NAME
Data::Domain - Data description and validation
SYNOPSIS
# some basic domains
my
$int_dom
= Int(
-min
=> -123,
-max
=> 456);
my
$nat_dom
= Nat(
-max
=> 100,
-default
=>
sub
{
int
(
rand
(100))});
my
$num_dom
= Num(
-min
=> 3.33,
-max
=> 18.5);
my
$string_dom
= String(
-min_length
=> 2);
my
$handle_dom
= Handle;
my
$enum_dom
= Enum(
qw/foo bar buz/
);
my
$int_list_dom
= List(
-min_size
=> 1,
-all
=> Int,
-default
=> [1, 2, 3]);
my
$mixed_list
= List(String, Int(
-min
=> 0), Date, True, Defined);
my
$struct_dom
= Struct(
foo
=> String,
bar
=> Int(
-optional
=> 1));
my
$obj_dom
= Obj(
-can
=>
'print'
);
my
$class_dom
= Class(
-can
=>
'print'
);
# using the domain to check data
my
$error_messages
=
$domain
->inspect(
$some_data
);
reject_form(
$error_messages
)
if
$error_messages
;
# or
die
$domain
->stringify_msg(
$error_messages
)
if
$error_messages
;
# using the domain to get back a tree of validated data
my
$valid_tree
=
$domain
->validate(
$initial_tree
);
# will return a copy with default values inserted;
# will die if there are validation errors
# using the domain for unpacking subroutine arguments
my
$sig
= List(Nat(
-max
=> 20), String(
-regex
=>
qr/^hello/
), Coderef)->func_signature;
sub
some_func {
my
(
$i
,
$s
,
$code
) =
&$sig
;
# or more verbose: = $sig->(@_);
...
}
# using the domain for unpacking method arguments
my
$sig
= List(Nat(
-max
=> 20), String(
-regex
=>
qr/^hello/
), Coderef)->meth_signature;
sub
some_method {
my
(
$self
,
$i
,
$s
,
$code
) =
&$meth_sig
;
# or more verbose: = $meth_sig->(@_);
...
}
# custom name and custom messages (2 different ways)
$domain
= Int(
-name
=>
'age'
,
-min
=> 3,
-max
=> 18,
-messages
=>
"only for people aged 3-18"
);
$domain
= Int(
-name
=>
'age'
,
-min
=> 3,
-max
=> 18,
-messages
=> {
TOO_BIG
=>
"not for old people over %d"
,
TOO_SMALL
=>
"not for babies under %d"
,
});
# examples of subroutines for specialized domains
sub
Phone { String(
-regex
=>
qr/^\+?[0-9() ]+$/
,
-messages
=>
"Invalid phone number"
,
@_
) }
sub
Email { String(
-regex
=>
qr/^[-.\w]+\@[\w.]+$/
,
-messages
=>
"Invalid email"
,
@_
) }
sub
Contact { Struct(
-fields
=> [
name
=> String,
phone
=> Phone,
mobile
=> Phone(
-name
=>
'Mobile'
,
-optional
=> 1),
emails
=> List(
-all
=> Email) ],
@_
) }
sub
UpdateContact { Contact(
-may_ignore
=>
'*'
,
@_
) }
# lazy subdomain
$domain
= Struct(
date_begin
=> Date(
-max
=>
'today'
),
date_end
=>
sub
{
my
$context
=
shift
;
Date(
-min
=>
$context
->{flat}{date_begin})},
);
# recursive domain
my
$expr_domain
;
$expr_domain
= One_of(Num, Struct(
operator
=> String(
qr(^[-+*/]$)
),
left
=>
sub
{
$expr_domain
},
right
=>
sub
{
$expr_domain
}));
# constants in deep datastructures
$domain
= Struct(
foo
=> 123,
# 123 becomes a domain
bar
=> List(Int,
'buz'
, Int) );
# 'buz' becomes a domain
# list with repetitive structure (here : triples)
my
$domain
= List(
-all
=> [String, Int, Obj(
-can
=>
'print'
)]);
DESCRIPTION
A data domain is a description of a set of values, either scalar or structured (arrays or hashes, possibly nested). The description can include many constraints, like minimal or maximal values, regular expressions, required fields, forbidden fields, and also contextual dependencies (for ex. one date must be posterior to another date). From that description, one can then invoke the domain's inspect
method to check if a given value belongs to the domain or not. In case of mismatch, a structured set of error messages is returned, giving detailed explanations about what was wrong.
The motivation for writing this package was to be able to express in a compact way some possibly complex constraints about structured data. Typically the data is a Perl tree (nested hashrefs or arrayrefs) that may come from XML, JSON, from a database through DBIx::DataModel, or from postprocessing an HTML form through CGI::Expand. Data::Domain
is a kind of tree parser on that structure, with some facilities for dealing with dependencies within the structure, and with several options to finely tune the error messages returned to the user.
The main usage for Data::Domain
is to check input from forms in interactive applications. Structured error messages returned by the domain give detailed information about which fields were rejected and why; this can be used to display a new form to the user, highlighting the wrong inputs.
A domain can also validate a datatree, instead of inspecting it. Instead of returning error messages, this returns a copy of the input data, where missing components are replaced by default values (if such defaults where specified within the domain). In case of failure, the validation operation dies with a stringified version of the error messages. This usage is quite similar to type systems like Type::Tiny or Specio, or to parameter validation modules like Params::ValidationCompiler; such systems are more focused on efficiency and on integration with Moose, while the present module is more focused on expressivity for describing constraints on deeply nested structures.
