File Coverage

=head1 NAME

Params::Validate::Strict - Validates a set of parameters against a schema

=head1 VERSION

Version 0.41

=cut

=head1 SYNOPSIS

    my $schema = {
        username => { type => 'string', min => 3, max => 50 },
        age => { type => 'integer', min => 0, max => 150 },
    };

    my $input = {
         username => 'john_doe',
         age => '30',        # Will be coerced to integer
    };

    my $validated_input = validate_strict(schema => $schema, input => $input);

    if(defined($validated_input)) {
        print "Example 1: Validation successful!\n";
        print 'Username: ', $validated_input->{username}, "\n";
        print 'Age: ', $validated_input->{age}, "\n";      # It's an integer now
    } else {
        print "Example 1: Validation failed: $@\n";
    }

Upon first reading this may seem overly complex and full of scope creep in a sledgehammer to crack a nut sort of way,
however two use cases make use of the extensive logic that comes with this code
and I have a couple of other reasons for writing it.

=over 4

=item * Black Box Testing

The schema can be plumbed into L<App::Test::Generator> to automatically create a set of black-box test cases.

=item * WAF

The schema can be plumbed into a WAF,
e.g., L<VWF|https://github.com/nigelhorne/VWF/>,
to protect from random user input.

=item * Improved API Documentation

Even if you don't use this module,
the specification syntax can help with documentation.

=item * I like it

I found it fun to write this,
even if nobody else finds it useful,
though I hope you will.

=back

=head1  METHODS

=head2 validate_strict

Validates a set of parameters against a schema.

This function takes two mandatory arguments:

=over 4

=item * C<schema> || C<members>

A reference to a hash that defines the validation rules for each parameter.
The keys of the hash are the parameter names, and the values are either a string representing the parameter type or a reference to a hash containing more detailed rules.

As an alternative the schema may be supplied as an B<arrayref of parameter hashrefs>,
where every element describes one parameter and carries a mandatory
C<name> key:

  $schema = [
    { name => 'username', type => 'string', min => 3, max => 50 },
    { name => 'age',      type => 'integer', min => 0, max => 150 },
    { name => 'role',     type => 'string', optional => 1, default => 'user' },
  ];

The arrayref form is normalised to the standard hashref form before any further
processing.  It is particularly useful when declaration order matters (e.g.
for positional or mixed calling conventions used by some CPAN modules).  The
C<name> key is consumed during normalisation and does not appear as a
validation rule.

For some sort of compatibility with L<Data::Processor>,
it is possible to wrap the schema within a hash like this:

  $schema = {
    description => 'Describe what this schema does',
    error_msg => 'An error message',
    schema => {
      # ... schema goes here
    }
  }

=item * C<args> || C<input>

A reference to a hash containing the parameters to be validated.
The keys of the hash are the parameter names, and the values are the parameter values.

=back

It takes optional arguments:

=over 4

=item * C<description>

What the schema does,
used in error messages.

=item * C<error_msg>

Overrides the default message when something doesn't validate.

=item * C<unknown_parameter_handler>

This parameter describes what to do when a parameter is given that is not in the schema of valid parameters.
It must be one of C<die>, C<warn>, or C<ignore>.

It defaults to C<die> unless C<carp_on_warn> is given, in which case it defaults to C<warn>.

=item * C<logger>

A logging object that understands messages such as C<error> and C<warn>.

=item * C<custom_types>

A reference to a hash that defines reusable custom types.
Custom types allow you to define validation rules once and reuse them throughout your schema,
making your validation logic more maintainable and readable.

