NAME

Valiant - Ruby on Rails-like validation framework.

Coverage

SYNOPSIS

    package Local::Person;

    use Moo;
    use Valiant::Validations;
    use Valiant::Filters;

    has name => (is=>'ro');
    has age => (is=>'ro');

    filters_with => 'Trim';

    validates name => (
      length => {
        maximum => 10,
        minimum => 3,
      }
    );

    validates age => (
      numericality => {
        is_integer => 1,
        less_than => 200,
      },
    );

    validates_with 'MySpecialValidator' => (arg1=>'foo', arg2=>'bar');

    my $person = Local::Person->new(
        name => 'Ja',
        age => 300,
      );

    $person->validate;
    $person->valid;     # FALSE
    $person->invalid;   # TRUE

    my %errors = $person->errors->to_hash(full_messages=>1);

    # \%errors = +{
    #   age => [
    #     "Age must be less than 200",
    #   ],
    #   name => [
    #     "Name is too short (minimum is 3 characters)',   
    #   ],
    # };

DESCRIPTION

WARNING: This is early release code. I've been using it for a while on personal projects and have a lot of test cases but there's probably corners I've not hit yet. All documented features should work as described. I will only change API and functionality in a breaking manner if its the only way to fix problems (caveat that I am reserving the right to change the way errors are listed and structured for nested validations, for now, as well as the way we resolve files for internationalization, whose current code may not be as performant as needed; additionally the Valiant::DBIx::Class packages are all currently subject to breaking changes as needed).

Domain level validations for Moo classes and related capabilities such as attribute filtering and internationalization. Provides a domain specific language which allows you to defined for a given class what a valid state for an instance of that class would be and to gather reportable error messages. Used to defined constraints related to business logic or for validating user input (for example via CGI forms).

When we say domain level or business logic validation, what we mean is that invalid data is a possible and expected state that needs to be evaluated and reported to the end user for correction. For example when writing a web application you might have a form that requests user profile information (such as name, DOB, address, etc). Its an expected condition that the user might submit form data that is invalid in some way (such as a DOB that is in the future) but is still 'well formed' and is able to be processed. In these cases your business logic would be to inform the user of the incorrect data and request fixes (rather than simply throw a 500 server error and giving up).

This differs from type constraints (such as Type::Tiny) that you might put on your Moo attributes which are used to express when attributes have values that are so unacceptable that no further work can be done and an exception must be thrown.

In fact you will note that when using validations that you generally won't add type constraints on your Moo attributes. That's because type constraints are applied when the object is instantiated and throw an exception when they fail. Validations on the other hand permit you to create the object and collect all the validation failure conditions. Also since you have a created object you can do more complex validations (such as those that involve the state of more than one attribute). You would only use attribute type constraints when the created object would be in such an invalid state that one could not correctly validate it anyway. An example of this might be when an attribute is assigned an array value when a scalar is expected.

Valiant fits into a similar category as HTML::Formhander and FormFu although its not HTML form specific. Prior art for this would be the validations system for ActiveRecords in Ruby on Rails and the Javascript library class-validator.js, both of which the author reviewed extensively when writing this code:

https://rubyonrails.org, https://github.com/typestack/class-validator

Documentation here details using Valiant with Moo or Moose based classes. If you want to use Valiant with DBIx::Class you will also wish to review DBIx::Class::Valiant which details how Valiant glues into DBIx::Class.

This document reviews all the bits of the Valiant system as a whole (validations, filters, internationalization, errores etc). You might also like to review API details from the following files:

Valiant::Validates, a Moo::Role which adds a validation API to your class, or Valiant::Validations which wraps Valiant::Validates in an easy to use DSL (domain specific language).

Valiant::Filterable, a Moo::Role which adds API to apply filtering to incoming attributes at object creation time, or Valiant::Filters, which wraps this in an easy to use DSL.

Valiant::I18N, API information on how we provide internationalized error messages for your validations.

Valiant::Validator and Valiant::Filter which provides details about validations and filters that are packaged with Valiant.

WHY OBJECT VALIDATION AS CLASS DATA?

Validating the state of things is one of the most common tasks we perform. For example a user might wish to change their profile information and you need to make sure that the new settings conform to acceptable limits (such as the user first and last name fits into the database and have acceptable characters, that a password is complex enough and that an address is complete, etc). This logic can get tricky over time as a system grows in complexity and edge cases need to be accounted for (for example for business reasons you might wish to allow pre-existing users to conform to different password complexity constraints or require newer users to supply more profile details).

