This guide teaches you how to hook into the life cycle of your Active Record objects. You will learn how to validate the state of objects before they go into the database, and how to perform custom operations at certain points in the object life cycle.
During the normal operation of a Rails application, objects may be created, updated, and destroyed. Active Record provides hooks into this <em>object life cycle</em> so that you can control your application and its data.
Validations allow you to ensure that only valid data is stored in your database. Callbacks and observers allow you to trigger logic before or after an alteration of an object's state.
Validations are used to ensure that only valid data is saved into your database. For example, it may be important to your application to ensure that every user provides a valid email address and mailing address.
There are several ways to validate data before it is saved into your database, including native database constraints, client-side validations, controller-level validations, and model-level validations:
* Database constraints and/or stored procedures make the validation mechanisms database-dependent and can make testing and maintenance more difficult. However, if your database is used by other applications, it may be a good idea to use some constraints at the database level. Additionally, database-level validations can safely handle some things (such as uniqueness in heavily-used tables) that can be difficult to implement otherwise.
* Client-side validations can be useful, but are generally unreliable if used alone. If they are implemented using JavaScript, they may be bypassed if JavaScript is turned off in the user's browser. However, if combined with other techniques, client-side validation can be a convenient way to provide users with immediate feedback as they use your site.
* Controller-level validations can be tempting to use, but often become unwieldy and difficult to test and maintain. Whenever possible, it's a good idea to [keep your controllers skinny](http://weblog.jamisbuck.org/2006/10/18/skinny-controller-fat-model), as it will make your application a pleasure to work with in the long run.
* Model-level validations are the best way to ensure that only valid data is saved into your database. They are database agnostic, cannot be bypassed by end users, and are convenient to test and maintain. Rails makes them easy to use, provides built-in helpers for common needs, and allows you to create your own validation methods as well.
There are two kinds of Active Record objects: those that correspond to a row inside your database and those that do not. When you create a fresh object, for example using the `new` method, that object does not belong to the database yet. Once you call `save` upon that object it will be saved into the appropriate database table. Active Record uses the `new_record?` instance method to determine whether an object is already in the database or not. Consider the following simple Active Record class:
Creating and saving a new record will send an SQL `INSERT` operation to the database. Updating an existing record will send an SQL `UPDATE` operation instead. Validations are typically run before these commands are sent to the database. If any validations fail, the object will be marked as invalid and Active Record will not perform the `INSERT` or `UPDATE` operation. This helps to avoid storing an invalid object in the database. You can choose to have specific validations run when an object is created, saved, or updated.
CAUTION: There are many ways to change the state of an object in the database. Some methods will trigger validations, but some will not. This means that it's possible to save an object in the database in an invalid state if you aren't careful.
The bang versions (e.g. `save!`) raise an exception if the record is invalid. The non-bang versions don't: `save` and `update_attributes` return `false`, `create` and `update` just return the objects.
To verify whether or not an object is valid, Rails uses the `valid?` method. You can also use this method on your own. `valid?` triggers your validations and returns true if no errors were found in the object, and false otherwise.
After Active Record has performed validations, any errors found can be accessed through the `errors` instance method, which returns a collection of errors. By definition, an object is valid if this collection is empty after running validations.
`invalid?` is simply the inverse of `valid?`. `invalid?` triggers your validations, returning true if any errors were found in the object, and false otherwise.
To verify whether or not a particular attribute of an object is valid, you can use `errors[:attribute]`. It returns an array of all the errors for `:attribute`. If there are no errors on the specified attribute, an empty array is returned.
This method is only useful _after_ validations have been run, because it only inspects the errors collection and does not trigger validations itself. It's different from the `ActiveRecord::Base#invalid?` method explained above because it doesn't verify the validity of the object as a whole. It only checks to see whether there are errors found on an individual attribute of the object.
We'll cover validation errors in greater depth in the [Working with Validation Errors](#working-with-validation-errors) section. For now, let's turn to the built-in validation helpers that Rails provides by default.
