Validating Data

Data is added to a database through update and insert queries. As part of these two operations, a ManagedObject<T> will ensure that its properties have valid values. For example, a Person object might ensure that its name starts with a capital letter and that its phone number has only numeric values. If one or more validation fails, the update or insert operation will fail and the data is not sent to the database. A validation failure will throw a QueryException that returns an HTTP response with error messaging to help the client correct their request.

The preferred way of setting a validation is to add Validate metadata to properties of a table definition. Here's an example of a validation that ensures a tweet is less than 140 characters:

class Tweet extends ManagedObject<_Tweet> implements _Tweet {}
class _Tweet {
  @primaryKey
  int id;

  @Validate.length(lessThan: 140)
  String message;
}

Built-in Validators

There are a handful of built-in validations for common operations. For example, it is common to apply a regular expression to a value to ensure it is formatted correctly or to restrict the possible values to a list of available options. Common validators are available as named constructors of the Validate class. Here is an example:

class _Story {
  @primaryKey
  int id;

  @Validate.oneOf(const ["started", "accepted", "rejected", "delivered"])
  String state;
}

A built-in validator is useful because it automatically generates an error message that is returned in an HTTP response. For example, the previous code snippet indicates that the state property must be one of the four listed strings. If a value other than one of those four strings is used, the error message returned to the HTTP client would be:

"The value `invalidValue` is not valid for `state`. Valid values are: 'started', 'accepted', 'rejected', 'delivered'.".

See the API reference for Validate and its named constructors for possible options.

Validate annotations on properties declared in a managed object subclass (transient properties) have no effect.

Validating Relationships

Validations are only valid for properties declared in a table definition. Validators applied to relationship properties are applied to the primary key of the related object (i.e. the foreign key value). Validation logic is only ran on the properties of the managed object being validated - validations on properties of a related object are not run. When validating a graph of managed objects, you must initiate validation on any related objects manually.

Validating Related Objects

The behavior of a validation is different when an object is being validated as a relationship. In other words, a validation applied to the primary key of an object likely requires different behavior when being applied to a foreign key reference to that object.

Custom Validators

There will be times where the built-in validators are not sufficient for your application's use case. You may create subclasses of Validate to provide custom validation behavior. For example, if there were a ValidatePhoneNumber class:

class _Person {
  @primaryKey
  int id;

  @ValidatePhoneNumber()
  String phoneNumber;
}

A subclass of Validate must override Validate.validate() and call the superclass' primary constructor when instantiated. Here's an example of that phone number validator:

class ValidatePhoneNumber extends Validate {
  ValidatePhoneNumber({bool onUpdate: true, bool onInsert: true}) :
    super(onUpdate: onUpdate, onInsert: onInsert);

  @override
  void validate(ValidationContext context, dynamic value) {  
    if (value.length != 15) {
      context.addError("must be 15 digits");      
    }

    if (containsNonNumericValues(value)) {
      context.addError("must contain characters 0-9 only.");      
    }
  }
}

If value is doesn't meet the validation criteria, this method adds an error string to the ValidationContext it is passed. Error messages should be brief and indicate the successful criteria that failed. Information about the property being validated will automatically be added to the error message, so you do not need to include that information. If the context has no errors at the end of validation, the validation succeeds; otherwise, it fails.

A ValidationContext also has information about the property being validated, and whether the validation is running for an object being inserted or an object being updated.

Validation Behavior

A property may have more than one Validate metadata. In this case, all of the validations for a property must pass. The order in which multiple validations are performed is undefined and subject to change. Here's an example of validations that ensure a property's value is 10 characters long and only contains 10 alphabetic capital letters:

@Validate.length(equalTo: 10)
@Validate.matches(r"$[A-Z]+^")
String tenCapitalLetters;

By default, validations are executed when a Query<T>'s insert or update method is invoked. A validator can be restricted to only run on insert or update by passing values for its optional constructor arguments onUpdate and onInsert:

@Validate.matches(r"^[A-Z]+$", onInsert: true, onUpdate: false)
String validateOnInsertOnly;

It is important to understand how validations work when a value for a property is not specified in an insert or update query. For example, consider a Person with a name and email property and then inserted in a query where email hasn't been set:

var query = new Query<Person>(context)
  ..values.name = "Bob";

await query.insert();

Because email was not set on Query.values, validations will not be run on that property.