One approach to this is to build a specific validation object that receives and processes data input. If the incoming data passes, you can proceed to send the data to your storage object (such as DBIx::Class) If it fails you then proceed to message the user the failure conditions. For example this is the approach taken by HTML::Formhandler.

The approach has much to merit because it clearly separates your validation logic from your storage logic. However when there is a big affinity between the two (such as when all your HTML forms are very similar to your database tables) this can lead to a lot of redundant code (such as defining the same field names in more than one class) which leads to a maintainance and understanding burden. Also it can be hard to do proper validation without access to a data object since often you will have validation logic that is dependent on the current state of your data. For example you might require that a new password not be one of the last three used; in this case you need access to the storage layer anyway.

Valiant offers a DSL (domain specific language) for adding validation meta data as class data to your business objects. This allows you to maintain separation of concerns between the job of validation and the rest of your business logic but also keeps the validation work close to the object that actually needs it, preventing action at a distance confusion. The actual validation code can be neatly encapsulated into standalone validator classes (subclasses based on Valiant::Validator or Valiant::Validator::Each) so they can be reused across more than one business object. To bootstrap your validation work, Valiant comes with a good number of validators which cover many common cases, such as validating string lengths and formats, date-time validation and numeric validations. Lastly, the validation meta data which is added via the DSL can aggregate across consumed roles and inherited classes. So you can create shared roles and base classes which defined validations that are used in many places.

Once you have decorated your business logic classes with Valiant validations, you can run those validations on blessed instances of those classes and inspect errors. There is also some introspection capability making it possible to do things like generate display UI from your errors.

EXAMPLES

The following are some example cases of how one can use Valiant to perform object validation

The simplest possible case

At its most simple, a validation can be just a reference to a subroutine which adds validation error messages based on conditions you code:

    package Local::Simple

    use Valiant::Validations;
    use Moo;

    has name => (is => 'ro');
    has age => (is => 'ro');

    validates_with sub {
      my ($self, $opts) = @_;
      $self->errors->add(name => "Name is too long") if length($self->name) > 20;
      $self->errors->add(age => "Age can't be negative") if  $self->age < 1;
    };

    my $simple = Local::Simple->new(
      name => 'A waaay too loooong name', # more than 20 characters
      age => -10, # less than 1
    );

    $simple->validate;
    $simple->valid;     # FALSE
    $simple->invalid;   # TRUE

    my %errors = $simple->errors->to_hash(full_messages=>1);

    #\%errors = {
    #  age => [
    #    "Age can't be negative",
    #  ],
    #  name => [
    #    "Name is too long",
    #  ],
    #}

The subroutine reference that the validates_with keyword accepts will receive the blessed instance as the first argument and a hash of options as the second. Options are added as additional arguments after the subroutine reference. This makes it easier to create parameterized validation methods:

    package Local::Simple2;

    use Valiant::Validations;
    use Moo;

    has name => (is => 'ro');
    has age => (is => 'ro');

    validates_with \&check_length, length_max => 20;
    validates_with \&check_age_lower_limit, min => 5;

    sub check_length {
      my ($self, $opts) = @_;
      $self->errors->add(name => "is too long") if length($self->name) > $opts->{length_max};
    }

    sub check_age_lower_limit {
      my ($self, $opts) = @_;
      $self->errors->add(age => "can't be lower than $opts->{min}") if $self->age < $opts->{min};
    }

    my $simple2 = Local::Simple2->new(
      name => 'A waaay too loooong name',
      age => -10,
    );

    $simple2->validate;
    $simple2->valid;     # FALSE
    $simple2->invalid;   # TRUE

    my %errors = $simple2->errors->to_hash(full_messages=>1);

    #\%errors = {
    #  age => [
    #    "Age can't be lower than 5",
    #  ],
    #  name => [
    #    "Name is too long",
    #  ],
    #}

The validation methods have access to the fully blessed instance so you can create complex validation rules based on your business requirements, including retrieving information from shared storage classes.