The validation operation can be encapsulates as a signature, which is a reference to an anonymous function that will unpack arguments passed to a subroutine or to a method, will validate them, and will return them in the form of an array or a hash, as demanded by the context. This is probably not as fast nor as elegant as the new "signature" feature introduced in Perl 5.20; but it is a convenient way for performing complex validity tests on parameters received from the caller.
The companion module Test::InDomain uses domains for checking datatrees in the context of automated tests.
There are several other packages in CPAN doing data validation; these are briefly listed in the "SEE ALSO" section.
COMPATIBILITY WARNING : API CHANGE FOR MESSAGE CODEREFS
Starting with version 1.13, the API for calling message coderefs has changed and is now in the form
$coderef
->(
$domain_name
,
$msg_id
,
@args
);
which is incompatible with previous versions of the module. See section "Backward compatibility for message coderefs" for a workaround.
EXPORTS
Domain constructors
# or
# or
Num Int Nat Date Time String
Enum List Struct One_of All_of/
;
Internally, domains are represented as Perl objects; however, it would be tedious to write
my
$domain
= Data::Domain::Struct->new(
anInt
=> Data::Domain::Int->new(
-min
=> 3,
-max
=> 18),
aDate
=> Data::Domain::Date->new(
-max
=>
'today'
),
...
);
so for each builtin domain constructor, Data::Domain
exports a plain function that just calls new
on the appropriate subclass; these functions are all exported in in a group called :constructors
, and allow us to write more compact code :
my
$domain
= Struct(
anInt
=> Int(
-min
=> 3,
-max
=> 18),
aDate
=> Date(
-max
=>
'today'
),
...
);
The list of available domain constructors is expanded below in "BUILTIN DOMAIN CONSTRUCTORS".
Shortcuts (domains with predefined options)
# or
# or
Obj Class/
;
The :shortcuts
export group contains a number of convenience functions that call the "Whatever" domain constructor with various pre-built options. Precise definitions for each of these functions are given below in "BUILTIN SHORTCUTS".
Renaming imported functions
Short function names like Int
, String
, List
, Obj
, True
, etc. are convenient but may cause name clashes with other modules. Thanks to the powerful features of Sub::Exporter, these functions can be renamed in various ways. Here is an example :
my
$domain
= dom_Struct(
anInt
=> dom_Int(
-min
=> 3,
-max
=> 18),
aDate
=> dom_Date(
-max
=>
'today'
),
...
);
There are a number of other ways to rename imported functions; see Sub::Exporter and Sub::Exporter::Tutorial.
Removing symbols from the import list
To preserve backwards compatibility with Exporter, the present module also supports exclamation marks to exclude some specific symbols from the import list. For example
will import everything except the Date
function.
METHODS COMMON TO ALL DOMAINS
new
The new
method creates a new domain object, from one of the domain constructors listed below (Num
, Int
, Date
, etc.). The Data::Domain
class itself has no new
method, because it is an abstract class.
This method is seldom called explicitly; it is usually more convenient to use the wrapper subroutines introduced above, i.e. to write Int(@args)
instead of Data::Domain::Int->new(@args)
. All examples below will use this shorter notation.
Arguments to the new
method may specify various options for the domain to be constructed. Option names always start with a dash. If no option name is given, parameters to the new
method are passed to the default option defined in each constructor subclass. For example the default option in Data::Domain::List
is -items
, so
my
$domain
= List(Int, String, Int);
is equivalent to
my
$domain
= List(
-items
=> [Int, String, Int]);
So in short, the "default option" is syntactic sugar for using positional parameters instead of named parameters.
Each domain constructor has its own list of available options; these will be presented with each subclass (for example options for setting minimal/maximal values, regular expressions, string length, etc.). However, there are also some generic options, available in every domain constructor; these are listed here, in several categories.
Options for customizing the domain behaviour
-optional
-
If true, the domain will accept
undef
, without generating an error message. -default
-
Specifies an default value to be inserted by the "validate" method if the input data is
undef
or nonexistent. For the "inspect" method, this option is equivalent to-optional
.If
-default
is a coderef, that subroutine will be called with the current context as parameter (see "Structure of context"); the resulting scalar value is inserted within the tree. -if_absent
-
Like
-default
except that it will only be applied when a data member does not exist in its parent structure (i.e. a missing field in a hash, or an element outside of the range of an array).This is useful for example when passing named arguments to a function, if you want to explicitly allow to pass
undef
to an argument :some_func(
arg1
=>
'foo'
,
arg2
=>
undef
)
# arg1 is defined, arg2 is undef but present, arg3 is absent
-name
-
Defines a name for the domain, that will be printed in error messages instead of the subclass name.
-messages
-
Defines ad hoc messages for that domain, instead of the builtin messages. The argument can be a string, a hashref or a coderef, as explained in the "CUSTOMIZING ERROR MESSAGES" section.
Options for checking boolean properties
Options in this category check if the data possesses, or does not possess, a given property; hence, the argument to each option must be a boolean. For example, here is a domain that accepts all blessed objects that are not weak references and are not readonly :
$domain
= Whatever(
-blessed
=> 1,
-weak
=> 0,
-readonly
=> 0);
Boolean property options are :
-true
-
Checks if the data is true.
