NAME

TUWF::Validate - Data and form validation and normalization

DESCRIPTION

This module provides an easy and simple interface for data validation. It can handle most types of data structures (scalars, hashes, arrays and nested data structures), and has some conveniences for validating form-like data.

This module requires no additional modules from CPAN, and can be used stand-alone, outside of the TUWF ecosystem. For integration with TUWF, see the compile() and validate() methods in TUWF::Misc.

Note that this module will not solve all your input validation problems. It can validate the format and the structure of the data, but it does not support validations that depend on other input values. For example, it is not possible to specify that the contents of a password field must be equivalent to that of a confirm_password field, but you can specify that both fields need to be filled out. Recursive data structures are not supported. There is also no built-in support for validating hashes with dynamic keys or arrays where not all elements conform to the same schema. These could technically still be validated with custom validations, but it won't be as convenient.

This module is designed to validate any kind of program input after it has been parsed into a Perl data structure. It should not be used to validate function parameters within Perl code. In fact, the correct answer to "how do I validate function parameters?" is "don't, document your assumptions instead".

API

Validation

TUWF::Validate provides two functions: compile() and validate(), these functions can be called with the full package name, but are also exported on request.

use TUWF::Validate;
state $validator = TUWF::Validate::compile($validations, $schema);
my $result = $validator->validate($input);

# Equivalent:
use TUWF::Validate qw/compile/;
state $validator = compile $validations, $schema;
my $result = $validator->validate($input);

validate() can also be used as a function with three arguments, so you can skip the compilation step:

use TUWF::Validate qw/validate/;
my $result = validate $validations, $schema, $input;

# Is equivalent to:
use TUWF::Validate qr/compile/;
my $result = compile($validations, $schema)->validate($input);

But if you are going to use the same schema to validate multiple inputs, it may be faster to call compile() only once and reuse the compiled $validator object.

In the above examples, $schema is the schema that describes the data to be validated (see "SCHEMA DEFINITION" below), $validations is a hashref containing custom validations that $schema can refer to, $input is the data to be validated, and the $result object is described below.

Both compile() and validate() may throw an error if the $validations or $schema are invalid. Errors in the $input should never cause an error to be thrown, these are always reported in the $result object.

This module takes great care that $input is not being modified in place, even if data normalization is being performed. The normalized data can be read from the $result object.

Result object

The $result object returned by validate() overloads boolean context, so you can check if the validation succeeded with a simple if statement:

my $result = TUWF::Validate::validate(..);
if($result) {
  # Success!
  my $data = $result->data;
} else {
  # Input failed to validate...
  my $error = $result->err;
}

In addition, the result object implements the following methods:

data()

Returns the validated and normalized data. This method will throw an error if validation failed, so if you're lazy and don't want to bother too much with proper error reporting, you can safely validate-and-die in a single step:

my $validated_data = validate(..)->data;

(Note regarding reference semantics: The returned data will usually be a (possibly modified) copy of $input, but may in some cases still have nested references to data in $input - so if you are working with nested hashrefs, arrayrefs or other objects and are going to make modifications to the values embedded within them, these changes may or may not also affect the values in the original $input. Make a deep copy of the data if you're concerned about this).

unsafe_data()

Same as data(), but does not throw an error if validation failed. Instead, it returns the partially validated/normalized data. Can be used to throw the data back at the user in a "Here, this is what I made of it, but I still don't like it so please fix it!" fashion.

err()

Returns undef if validation succeeded, an error object otherwise.

An error object is a hashref containing at least one key: validation, which indicates the name of the validation that failed. Additional keys with more detailed information may be present, depending on the validation. These are documented in "SCHEMA DEFINITION" below.

SCHEMA DEFINITION

A schema is a hashref, each key is the name of a built-in option or of a validation to be performed. None of the options or validations are required, but some built-ins have default values. This means that the empty schema {} is actually equivalent to:

{ type         => 'scalar',
  rmwhitespace => 1,
  required     => 1
}

Built-in options

type => $type

Specify the required type of the input, this can be scalar, array, hash or any. If no type is specified or implied by other validations, the default type is scalar.