Since many of your validations will be directly on attributes of your object, you can use the validates keyword which offers some shortcuts and better code reusability for attributes. We can rewrite the last class as follows:

    package Local::Simple3;

    use Valiant::Validations;
    use Moo;

    has name => (is => 'ro');
    has age => (is => 'ro');

    validates name => ( \&check_length => { length_max => 20 } );
    validates age => ( \&check_age_lower_limit => { min => 5 } );

    sub check_length {
      my ($self, $attribute, $value, $opts) = @_;
      $self->errors->add($attribute => "is too long", $opts) if length($value) > $opts->{length_max};
    }

    sub check_age_lower_limit {
      my ($self, $opts) = @_;
      $self->errors->add($attribute => "can't be lower than $opts->{min}", $opts) if $value < $opts->{min};
    }

    my $simple3 = Local::Simple2->new(
      name => 'A waaay too loooong name',
      age => -10,
    );

    $simple3->validate;

    my %errors = $simple3->errors->to_hash(full_messages=>1);

    #\%errors = {
    #  age => [
    #    "Age can't be lower than 5",
    #  ],
    #  name => [
    #    "Name is too long",
    #  ],
    #}

Using the validates keyword allows you to name the attribute for which the validations are intended. When you do this the signature of the arguments for the subroutine reference changes to included both the attribute name (as a string) and the current attribute value. This is useful since you can now use the validation method across different attributes, avoiding hardcoding its name into your validation rule. One difference from validates_with you will note is that if you want to pass arguments as parameter options you need to use a hashref and not a hash. This is due to the fact that validates can take a list of validators, each with its own arguments. For example you could have the following:

    validates name => (
      \&check_length => { length_max => 20 },
      \&looks_like_a_name,
      \&is_unique_name_in_database,
    );

Also, similiar to the has keyword that Moo imports, you can use an arrayref of attribute name for grouping those with the same validation rules:

    validates ['first_name', 'last_name'] => ( \&check_length => { length_max => 20 } );

At this point you can see how to write fairly complex and parameterized validations on your attributes directly or on the object as a whole (using validates for attributes and validates_with for validations that are not directly tied to an attribute but instead validate the object as a whole). However it is often ideal to isolate your validation logic into a stand alone class to promote code reuse as well as better separate your valiation logic from your classes.

Using a validator class

Although you could use subroutine references for all your validation if you did so you'd likely end up with a lot of repeated code across your classes. This is because a lot of validations are standard (such as string length and allowed characters, numeric ranges and so on). As a result you will likely build at least some custom validators and make use of the prepacked ones that ship with Valiant (see Valiant::Validator). Lets return to one of the earlier examples that used valiates_with but instead of using a subroutine reference we will rewrite it as a custom validator:

    package Local::Person::Validator::Custom;

    use Moo;
    with 'Valiant::Validator';

    has 'max_name_length' => (is=>'ro', required=>1);
    has 'min_age' => (is=>'ro', required=>1);

    sub validate {
      my ($self, $object, $opts) = @_;
      $object->errors->add(name => "is too long", $opts) if length($object->name) > $self->max_name_length;
      $object->errors->add(age => "can't be lower than @{[ $self->min_age ]}", $opts) if $object->age < $self->min_age;
    }

And use it in a class:

    package Local::Person;

    use Valiant::Validations;
    use Moo;

    has name => (is => 'ro');
    has age => (is => 'ro');

    validates_with Custom => (
      max_name_length => 20, 
      min_age => 5,
    );

    my $person = Local::Person->new(
      name => 'A waaay too loooong name',
      age => -10,
    );

    $person->validate;
    $person->invalid; # TRUE

    my %errors = $person->errors->to_hash(full_messages=>1) };

    #\%errors =  +{
    #  age => [
    #    "Age can't be lower than 5",
    #  ],
    #  name => [
    #    "Name is too long",
    #  ],
    #}; 

A custom validator is just a class that does the validate method (although I recommend that you consume the Valiant::Validator role as well; this might be required at some point). When this validator is added to a class, it is instantiated once with any provided arguments (which are passed to new as init_args). Each time you call validate, it runs the validate method with the following signature:

    sub validate {
      my ($self, $object, $opts) = @_;
      $object->errors->add(...) if ...
    }

Where $self is the validator object, $object is the current instance of the class you are validating and $opts is the options hashref.

Within this method you can do any special or complex validation and add error messages to the $object based on its current state.

Custom Validator Namespace Resolution

When you use a custom validator class namepart (either via validates or validates_with) we search thru a number of namespaces to find a match. This is done to allow you to create increasingly custom valiators for your classes. Basically we start with the package name of the class which is adding the validator, add "::Validator::${namepart}" and then look down the namespace tree for a loadable file. If we don't find a match in your project package namespace we then also look in the two globally shared namespaces Valiant::ValidatorX and Valiant::Validator. If we still don't find a match we then throw an exception. For example if your package is named Local::Person as in the class above and you specify the Custom validator we will search for it in all the namespaces below, in order written:

    Local::Person::Validator::Custom
    Local::Validator::Custom
    Validator::Custom
    Valiant::ValidatorX::Custom
    Valiant::Validator::Custom

This lookup only happens once when your classes are first loaded, so this will cause a a delay in startup but not at runtime. However the delay probably makes Valiant unsuitable for non persistant applications such as CGI web applications or possibly scripts that run as part of a cron job.