-blessed
-
Checks if the data is blessed, according to "blessed" in Scalar::Util.
-package
-
Checks if the data is a package. This is considered true whenever the data is not a reference and satisfies
$data->isa($data)
. -ref
-
Checks if the data is a reference.
-isweak
-
Checks if the data is a weak reference, according to "isweak" in Scalar::Util.
-readonly
-
Checks if the data is readonly, according to "readonly" in Scalar::Util.
-tainted
-
Checks if the data is tainted, according to "tainted" in Scalar::Util.
Options for checking other general properties
Options in this category do not take a boolean argument, but a class name, method name, role or smart match operand.
-isa
-
Checks if the data is an object or a subclass of the specified class; this is checked through
eval {$data->isa($class)}
. -can
-
Checks if the data implements the listed methods, supplied either as an arrayref (several methods) or as a scalar (just one method); this is checked through
eval {$data->can($method)}
. -does
-
Checks if the data does the supplied role; this is checked through Scalar::Does. Used for example by the "Regexp" and "Coderef" domain shortcuts.
-matches
-
Was originally designed for the smart match operator in Perl 5.10. Smart mach is now deprecated, so this option is now implemented through match::simple.
Options for checking return values
Options in this category call methods or coderefs within the data, and then check the results against the supplied domains. This is somehow contrary to the principle of "domains", because a function call or method call not only inspects the data : it might also alter the data. However, one could also argue that peeking into an object's internals is contrary to the principle of encapsulation, so in this sense, method calls are more appropriate. You decide ... but beware of side-effects in your data!
-has
-
$domain
= Obj(
-has
=> [
foo
=> String,
# ->foo() must return a String
foo
=> [
-all
=> String],
# ->foo() in list context must
# return a list of Strings
[
bar
=> 123] => Obj(
-can
=>
'print'
),
# ->bar(123) must return a printable obj
]);
The
-has
option takes an arrayref argument; that arrayref must contain pairs of($method_spec => $expected_result)
, where$method_spec
is either a method name, or an arrayref containing the method name followed by the list of arguments for calling the method.$expected_result
is either a domain, or an arrayref containing arguments for aList(...)
domain. In the former case, the method call will be performed in scalar context; in the latter case, it will be performed in list context, and the resulting list will be checked against aList
domain built from the given arguments.
Note that this property can be invoked not only on
Obj
, but on any domain; hence, it is possible to simultaneously check if an object has some given internal structure, and also answers to some method calls :$domain
= Struct(
# must be a hashref
-fields
=> {
foo
=> String}
# must have a {foo} key with a String value
-has
=> [
foo
=> String],
# must have a ->foo method that returns a String
);
-returns
-
$domain
= Whatever(
-returns
=> [
[] => String,
[123, 456] => Int,
]);
The
-returns
option treats the data as a coderef. It takes an arrayref argument; that arrayref must contain pairs of($call_spec => $expected_result)
, where$call_spec
is an arrayref containing the list of arguments for calling the subroutine.$expected_result
is either a domain, or an arrayref containing arguments for aList(...)
domain. In the former case, the method call will be performed in scalar context; in the latter case, it will be performed in list context.
inspect
my
$messages
=
$domain
->inspect(
$some_data
);
This method inspects the supplied data, and returns an error message (or a structured collection of messages) if anything is wrong. If the data successfully passed all domain tests, the method returns undef
.
For scalar domains (Num
, String
, etc.), the error message is just a string. For structured domains (List
, Struct
), the return value is an arrayref or hashref of the same structure, like for example
{
anInt
=>
"smaller than mimimum 3"
,
aDate
=>
"not a valid date"
,
aList
=> [
"message for item 0"
,
undef
,
undef
,
"message for item 3"
]}
The client code can then exploit this structure to dispatch error messages to appropriate locations (like for example the form fields from which the data was gathered).
validate
my
$valid_data
=
$domain
->validate(
$some_data
);
This method builds a copy of the supplied data, where missing items are replaced by default values (if such defaults where specified within the domain). If the data is invalid, an error is thrown with a stringified version of the error message.
The returned value is either a scalar or a reference to a nested datastructure (arrayref or hashref).
func_signature
my
$sig_list
= List(...)->func_signature;
sub
some_func {
my
(
$x
,
$y
,
$z
) =
&$sig_list
;
# or $sig_list->(@_);
...
}
my
$sig_hash
= Struct(...)->func_signature;
sub
some_other_func {
my
%args
=
&$sig_hash
;
# or $sig_hash->(@_);
...
}
Returns a reference to an anonymous function that can be used for unpacking arguments passed to a subroutine. The arguments array be encapsulated as an arrayref or hashref, depending on what is expected by the domain, and will be passed to the "validate" method; the result is dereferenced and returned as a list, so that it can be used on the right-hand side of a assignment to variables.
Signatures can be invoked on any list, but in most cases it makes sense to invoke them on the parameters array @_
. This can be done either explicitly :
$sig
->(
@_
);
or it can be done implicitly through Perl's arcane syntax for function calls
&$sig
;
# current @_ is made visible to the $sig subroutine
Arguments unpacking may not work properly for domains that have varying datastructures, like for example Any_of(List(...), Struct(...))
. Such a domain would accept either an arrayref or a hashref, but this cannot be unpacked deterministically by the func_signature
method.
meth_signature
my
$sig_list
= List(...)->meth_signature;
sub
some_meth {
my
(
$self
,
$x
,
$y
,
$z
) =
&$sig_list
;
...