Each custom type is defined as a hash reference containing the same validation rules available for regular parameters
(C<type>, C<min>, C<max>, C<matches>, C<memberof>, C<values>, C<enum>, C<notmemberof>, C<callback>, etc.).

  my $custom_types = {
    email => {
      type => 'string',
      matches => qr/^[^@\s]+\@[^@\s]+\.[^@\s]+$/,
      error_msg => 'Invalid email address format'
    }, phone => {
      type => 'string',
      matches => qr/^\+?[1-9]\d{1,14}$/,
      min => 10,
      max => 15
    }, percentage => {
      type => 'number',
      min => 0,
      max => 100
    }, status => {
      type => 'string',
      memberof => ['draft', 'published', 'archived']
    }
  };

  my $schema = {
    user_email => { type => 'email' },
    contact_number => { type => 'phone', optional => 1 },
    completion => { type => 'percentage' },
    post_status => { type => 'status' }
  };

  my $validated = validate_strict(
    schema => $schema,
    input => $input,
    custom_types => $custom_types
  );

Custom types can be extended or overridden in the schema by specifying additional constraints:

  my $schema = {
    admin_username => {
      type => 'username',  # Uses custom type definition
      min => 5,            # Overrides custom type's min value
      max => 15            # Overrides custom type's max value
    }
  };

Custom types work seamlessly with nested schema, optional parameters, and all other validation features.

=back

The schema can define the following rules for each parameter:

=over 4

=item * C<type>

The data type of the parameter.
Valid types are C<string>, C<integer>, C<number>, C<float>, C<boolean>, C<scalar>, C<scalarref>, C<stringref>, C<hashref>, C<arrayref>, C<object>, C<coderef>, C<regex>, C<handle>, C<arraylike>, C<hashlike>, C<codelike>, C<invocant> and C<void>.
C<scalar> accepts any plain scalar value (string, number, boolean, etc.) but rejects references (arrayrefs, hashrefs, coderefs, objects).
C<scalarref> accepts a reference to a scalar value (e.g. C<\$var>) but rejects plain scalars, arrayrefs, hashrefs, coderefs, and objects.
C<stringref> accepts a reference to a scalar that contains a plain string (e.g. C<\$str>) and rejects plain scalars, references-to-references, arrayrefs, hashrefs, coderefs, and objects.
C<void> asserts that the parameter value is C<undef> (the parameter represents a void return or absent output).
When C<void> is used the schema must contain exactly one parameter.
The C<min>/C<max> constraints apply to the B<length> (in characters) of the referenced string.
All other string rules (C<matches>, C<nomatch>, C<memberof>, etc.) operate on the dereferenced string value.
The validated return value is the dereferenced plain string.
C<regex> accepts a compiled regular expression (C<qr//> object); the value is returned unchanged.
C<handle> accepts a file handle: a glob reference with a defined C<fileno>, an C<IO::Handle> subclass instance, or any value for which C<fileno> returns a defined value.
C<arraylike> accepts an array reference or a blessed object that overloads C<@{}> array dereferencing.
C<hashlike> accepts a hash reference or a blessed object that overloads C<%{}> hash dereferencing.
C<codelike> accepts a code reference or a blessed object that overloads C<&{}> code dereferencing.
C<invocant> accepts either a blessed object instance or a plain string that is a syntactically valid Perl class name (e.g. C<'MyApp::Widget'>).

A type can be an arrayref when a parameter could have different types (e.g. a string or an object).

  $schema = {
    username => [
      { type => 'string', min => 3, max => 50 },       # Name
      { type => 'integer', 'min' => 1 },  # UID that isn't root
    ]
  };

As a shorthand, C<type> itself may be an arrayref of type name strings (a I<union type>),
or a pipe-separated string, when all other constraints are shared between the alternatives:

  $schema = {
    data => { type => ['string', 'arrayref'] },
    id   => { type => 'string|integer', optional => 1 },
  };

This is equivalent to the full array-of-rules form but more concise.
Whitespace around the C<|> is ignored, so C<'string | arrayref'> is the same as C<'string|arrayref'>.
Every other key in the rule hash (C<optional>, C<min>, C<max>, C<matches>, etc.)
is inherited by each candidate type and validated independently against it.
Type names are tried left-to-right; the first match wins and its coercion
(e.g. numeric types) is propagated back to the caller.
If the value fails all candidate types, validation croaks with a message
listing the union members.