NOTE: The namespace Valiant::Validator is reserved for validators that ship with Valiant. The Valiant::ValidatorX namespace is reserved for additional validators on CPAN that are packaged separately from Valiant. If you wish to share a custom validator that you wrote the proper namespace to use on CPAN is Valiant::ValidatorX.

You can also prepend your validator name with '+' which will cause Valiant to ignore the namespace resolution and try to load the class directly. For example:

    validates_with '+App::MyValidator';

Will try to load the class App::MyValidator and use it as a validator directly (or throw an exception if it fails to load).

Validator classes and attributes

Since many of your validations will be on your class's attributes, Valiant makes it easy to use custom and prepackaged validator classes directly on attributes. All validator classes which operate on attributes must consume the role Valiant::Validator::Each. Here's an example of a class which is using several of the prepackaged attribute validator classes that comes with Valiant.

    package Local::Task;

    use Valiant::Validations;
    use Moo;

    has priority => (is => 'ro');
    has description => (is => 'ro');
    has due_date => (is => 'ro');

    validates priority => (
      presence => 1,
      numericality => { only_integer => 1, between => [1,10] },
    );

    validates description => (
      presence => 1,
      length => [10,60],
    );

    validates due_date => (
      presence => 1,
      date => 'is_future',
    );

In this case our class defines three attributes, 'priority' (which defined how important a task is), 'description' (which is a human read description of the task that needs to happen) and a 'due_date' (which is when the task should be completed). We then have validations which place some constraints on the allowed values for these attributes. Our validations state that:

    'priority' must be defined, must be an integer and the number must be from 1 thru 10.
    'description' must be defined and a string that is longer than 10 characters but less than 60.
    'due_date' must be in a date format (YYYY-MM-DD or eg. '2000-01-01') and also must be a future date.

This class uses the following validators: Valiant::Validator::Presence, to verify that the attribute has a meaningful defined value; Valiant::Validator::Numericality, to verify the value is an integer and is between 1 and 10; Valiant::Validator::Length, to check the length of a string and Valiant::Validator::Date to verify that the value looks like a date and is a date in the future.

Canonically a validator class accepts a hashref of options, but many of the packaged validators also accept shortcut forms for the most common use cases. For example since its common to require a date be sometime in the future you can write "date => 'is_future'". Documentation for these shortcut forms are detailed in each validator class.

Creating a custom attribute validator class

Creating your own custom attribute validator classes is just as easy as it was for creating a general validator class. You need to write a Moo class that consumes the Valiant::Validator::Each role and provides a validates_each method with the following signature:

    sub validates_each {
      my ($self, $object, $attribute, $value, $opts) = @_; 
    }

Where $self is the validator class instance (this is created once when the validator is added to the class), $object is the instance of the class you are validating, $attribute is the string name of the attribute this validation is running on, $value is the current attribute's value and $opts is a hashref of options passed to the class. For example, here is simple Boolean truth validator:

    package MyApp::Validator::True;

    use Moo;

    with 'Valiant::Validator::Each';

    sub validate_each {
      my ($self, $object, $attribute, $value, $opts) = @_;
      $object->errors->add($attribute, 'is not a true value', $opts) unless $value;
    }

And example of using it in a class:

    package MyApp::Foo;

    use Moo;
    use Valiant::Validations;

    has 'bar' => ('ro'=>1);

    validates bar => (True => +{}); # No arguments passed to MyApp::Validator::True->new()

Two things to note: There is no meaning assigned to the return value of validate_each (or of validates). Also you should remember to pass $opts as the third argument to the add method. Even if you are not using the options hashref in your custom validator, it might contain values that influence other aspects of the framework, such as how the error message is formatted.

When resolving an attribute validator namepart, the same rules described above for general validator classes apply.

PREPACKAGED VALIDATOR CLASSES

Please see Valiant::Validator for a list of all the valiators that are shipped with Valiant and/or search CPAN for validators in the Valiant::ValidatorX namespace.

TYPE CONSTRAINT SUPPORT

If you are comfortable using common type contraint libaries such as Type::Tiny you can use those directly as validation rules much in the same way you can use subroutine references.

    package Local::Test::Check;

    use Moo;
    use Valiant::Validations;
    use Types::Standard 'Int';

    has drinking_age => (is=>'ro');

    validates drinking_age => (
      Int->where('$_ >= 21'), +{
        message => 'is too young to drink!',
      },
    );

When using this option you should use a library system such as Type::Tiny that provides an object with a check method that returns a boolean indicating if the constraint was passed or not. There are several such library systems on CPAN that you might find useful in helping you to write validations.