}
This is like "func_signature", except that the first item in @_
is kept apart, since it is a reference to the invocant object or class, and therefore should not be passed to the domain for validation.
stringify_msg
my
$string_msg
=
$domain
->stringify_msg(
$messages
);
die
$string_msg
;
For clients that need a string instead of a datastructur of error messages, method stringify_msg
collects all error information into a single string.
domain stringification
When printed, domains stringify to a compact Data::Dumper representation of their internal attributes; these details can be useful for debugging or logging purposes.
BUILTIN DOMAIN CONSTRUCTORS
Whatever
my
$just_anything
= Whatever;
my
$is_defined
= Whatever(
-defined
=> 1);
my
$is_undef
= Whatever(
-defined
=> 0);
my
$is_true
= Whatever(
-true
=> 1);
my
$is_false
= Whatever(
-true
=> 0);
my
$is_of_class
= Whatever(
-isa
=>
'Some::Class'
);
my
$does_role
= Whatever(
-does
=>
'Some::Role'
);
my
$has_methods
= Whatever(
-can
=> [
qw/jump swim dance sing/
]);
my
$is_coderef
= Whatever(
-does
=>
'CODE'
);
The Data::Domain::Whatever
domain can contain any kind of Perl value, including undef
(actually this is the only domain that contains undef
). The only specific option is :
The Whatever
domain is mostly used together with some of the general options described above, like -true
, -does
, -can
, etc. The most common combinations are encapsulated under their own domain names : see "BUILTIN SHORTCUTS".
Empty
The Data::Domain::Empty
domain always fails when inspecting any data. This is sometimes useful within lazy constructors, like in this example :
Struct(
foo
=> String,
bar
=>
sub
{
my
$context
=
shift
;
if
(some_condition(
$context
)) {
return
Empty(
-messages
=>
'your data is wrong'
)
}
else
{
...
}
}
)
The "LAZY CONSTRUCTORS" section gives more explanations about lazy domains.
Num
my
$domain
= Num(
-range
=>[-3.33, 999],
-not_in
=> [2, 3, 5, 7, 11]);
Domain for numbers (including floats). Numbers are recognized through "looks_like_number" in Scalar::Util. Options for the domain are :
- -min
-
The data must be greater or equal to the supplied value.
- -max
-
The data must be smaller or equal to the supplied value.
- -range
-
-range => [$min, $max]
is equivalent to-min => $min, -max => $max
. - -not_in
-
The data must be different from all values in the exclusion set, supplied as an arrayref.
Int
my
$domain
= Int(
-min
=> -999,
-max
=> 999,
-not_in
=> [2, 3, 5, 7, 11]);
Domain for integers. Integers are recognized through the regular expression /^-?\d+$/
. This domain accepts the same options as Num
and returns the same error messages.
Nat
my
$domain
= Nat(
-max
=> 999);
Domain for natural numbers (i.e. positive integers). Natural numbers are recognized through the regular expression /^\d+$/
. This domain accepts the same options as Num
and returns the same error messages.
Date
Data::Domain::Date->parser(
'EU'
);
# default
my
$domain
= Date(
-min
=>
'01.01.2001'
,
-max
=>
'today'
,
-not_in
=> [
'02.02.2002'
,
'03.03.2003'
,
'yesterday'
]);
Domain for dates, implemented via the Date::Calc module. By default, dates are parsed according to the European format, i.e. through the Decode_Date_EU method; this can be changed by setting
Data::Domain::Date->parser(
'US'
);
# will use Decode_Date_US
or
Data::Domain::Date->parser(\
&your_own_date_parsing_function
);
# that func. should return an array ($year, $month, $day)
Options to this domain are:
- -min
-
The data must be greater or equal to the supplied value. That value can be either a regular date, or one of the special keywords
today
,yesterday
ortomorrow
; these will be replaced by the appropriate date when performing comparisons. - -max
-
The data must be smaller or equal to the supplied value. Of course the same special keywords (as for
-min
) are also admitted. - -range
-
-range => [$min, $max]
is equivalent to-min => $min, -max => $max
. - -not_in
-
The data must be different from all values in the exclusion set, supplied as an arrayref.
When outputting error messages, dates will be printed according to Date::Calc's current language (english by default); see that module's documentation for changing the language.
Time
my
$domain
= Time(
-min
=>
'08:00'
,
-max
=>
'now'
);
Domain for times in format hh:mm:ss
(minutes and seconds are optional).
Options to this domain are:
- -min
-
The data must be greater or equal to the supplied value. The special keyword
now
may be used as a value, and will be replaced by the current local time when performing comparisons. - -max
-
The data must be smaller or equal to the supplied value. The special keyword
now
may also be used as a value. - -range
-
-range => [$min, $max]
is equivalent to-min => $min, -max => $max
.
String
my
$domain
= String(
qr/^[A-Za-z0-9_\s]+$/
);
my
$domain
= String(
-regex
=>
qr/^[A-Za-z0-9_\s]+$/
,
-antiregex
=>
qr/$RE{profanity}/
,
# see Regexp::Common
-range
=> [
'AA'
,
'zz'
],
-length
=> [1, 20],
-not_in
=> [
qw/foo bar/
]);
Domain for strings. Things considered as strings are either scalar values, or objects with an overloaded stringification method; by contrast, a hash reference is not considered to be a string, even if it can stringify to something like "HASH(0x3f9fc4)" or "Some::Class=HASH(0x3f9fc4)" through Perl's internal rules.
