NAME Validator::Declarative - Declarative parameters validation VERSION version 1.20130722.2105 SYNOPSIS sub MakeSomethingCool { my $serialized_parameters; my ( $ace_id, $upm_id, $year, $week, $timestamp_ms ) = Validator::Declarative->validate( \@_ => [ ace_id => 'id', upm_id => 'id', year => 'year', week => 'week', timestamp_ms => [ 'to_msec', 'mdy', 'timestamp' ], ], ); # here all parameters are validated # ....... } DESCRIPTION Almost every function checks the input parameters, in one or other manner. But often checking of some parameters are not made at all or are made not properly. In most cases, checking is done by means of one or more conditional statements for each parameter individually. This reduces the readability of the code and makes it difficult to maintain. Often checking is done using "unless" with several conditions, which make things even worse. Also, lot of conditional statements increases the cyclomatic complexity of the function, which makes it impossible to use automated tests to check the quality and complexity of the code. To solve these problems, we can use declarative description of function parameters. IMPLEMENTATION In general, code for declarative validation definition looks like this: my ($param1_name, $param2_name) = Validator::Declarative->validate( \@_ => [ param1_name => [ validation_definition1 ], param2_name => [ validation_definitions2 ], .... ]); This is usual key=>value pairs, but it should be written as array, not as hash, because order does matter: one pair represents one parameter, and order of pairs should be same as order of parameters in @_. Each validation definition is an array ref. For simple cases, when validation definition is represented by only one rule, we can type less and skip surrounding brackets: my ($param1_name, $param2_name, $param3_name, $param4_name) = Validator::Declarative->validate( \@_ => [ param1_name => 'name_of_rule1', param2_name => { 'name_of_rule2' => param_for_rule2 }, param3_name => { 'name_of_rule3' => [ params_for_rule3 ] }, param4_name => { 'name_of_rule4' => { hash_of_params_for_rule4 } }, .... ]); These are shortcuts for: my ($param1_name, $param2_name, $param3_name, $param4_name) = Validator::Declarative->validate( \@_ => [ param1_name => [ 'name_of_rule1' ], param2_name => [ { 'name_of_rule2' => param_for_rule2 } ], param3_name => [ { 'name_of_rule3' => [ params_for_rule3 ] } ], param4_name => [ { 'name_of_rule4' => { hash_of_params_for_rule4 } } ], .... ]); Grammars Grammars for validation rules are: simple validation_rule ::= 'rule_name' parameterized validation_rule ::= { 'rule_name' => 'parameter' } validation_rule ::= { 'rule_name' => [ 'parameter' ] } validation_rule ::= { 'rule_name' => [ 'param1', 'param2', ... ] } validation_rule ::= { 'rule_name' => { 'param1' => 'param2', ... } } set of rules validation_rule ::= validation_rule, validation_rule, .... Rules Possible kinds of rules are: types (simple and parametrized), converters, constraints. Simple and parametrized rules works only on defined values, for undef all of them return OK (this is needed to support declarations of optional parameters). Simple types any always true, aliases: string bool qr/^(1|true|yes|0|false|no|)$/i, empty string accepted as false, arbitrary data is not allowed float qr/^[+-]?\d+(:?\.\d*)?$/ int qr/^[+-]?\d+$/, aliases: integer positive >0 negative <0 id int && positive email result of Email::Valid->address($_) Simple types (date-like) year int && [1970 .. 3000] week int && [1 .. 53] month int && [1 .. 12] day int && [1 .. 31] ymd like YYYY-MM-DD mdy like M/D/Y (M and D can be 1 or 2 digits, Y can be 2 or 4 digits) time like HH:MM:SS, 00:00:00 ... 23:59:59 hhmm like HH:MM, 00:00 ... 23:59 timestamp almost same as float (because of Time::HiRes), but can't have sign msec timestamp in milliseconds (ts/1000), alias to timestamp Parametrized types min => value minimal accepted value for parameter max => value maximal accepted value for parameter ref => ref_type | [ref_types] ref($_) && ref($_) eq (any of @ref_types) class => class_name | [class_names] blessed($_) && $_->isa(any of @class_names) can => method_name | [method_names] blessed($_) && $_->can(all of @method_names), aliases: ducktype can_any => method_name | [method_names] blessed($_) && $_->can(any of @method_names) any_of => [values] anything from values provided in array ref, aliases: enum list_of => validation_rule list of "values with specific validation check", recursive hash_of => { simple_type => validation_rule } hash of "keys with specific simple type" to "values with specific validation check", recursive hash_of => [ validation_rule => validation_rule ] hash of "keys with specific validation check" to "values with specific validation check", recursive hash => { key => validation_rule, .... } hash with specified key names (not required to exists) and "values with specific validation check", recursive date => format | [formats] date/time in specific format Types ref and class can be used as simple (without parameter), in this case they check whether ref($_) and blessed($_) returns true. Type date can be used as simple (without parameter), in this case it accept all same formats that accepted by any_to_mdy(): /^20\d\d\d\d\d\d$/ /^[+-]?