=item * C<can>

The parameter must be an object that understands the method C<can>.
C<can> can be a simple scalar string of a method name,
or an arrayref of a list of method names, all of which must be supported by the object.

   $schema = {
     gedcom => { type => object, can => 'get_individual' }
   }

=item * C<isa>

The parameter must be an object of type C<isa>.
Requires C<type =E<gt> 'object'>.

=item * C<does>

The parameter must be a blessed object that satisfies the role via C<-E<gt>DOES>.
Requires C<type =E<gt> 'object'>.

  handler => { type => 'object', does => 'My::Role::Printable' }

=item * C<classisa>

The parameter must be a string holding a syntactically valid Perl class name
that passes C<-E<gt>isa('Base::Class')>.
The class must already be loaded (its C<@ISA> must be reachable).
Does not accept blessed object references; use C<isa> for those.

  backend => { type => 'string', classisa => 'My::Backend::Base' }

=item * C<subclass>

Like C<classisa>, but requires a I<strict> subclass: the value must not equal
the base class name itself.

  plugin => { type => 'string', subclass => 'My::Plugin::Base' }

=item * C<classdoes>

Like C<classisa>, but tests C<-E<gt>DOES> (role consumption) instead of C<-E<gt>isa>.

  consumer => { type => 'string', classdoes => 'My::Role::Loggable' }

=item * C<driver>

The parameter must be a valid class name that: (1) can be loaded via C<require>,
and (2) passes C<-E<gt>isa('Base::Class')>.
The module is actually loaded as a side effect of validation.

  store => { type => 'string', driver => 'Cache::Store' }

=item * C<memberof>

The parameter must be a member of the given arrayref.

  status => {
    type => 'string',
    memberof => ['draft', 'published', 'archived']
  }

  priority => {
    type => 'integer',
    memberof => [1, 2, 3, 4, 5]
  }

For string types, the comparison is case-sensitive by default. Use the C<case_sensitive>
flag to control this behavior:

  # Case-sensitive (default) - must be exact match
  code => {
    type => 'string',
    memberof => ['ABC', 'DEF', 'GHI']
    # 'abc' will fail
  }

  # Case-insensitive - any case accepted
  code => {
    type => 'string',
    memberof => ['ABC', 'DEF', 'GHI'],
    case_sensitive => 0
    # 'abc', 'Abc', 'ABC' all pass, original case preserved
  }

For numeric types (C<integer>, C<number>, C<float>), the comparison uses numeric
equality (C<==> operator):

  rating => {
    type => 'number',
    memberof => [0.5, 1.0, 1.5, 2.0]
  }

Note that C<memberof> cannot be combined with C<min> or C<max> constraints as they
serve conflicting purposes - C<memberof> defines an explicit whitelist while C<min>/C<max>
define ranges.

=item * C<enum>

Same as C<memberof>.

=item * C<values>

Same as C<memberof>.

=item * C<notmemberof>

The parameter must not be a member of the given arrayref (blacklist).
This is the inverse of C<memberof>.

  username => {
    type => 'string',
    notmemberof => ['admin', 'root', 'system', 'administrator']
  }

  port => {
    type => 'integer',
    notmemberof => [22, 23, 25, 80, 443]  # Reserved ports
  }

Like C<memberof>, string comparisons are case-sensitive by default but can be controlled
with the C<case_sensitive> flag:

  # Case-sensitive (default)
  username => {
    type => 'string',
    notmemberof => ['Admin', 'Root']
    # 'admin' would pass, 'Admin' would fail
  }

  # Case-insensitive
  username => {
    type => 'string',
    notmemberof => ['Admin', 'Root'],
    case_sensitive => 0
    # 'admin', 'ADMIN', 'Admin' all fail
  }

The blacklist is checked after any C<transform> rules are applied, allowing you to
normalize input before checking:

  username => {
    type => 'string',
    transform => sub { lc($_[0]) },  # Normalize to lowercase
    notmemberof => ['admin', 'root', 'system']
  }

C<notmemberof> can be combined with other validation rules:

  username => {
    type => 'string',
    notmemberof => ['admin', 'root', 'system'],
    min => 3,
    max => 20,
    matches => qr/^[a-z0-9_]+$/
  }

=item * C<case_sensitive>

A boolean value indicating whether string comparisons should be case-sensitive.
This flag affects the C<memberof> and C<notmemberof> validation rules.
The default value is C<1> (case-sensitive).