Please note this is just wrapping Valiant::Validator::Check so if you need more control you might prefer to use the validator class.

INHERITANCE AND ROLES

You can aggregate validation rules via inheritance and roles. This makes it so that if you have a number of classes with similar validation rules you can avoid repeating yourself. Example:

    package Person;

    use Moo;
    use Valiant::Validations;

    has 'name' => (is=>'ro',);
    has 'age' => (is=>'ro');

    validates name => (
      presence => 1,
      length => [3,20],
    );

    validates age => (
      numericality => {
        only_integer => 1,
        greater_than => 0,
        less_than => 150,
      },
    );

    package IsRetirementAge;

    use Moo::Role;
    use Valiant::Validations;

    requires 'age';

    validates age => (
      numericality => {
        greater_than => 64,
      },
    );

    package Retiree;

    use Moo;
    use Valiant::Validations;

    extends 'Person';
    with 'IsRetirementAge';

    1;

    my $retiree = Retiree->new(name=>'Molly Millions', age=>24);

    $retiree->invalid; # true

    my %errors = $object->errors->to_hash(full_messages=>1);
    # \%errors = +{
    #   age => ["Age must be greater than 64" ]
    # }

GLOBAL ATTRIBUTE VALIDATOR OPTIONS

All attribute validators can accept the following options. Options can be added to each validator separately (if you have several) or can be added globally to the end of the validator rules. Global rules run first, followed by those applied to each validator.

allow_undef

If the attribute value is undef, skip validation and allow it.

    package Person;

    use Moo;
    use Valiant::Validations;

    has 'name' => (is=>'ro',);
    has 'age' => (is=>'ro');

    validates name => (
      format => 'alphabetic',
      length => [3,20],
      allow_undef => 1, # Skip BOTH 'format' and 'length' validators if 'name' is undefined.
    );

    validates age => (
      numericality => {
        only_integer => 1,
        greater_than => 0,
        less_than => 150,
        allow_undef => 1,  # Skip only the 'numericality' validator if 'age' is undefined
      },
    );
 

allow_blank

If the attribute is blank (that is its one of undef, '', or a scalar composing only whitespace) skip validation and allow it. This is similar to the allow_undef option except it allows a broader definition of 'blank'. Useful for form validations.

if / unless

Accepts a coderef or the name of a method which executes and is expected to return true or false. If false we skip the validation (or true for unless). Recieves the object, the attribute name, value to be checked and the options hashref as arguments.

You can set more than one value to these with an arrayref:

    if => ['is_admin', sub { ... }, 'active_account'],

Example:

    package Person;

    use Moo;
    use Valiant::Validations;

    has 'name' => (is=>'ro',);
    has 'password' => (is=>'ro'); 

    validates name => (
      format => 'alphabetic',
      length => [3,20],
    );

    validates password => (
      length => [12,32],
      with => \&is_a_secure_looking_password,
      unless => sub {
        my ($self, $attr, $value, $opts) = @_;
        $self->name eq 'John Napiorkowski';  # John can make an insecure password if he wants!
      },
    );

message

Provide a global error message override for the constraint. Message can be formated in the same way as the second argument to 'errors->add' and will accept a string, a translation tag, a reference to a string or a reference to a function. Using this will override the custom error message provided by the validator.

    package Person;

    use Moo;
    use Valiant::Validations;

    has 'name' => (is=>'ro',);
    has 'age' => (is=>'ro');

    validates name => (
      format => 'alphabetic',
      length => [3,20],
      message => 'Unacceptable Name', # Overrides both the 'format' and 'length' messages.
    );

    validates age => (
      presence => 1,
      numericality => {
        only_integer => 1,
        greater_than => 0,
        less_than => 150,
        message => 'Age given is not valid',  # overrides just the 'numericality' messages
                                              # and not the 'presence' message.
      },
    );

Please not that many validators also provide error type specific messages for custom errors (as well as the ability to setup your own errors in a localization file.) Using this attribute is the easiest but probably not always your best option.

Messages can be:

A string

If a string this is the error message recorded.

    validates name => (
      format => 'alphabetic',
      length => [3,20],
      message => 'Unacceptable Name',
    );
A translation tag
    use Valiant::I18N;

    validates name => (
      format => 'alphabetic',
      length => [3,20],
      message => _t('bad_name'),
    );

This will look up a translated version of the tag. See Valiant::I18N for more details.

