update, and destroy entries in the versions table. As well, these hints are only needed in situations where the inheritance_column has changed at some point in time, or in cases where the entire STI structure is modified or is abandoned. It ultimately facilitates better reporting for historic subclassed items, and allows PT-AT to properly reify these historic objects through associations.

Once you have run the migration, please inform PaperTrail, so that it won't warn you about this.

# config/initializers/paper_trail.rb
::PaperTrail.config.i_have_updated_my_existing_item_types = true

5.d. Protected Attributes

As of version 6, PT no longer supports rails 3 or the protected_attributes gem. If you are still using them, you may use PT 5 or lower. We recommend upgrading to strong_parameters as soon as possible.

If you must use protected_attributes for now, and want to use PT > 5, you can reopen PaperTrail::Version and add the following attr_accessible fields:

# app/models/paper_trail/version.rb
module PaperTrail
  class Version < ActiveRecord::Base
    include PaperTrail::VersionConcern
    attr_accessible :item_type, :item_id, :event, :whodunnit, :object, :object_changes, :created_at

This unsupported workaround has been tested with protected_attributes 1.0.9 / rails 4.2.8 / paper_trail 7.0.3.

6. Extensibility

6.a. Custom Version Classes

You can specify custom version subclasses with the :class_name option:

class PostVersion < PaperTrail::Version
  # custom behaviour, e.g:
  self.table_name = :post_versions

class Post < ActiveRecord::Base
  has_paper_trail class_name: 'PostVersion'

Unlike ActiveRecord's class_name, you'll have to supply the complete module path to the class (e.g. Foo::BarVersion if your class is inside the module Foo).


  1. For models which have a lot of versions, storing each model's versions in a separate table can improve the performance of certain database queries.
  2. Store different version metadata for different models.


If you are using Postgres, you should also define the sequence that your custom version class will use:

class PostVersion < PaperTrail::Version
  self.table_name = :post_versions
  self.sequence_name = :post_versions_id_seq

If you only use custom version classes and don't have a versions table, you must let ActiveRecord know that the PaperTrail::Version class is an abstract_class.

# app/models/paper_trail/version.rb
module PaperTrail
  class Version < ActiveRecord::Base
    include PaperTrail::VersionConcern
    self.abstract_class = true

You can also specify custom names for the versions and version associations. This is useful if you already have versions or/and version methods on your model. For example:

class Post < ActiveRecord::Base
  has_paper_trail versions: :paper_trail_versions,
                  version:  :paper_trail_version

  # Existing versions method.  We don't want to clash.
  def versions
    # ...

  # Existing version method.  We don't want to clash.
  def version
    # ...

6.b. Custom Serializer

By default, PaperTrail stores your changes as a YAML dump. You can override this with the serializer config option:

PaperTrail.serializer = MyCustomSerializer

A valid serializer is a module (or class) that defines a load and dump method. These serializers are included in the gem for your convenience:

PostgreSQL JSON column type support

If you use PostgreSQL, and would like to store your object (and/or object_changes) data in a column of type json or type jsonb, specify json instead of text for these columns in your migration:

create_table :versions do |t|
  # ...
  t.json :object          # Full object changes
  t.json :object_changes  # Optional column-level changes
  # ...

If you use the PostgreSQL json or jsonb column type, you do not need to specify a PaperTrail.serializer.

Convert existing YAML data to JSON

If you've been using PaperTrail for a while with the default YAML serializer and you want to switch to JSON or JSONB, you're in a bit of a bind because there's no automatic way to migrate your data. The first (slow) option is to loop over every record and parse it in Ruby, then write to a temporary column:

add_column :versions, :new_object, :jsonb # or :json

PaperTrail::Version.find_each do |version|
  version.update_column :new_object, YAML.load(version.object) if version.object

remove_column :versions, :object
rename_column :versions, :new_object, :object

This technique can be very slow if you have a lot of data. Though slow, it is safe in databases where transactions are protected against DDL, such as Postgres. In databases without such protection, such as MySQL, a table lock may be necessary.