Active Record offers many pre-defined validation helpers that you can use directly inside your class definitions. These helpers provide common validation rules. Every time a validation fails, an error message is added to the object's `errors` collection, and this message is associated with the attribute being validated.
Each helper accepts an arbitrary number of attribute names, so with a single line of code you can add the same kind of validation to several attributes.
All of them accept the `:on` and `:message` options, which define when the validation should be run and what message should be added to the `errors` collection if it fails, respectively. The `:on` option takes one of the values `:save` (the default), `:create` or `:update`. There is a default error message for each one of the validation helpers. These messages are used when the `:message` option isn't specified. Let's take a look at each one of the available helpers.
Validates that a checkbox on the user interface was checked when a form was submitted. This is typically used when the user needs to agree to your application's terms of service, confirm reading some text, or any similar concept. This validation is very specific to web applications and this 'acceptance' does not need to be recorded anywhere in your database (if you don't have a field for it, the helper will just create a virtual attribute).
You should use this helper when your model has associations with other models and they also need to be validated. When you try to save your object, `valid?` will be called upon each one of the associated objects.
The default error message for `validates_associated` is "_is invalid_". Note that each associated object will contain its own `errors` collection; errors do not bubble up to the calling model.
You should use this helper when you have two text fields that should receive exactly the same content. For example, you may want to confirm an email address or a password. This validation creates a virtual attribute whose name is the name of the field that has to be confirmed with "_confirmation" appended.
This check is performed only if `email_confirmation` is not `nil`. To require confirmation, make sure to add a presence check for the confirmation attribute (we'll take a look at `presence` later on this guide):
The `exclusion` helper has an option `:in` that receives the set of values that will not be accepted for the validated attributes. The `:in` option has an alias called `:within` that you can use for the same purpose, if you'd like to. This example uses the `:message` option to show how you can include the attribute's value.
The `inclusion` helper has an option `:in` that receives the set of values that will be accepted. The `:in` option has an alias called `:within` that you can use for the same purpose, if you'd like to. The previous example uses the `:message` option to show how you can include the attribute's value.
The default error messages depend on the type of length validation being performed. You can personalize these messages using the `:wrong_length`, `:too_long`, and `:too_short` options and `%{count}` as a placeholder for the number corresponding to the length constraint being used. You can still use the `:message` option to specify an error message.
Note that the default error messages are plural (e.g., "is too short (minimum is %{count} characters)"). For this reason, when `:minimum` is 1 you should provide a personalized message or use `validates_presence_of` instead. When `:in` or `:within` have a lower limit of 1, you should either provide a personalized message or call `presence` prior to `length`.
This helper validates that your attributes have only numeric values. By default, it will match an optional sign followed by an integral or floating point number. To specify that only integral numbers are allowed set `:only_integer` to true.
*`:greater_than` - Specifies the value must be greater than the supplied value. The default error message for this option is "_must be greater than %{count}_".
*`:greater_than_or_equal_to` - Specifies the value must be greater than or equal to the supplied value. The default error message for this option is "_must be greater than or equal to %{count}_".
*`:equal_to` - Specifies the value must be equal to the supplied value. The default error message for this option is "_must be equal to %{count}_".
*`:less_than` - Specifies the value must be less than the supplied value. The default error message for this option is "_must be less than %{count}_".
*`:less_than_or_equal_to` - Specifies the value must be less than or equal the supplied value. The default error message for this option is "_must be less than or equal to %{count}_".
*`:odd` - Specifies the value must be an odd number if set to true. The default error message for this option is "_must be odd_".
*`:even` - Specifies the value must be an even number if set to true. The default error message for this option is "_must be even_".
This helper validates that the specified attributes are not empty. It uses the `blank?` method to check if the value is either `nil` or a blank string, that is, a string that is either empty or consists of whitespace.
If you want to be sure that an association is present, you'll need to test whether the foreign key used to map the association is present, and not the associated object itself.
If you validate the presence of an object associated via a `has_one` or `has_many` relationship, it will check that the object is neither `blank?` nor `marked_for_destruction?`.
Since `false.blank?` is true, if you want to validate the presence of a boolean field you should use `validates :field_name, :inclusion => { :in => [true, false] }`.