Upon failure, the error object will look something like:

{ validation => 'type',
  expected   => 'hash',
  got        => 'scalar'
}

required => 0/1

Whether this input is required to have a value. Specifically, this means exists($x) && defined($x) && $x ne ''. If the input is empty and this option is disabled, the default option is returned when it is set, otherwise the input is simply returned as-is.

As a corollary: Other validations will never get to validate undef or an empty string, these values are either rejected or substituted with a default.

Note that this option is checked after rmwhitespace and before any other validation. So a string containing only whitespace is considered an empty string, and will fail the required test.

Default: true.

default => $val

The value to return if required is false and the input is empty or undef. If $val is a CODE reference, the subroutine will be called with the original input as first argument and its return value will be used for this validation.

onerror => $val

Instead of reporting an error, return $val if this input fails validation for whatever reason. Setting this option in the top-level schema ensures that the validation will always succeed regardless of the input.

If $val is a CODE reference, the subroutine will be called with the result object for this validation as its first argument. The return value of the subroutine will be returned for this validation.

rmwhitespace => 0/1

By default, any whitespace around scalar-type input is removed before testing any other validations. Setting rmwhitespace to a false value will disable this behavior.

keys => $hashref

For type => 'hash', this option specifies which keys are permitted, and how to validate the values. Each key in $hashref corresponds to a key with the same name in the input. Each value is a schema definition by which the value in the input will be validated. The schema definition may be a bare hashref or a validator returned by compile(). If a value with required => 0 is not present in the input hash, it will be created in the output with the default value (or undef).

For example, the following schema specifies that the input must be a hash with three keys:

{ type => 'hash',
  keys => {
    username => { maxlength => 16 },
    password => { minlength => 8 },
    email    => { required => 0, email => 1 }
  }
}

If validation on one or more keys fail, the error object that is returned looks like:

{ validation => 'keys',
  errors => [
    # List of error objects, each with an additional 'key' field.
    { key => 'username', validation => 'required' }
    # In this case, the username was required but either absent or empty.
  ]
}

unknown => $option

For type => 'hash', this option specifies what to do with keys in the input data that have not been defined in the keys option. Possible values are remove to remove unknown keys from the output data (this is the default), reject to return an error if there are unknown keys in the input, or pass to pass through any unknown keys to the output data. Note that the values for passed-through keys will not be validated against any schema!

In the case of reject, the error object will look like:

{ validation => 'unknown',
  # List of unknown keys present in the input
  keys       => ['unknown1', .. ],
  # List of known keys (which may or may not be present
  # in the input - that is checked at a later stage)
  expected   => ['known1', .. ]
}

values => $schema

For type => 'array', this defines the schema that applies to all items in the array. The schema definition may be a bare hashref or a validator returned by compile().

Failure is reported in a similar fashion to keys:

{ validation => 'values',
  errors => [
    { index => 1, validation => 'required' }
  ]
}

scalar => 0/1

For type => 'array', this option will also permit the input to be a scalar. In this case, the input is interpreted and returned as an array with only one element. This option exists to make it easy to validate multi-value form inputs. For example, suppose that we wanted to parse a query string where an option may be present multiple times with different values, like in a=1&b=2&a=3, and suppose that we have a query string parser that, given such a string, would parse that into the following hash:

{ a => [1, 3], b => 1 }

But if a is only specified once, it would parse into a scalar instead of an array. With the scalar option, we can permit a to be a scalar and force it into a single-element array. The following schema definition will validate the above hash:

{ type => 'hash',
  keys => {
    a => { type => 'array', scalar => 1 },
    b => { }
  }
}

sort => $option

For type => 'array', sort the array after validating its elements. $option determines how the array is sorted, possible values are str for string comparison, num for numeric comparison, or a subroutine reference for custom comparison function. The subroutine must be similar to the one given to Perl's sort() function, except it should compare $_[0] and $_[1] instead of $a and $b.

unique => $option

For type => 'array', require elements to be unique. That is, don't allow duplicate elements. There are several ways to specify what uniqueness means in this context:

If $option is a subroutine reference, then the subroutine is given an element as first argument, and it should return a string that is used to check for uniqueness. For example, if array elements are hashes, and you want to check for uniqueness of a hash key named id, you can specify this as unique => sub { $_[0]{id} }.