If the above technique is too slow for your needs, and you're okay doing without PaperTrail data temporarily, you can create the new column without converting the data.

rename_column :versions, :object, :old_object
add_column :versions, :object, :jsonb # or :json

After that migration, your historical data still exists as YAML, and new data will be stored as JSON. Next, convert records from YAML to JSON using a background script.

PaperTrail::Version.where.not(old_object: nil).find_each do |version|
  version.update_columns old_object: nil, object: YAML.load(version.old_object)

Finally, in another migration, remove the old column.

remove_column :versions, :old_object

If you use the optional object_changes column, don't forget to convert it also, using the same technique.

Convert a Column from Text to JSON

If your object column already contains JSON data, and you want to change its data type to json or jsonb, you can use the following DDL. Of course, if your object column contains YAML, you must first convert the data to JSON (see above) before you can change the column type.

Using SQL:

alter table versions
alter column object type jsonb
using object::jsonb;

Using ActiveRecord:

class ConvertVersionsObjectToJson < ActiveRecord::Migration
  def up
    change_column :versions, :object, 'jsonb USING object::jsonb'

  def down
    change_column :versions, :object, 'text USING object::text'

6.c. Custom Object Changes

By default, PaperTrail stores object changes in a before/after array of objects containing keys of columns that have changed in that particular version. You can override this behaviour by using the object_changes_adapter config option:

PaperTrail.config.object_changes_adapter =

A valid adapter is a class that contains the following methods:

  1. diff: Returns the changeset in the desired format given the changeset in the original format
  2. load_changeset: Returns the changeset for a given version object
  3. where_object_changes: Returns the records resulting from the given hash of attributes.

For an example of such an implementation, see paper_trail-hashdiff

6.d. Excluding the Object Column

The object column ends up storing a lot of duplicate data if you have models that have many columns, and that are updated many times. You can save ~50% of storage space by removing the column from the versions table. It's important to note that this will disable reify and where_object.

7. Testing

You may want to turn PaperTrail off to speed up your tests. See Turning PaperTrail Off above.

7.a. Minitest

First, disable PT for the entire ruby process.

# in config/environments/test.rb
config.after_initialize do
  PaperTrail.enabled = false

Then, to enable PT for specific tests, you can add a with_versioning test helper method.

# in test/test_helper.rb
def with_versioning
  was_enabled = PaperTrail.enabled?
  was_enabled_for_request = PaperTrail.request.enabled?
  PaperTrail.enabled = true
  PaperTrail.request.enabled = true
    PaperTrail.enabled = was_enabled
    PaperTrail.request.enabled = was_enabled_for_request

Then, use the helper in your tests.

test 'something that needs versioning' do
  with_versioning do
    # your test

7.b. RSpec

PaperTrail provides a helper, paper_trail/frameworks/rspec.rb, that works with RSpec to make it easier to control when PaperTrail is enabled during testing.

# spec/rails_helper.rb
ENV["RAILS_ENV"] ||= 'test'
require 'spec_helper'
require File.expand_path("../../config/environment", __FILE__)
require 'rspec/rails'
# ...
require 'paper_trail/frameworks/rspec'

With the helper loaded, PaperTrail will be turned off for all tests by default. To enable PaperTrail for a test you can either wrap the test in a with_versioning block, or pass in versioning: true option to a spec block.

describe 'RSpec test group' do
  it 'by default, PaperTrail will be turned off' do
    expect(PaperTrail).to_not be_enabled

  with_versioning do
    it 'within a `with_versioning` block it will be turned on' do
      expect(PaperTrail).to be_enabled

  it 'can be turned on at the `it` or `describe` level', versioning: true do
    expect(PaperTrail).to be_enabled

The helper will also reset whodunnit to nil before each test to help prevent data spillover between tests. If you are using PaperTrail with Rails, the helper will automatically set the PaperTrail.controller_info value to {} as well, again, to help prevent data spillover between tests.