This helper validates that the attribute's value is unique right before the object gets saved. It does not create a uniqueness constraint in the database, so it may happen that two different database connections create two records with the same value for a column that you intend to be unique. To avoid that, you must create a unique index in your database.
There is also a `:case_sensitive` option that you can use to define whether the uniqueness constraint will be case sensitive or not. This option defaults to true.
The `validates_with` helper takes a class, or a list of classes to use for validation. There is no default error message for `validates_with`. You must manually add errors to the record's errors collection in the validator class.
Like all other validations, `validates_with` takes the `:if`, `:unless` and `:on` options. If you pass any other options, it will send those options to the validator class as `options`:
This helper validates attributes against a block. It doesn't have a predefined validation function. You should create one using a block, and every attribute passed to `validates_each` will be tested against it. In the following example, we don't want names and surnames to begin with lower case.
The block receives the record, the attribute's name and the attribute's value. You can do anything you like to check for valid data within the block. If your validation fails, you should add an error message to the model, therefore making it invalid.
The `:allow_blank` option is similar to the `:allow_nil` option. This option will let validation pass if the attribute's value is `blank?`, like `nil` or an empty string for example.
As you've already seen, the `:message` option lets you specify the message that will be added to the `errors` collection when validation fails. When this option is not used, Active Record will use the respective default error message for each validation helper.
The `:on` option lets you specify when the validation should happen. The default behavior for all the built-in validation helpers is to be run on save (both when you're creating a new record and when you're updating it). If you want to change it, you can use `:on => :create` to run the validation only when a new record is created or `:on => :update` to run the validation only when a record is updated.
Sometimes it will make sense to validate an object just when a given predicate is satisfied. You can do that by using the `:if` and `:unless` options, which can take a symbol, a string, a `Proc` or an `Array`. You may use the `:if` option when you want to specify when the validation **should** happen. If you want to specify when the validation **should not** happen, then you may use the `:unless` option.
You can associate the `:if` and `:unless` options with a symbol corresponding to the name of a method that will get called right before validation happens. This is the most commonly used option.
You can also use a string that will be evaluated using `eval` and needs to contain valid Ruby code. You should use this option only when the string represents a really short condition.
Finally, it's possible to associate `:if` and `:unless` with a `Proc` object which will be called. Using a `Proc` object gives you the ability to write an inline condition instead of a separate method. This option is best suited for one-liners.
On the other hand, when multiple conditions define whether or not a validation should happen, an `Array` can be used. Moreover, you can apply both `:if` and `:unless` to the same validation.
Custom validators are classes that extend `ActiveModel::Validator`. These classes must implement a `validate` method which takes a record as an argument and performs the validation on it. The custom validator is called using the `validates_with` method.
The easiest way to add custom validators for validating individual attributes is with the convenient `ActiveModel::EachValidator`. In this case, the custom validator class must implement a `validate_each` method which takes three arguments: record, attribute and value which correspond to the instance, the attribute to be validated and the value of the attribute in the passed instance.
You can also create methods that verify the state of your models and add messages to the `errors` collection when they are invalid. You must then register these methods by using the `validate` class method, passing in the symbols for the validation methods' names.
By default such validations will run every time you call `valid?`. It is also possible to control when to run these custom validations by giving an `:on` option to the `validate` method, with either: `:create` or `:update`.
You can even create your own validation helpers and reuse them in several different models. For example, an application that manages surveys may find it useful to express that a certain field corresponds to a set of choices:
Simply reopen `ActiveRecord::Base` and define a class method like that. You'd typically put this code somewhere in `config/initializers`. You can use this helper like this:
In addition to the `valid?` and `invalid?` methods covered earlier, Rails provides a number of methods for working with the `errors` collection and inquiring about the validity of objects.
The following is a list of the most commonly used methods. Please refer to the `ActiveModel::Errors` documentation for a list of all the available methods.
Returns an instance of the class `ActiveModel::Errors` containing all errors. Each key is the attribute name and the value is an array of strings with all errors.