When set to C<0>, string comparisons are performed case-insensitively, allowing values
with different casing to match. The original case of the input value is preserved in
the validated output.

  # Case-sensitive (default)
  status => {
    type => 'string',
    memberof => ['Draft', 'Published', 'Archived'] # Input 'draft' will fail - must match exact case
  }

  # Case-insensitive
  status => {
    type => 'string',
    memberof => ['Draft', 'Published', 'Archived'],
    case_sensitive => 0 # Input 'draft', 'DRAFT', or 'DrAfT' will all pass
  }

  country_code => {
    type => 'string',
    memberof => ['US', 'UK', 'CA', 'FR'],
    case_sensitive => 0  # Accept 'us', 'US', 'Us', etc.
  }

This flag has no effect on numeric types (C<integer>, C<number>, C<float>) as numbers
do not have case.

=item * C<min>/C<minimum>

The minimum length (for strings in characters not bytes), value (for numbers) or number of keys (for hashrefs).

=item * C<max>

The maximum length (for strings in characters not bytes), value (for numbers) or number of keys (for hashrefs).

=item * C<matches>

A regular expression that the parameter value must match.
Checks all members of arrayrefs.

=item * C<nomatch>

A regular expression that the parameter value must not match.
Checks all members of arrayrefs.

=item * C<bnf>

An arrayref of BNF grammar lines that defines the set of strings the
parameter value must belong to.
The first rule in the grammar is the start rule; the value must match it
exactly (anchored).

Each element is either a rule definition (C<< <name> ::= ... >>) or a
continuation of the previous rule.  Terminals are double-quoted; non-terminals
use angle brackets.  Alternatives are separated by C<|>.

  $schema = {
    na_tel_no => {
      type => 'string',
      bnf  => [
        '<telephone-number> ::= <country-code-opt> <area-code> <separator-opt>',
        '<central-office-code> <separator-opt> <station-code>',
        '<country-code-opt> ::= "" | "+1" | "1"',
        '<separator-opt>    ::= "" | "-" | " " | "."',
        '<area-code>        ::= <digit2-9> <digit0-9> <digit0-9>',
        '<central-office-code> ::= <digit2-9> <digit0-9> <digit0-9>',
        '<station-code>     ::= <digit0-9> <digit0-9> <digit0-9> <digit0-9>',
        '<digit0-9> ::= "0"|"1"|"2"|"3"|"4"|"5"|"6"|"7"|"8"|"9"',
        '<digit2-9> ::= "2"|"3"|"4"|"5"|"6"|"7"|"8"|"9"',
      ],
    },
  };

Implemented by L<Params::Validate::Strict::BNF>.  Recursive grammars are not
supported.

=item * C<position>

For routines and methods that take positional args,
this integer value defines which position the argument will be in.
If this is set for all arguments,
C<validate_strict> will return a reference to an array, rather than a reference to a hash.

=item * C<slurp>

Valid only in positional-argument schemas (those where every parameter has a
C<position> value).  When C<slurp =E<gt> 1> is set, this parameter collects
I<all> remaining positional arguments starting from C<position> into an
arrayref, rather than taking only the single element at that index.

  # sub log_message($level, @messages)
  my $schema = {
    level    => { type => 'string',   position => 0 },
    messages => { type => 'arrayref', position => 1, slurp => 1 },
  };

The slurp parameter is implicitly optional: if there are no arguments at or
beyond C<position>, the value is an empty arrayref.  Combine with C<min =E<gt>
1> to require at least one element:

  messages => { type => 'arrayref', position => 1, slurp => 1, min => 1 }

At most one slurp parameter may be defined per schema, and it must have the
highest C<position> value.  The return value at that position is an arrayref.