There is also a be_versioned matcher provided by PaperTrail's RSpec helper which can be leveraged like so:

class Widget < ActiveRecord::Base

describe Widget do
  it 'is not versioned by default' do
    is_expected.to_not be_versioned

  describe 'add versioning to the `Widget` class' do
    before(:all) do
      class Widget < ActiveRecord::Base

    it 'enables paper trail' do be_versioned


The have_a_version_with matcher makes assertions about versions using where_object, based on the object column.

describe '`have_a_version_with` matcher' do
  it 'is possible to do assertions on version attributes' do
    widget.update_attributes!(name: 'Leonard', an_integer: 1)
    widget.update_attributes!(name: 'Tom')
    widget.update_attributes!(name: 'Bob')
    expect(widget).to have_a_version_with name: 'Leonard', an_integer: 1
    expect(widget).to have_a_version_with an_integer: 1
    expect(widget).to have_a_version_with name: 'Tom'

The have_a_version_with_changes matcher makes assertions about versions using where_object_changes, based on the optional object_changes column.

describe '`have_a_version_with_changes` matcher' do
  it 'is possible to do assertions on version changes' do
    widget.update_attributes!(name: 'Leonard', an_integer: 1)
    widget.update_attributes!(name: 'Tom')
    widget.update_attributes!(name: 'Bob')
    expect(widget).to have_a_version_with_changes name: 'Leonard', an_integer: 2
    expect(widget).to have_a_version_with_changes an_integer: 2
    expect(widget).to have_a_version_with_changes name: 'Bob'

For more examples of the RSpec matchers, see the Widget spec

7.c. Cucumber

PaperTrail provides a helper for Cucumber that works similar to the RSpec helper. If you want to use the helper, you will need to require in your cucumber helper like so:

# features/support/env.rb

ENV["RAILS_ENV"] ||= 'cucumber'
require File.expand_path(File.dirname(__FILE__) + '/../../config/environment')
# ...
require 'paper_trail/frameworks/cucumber'

When the helper is loaded, PaperTrail will be turned off for all scenarios by a before hook added by the helper by default. When you want to enable PaperTrail for a scenario, you can wrap code in a with_versioning block in a step, like so:

Given /I want versioning on my model/ do
  with_versioning do
    # PaperTrail will be turned on for all code inside of this block

The helper will also reset the whodunnit value to nil before each test to help prevent data spillover between tests. If you are using PaperTrail with Rails, the helper will automatically set the PaperTrail.controller_info value to {} as well, again, to help prevent data spillover between tests.

7.d. Spork

If you want to use the RSpec or Cucumber helpers with Spork, you will need to manually require the helper(s) in your prefork block on your test helper, like so:

# spec/rails_helper.rb

require 'spork'

Spork.prefork do
  # This file is copied to spec/ when you run 'rails generate rspec:install'
  ENV["RAILS_ENV"] ||= 'test'
  require 'spec_helper'
  require File.expand_path("../../config/environment", __FILE__)
  require 'rspec/rails'
  require 'paper_trail/frameworks/rspec'
  require 'paper_trail/frameworks/cucumber'
  # ...

7.e. Zeus or Spring

If you want to use the RSpec or Cucumber helpers with Zeus or Spring, you will need to manually require the helper(s) in your test helper, like so:

# spec/rails_helper.rb

ENV["RAILS_ENV"] ||= 'test'
require 'spec_helper'
require File.expand_path("../../config/environment", __FILE__)
require 'rspec/rails'
require 'paper_trail/frameworks/rspec'

8. PaperTrail Plugins

9. Integration with Other Libraries



Please use GitHub's issue tracker.


Created by Andy Stewart in 2010, maintained since 2012 by Ben Atkins, since 2015 by Jared Beck, with contributions by over 150 people.


See our contribution guidelines


Intellectual Property

Copyright (c) 2011 Andy Stewart ([email protected]). Released under the MIT licence.