Options to this domain are:
- -regex
-
The data must match the supplied compiled regular expression. Don't forget to put
^
and$
anchors if you want your regex to check the whole string.-regex
is the default option, so you may just pass the regex as a single unnamed argument toString()
. - -antiregex
-
The data must not match the supplied regex.
- -min
-
The data must be stringwise greater or equal to the supplied value.
- -max
-
The data must be stringwise smaller or equal to the supplied value.
- -range
-
-range => [$min, $max]
is equivalent to-min => $min, -max => $max
. - -min_length
-
The string length must be greater or equal to the supplied value.
- -max_length
-
The string length must be smaller or equal to the supplied value.
- -length
-
-length => [$min, $max]
is equivalent to-min_length => $min, -max_length => $max
. - -not_in
-
The data must be different from all values in the exclusion set, supplied as an arrayref.
Handle
my
$domain
= Handle();
Domain for filehandles. This domain has no options. Domain membership is checked through "openhandle" in Scalar::Util.
Enum
my
$domain
= Enum(
qw/foo bar buz/
);
Domain for a finite set of scalar values. Options are:
- -values
-
Ref to an array of values admitted in the domain. This would be called as
Enum(-values => [qw/foo bar buz/])
, but since this it is the default option, it can be simply written asEnum(qw/foo bar buz/)
.Undefined values are not allowed in the list (use the
-optional
argument instead).
List
my
$domain
= List(String, Int, String, Num);
my
$domain
= List(
-items
=> [String, Int, String, Num]);
# same as above
my
$domain
= List(
-all
=> String(
qr/^[A-Z]+$/
),
-any
=> String(
-min_length
=> 3),
-size
=> [3, 10]);
my
$domain
= List(
-all
=> [String, Int, Whatever(
-can
=>
'print'
)]);
Domain for lists of values (stored as Perl arrayrefs). Options are:
- -items
-
Ref to an array of domains; then the first n items in the data must match those domains, in the same order.
This is the default option, so item domains may be passed directly to the
new
method, without the-items
keyword. - -min_size
-
The data must be a ref to an array with at least that number of entries.
- -max_size
-
The data must be a ref to an array with at most that number of entries.
- -size
-
-size => [$min, $max]
is equivalent to-min_size => $min, -max_size => $max
. - -all
-
All remaining entries in the array, after the first n entries as specified by the
-items
option (if any), must satisfy the-all
specification. That specification can bea single domain : in that case, all remaining items in the data must belong to that domain
an arrayref of domains : in that case, remaining items in the data are grouped into tuples, and each tuple must satisfy the specification. So the last example above says that the list must contain triples where the first item is a string, the second item is an integer and the third item is an object with a
print
method.
This can also be used for ensuring that the list will not contain any other items after the required items :
List(
-items
=> [Int, Bool, String],
-all
=> Empty);
# cannot have anything after the third item
- -any
-
At least one remaining entry in the array, after the first n entries as specified by the
-items
option (if any), must satisfy that domain specification. A list domain can have both an-all
and an-any
constraint.The argument to
-any
can also be an arrayref of domains, as inList(
-any
=> [String(
qr/^foo/
), Num(
-range
=> [1, 10]) ])
This means that one member of the list must be a string starting with
foo
, and one member of the list must be a number between 1 and 10. Note that this is different fromList(
-any
=> One_of(String(
qr/^foo/
), Num(
-range
=> [1, 10]))
which says that one member of the list must be either a string starting with
foo
or a number between 1 and 10.
Struct
my
$domain
= Struct(
foo
=> Int,
bar
=> String);
my
$domain
= Struct(
-fields
=> {
foo
=> Int,
bar
=> String});
# same as above
my
$domain
= Struct(
-fields
=> [
foo
=> Int,
bar
=> String],
-exclude
=>
'*'
);
# only 'foo' and 'bar', nothing else
my
$domain
= Struct(
-fields
=> [
foo
=> Int,
bar
=> String],
-may_ignore
=>
'*'
);
# will not complain for missing fields
my
$domain
= Struct(
-keys
=> List(
-all
=> String(qr/^[abc])),
-values
=> List(
-all
=> Int));
Domain for associative structures (stored as Perl hashrefs). Options are:
- -fields
-
Supplies a list of fields (hash keys) with their associated domains. The list might be given either as a hashref or as an arrayref. Specifying it as an arrayref is useful for controlling the order in which field checks will be performed; this may make a difference when there are context dependencies (see "LAZY CONSTRUCTORS" below ).
- -exclude
-
Specifies which fields are not allowed in the structure. The exclusion may be specified as an arrayref of field names, as a compiled regular expression, or as the string constant '
*
' or 'all
' (meaning that no hash key will be allowed except those explicitly listed in the-fields
option. The "Struict" domain described below is syntactic sugar for aStruct
domain with option-exclude => '*'
automatically enabled. - -may_ignore
-
Specifies which fields may be ignored by the domain, i.e. may not exist in the inspected structure. Like for
-exclude
, this option can be specified as an arrayref of field names, as a compiled regular expression, or as the string constant '*
' or 'all
'. Absent fields will not generate errors if their name matches this specification. This is especially useful when your application needs to distinguish between an INSERT operation, where all fields must be present, and an UPDATE operation, where only a subset of fields are updated -- see the example in the "SYNOPSIS".Another way is to use the
-optional
flag in domains associated with fields; but there is a subtle difference :-optional
accepts both missing keys or keys containingundef
, while-may_ignore
only accepts missing keys. Consider :Struct(
-fields
=> {
a
=> Int,
b
=> Int(
-optional
=> 1),
c
=> Int,
d
=> Str},
-may_ignore
=> [
qw/c d/
],
)
In this domain,
a
must always be present,b
may be absent or may be undef,c
andd
may be absent but if present cannot be undef. - -keys
-
Specifies a List domain, for inspecting the list of keys in the hash.