`errors[]` is used when you want to check the error messages for a specific attribute. It returns an array of strings with all error messages for the given attribute, each string with one error message. If there are no errors related to the attribute, it returns an empty array.
The `add` method lets you manually add messages that are related to particular attributes. You can use the `errors.full_messages` or `errors.to_a` methods to view the messages in the form they might be displayed to a user. Those particular messages get the attribute name prepended (and capitalized). `add` receives the name of the attribute you want to add the message to, and the message itself.
You can add error messages that are related to the object's state as a whole, instead of being related to a specific attribute. You can use this method when you want to say that the object is invalid, no matter the values of its attributes. Since `errors[:base]` is an array, you can simply add a string to it and it will be used as an error message.
The `clear` method is used when you intentionally want to clear all the messages in the `errors` collection. Of course, calling `errors.clear` upon an invalid object won't actually make it valid: the `errors` collection will now be empty, but the next time you call `valid?` or any method that tries to save this object to the database, the validations will run again. If any of the validations fail, the `errors` collection will be filled again.
When creating a form with the `form_for` helper, you can use the `error_messages` method on the form builder to render all failed validation messages for the current model instance.
NOTE: The appearance of the generated HTML will be different from the one shown, unless you have used scaffolding. See [Customizing the Error Messages CSS](#customizing-the-error-messages-css).
You can also use the `error_messages_for` helper to display the error messages of a model assigned to a view template. It is very similar to the previous example and will achieve exactly the same result.
The displayed text for each error message will always be formed by the capitalized name of the attribute that holds the error, followed by the error message itself.
Both the `form.error_messages` and the `error_messages_for` helpers accept options that let you customize the `div` element that holds the messages, change the header text, change the message below the header, and specify the tag used for the header element. For example,
If scaffolding was used, file `app/assets/stylesheets/scaffolds.css.scss` will have been generated automatically. This file defines the red-based styles you saw in the examples above.
By default, form fields with errors are displayed enclosed by a `div` element with the `field_with_errors` CSS class. However, it's possible to override that.
Below is a simple example where we change the Rails behavior to always display the error messages in front of each of the form fields in error. The error messages will be enclosed by a `span` element with a `validation-error` CSS class. There will be no `div` element enclosing the `input` element, so we get rid of that red border around the text field. You can use the `validation-error` CSS class to style it anyway you want.
Callbacks are methods that get called at certain moments of an object's life cycle. With callbacks it is possible to write code that will run whenever an Active Record object is created, saved, updated, deleted, validated, or loaded from the database.
In order to use the available callbacks, you need to register them. You can implement the callbacks as ordinary methods and use a macro-style class method to register them as callbacks:
The macro-style class methods can also receive a block. Consider using this style if the code inside your block is so short that it fits in a single line:
It is considered good practice to declare callback methods as protected or private. If left public, they can be called from outside of the model and violate the principle of object encapsulation.
Here is a list with all the available Active Record callbacks, listed in the same order in which they will get called during the respective operations:
WARNING. `after_save` runs both on create and update, but always _after_ the more specific callbacks `after_create` and `after_update`, no matter the order in which the macro calls were executed.
The `after_initialize` callback will be called whenever an Active Record object is instantiated, either by directly using `new` or when a record is loaded from the database. It can be useful to avoid the need to directly override your Active Record `initialize` method.
The `after_find` callback will be called whenever Active Record loads a record from the database. `after_find` is called before `after_initialize` if both are defined.
The `after_initialize` and `after_find` callbacks have no `before_*` counterparts, but they can be registered just like the other Active Record callbacks.
Just as with validations, it is also possible to skip callbacks. These methods should be used with caution, however, because important business rules and application logic may be kept in callbacks. Bypassing them without understanding the potential implications may lead to invalid data.
As you start registering new callbacks for your models, they will be queued for execution. This queue will include all your model's validations, the registered callbacks, and the database operation to be executed.
The whole callback chain is wrapped in a transaction. If any <em>before</em> callback method returns exactly `false` or raises an exception, the execution chain gets halted and a ROLLBACK is issued; <em>after</em> callbacks can only accomplish that by raising an exception.
