Steven Haryanto


Data::Schema - (DEPRECATED) Validate nested data structures with nested structure


version 0.136


    # OO interface
    use Data::Schema;
    my $ds = Data::Schema->new();
    my $schema = [array => {min_len=>2, max_len=>4}];
    my $data = [1, 2, 3];
    my $res = $ds->validate($data, $schema);
    print "valid!" if $res->{success}; # prints 'valid!'

    # procedural interface
    use Data::Schema;
    my $sch = ["hash",
               {keys =>
                    {name => "str",
                     age  => ["int", {required=>1, min=>18}]
    my $r;
    $r = ds_validate({name=>"Lucy", age=>18}, $sch); # success
    $r = ds_validate({name=>"Lucy"         }, $sch); # fail: missing age
    $r = ds_validate({name=>"Lucy", age=>16}, $sch); # fail: underage

    # some schema examples

    # -- array

    # -- array of ints
    [array => {of=>"int"}]

    # -- array of positive, even ints
    [array => {of=>[int => {min=>0, divisible_by=>2}]}]

    # -- 3x3x3 "multi-dim" arrays
    [array => {len=>3, of=>
        [array => {len=>3, of=>
            [array => {len=>3}]}]}]

    # -- HTTP headers, each header can be a string or array of strings
    [hash => {
        required => 1,
        keys_match => '^\w+(-w+)*$',
        values_of => [either => {of=>[
            [array=>{of=>"str", minlen=>1}],

    # -- records (demonstrates subschema and attribute merging). Note:
    # I am not sexist or anything, just that for the love of g*d I
    # can't think of a better example atm. it's late...
    {def => {
        person => [hash => {
            keys => {
                name       => "str",
                race       => "str",
                age        => [int => {min=>0, max=>100}],

        # women are like people, but they have additional keys
        # 'husband' and 'cup_size' (additive) and different age
        # restriction (replace).

        woman => [person => {
            '*keys' => {
                husband    => "str",
                cup_size   => [str => {one_of=>[qw/AA A B C D DD/]}],
                '*age'     => [int => {min=>0, max=>120}],

        # girls are like women, but they do not have husbands yet
        # (remove keys)

        girl => [woman => {
            '*keys' => {
                '!husband' => undef,

        girls  => [array => {of=>"girl"}],
    type => "girls",



Data::Schema (DS) is a schema system for data validation. It lets you write schemas as data structures, ranging from very simple (a scalar) to fairly complex (nested hashes/arrays with various criteria).

Writing schemas as data structures themselves has several advantages. First, it is more portable across languages (e.g. using YAML to share schemas between Perl, Python, PHP, Ruby). Second, you can validate the schema using the schema system itself. Third, it is easy to generate code, help message (e.g. so-called "usage" for function/command line script), etc. from the schema.

Potential application of DS: validating configuration, function parameters, command line arguments, etc.

To get started, see Data::Schema::Manual::Tutorial.


When importing this module, you can pass a list of module names.

 use Data::Schema qw(Plugin::Foo Type::Bar Schema::Baz ...);
 my $ds = Data::Schema->new; # foo, bar, baz will be loaded by default

This is a shortcut to the more verbose form:

 use Data::Schema;
 my $ds = Data::Schema->new;


 $ds->register_type('bar', 'Data::Schema::Type::Bar');

 use Data::Schema::Schema::Baz;
 $ds->register_schema_as_type($_, $Data::Schema::Schema::Baz::DS_SCHEMAS->{$_})
    for keys %$Data::Schema::Schema::Baz::DS_SCHEMAS;


ds_validate($data, $schema)

Non-OO wrapper for validate(). Exported by default. See validate() method.



Configuration object. See Data::Schema::Config.



Merge several attribute hashes if there are hashes that can be merged (i.e. contains merge prefix in its keys). Used by DST::Base and DST::Schema. As DS user, normally you wouldn't need this.


Initialize validation state. Used internally by validate(). As DS user, normally you wouldn't need this.


Save validation state (position in data, position in schema, number of errors, etc) into a stack, so that you can start using the validator to validate a new data with a new schema, even in the middle of validating another data/schema. Used internally by validate() and DST::Schema. As DS user, normally you wouldn't need this.

See also: restore_validation_state().


Restore the last validation state from the stack. Used internally by validate() and DST::Schema. As DS user, normally you wouldn't need this.

See also: save_validation_state().


Initialize compilation state. Used internally by emit_perl(). As DS user, normally you wouldn't need this.


Save compilation state. Used internally by emit_perl() and DST::Schema. As DS user, normally you wouldn't need this.

See also: restore_compilation_state().


Restore the last compilation state from the stack. Used internally by emit_perl() and DST::Schema. As DS user, normally you wouldn't need this.

See also: save_compilation_state().


Add a data error when in validation process. Will not add if there are already too many errors (too_many_errors attribute is true). Used by type handlers. As DS user, normally you wouldn't need this.