Otherwise, if $option is true and the sort option is set, then the comparison function used for sorting is also used as uniqueness check. Two elements are the same if the comparison function returns 0.

If $option is true and sort is not set, then the elements will be interpreted as strings, similar to setting unique => sub { $_[0] }.

All of that may sound complicated, but it's quite easy to use. Here's a few examples:

# This describes an array of hashes with keys 'id' and 'name'.
{ type => 'array',
  values => {
    type => 'hash',
    keys => {
      id   => { uint => 1 },
      name => {}
    }
  },
  # Sort the array on 'id'
  sort => sub { $_[0]{id} <=> $_[1]{id} },
  # And require that 'id' fields are unique
  unique => 1
}

# Contrived example: An array of strings, and we want
# each string to start with a different character.
{ type => 'array',
  values => { minlength => 1 },
  unique => sub { substr $_[0], 0, 1 }
}

On failure, this validation returns the following error object. This output assumes the first schema from the previous example.

{ validation => 'unique',
  # Index and value of element a
  index_a => 1,
  value_a => { id => 3, name => 'whatever' }
  # Index and value of duplicate element b
  index_b => 4,
  value_b => { id => 3, name => 'something else' },
  # If string-based uniqueness was used, this is included as well:
  # key => '..'
}

func => $sub

Run the input through a subroutine to perform additional validation or normalization. The subroutine is only called after all other validations have been checked. The subroutine is called with the input as its only argument. Normalization of the input can be done by assigning to the first argument or modifying its value in-place.

On success, the subroutine should return a true value. On failure, it should return either a false value or a hashref. The hashref will have the validation key set to func, and this will be returned as error object.

When func is used inside a custom validation, the returned error object will have its validation field set to the name of the custom validation. This makes custom validations to behave as first-class validations in terms of error reporting.

Standard validations

Standard validations are provided by the module. It is possible to override, re-implement and supplement these with custom validations. Internally, these are, in fact, implemented as custom validations.

regex => $re

Implies type => 'scalar'. Validate the input against a regular expression.

enum => $options

Implies type => 'scalar'. Validate the input against a list of known values. $options can be either a scalar (in which case that is the only permitted input), an array (listing all possible inputs) or a hash (where the hash keys are considered to be the list of permitted inputs).

minlength => $num

Minimum length of the input. The length is the string length() if the input is a scalar, the number of elements if the input is an array, or the number of keys if the input is a hash.

maxlength => $num

Maximum length of the input.

length => $option

If $option is a number, then this specifies the exact length of the input. If $option is an array, then this is a shorthand for [$minlength,$maxlength].

anybool => 1

Accept any value of any type as input, and normalize it to either a 0 or a 1 according to Perl's idea of truth.

undefbool => 1

Like anybool, but missing or empty values are normalized to undef. All other values are normalized to either 0 or 1 according to Perl's idea of truth.

jsonbool => 1

Require the input to be a boolean type returned by a JSON parser. Supported types are JSON::PP, JSON::XS, Types::Serialiser, Cpanel::JSON::XS and boolean.

num => 1

Implies type => 'scalar'. Require the input to be a number formatted using the format permitted by JSON. Note that this is slightly more restrictive from Perl's number formatting, in that 'NaN', 'Inf' and thousand separators are not permitted.

int => 1

Implies type => 'scalar'. Require the input to be an (arbitrarily large) integer.

uint => 1

Implies type => 'scalar'. Require the input to be an (arbitrarily large) positive integer.

min => $num

Implies num => 1. Require the input to be larger than or equal to $num.

max => $num

Implies num => 1. Require the input to be smaller than or equal to $num.

range => [$min,$max]

Equivalent to min => $min, max => $max.

ascii => 1

Implies type => 'scalar'. Require the input to wholly consist of printable ASCII characters.

ipv4 => 1

Implies type => 'scalar'. Require the input to be an IPv4 address.

ipv6 => 1

Implies type => 'scalar'. Require the input to be an IPv6 address. Note that the IP address is not normalized, and fancy features such as IPv4-manned-IPv6 addresses are not permitted.

ip => 1

Require either ipv4 => 1 or ipv6 => 1.

email => 1

Implies type => 'scalar'. Validate the email address against a monstrosity of a regular expression. This email validation is designed to catch obviously invalid addresses and addresses that, while compliant with some RFCs, will not be accepted by most actual SMTP implementations.