=item * C<aliases>

An arrayref of alternative input-key names that are also accepted for this
parameter.  When any alias is found in the input the parameter is stored
under its canonical schema key; if both the canonical name and an alias are
present the canonical name takes precedence.  Aliases are not treated as
unknown parameters regardless of the C<unknown_parameter_handler> setting.

  colour => {
    type    => 'string',
    aliases => ['color'],
    memberof => ['red', 'green', 'blue'],
  }

Only named (hashref) input supports aliases; positional (arrayref) input
ignores them.

=item * C<regex>

Synonym of matches

=item * C<description>

The description of the rule

=item * C<callback>

A code reference to a subroutine that performs custom validation logic.
The subroutine should accept the parameter value, the argument list and the schema as arguments and return true if the value is valid, false otherwise.

Use this to test more complex examples:

  my $schema = {
    even_number => {
      type => 'integer',
      callback => sub { $_[0] % 2 == 0 }
  };

  # Specify the arguments for a routine which has a second, optional argument, which, if given, must be less than or equal to the first
  my $schema = {
    first => {
      type => 'integer'
    }, second => {
      type => 'integer',
      optional => 1,
      callback => sub {
        my($value, $args) = @_;
        # The 'defined' is needed in case 'second' is evaluated before 'first'
        return (defined($args->{first}) && $value <= $args->{first}) ? 1 : 0
      }
    }
  };

=item * C<optional>

A boolean value indicating whether the parameter is optional.
If true, the parameter is not required.
If false or omitted, the parameter is required.

It can be a reference to a code snippet that will return true or false,
to determine if the parameter is optional or not.
The code will be called with two arguments: the value of the parameter and hash ref of all parameters:

  my $schema = {
    optional_field => {
      type => 'string',
      optional => sub {
        my ($value, $all_params) = @_;
        return $all_params->{make_optional} ? 1 : 0;
      }
    },
    make_optional => { type => 'boolean' }
  };

  my $result = validate_strict(schema => $schema, input => { make_optional => 1 });

If the parameter is not optional, it can be passed an undef value, which will not flag an error.
This is by design.
So this will not say that the required parameter 's' is missing:

    validate_strict(
        schema => { s => { type => 'string' } },
        input  => { s => undef },
    );

=item * C<default>

Populate missing optional parameters with the specified value.
Note that this value is not validated.

  username => {
    type => 'string',
    optional => 1,
    default => 'guest'
  }

=item * C<element_type>

Extends the validation to individual elements of arrays.

  tags => {
    type => 'arrayref',
    element_type => 'number',        # Float means the same
    min => 1,        # this is the length of the array, not the min value for each of the numbers. For that, add a C<schema> rule
    max => 5
  }

=item * C<error_msg>

The custom error message to be used in the event of a validation failure.

  age => {
    type => 'integer',
    min => 18,
    error_msg => 'You must be at least 18 years old'
  }

=item * C<nullable>

Like optional,
though this cannot be a coderef,
only a flag.

=item * C<schema>

You can validate nested hashrefs and arrayrefs using the C<schema> property:

    my $schema = {
        user => {    # 'user' is a hashref
            type => 'hashref',
            schema => {      # Specify what the elements of the hash should be
                name => { type => 'string' },
                age => { type => 'integer', min => 0 },
                hobbies => { # 'hobbies' is an array ref that this user has
                    type => 'arrayref',
                    schema => { type => 'string' }, # Validate each hobby
                    min => 1 # At least one hobby
                }
            }
        }, metadata => {
            type => 'hashref',
            schema => {
                created => { type => 'string' },
                tags => {
                    type => 'arrayref',
                    schema => {
                        type => 'string',
                        matches => qr/^[a-z]+$/      # Or you can say matches => '^[a-z]+$'
                    }
                }
            }
        }
    };