There are two special validators that can require a property to be set, or require that a property not be set. Validate.present() requires that the associated property must have a value. A property with this validator must be provided each time the object is inserted or updated. For example, the following declaration requires that email is set on insertion, but doesn't have to be for updates:

@Validate.present(onUpdate: false, onInsert: true)
String email;

The inverse of Validate.present() is Validate.absent(). This validation prevents a property from being set. This is useful when a value should be included during insertion, but can't be updated. Here's an example:

@Validate.absent(onUpdate: true, onInsert: false)
String canOnlyBeSetOnce;

In the above declaration, the validator is only run on update operations and ensures that the property canOnlyBeSetOnce does not have a value. Because this validator is not run on insert operations, there is no restriction when the object is first inserted.

Validators are not run when a value is null. For example, the following insertion explicitly inserts null for the property email:

var query = new Query<Person>(context)
  ..values.email = null
  ..values.name = "Bob";

await query.insert();

Nullability is enforced by Column.isNullable property. Consider the following declaration:

@Column(nullable: false)
@Validate.length(greaterThan: 10)
String name;

Here, the property name must not be null and must be greater than 10 characters long. The behavior of inserting or updating this property is shown in the following table.

Input Value for Name	Validation Runs?	Outcome
Insert value longer than 10 characters	Yes	Successful database insert
Insert value shorter than 10 characters	Yes	Database insert not executed, exception thrown
Insert value not specified	No	Database insert fails with non-null violation, exception thrown
Insert value is null	No	Database insert fails with non-null violation, exception thrown
Update value longer than 10 characters	Yes	Successful database update
Update value shorter than 10 characters	Yes	Database update not executed, exception thrown
Update value not specified	No	Successful database update
Update value is explicit null	No	Successful database update

This behavior allows ManagedObject<T> instances to be partially updated, which also allows for partial PUTs. Partial PUTs can be prevented by adding Validate.present() metadata to all properties.

This also means that any custom validator can assert that a value passed to Validate.validate() is non-null.

Other Validator Behavior

For validators that can't be built by subclassing Validate, you may override ManagedObject<T>.validate(). This method is useful when a validation involves more than one property. Here's an example:

class Person extends ManagedObject<_Person> implements _Person {
  @override
  ValidationContext validate({Validating forEvent: Validating.insert}) {
   final ctx = super.validate(forEvent: forEvent);

    if (a + b > 10) {
      ctx.addError("a + b must be greater than 10");
    }

    return ctx;
  }
}

When overriding this method, the super implementation must be invoked to run validations managed by annotations. You must return the ValidationContext created by the superclass' implementation.

Skipping Validations

Validations are only run when values are set via Query<T>.values. Values set via Query<T>.valueMap are not validated and is useful for inserting data without validation. Here's an example of skipping validation:

var query = new Query<Person>(context)
  ..valueMap = {
    "name" : "xyz",
    "email" : "whatever"
  };

Update and Insert Callbacks

ManagedObject<T> subclasses may override willUpdate and willInsert to make changes prior to being updated or inserted. For example, a managed object may have updated and created dates that can be guaranteed to be set when inserted or updated:

class Person extends ManagedObject<_Person> implements _Person {
  @override
  void willUpdate() {
    updatedAt = new DateTime.now().toUtc();
  }

  @override
  void willInsert() {
    createdAt = new DateTime.now().toUtc();
  }
}
class _Person {
  @primaryKey
  int id;

  String name;
  DateTime createdAt;
  DateTime updatedAt;
}

Both willUpdate and willInsert are run before any validation occurs. Like validations, willUpdate and willInsert are skipped when using Query.valueMap.