A scalar reference
    validates name => (
      format => 'alphabetic',
      length => [3,20],
      message => \'Unacceptable {{attribute}}',
    );

Similar to string but we will expand any placeholder variables which are indicated by the '{{' and '}}' tokens (which are removed from the final string). You can use any placeholder that is a key in the options hash (and you can pass additional values when you add an error). By default the following placeholder expansions are available attribute (the attribute name), value (the current attribute value), model (the human name of the model containing the attribute and object (the actual object instance that has the error).

A subroutine reference
    validates name => (
      format => 'alphabetic',
      length => [3,20],
      message => sub {
        my ($self, $attr, $value, $opts) = @_;
        return "Unacceptable $attr!";
      }
    );

Similar to the scalar reference option just more flexible since you can write custom code to build the error message. For example you could return different error messages based on the identity of a person. Also if you return a translation tag instead of a simple string we will attempt to resolve it to a translated string (see Valiant::I18N).

strict

When true instead of adding a message to the errors list, will throw exception with the error instead. If the true value is a string (basically anything other than a 1) we use the string given as the message. If the true value is the name of a class that provides a throw message, will use that instead. Lastly if the value is a coderef we call that.

on

A scalar or array reference of contexts that can be used to control the situation ('context') under which the validation is executed. If you specify an on context that validation will only run if you pass that context via validate. However if you don't set a context for the validate (in other words you don't set an on value) then that validation ALWAYS runs (whether or not you set a context via validate. Basically not setting a context means validation runs in all contexts and none. Examples:

  package Local::Test::Person;

  use Moo;
  use Valiant::Validations;
  use Valiant::I18N;

  has age => (is=>'ro');

  validates age => (
    numericality => {
      is_integer => 1,
      less_than => 200,
    },
    numericality => {
      greater_than_or_equal_to => 18,
      on => 'voter',
    },
    numericality => {
      greater_than_or_equal_to => 65,
      on => 'retiree',
    },
    numericality => {
      greater_than_or_equal_to => 100,
      on => 'centarion',
    },
  );

  my $person = Local::Test::Person->new(age=>50);

  $person->validate();
  $person->valid; # True.

  $person->validate(context=>'retiree');
  $person->valid; # False; "Age must be greater than or equal to 65"

  $person->validate(context=>'voter');
  $person->valid; # True.

  $person->validate(context=>'centarion');
  $person->valid; # False; "Age must be greater than or equal to 100"

  my $another_person = Local::Test::Person->new(age=>"not a number");

  $another_person->validate();
  $another_person->valid; # False "Age does not look like an integer"

In this example you can see that since the first validation does not set an on context it always runs no matter what context you set via validate (even when you don't set one). So we always check that the value is an integer.

Basically the rule to remember is validations with no on option will run no matter the context is set via validation options (or set automatically). Validations with some on option will only run in the specified context.

So if your validation requests one or more contexts via on they only run when at least one of the passed contexts is matching. If your validation does not request a context via on then they match ANY or NONE contexts!

GLOBAL MODEL VALIDATOR OPTIONS

Model validators (added via validates_with) support a subset of the the same options as "GLOBAL ATTRIBUTE VALIDATOR OPTIONS". These options work identically as described in that section:

    if/unless
    on
    message
    strict

ERROR MESSAGES

When you create an instance of a class that consumes the Valiant::Validates role (typically by using Valiant::Validations to import the standard methods into the class) that role will add an attribute called errors to your class. This attribute will contain an instance of Valiant::Errors, which is a collection class that contains a list of errors (instances of Valiant::Error) along with methods to add, retrieve and introspect errors. You should review Valiant::Errors for the full class API since we will only coverage the most commonly used methods in our example.

The most common use cases you will have is adding errors and checking for and recovering error messages.

Adding an error message.

You can add an error message anywhere in your code, although most commonly you will do so in you validation methods or validation callbacks. In all cases the method signature is the same:

    $object->errors->add($attribute_name, $error_message, \%opts);

Where $attribute_name is the string name of the attribute for which a validation error is being recorded, $error_message is one of the allowed error message types (see below for details) and \%opts is a hashref of options and/or error message template variable expansions which is used to influence how the error is processed.

When the error is scoped to the $object and not a particular attribute you can just use undef instead of an attribute name. This will record the error as a model error:

    $object->errors->add(undef, $error_message, \%opts);

Lastly the %opts hashref can be left off the method call if you don't have it. Generally its passed as the last argument to validate or any validation subroutine references but if you are adding an error outside those methods you won't have it. For example you might wrap a database call insidean eval and wish to add a model error if there's an exception.