=item * C<validate>

A snippet of code that validates the input.
It's passed the input arguments,
and return a string containing a reason for rejection,
or undef if it's allowed.

    my $schema = {
      user => {
        type => 'string',
        validate => sub {
          if($_[0]->{'password'} eq 'bar') {
            return undef;
          }
          return 'Invalid password, try again';
        }
      }, password => {
         type => 'string'
      }
    };

=item * C<transform>

A code reference to a subroutine that transforms/sanitizes the parameter value before validation.
The subroutine should accept the parameter value as an argument and return the transformed value.
The transformation is applied before any validation rules are checked, allowing you to normalize
or clean data before it is validated.

Common use cases include trimming whitespace, normalizing case, formatting phone numbers,
sanitizing user input, and converting between data formats.

  # Simple string transformations
  username => {
    type => 'string',
    transform => sub { lc(trim($_[0])) },  # lowercase and trim
    matches => qr/^[a-z0-9_]+$/
  }

  email => {
    type => 'string',
    transform => sub { lc(trim($_[0])) },  # normalize email
    matches => qr/^[^@\s]+\@[^@\s]+\.[^@\s]+$/
  }

  # Array transformations
  tags => {
    type => 'arrayref',
    transform => sub { [map { lc($_) } @{$_[0]}] },  # lowercase all elements
    element_type => 'string'
  }

  keywords => {
    type => 'arrayref',
    transform => sub {
      my @arr = map { lc(trim($_)) } @{$_[0]};
      my %seen;
      return [grep { !$seen{$_}++ } @arr];  # remove duplicates
    }
  }

  # Numeric transformations
  quantity => {
    type => 'integer',
    transform => sub { int($_[0] + 0.5) },  # round to nearest integer
    min => 1
  }

  # Sanitization
  slug => {
    type => 'string',
    transform => sub {
      my $str = lc(trim($_[0]));
      $str =~ s/[^\w\s-]//g;  # remove special characters
      $str =~ s/\s+/-/g;      # replace spaces with hyphens
      return $str;
    },
    matches => qr/^[a-z0-9-]+$/
  }

  phone => {
    type => 'string',
    transform => sub {
      my $str = $_[0];
      $str =~ s/\D//g;  # remove all non-digits
      return $str;
    },
    matches => qr/^\d{10}$/
  }

The C<transform> function is applied to the value before any validation checks (C<min>/C<minimum>, C<max>,
C<matches>, C<callback>, etc.), ensuring that validation rules are checked against the cleaned data.

Transformations work with all parameter types including nested structures:

  user => {
    type => 'hashref',
    schema => {
      name => {
        type => 'string',
        transform => sub { trim($_[0]) }
      }, email => {
        type => 'string',
        transform => sub { lc(trim($_[0])) }
      }
    }
  }

Transformations can also be defined in custom types for reusability:

  my $custom_types = {
    email => {
      type => 'string',
      transform => sub { lc(trim($_[0])) },
      matches => qr/^[^@\s]+\@[^@\s]+\.[^@\s]+$/
    }
  };

Note that the transformed value is what gets returned in the validated result and is what
subsequent validation rules will check against. If a transformation might fail, ensure it
handles edge cases appropriately.
It is the responsibility of the transformer to ensure that the type of the returned value is correct,
since that is what will be validated.

Many validators also allow a code ref to be passed so that you can create your own, conditional validation rule, e.g.:

  $schema = {
    age => {
      type => 'integer',
      min => sub {
          my ($value, $all_params) = @_;
          return $all_params->{country} eq 'US' ? 21 : 18;
      }
    }
  }

=item * C<validator>

A synonym of C<validate>, for compatibility with L<Data::Processor>.

=item * C<cross_validation>

A reference to a hash that defines validation rules that depend on more than one parameter.
Cross-field validations are performed after all individual parameter validations have passed,
allowing you to enforce business logic that requires checking relationships between different fields.