WARNING. Raising an arbitrary exception may break code that expects `save` and its friends not to fail like that. The `ActiveRecord::Rollback` exception is thought precisely to tell Active Record a rollback is going on. That one is internally captured but not reraised.
Callbacks work through model relationships, and can even be defined by them. Suppose an example where a user has many posts. A user's posts should be destroyed if the user is destroyed. Let's add an `after_destroy` callback to the `User` model by way of its relationship to the `Post` model:
As with validations, we can also make the calling of a callback method conditional on the satisfaction of a given predicate. We can do this using the `:if` and `:unless` options, which can take a symbol, a string, a `Proc` or an `Array`. You may use the `:if` option when you want to specify under which conditions the callback **should** be called. If you want to specify the conditions under which the callback **should not** be called, then you may use the `:unless` option.
You can associate the `:if` and `:unless` options with a symbol corresponding to the name of a predicate method that will get called right before the callback. When using the `:if` option, the callback won't be executed if the predicate method returns false; when using the `:unless` option, the callback won't be executed if the predicate method returns true. This is the most common option. Using this form of registration it is also possible to register several different predicates that should be called to check if the callback should be executed.
You can also use a string that will be evaluated using `eval` and hence needs to contain valid Ruby code. You should use this option only when the string represents a really short condition:
Finally, it is possible to associate `:if` and `:unless` with a `Proc` object. This option is best suited when writing short validation methods, usually one-liners:
Sometimes the callback methods that you'll write will be useful enough to be reused by other models. Active Record makes it possible to create classes that encapsulate the callback methods, so it becomes very easy to reuse them.
When declared inside a class, as above, the callback methods will receive the model object as a parameter. We can now use the callback class in the model:
Note that we needed to instantiate a new `PictureFileCallbacks` object, since we declared our callback as an instance method. This is particularly useful if the callbacks make use of the state of the instantiated object. Often, however, it will make more sense to declare the callbacks as class methods:
Observers are similar to callbacks, but with important differences. Whereas callbacks can pollute a model with code that isn't directly related to its purpose, observers allow you to add the same functionality without changing the code of the model. For example, it could be argued that a `User` model should not include code to send registration confirmation emails. Whenever you use callbacks with code that isn't directly related to your model, you may want to consider creating an observer instead.
For example, imagine a `User` model where we want to send an email every time a new user is created. Because sending emails is not directly related to our model's purpose, we should create an observer to contain the code implementing this functionality.
Observers are conventionally placed inside of your `app/models` directory and registered in your application's `config/application.rb` file. For example, the `UserObserver` above would be saved as `app/models/user_observer.rb` and registered in `config/application.rb` this way:
As usual, settings in `config/environments` take precedence over those in `config/application.rb`. So, if you prefer that an observer doesn't run in all environments, you can simply register it in a specific environment instead.
By default, Rails will simply strip "Observer" from an observer's name to find the model it should observe. However, observers can also be used to add behavior to more than one model, and thus it is possible to explicitly specify the models that our observer should observe:
In this example, the `after_create` method will be called whenever a `Registration` or `User` is created. Note that this new `MailerObserver` would also need to be registered in `config/application.rb` in order to take effect:
There are two additional callbacks that are triggered by the completion of a database transaction: `after_commit` and `after_rollback`. These callbacks are very similar to the `after_save` callback except that they don't execute until after database changes have either been committed or rolled back. They are most useful when your active record models need to interact with external systems which are not part of the database transaction.
Consider, for example, the previous example where the `PictureFile` model needs to delete a file after the corresponding record is destroyed. If anything raises an exception after the `after_destroy` callback is called and the transaction rolls back, the file will have been deleted and the model will be left in an inconsistent state. For example, suppose that `picture_file_2` in the code below is not valid and the `save!` method raises an error.
The `after_commit` and `after_rollback` callbacks are guaranteed to be called for all models created, updated, or destroyed within a transaction block. If any exceptions are raised within one of these callbacks, they will be ignored so that they don't interfere with the other callbacks. As such, if your callback code could raise an exception, you'll need to rescue it and handle it appropriately within the callback.