- -values
-
Specifies a List domain, for inspecting the list of values in the hash.
In case of errors, the inspect()
method returns a hashref. Errors with specific fields are reported under that field's name; errors with the -exclude
, -keys
or -values
constraints are reported under the constraint's name. So for example in
my
$dom
= Struct(
-fields
=> [
age
=> Int],
-exclude
=>
'*'
);
my
$err
=
$dom
->inspect({
age
=>
'canonical'
,
foo
=> 123,
bar
=> 456});
$err
will contain :
{
age
=>
"Int: invalid number"
,
-exclude
=>
"Struct: contains forbidden field(s): 'bar', 'foo'"
,
}
Struict
my
$domain
= Struict(
foo
=> Int,
bar
=> String);
This is a pun for a "strict Struct" domain : it behaves exactly like /Struct
, except that the option -exclude => '*'
is automatically enabled : therefore the domain is "strict" in the sense that it does not accept any additional key in the input hashref.
One_of
my
$domain
= One_of(
$domain1
,
$domain2
, ...);
Union of domains : successively checks the member domains, until one of them succeeds. Options are:
All_of
my
$domain
= All_of(
$domain1
,
$domain2
, ...);
Intersection of domains : checks all member domains, and requires that all of them succeed. Options are:
BUILTIN SHORTCUTS
Below are the precise definition for the shortcut functions exported in the :shortcuts
group. Each of these functions sets some initial options, but also accepts further options as arguments, so for example it is possible to write something like Obj(-does => 'Storable', -optional => 1)
, which is equivalent to Whatever(-blessed => 1, -does => 'Storable', -optional => 1)
.
True
Whatever(-true => 1)
False
Whatever(-true => 0)
Defined
Whatever(-defined => 1)
Undef
Whatever(-defined => 0)
Blessed
Whatever(-blessed => 1)
Unblessed
Whatever(-blessed => 0)
Regexp
Whatever(-does => 'Regexp')
Obj
Whatever(-blessed => 1)
(synonym to Blessed
)
Class
Whatever(-blessed => 0, -isa => 'UNIVERSAL')
Coderef
Whatever(-does => 'CODE')
LAZY CONSTRUCTORS (CONTEXT DEPENDENCIES)
Principle
If an element of a structured domain (List
or Struct
) depends on another element, then we need to lazily construct that subdomain. Consider for example a struct in which the value of field date_end
must be greater than date_begin
: the subdomain for date_end
can only be constructed when the argument to -min
is known, namely when the domain inspects an actual data structure.
Lazy domain construction is achieved by supplying a subroutine reference instead of a domain object. That subroutine will be called with some context information, and should return the domain object. So our example becomes :
my
$domain
= Struct(
date_begin
=> Date,
date_end
=>
sub
{
my
$context
=
shift
;
Date(
-min
=>
$context
->{flat}{date_begin})}
);
Structure of context
The supplied context is a hashref containing the following information:
- root
-
the overall root of the inspected data
- path
-
the sequence of keys or array indices that led to the current data node. With that information, the subdomain is able to jump to other ancestor or sibling data nodes within the tree (Data::Reach is your friend for doing that).
- flat
-
a flat hash containing an entry for any hash key met so far while traversing the tree. In case of name clashes, most recent keys (down in the tree) override previous keys.
- list
-
a reference to the last list (arrayref) encountered while traversing the tree.
To illustrate this, the following code :
my
$domain
= Struct(
foo
=> List(Whatever,
Whatever,
Struct(
bar
=>
sub
{
my
$context
=
shift
;
Dumper(
$context
);
String;})
)
);
my
$data
= {
foo
=> [
undef
, 99, {
bar
=>
"hello, world"
}]};
$domain
->inspect(
$data
);
will print :
$VAR1
= {
'root'
=> {
'foo'
=> [
undef
, 99, {
'bar'
=>
'hello, world'
}]},
'path'
=> [
'foo'
, 2,
'bar'
],
'list'
=>
$VAR1
->{
'root'
}{
'foo'
},
'flat'
=> {
'bar'
=>
'hello, world'
,
'foo'
=>
$VAR1
->{
'root'
}{
'foo'
}
}
};
Examples of lazy domains
Contextual sets
The domain below accepts hashrefs with a country
and a city
, but also checks that the city actually belongs to the given country :
%SOME_CITIES
= {
Switzerland
=> [
qw/Genève Lausanne Bern Zurich Bellinzona/
],
France
=> [
qw/Paris Lyon Marseille Lille Strasbourg/
],
Italy
=> [
qw/Milano Genova Livorno Roma Venezia/
],
};
my
$domain
= Struct(
country
=> Enum(
keys
%SOME_CITIES
),
city
=>
sub
{
my
$context
=
shift
;
Enum(
-values
=>
$SOME_CITIES
{
$context
->{flat}{country}});
});
Ordered lists
A domain for ordered lists of integers:
my
$domain
= List(
-all
=>
sub
{
my
$context
=
shift
;
my
$index
=
$context
->{path}[-1];
return
$index
== 0 ? Int
: Int(
-min
=>
$context
->{list}[
$index
-1]);
});
The subdomain for the first item in the list has no specific constraint; but the next subdomains have a minimal bound that comes from the previous list item.