Error message types.

When adding an error there's four options for what the value of <$error_message> can be and are described above "'GLOBAL ATTRIBUTE VALIDATOR OPTIONS'"

Message priority

Valiant allows you to set messages at various points to give you a lot of flexibility in customizing your response. You can add errors messages at the point you add it to the errors collection, in the options for the validator and globally for all validators in a chain. For example:

    package MyApp::Errors;

    use Moo;
    use Valiant::Validations;

    has name => (is=>'ro');

    validates name => (
      with => {
        cb => sub {
          my ($self, $attr, $value, $opts) = @_;
          $self->errors->add($attr, 'is always in error!', $opts);  #LAST
        },
        message => 'has wrong value', #FIRST
      },
      message => 'has some sort of error', #SECOND
    );

Here you can see error messages at three levels. Here's the resolution order:

    Messages added to the errors collection via ->add are last
    Messages added globally to a validation clause are second
    Messages added via the 'message' option for a validator is first

In this case the error would be "Name has wrong value".

Note for the outermost global message, please keep in mind that it will override all the error messages of any of the validators in the clause.

Accessing and displaying error messages

Once you have validated your object (via the ->validate method) you can check for for validate state and review or display any errors.

Checking for errors on a validated object

You can use the valid or invalid methods on your object to check for its validation state. These methods won't run validations, unless they have not yet been run, so you can call them as often as you want without incurring a runtime performance penalty.

NOTE However if you pass arguments such as context then any existing validations are cleared and validations are re run.

    $object->validate;
    $object->valid; # TRUE if there are no errors, FALSE otherwise
    $object->invalid # Opposite of 'valid'

You can also just check the size of the errors collection (size of 0 means no errors):

    $object->validate;
    $object->errors->size;

To make this a bit easier the validate method returns its calling object so you can chain methods:

    $object->validate->valid;

Retrieving error messages

The Valiant::Errors collection object gives you a few ways to retrieve error messages. Assuming there is a model with errors like the following for these examples and discussion:

    $model->errors->add(undef, "Your Form is invalid");
    $model->errors->add(name => "is too short");
    $model->errors->add(name => "has disallowed characters");
    $model->errors->add(age => "must be above 5");
    $model->errors->add(email => "does not look like an email address");
    $model->errors->add(password => "is too short");
    $model->errors->add(password => "can't look like your name");
    $model->errors->add(password => "needs to contain both numbers and letters");
Getting all the errors at once, or groups of errors

You can get all the error messages in a simple array with either the messages or full_messages method on the errors collection:

For messages:

    is_deeply [$model->errors->messages], [
      "Your Form is invalid",
      "is too short",
      "has disallowed characters",
      "must be above 5",
      "does not look like an email address",
      "is too short",
      "can't look like your name",
      "needs to contain both numbers and letters",
    ];

For full_messages:

    is_deeply [$model->errors->full_messages], [
      "Your Form is invalid",
      "Name is too short",
      "Name has disallowed characters",
      "Age must be above 5",
      "Email does not look like an email address",
      "Password is too short",
      "Password can't look like your name",
      "Password needs to contain both numbers and letters",
    ];

This combines all the attribute and model message into a flat list. Please not that the current order is the order in which messages are added as errors to the errors collection.

The only difference between messages and full_messages is that the latter will combine your error message with a human readable version of your attribute name. By default this is just a title cased version of the attribute name but you can customize this via setting a translation (see "INTERNATONALIZATION"). full_messages by default uses the following expansion template: "{{attribute}} {{message}}" however you can customize this by setting the format key in your translation file (again see "INTERNATONALIZATION").

If you just want the model level errors you can use model_messages:

    is_deeply [$model->errors->model_messages], [
      "Your Form is invalid",
    ];

There is a similar pair of methods for just getting messages associated with attributes: attribute_messages and full_attribute_messages:

    is_deeply [$model->errors->attribute_messages], [
      "is too short",
      "has disallowed characters",
      "must be above 5",
      "does not look like an email address",
      "is too short",
      "can't look like your name",
      "needs to contain both numbers and letters",
    ];

    is_deeply [$model->errors->full_attribute_messages], [
      "Name is too short",
      "Name has disallowed characters",
      "Age must be above 5",
      "Email does not look like an email address",
      "Password is too short",
      "Password can't look like your name",
      "Password needs to contain both numbers and letters",
    ];