Email validation is quite a minefield, see Data::Validate::Email for an alternative solution.

weburl => 1

Implies type => 'scalar'. Requires the input to be a http:// or https:// url.

Custom validations

Custom validations can be passed to compile() and validate() as the $validations hashref argument. A custom validation is, in simple terms, either a schema or a subroutine that returns a schema. The custom validation can then be referenced from other schemas.

Here's a simple example that defines and uses a custom validation named stringbool, which accepts either the string true or false.

my $validations = {
  stringbool => { enum => ['true', 'false'] }
};
my $schema = { stringbool => 1 };
my $result = validate $validations, $schema, 'true';
# $result->data() eq 'true'

A custom validation can also be defined as a subroutine, in which case it can accept options. Here is an example of a prefix custom validation, which requires that the string starts with the given prefix. The subroutine returns a schema that contains the func built-in option to do the actual validation.

my $validations = {
  prefix => sub {
    my $prefix = shift;
    return {
      func => sub { $_[0] =~ /^\Q$prefix/ }
    }
  }
};
my $schema = { prefix => 'Hello, ' };
my $result = validate $validations, $schema, 'Hello, World!';

Custom validations and built-in options

Custom validations can also set built-in options, but the semantics differ a little depending on the option. First, be aware that many of the built-in options apply to the whole schema and not just to the custom validation. For example, if the top-level schema sets rmwhitespace => 0, then all of the validations used in that schema may get input with whitespace around it.

All validations used in a schema need to agree upon a single type option. If a custom validation does not specify a type option (and no type is implied by another validation such as enum or regex), then the validation should work with every type. It is an error to define a schema that mixes validations of different types. For example, the following will throw an error:

compile {}, {
  # top-level schema says we expect a hash
  type => 'hash',
  # but the 'int' validation implies that the type is a scalar
  int => 1
};

The keys, values and func built-in options will be validated separately for each custom validation. So if you have multiple custom validations that set the values option, then the array elements must validate all the listed schemas. The same applies to keys: If the same key is listed in multiple custom validations, then the key must conform to all schemas. With respect to the unknown option, a key that is mentioned in any of the keys options is considered "known".

All other built-in options follow inheritance semantics: These options can be set in a custom validation, and they will be inherited by the top-level schema. If the same option is set in multiple validations, only the first one (in alphabetic order by the name of the validation) will be inherited. The top-level schema can always override options set by custom validations.

SEE ALSO

TUWF.

TUWF::Validate has drawn inspiration from Brannigan. Brannigan is very similar, but slightly more complex and more buggy (and, unfortunately, unmaintained). TUWF::Validate has more detailed error types and more powerful custom validations, but lacks grouping, inheritance and wildcard hash keys.

Sah and Data::Sah provide a more advanced interface for data validation. I have found Sah schemas to not be terribly convenient for form validation. I haven't done any benchmarks, but I suspect that Sah is a bit faster than TUWF::Validate, at the cost of higher memory usage and a large dependency tree.

JSON::Schema is similar to Sah: It features more advanced data structure validation, but the schema is not terribly convenient for form validation, and the module has more dependencies than I'd prefer.

COPYRIGHT

Copyright (c) 2008-2019 Yoran Heling.

This module is part of the TUWF framework and is free software available under the liberal MIT license. See the COPYING file in the TUWF distribution for the details.

AUTHOR

Yoran Heling <projects@yorhel.nl>

cpanm TUWF

perl -MCPAN -e shell
install TUWF