Add a data warning when in validation process. Will not add if there are already too many warnings (too_many_warnings attribute is true). Used by type handlers. As DS user, normally you wouldn't need this.

debug($message[, $level])

Log debug messages. Used by type handlers when validating. As DS user, normally you wouldn't need this.


Method to call when encountering schema error during validation/compilation. Used by type handlers. As DS user, normally you wouldn't need this.


Checks whether $name is a valid type name. Returns true if valid, false if invalid. By default it requires that type name starts with a lowercase letter and contains only lowercase letters, numbers, and underscores. Maximum length is 64.

You can override this method if you want stricter/looser type name criteria.

register_type($name, $class|$obj)

Register a new type, along with a class name ($class) or the actual object ($obj) to handle the type. If $class is given, the class will be require'd and instantiated to become object later when needed via get_type_handler.

Any object can become a type handler, as long as it has:

  • a validator() rw property to store/set validator object;

  • handle_type() method to handle type checking;

  • zero or more handle_attr_*() methods to handle attribute checking.

See Data::Schema::Manual::TypeHandler for more details on writing a type handler.


Register a new plugin. Accept a plugin object or class. If $class is given, the class will be require'd (if not already require'd) and instantiated to become object.

Any object can become a plugin, you don't need to subclass from anything, as long as it has:

  • a validator() rw property to store/set validator object;

  • zero or more handle_*() methods to handle some events/hooks.

See Data::Schema::Manual::Plugin for more details on writing a plugin.

call_handler($name, [@args])

Try handle_*() method from each registered plugin until one returns 0 or 1. If a plugin return -1 (decline) then we continue to the next plugin. Returns the status of the last plugin. Returns -1 if there's no handler to invoke.


Try to get type handler for a certain type. If type handler is not an object (a class name), instantiate it first. If type is not found, invoke handle_unknown_type() in plugins to give plugins a chance to load the type. If type is still not found, return undef.


Normalize a schema into the third form (hash form) ({type=>..., attr_hashes=>..., def=>...) as well as do some sanity checks on it. Returns an error message string if fails.

register_schema_as_type($schema, $name)

Register schema as new type. $schema is a normalized schema. Return {success=>(0 or 1), error=>...}. Fails if type with name $name is already defined, or if $schema cannot be parsed. Might actually register more than one type actually, if the schema contains other types in it (hash form of schema can define types).

validate($data[, $schema])

Validate a data structure. $schema must be given unless you already give the schema via the schema attribute.

Returns {success=>0 or 1, errors=>[...], warnings=>[...]}. The 'success' key will be set to 1 if the data validates, otherwise 'errors' will be filled with the details.


Return formatted errors in an array of strings.


Return formatted warnings in an array of strings.


Return formatted logs in an array of strings.


Return Perl code equivalent to schema $schema.

If you want to get the compiled code (as a coderef) directly, use compile.


Compile the schema into Perl code and return a 2-element list: ($coderef, $subname). $coderef is the resulting subroutine and $subname is the subroutine name in the compilation namespace (Data::Schema::__compiled).

If the same schema is already compiled, the existing compiled subroutine is returned instead.

Dies if code can't be generated, or an error occured when compiling the code.

If you just want to get the Perl code in a string, use emit_perl.


There are already a lot of data validation modules on CPAN. However, most of them do not validate nested data structures. Many seem to focus only on "form" (which is usually presented as shallow hash in Perl).

And of the rest which do nested data validation, either I am not really fond of the syntax, or the validator/schema system is not simple/flexible/etc enough for my taste. For example, other data validation modules might require you to always write:

 { type => "int" }

even when all you want is just validating an int with no other extra requirements. With DS you can just write:


Another design consideration for DS is, I want to maximize reusability of my schemas. And thus DS allows you to define schemas in terms of other schemas. External schemas can be "require"-d from Perl variables or loaded from YAML files. Of course, you can also extend with Perl as usual (e.g. writing new types and new attributes).






Some other data validation modules on CPAN: Data::FormValidator, Data::Rx, Kwalify.

Config::Tree uses Data::Schema to check command-line options and makes it easy to generate --help/usage information.

LUGS::Events::Parser by Steven Schubiger is apparently one of the first modules (outside my own of course) which use Data::Schema.

Data::Schema::Schema:: namespace is reserved for modules that contain DS schemas. For example, Data::Schema::Schema::CPANMeta validates CPAN META.yml. Data::Schema::Schema::Schema contains the schema for DS schema itself.


Please report any bugs or feature requests to bug-data-schema at, or through the web interface at I will be notified, and then you'll automatically be notified of progress on your bug as I make changes.


You can find documentation for this module with the perldoc command.

    perldoc Data::Schema

You can also look for information at:



  Steven Haryanto <>


This software is copyright (c) 2009 by Steven Haryanto.

This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.

Hosting generously
sponsored by Bytemark