Lastly you can retrieve all the error messages as a hash using the to_hash method. This return a hash where the hash keys refer to attributes with errors (or in the case of model errors the key is '*') and the value is an arrayref of the error(s) associated with that key. The to_hash method accepts an argument to control if the error messages return using the full_messages value or the messages value:

    is_deeply +{ $model->errors->to_hash }, {
      "*" => [
        "Your Form is invalid",
      ],
      age => [
        "must be above 5",
      ],
      email => [
        "does not look like an email address",
      ],
      name => [
        "is too short",
        "has disallowed characters",
      ],
      password => [
        "is too short",
        "can't look like your name",
        "needs to contain both numbers and letters",
      ],
    };

    is_deeply +{ $model->errors->to_hash(full_messages=>1) }, {
      "*" => [
        "Your Form is invalid",
      ],
      age => [
        "Age must be above 5",
      ],
      email => [
        "Email does not look like an email address",
      ],
      name => [
        "Name is too short",
        "Name has disallowed characters",
      ],
      password => [
        "Password is too short",
        "Password can't look like your name",
        "Password needs to contain both numbers and letters",
      ],
    };
Getting errors for individual attributes

Similar to the method that allows you to get the errors just for the model you can get errors for individual attibutes with the messages_for and full_messages_for methods:

    is_deeply [$model->errors->full_messages_for('password')], [
        "Password is too short",
        "Password can't look like your name",
        "Password needs to contain both numbers and letters",
      ];

    is_deeply [$model->errors->messages_for('password')], [
        "is too short",
        "can't look like your name",
        "needs to contain both numbers and letters",
      ];

If you request errors for an attribute that has none you will get an empty array.

Please note that these methods always return arrays even if the case where you have only a single error.

Searching for errors, interating and introspection.

Valiant::Errors contains additional methods for iterating over error collections, searching for errors and introspecting errors. Please refer to that package for full documentation and examples.

NESTED OBJECTS AND ARRAYS

In some cases you may have complex, nested objects or objects that contain arrays of values which need validation. When an object is nested as a attribute under another object it may itself contain validations. For these more complex cases we provide two validator classes Valiant::Validator::Object and Valiant::Validator::Array. You should refer to documentation in each of those validators for API level overview and examples.

INTERNATONALIZATION

Internationalization for Valiant will concern our ability to create tags that represent human readable strings for different languages. Generally we will create tags, which are abstract labels representing a message, and then map those lables to various human languages which we wish to support. In using Valiant::I18N with Valiant there are generally three things that we will internationalize

error messages
attribute names
model names

In the case of error messages there is an additional complication in that often we need to customize the message base on the value of the attributes. For example when the attribute represents a number of items often the message for zero items will be different than for many (think "You have 3 items in you bag, the minimum is 5" versus "You have no items in your bag, the minimum is 5"). The rules for this can be complex depending on the language. Therefore in the case of error messages you will need the ability to return a different string for those cases.

For now please see Valiant::I18N for more, and the test suite. This will need more documentation (volunteers welcomed).

FILTERING

Quite often you will wish to allow your users a bit of leeway when providing information. For example you might want incoming data for a field to be in all capitals, and to be free from any extra post or trailing whitespace. You could test for these in a validation and return error conditions when they are present. However that is not always the best user experience. In cases where you are willing to accept such input from users but you want to 'clean up' the data before trying to validate it you can use filters.

For now please see Valiant::Filters and Valiant::Filters for API level documentations on filters as well as some examples. Also see Valiant::Filter for a list of the prepackaged filter that ship with Valiant

DEBUGGING

You can set the %ENV variable VALIANT_DEBUG to a number ranging 1 to 3 which will give increasing more detailed debugging output that should assist you if things are not working as expected. Debug level 1 only returns messages during the startup / compile stage so its reasonable safe to run even in a production environment since it should not impact run time performance.

DEDICATIONS

This module is eternally dedicated to the memory of my Bernese Mountain Dog 'Tornado' who we lost to cancer in 16 August 2020. If you find this code useful, if it helps your company or makes you money please consider a donation to help other owners of this dog breed or to help the dogs themselves: http://www.berner.org/pages/charities.php

SEE ALSO

There's no lack of validation systems on CPAN. I've used (and contributed) to HTML::FormHandler and <Data::MuForm>. I've also used HTML::FormFu. Recently I spotted Form::Tiny which is a similar DSL style system as Valiant but with a smaller footprint and sane looking sort code. This list is not exhaustive, just stuff I've either used or reviewed.

Valiant, Valiant::Validations, Valiant::Validates, Valiant::Filters, Valiant::Filterable

COPYRIGHT & LICENSE

Copyright 2021, John Napiorkowski email:jjnapiork@cpan.org

This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.