Each cross-validation rule is a key-value pair where the key is a descriptive name for the validation
and the value is a code reference that accepts a hash reference of all validated parameters.
The subroutine should return C<undef> if the validation passes, or an error message string if it fails.

  my $schema = {
    password => { type => 'string', min => 8 },
    password_confirm => { type => 'string' }
  };

  my $cross_validation = {
    passwords_match => sub {
      my $params = shift;
      return $params->{password} eq $params->{password_confirm}
        ? undef : "Passwords don't match";
    }
  };

  my $validated = validate_strict(
    schema => $schema,
    input => $input,
    cross_validation => $cross_validation
  );

Common use cases include password confirmation, date range validation, numeric comparisons,
and conditional requirements:

  # Date range validation
  my $cross_validation = {
    date_range_valid => sub {
      my $params = shift;
      return $params->{start_date} le $params->{end_date}
        ? undef : "Start date must be before or equal to end date";
    }
  };

  # Price range validation
  my $cross_validation = {
    price_range_valid => sub {
      my $params = shift;
      return $params->{min_price} <= $params->{max_price}
        ? undef : "Minimum price must be less than or equal to maximum price";
    }
  };

  # Conditional required field
  my $cross_validation = {
    address_required_for_delivery => sub {
      my $params = shift;
      if ($params->{shipping_method} eq 'delivery' && !$params->{delivery_address}) {
        return "Delivery address is required when shipping method is 'delivery'";
      }
      return undef;
    }
  };

Multiple cross-validations can be defined in the same hash, and they are all checked in order.
If any cross-validation fails, the function will C<croak> with the error message returned by the validation:

  my $cross_validation = {
    passwords_match => sub {
      my $params = shift;
      return $params->{password} eq $params->{password_confirm}
        ? undef : "Passwords don't match";
    },
    emails_match => sub {
      my $params = shift;
      return $params->{email} eq $params->{email_confirm}
        ? undef : "Email addresses don't match";
    },
    age_matches_birth_year => sub {
      my $params = shift;
      my $current_year = (localtime)[5] + 1900;
      my $calculated_age = $current_year - $params->{birth_year};
      return abs($calculated_age - $params->{age}) <= 1
        ? undef : "Age doesn't match birth year";
    }
  };

Cross-validations receive the parameters after individual validation and transformation have been applied,
so you can rely on the data being in the correct format and type:

  my $schema = {
    email => {
      type => 'string',
      transform => sub { lc($_[0]) }  # Lowercased before cross-validation
    },
    email_confirm => {
      type => 'string',
      transform => sub { lc($_[0]) }
    }
  };

  my $cross_validation = {
    emails_match => sub {
      my $params = shift;
      # Both emails are already lowercased at this point
      return $params->{email} eq $params->{email_confirm}
        ? undef : "Email addresses don't match";
    }
  };

Cross-validations can access nested structures and optional fields:

  my $cross_validation = {
    guardian_required_for_minors => sub {
      my $params = shift;
      if ($params->{user}{age} < 18 && !$params->{guardian}) {
        return "Guardian information required for users under 18";
      }
      return undef;
    }
  };

=item * metadata

Fields starting with <_> are generated by L<App::Test::Generator::SchemaExtractor>,
and are currently ignored.

=item * C<semantic>

A hint about the semantic meaning of the parameter value.
Supported values: C<unix_timestamp>, C<identifier>, C<class_name>.

  ts     => { type => 'integer', semantic => 'unix_timestamp' }
  func   => { type => 'string',  semantic => 'identifier' }
  module => { type => 'string',  semantic => 'class_name' }

When C<semantic> is C<unix_timestamp>, the value must be a non-negative integer no greater than
C<2147483647> (i.e. a valid 32-bit Unix epoch timestamp).
Values outside this range cause the function to C<croak>.

When C<semantic> is C<identifier>, the value must match C</\A[A-Za-z_]\w*\z/>, a single
valid Perl bareword identifier.  Package separators (C<::>) are not permitted; use
C<class_name> for those.