Recursive domain
A domain for expression trees, where leaves are numbers, and intermediate nodes are binary operators on subtrees :
my
$expr_domain
;
$expr_domain
= One_of(Num, Struct(
operator
=> String(
qr(^[-+*/]$)
),
left
=>
sub
{
$expr_domain
},
right
=>
sub
{
$expr_domain
}));
Observe that recursive calls to the domain are encapsulated within sub {...}
so that they are treated as lazy domains.
WRITING NEW DOMAIN CONSTRUCTORS
Implementing new domain constructors is fairly simple : create a subclass of Data::Domain
and implement a new
method and an _inspect
method. See the source code of Data::Domain::Num
or Data::Domain::String
for short examples.
However, before writing such a class, consider whether the existing mechanisms are not enough for your needs. For example, many domains could be expressed as a String
constrained by a regular expression; therefore it is just a matter of writing a subroutine that wraps a call to the domain constructor, while supplying some of its arguments :
sub
Phone { String(
-regex
=>
qr/^\+?[0-9() ]+$/
,
-messages
=>
"Invalid phone number"
,
@_
) }
sub
Email { String(
-regex
=>
qr/^[-.\w]+\@[\w.]+$/
,
-messages
=>
"Invalid email"
,
@_
) }
sub
Contact { Struct(
-fields
=> [
name
=> String,
phone
=> Phone,
mobile
=> Phone(
-optional
=> 1),
emails
=> List(
-all
=> Email) ],
@_
) }
Observe that these examples always pass @_
to the domain call : this is so that the client can still add its own arguments to the call, like
$domain
= Phone(
-name
=>
'private phone'
,
-optional
=> 1,
-not_in
=> [ 1234567, 9999999 ]);
CONSTANT SUBDOMAINS
For convenience, elements of List()
or Struct()
may be plain scalar constants, and are automatically translated into constant domains :
$domain
= Struct(
foo
=> 123,
bar
=> List(Int,
'buz'
, Int));
This is exactly equivalent to
$domain
= Struct(
foo
=> Int(
-min
=> 123,
-max
=> 123),
bar
=> List(Int, String(
-min
=>
'buz'
,
-max
=>
'buz'
), Int));
CUSTOMIZING ERROR MESSAGES
Messages returned by validation rules have default values, but can be customized in several ways.
General structure of error messages
Each error message has an internal string identifier, like TOO_SHORT
, NOT_A_HASH
, etc. The section "Message identifiers" below tells which message identifiers may be generated by each domain constructor.
Message identifiers are then associated with user-friendly strings, either within the domain itself, or via a global table. Such strings are actually sprintf format strings, with placeholders for printing some specific details about the validation rule : for example the String
domain defines default messages such as
TOO_SHORT
=>
"less than %d characters"
,
SHOULD_MATCH
=>
"should match '%s'"
,
The -messages
option to domain constructors
Any domain constructor may receive a -messages
option to locally override the messages for that domain. The argument may be
a plain string : that string will be returned for any kind of validation error within the domain
a hashref : keys of the hash should be message identifiers, and values should be the associated error strings. Here is an example :
sub
Phone {
String(
-regex
=>
qr/^\+?[0-9() ]+$/
,
-min_length
=> 7,
-messages
=> {
TOO_SHORT
=>
"phone number should have at least %d digits"
,
SHOULD_MATCH
=>
"invalid chars in phone number"
,
},
@_
);
}
a coderef : the referenced subroutine is called, and should return the error string. The called subroutine receives as arguments:
($domain_name, $message_id, @optional_domain_args)
The messages
class method
Default strings associated with message identifiers are stored in a global table. The Data::Domain
distribution contains builtin tables for english (the default) and for french : these can be chosen through the messages
class method :
Data::Domain->messages(
'english'
);
# the default
Data::Domain->messages(
'français'
);
The same method can also receive a custom table.
my
$custom_table
= {...};
Data::Domain->messages(
$custom_table
);
This should be a two-level hashref : first-level entries in the hash correspond to Data::Domain
subclasses (i.e Num => {...}
, String => {...}
), or to the constant Generic
; for each of those, the second-level entries should correspond to message identifiers as specified in the doc for each subclass (for example TOO_SHORT
, NOT_A_HASH
, etc.). Values should be either strings suitable to be fed to sprintf, or coderefs. Look at $builtin_msgs
in the source code to see an example.
Finally, it is also possible to write your own message generation handler :
Data::Domain->messages(
sub
{
my
(
$domain_name
,
$msg_id
,
@args
) =
@_
;
return
"you just got it wrong ($msg_id)"
});
What is received in @args
depends on which validation rule is involved; it can be for example the minimal or maximal bounds, or the regular expression being checked.