\d{1,10}$/ /^[+-]?\d{11,13}$/ /^\d\d\d\d-?\d\d-?\d\d(?:t\d\d:?\d\d:?\d\d(?:z|\+00)?)?$/i /\d+\D+\d+\D+\d+/ When parameter to date is not skipped, it should be name of any of date-like simple type ('year', 'week', 'mdy' etc) or formatting string for DateTime::Format::Strptime::parse_datetime (example: '%e/%b/%Y:%H:%M:%S %z', see DateTime::Format::Strptime for details). There is no strict requirement for installed DateTime::Format::Strptime - if module can't be loaded, checking with the appropriate format will always lead to a positive result. Converters default => value substitute $_ with provided value (only when actual parameter is undef) assume_true substitute $_ with 0 if it looks like false value (see bool, except for empty string), and 1 otherwise assume_false substitute $_ with 1 if it looks like true value (see bool, except for empty string), and 0 otherwise Constraints required result of defined($_), applied by default optional OK if !defined($_) not_empty for list_of/hash_of/hash: has at least one element for any/string: length($_) > 0 Order of execution Order of rules in validation definition doesn't matters. All specified rules will be executed in this order: 1. Actual parameter is checked to satisfy all constraints. It's error to specify both required and optional at the same time. If none of required and optional were specified, then required is implied. 2. Actual parameter is passed thru converter (if any). It's error to specify more than one converter, except for default. If present, default will be executed at first place. It's error to specify default if there is no optional constraint. 3. Parameter (actual or modified by converter, if any) is checked to satisfy any type (simple or parametrized). If no one type were specified, then any is implied. Order of types in checking is not defined and doesn't matter. First successful check will finish entire validation for this parameter. Errors and logging For any calls all parameters will be checked, and in case of any errors exception should be thrown. Description of all errors will be included into exception text message. METHODS validate(\@params => \@rules) register_type( $name => $code, ...) register_converter( $name => $code, ...) register_constraint( $name => $code, ...) EXAMPLES # Parameter is optional, and can be any type field_name => [ 'any', 'optional' ] # Parameter is optional, and it's id in database field_name => [ 'id', 'optional' ] # Parameter is optional, and it's id in database, with default value field_name => [ 'id', 'optional', {default => undef} ] # Parameter is optional, and it's id or list of ids in database field_name => [ 'id', 'optional', {list_of => 'id'} ] # Parameter is mandatory, and can be any type field_name => 'any' # full form: [ 'required', 'any' ] # Parameter is mandatory, and it's id in database field_name => 'id' # full form: [ 'required', 'id' ] # Parameter is mandatory, and it's id or list of ids in database field_name => [ 'id', {list_of => 'id'} ] # full form: [ 'required', 'id', {list_of => 'id'} ] # Parameter is bool and optional field_name => [ 'bool', 'optional' ] # Parameter is bool and optional, and default is true field_name => [ 'bool', 'optional', {default => 1} ] # Parameter args is mandatory, and it's hash with keys: # - suspensions: not required, hash with keys: # - cssnote_ref: not required, id # - review_deadline: not required, timestamp # - reasons: required, can be id or list of ids # - resumptions: not required, hash with keys: # - cssnote_ref: not required, id # - reasons: required, can be id, list of ids or hash "id to id" # At least one key (suspensions or resumptions) should exists in args. args => [ 'not_empty', { hash => { suspensions => { hash => { cssnote_ref => [ 'optional', 'id' ], review_deadline => [ 'optional', 'timestamp' ], reasons => [ 'id', {list_of => 'id'} ], }}, resumptions => { hash => { cssnote_ref => [ 'optional', 'id' ], reasons => [ 'id', {list_of => 'id'}, {hash_of => {'id' => 'id'}} ], }}, }}] SEE ALSO Inspired by Validator::LIVR - AUTHOR Oleg Kostyuk, "" TODO Implement types list_of, hash_of, hash and date. Implement additional converters, like to_ts, to_mdy and several others. BUGS Please report any bugs or feature requests to Github AUTHOR Oleg Kostyuk COPYRIGHT AND LICENSE This software is Copyright (c) 2013 by Oleg Kostyuk. This is free software, licensed under: The (three-clause) BSD License