When C<semantic> is C<class_name>, the value must match
C</\A[A-Za-z_]\w*(?:::[A-Za-z_]\w*)*\z/>, a syntactically valid Perl class name such
as C<'Foo'> or C<'Foo::Bar::Baz'>.  The class does not need to be loaded.

Unknown semantic values emit a warning but do not cause an error.

=item * schematic

TODO: gives an idea of what the field will be, e.g. C<filename>.

All cross-validations must pass for the overall validation to succeed.

=item * C<relationships>

A reference to an array that defines validation rules based on relationships between parameters.
Relationship validations are performed after all individual parameter validations have passed,
but before cross-validations.

Each relationship is a hash reference with a C<type> field and additional fields depending on the type:

=over 4

=item * B<mutually_exclusive>

Parameters that cannot be specified together.

  relationships => [
    {
      type => 'mutually_exclusive',
      params => ['file', 'content'],
      description => 'Cannot specify both file and content'
    }
  ]

=item * B<required_group>

At least one parameter from the group must be specified.

  relationships => [
    {
      type => 'required_group',
      params => ['id', 'name'],
      logic => 'or',
      description => 'Must specify either id or name'
    }
  ]

=item * B<conditional_requirement>

If one parameter is specified, another becomes required.

  relationships => [
    {
      type => 'conditional_requirement',
      if => 'async',
      then_required => 'callback',
      description => 'When async is specified, callback is required'
    }
  ]

=item * B<dependency>

One parameter requires another to be present.

  relationships => [
    {
      type => 'dependency',
      param => 'port',
      requires => 'host',
      description => 'port requires host to be specified'
    }
  ]

=item * B<value_constraint>

Specific value requirements between parameters.

  relationships => [
    {
      type => 'value_constraint',
      if => 'ssl',
      then => 'port',
      operator => '==',
      value => 443,
      description => 'When ssl is specified, port must equal 443'
    }
  ]

=item * B<value_conditional>

Parameter required when another has a specific value.

  relationships => [
    {
      type => 'value_conditional',
      if => 'mode',
      equals => 'secure',
      then_required => 'key',
      description => "When mode equals 'secure', key is required"
    }
  ]

=back

If a parameter is optional and its value is C<undef>,
validation will be skipped for that parameter.

If the validation fails, the function will C<croak> with an error message describing the validation failure.

If the validation is successful, the function will return a reference to a new hash containing the validated and (where applicable) coerced parameters.  Integer and number parameters will be coerced to their respective types.

The C<description> field is optional but recommended for clearer error messages.

=back

=head2 Example Usage

  my $schema = {
    host => { type => 'string' },
    port => { type => 'integer' },
    ssl => { type => 'boolean' },
    file => { type => 'string', optional => 1 },
    content => { type => 'string', optional => 1 }
  };

  my $relationships = [
    {
      type => 'mutually_exclusive',
      params => ['file', 'content']
    }, {
      type => 'required_group',
      params => ['host', 'file']
    },
    {
      type => 'dependency',
      param => 'port',
      requires => 'host'
    },
    {
      type => 'value_constraint',
      if => 'ssl',
      then => 'port',
      operator => '==',
      value => 443
    }
  ];

  my $validated = validate_strict(
    schema => $schema,
    input => $input,
    relationships => $relationships
  );

=head1 MIGRATION FROM LEGACY VALIDATORS

=head2 From L<Params::Validate>

    # Old style
    validate(@_, {
        name => { type => SCALAR },
        age => { type => SCALAR, regex => qr/^\d+$/ }
    });

    # New style
    validate_strict(
        schema => {  # or "members"
            name => 'string',
            age => { type => 'integer', min => 0 }
        },
        args => { @_ }
    );

=head2 From L<Type::Params>

    # Old style
    my ($name, $age) = validate_positional \@_, Str, Int;

    # New style - requires converting to named parameters first
    my %args = (name => $_[0], age => $_[1]);
    my $validated = validate_strict(
        schema => { name => 'string', age => 'integer' },
        args => \%args
    );

=cut