Clearly this class method has a global side-effect. In most cases this is exactly what is expected. However it is possible to limit the impact by localizing the $msgs
class variable :
{
local
$Data::Domain::GLOBAL_MSGS
;
Data::Domain->messages(
$custom_table
);
check_my_data(...);
}
# end of block; Data::Domain is back to the original messages table
Backward compatibility for message coderefs
In the current version of this module, message coderefs are called as
$coderef
->(
$domain_name
,
$msg_id
,
@args
);
Versions prior to 1.13 used a different API where the $domain_name was not available :
$coderef
->(
$msg_id
,
@args
);
So for clients that were using message coderefs in versions prior to 1.13, this is an incompatible change. Backward compatibility can be restored by setting a global variable to a true value :
$Data::Domain::USE_OLD_MSG_API
= 1;
The -name
option to domain constructors
The name of the domain is prepended in front of error messages. The default name is the subclass of Data::Domain
, so a typical error message for a string would be
String: less than 7 characters
However, if a -name
is supplied to the domain constructor, that name will be printed instead;
my
$dom
= String(
-min_length
=> 7,
-name
=>
'Phone'
);
# now error would be: "Phone: less than 7 characters"
Message identifiers
This section lists all possible message identifiers generated by the builtin constructors.
Whatever
-
MATCH_DEFINED
,MATCH_TRUE
,MATCH_ISA
,MATCH_CAN
,MATCH_DOES
,MATCH_BLESSED
,MATCH_SMART
. Num
-
INVALID
,TOO_SMALL
,TOO_BIG
,EXCLUSION_SET
. Date
-
INVALID
,TOO_SMALL
,TOO_BIG
,EXCLUSION_SET
. Time
-
INVALID
,TOO_SMALL
,TOO_BIG
. String
-
TOO_SHORT
,TOO_LONG
,TOO_SMALL
,TOO_BIG
,EXCLUSION_SET
,SHOULD_MATCH
,SHOULD_NOT_MATCH
. Enum
-
NOT_IN_LIST
. List
-
The domain will first check if the supplied array is of appropriate shape; in case of of failure, it will return one of the following scalar messages :
NOT_A_LIST
,TOO_SHORT
,TOO_LONG
.Then it will check all items in the supplied array according to the
-items
and-all
specifications; in case of failure, an arrayref of messages is returned, where message positions correspond to the positions of offending data items.Finally, the domain will check the
-any
constraint; in case of failure, it returns anANY
scalar message. Since that message contains the name of the missing domain, it is a good idea to use the-name
option so that the message is easily comprehensible, as for example inList(
-any
=> String(
-name
=>
"uppercase word"
,
-regex
=>
qr/^[A-Z]$/
))
Here the error message would be : should have at least one uppercase word.
Struct
-
The domain will first check if the supplied hash is of appropriate shape; in case of of failure, it will return one of the following scalar messages :
NOT_A_HASH
,FORBIDDEN_FIELD
.Then it will check all entries in the supplied hash according to the
-fields
specification, and return a hashref of messages, where keys correspond to the keys of offending data items. One_of
-
If all member domains failed to accept the data, an arrayref or error messages is returned, where the order of messages corresponds to the order of the checked domains.
All_of
-
If any member domain failed to accept the data, an arrayref or error messages from all failing subdomains is returned, where the order of messages corresponds to the order of the checked domains.
INTERNALS
Variables
MAX_DEEP
In order to avoid infinite loops, the "inspect" method will raise an exception if $MAX_DEEP
recursive calls were exceeded. The default limit is 100, but it can be changed like this :
local
$Data::Domain::MAX_DEEP
= 999;
Methods
node_from_path (DEPRECATED)
my
$node
= node_from_path(
$root
,
@path
);
Convenience function to find a given node in a data tree, starting from the root and following a path (a sequence of hash keys or array indices). Returns undef
if no such path exists in the tree. Mainly useful for contextual constraints in lazy constructors. Now superseded by Data::Reach.
msg
Internal utility method for generating an error message.
subclass
Method that returns the short name of the subclass of Data::Domain
(i.e. returns 'Int' for Data::Domain::Int
).
name
Returns the -name
domain parameter, or, if absent, the subclass.
_expand_range
Internal utility method for converting a "range" parameter into "min" and "max" parameters.
_build_subdomain
Internal utility method for dynamically converting lazy domains (coderefs) into domains.
SEE ALSO
Doc and tutorials on complex Perl data structures: perlref, perldsc, perllol.
Other CPAN modules doing data validation : Data::FormValidator, CGI::FormBuilder, HTML::Widget::Constraint, Jifty::DBI, Data::Constraint, Declare::Constraints::Simple, Moose::Manual::Types, Smart::Match, Test::Deep, Params::Validate, Validation::Class.
Among those, Declare::Constraints::Simple
is the closest to Data::Domain
, because it is also designed to deal with substructures; yet it has a different approach to combinations of constraints and scope dependencies.
Some inspiration for Data::Domain
came from the wonderful Parse::RecDescent module, especially the idea of passing a context where individual rules can grab information about neighbour nodes. Ideas for some features were borrowed from Test::Deep and from Moose::Manual::Types.
ACKNOWLEDGEMENTS
Thanks to
David Cantrell and Gabor Szabo for their help on issues related to smartmatch deprecation.
David Schmidt (davewood) for suggesting extensions to the Struct() domain.
AUTHOR
Laurent Dami, <dami at cpan.org>
COPYRIGHT AND LICENSE
Copyright 2006-2024 by Laurent Dami.
This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.