Params::Validate - Validate method/function parameters
use Params::Validate qw(:all);
# takes named params (hash or hashref)
sub foo
{
validate( @_, { foo => 1, # mandatory
bar => 0, # optional
}
);
}
# takes positional params
sub bar
{
# first two are mandatory, third is optional
validate_pos( @_, 1, 1, 0 );
}
sub foo2
{
validate( @_,
{ foo =>
# specify a type
{ type => ARRAYREF },
bar =>
# specify an interface
{ can => [ 'print', 'flush', 'frobnicate' ] },
baz =>
{ type => SCALAR, # a scalar ...
# ... that is a plain integer ...
regex => qr/^\d+$/,
callbacks =>
{ # ... and smaller than 90
'less than 90' => sub { shift() < 90 },
},
}
}
);
}
sub with_defaults
{
my %p = validate( @_, { foo => 1, # required
# $p{bar} will be 99 if bar is not
# given. bar is now optional.
bar => { default => 99 } } );
}
sub pos_with_defaults
{
my @p = validate_pos( @_, 1, { default => 99 } );
}
sub sets_options_on_call
{
my %p = validate_with
( params => \@_,
spec => { foo => { type SCALAR, default => 2 } },
normalize_keys => sub { $_[0] =~ s/^-//; lc $_[0] },
);
}
The Params::Validate module allows you to validate method or function call parameters to an arbitrary level of specificity. At the simplest level, it is capable of validating the required parameters were given and that no unspecified additional parameters were passed in.
It is also capable of determining that a parameter is of a specific type, that it is an object of a certain class hierarchy, that it possesses certain methods, or applying validation callbacks to arguments.
The module always exports the validate() and validate_pos()
functions.
It also has an additional function available for export,
validate_with, which can be used to validate any type of
parameters, and set various options on a per-invocation basis.
In addition, it can export the following constants, which are used as
part of the type checking. These are SCALAR, ARRAYREF,
HASHREF, CODEREF, GLOB, GLOBREF, and SCALARREF,
UNDEF, OBJECT, BOOLEAN, and HANDLE. These are explained
in the section on Type Validation.
The constants are available via the export tag :types. There is
also an :all tag which includes all of the constants as well as the
validation_options() function.
The validation mechanisms provided by this module can handle both named or positional parameters. For the most part, the same features are available for each. The biggest difference is the way that the validation specification is given to the relevant subroutine. The other difference is in the error messages produced when validation checks fail.
When handling named parameters, the module will accept either a hash or a hash reference.
Subroutines expecting named parameters should call the validate()
subroutine like this:
validate( @_, { parameter1 => validation spec,
parameter2 => validation spec,
...
} );
Subroutines expecting positional parameters should call the
validate_pos() subroutine like this:
validate_pos( @_, { validation spec }, { validation spec } );
If you just want to specify that some parameters are mandatory and others are optional, this can be done very simply.
For a subroutine expecting named parameters, you would do this:
validate( @_, { foo => 1, bar => 1, baz => 0 } );
This says that the "foo" and "bar" parameters are mandatory and that the "baz" parameter is optional. The presence of any other parameters will cause an error.
For a subroutine expecting positional parameters, you would do this:
validate_pos( @_, 1, 1, 0, 0 );
This says that you expect at least 2 and no more than 4 parameters. If you have a subroutine that has a minimum number of parameters but can take any maximum number, you can do this:
validate_pos( @_, 1, 1, (0) x (@_ - 2) );
This will always be valid as long as at least two parameters are given. A similar construct could be used for the more complex validation parameters described further on.
Please note that this:
validate_pos( @_, 1, 1, 0, 1, 1 );
makes absolutely no sense, so don't do it. Any zeros must come at the end of the validation specification.
In addition, if you specify that a parameter can have a default, then it is considered optional.
This module supports the following simple types, which can be exported as constants:
10 or 'hello'. A
parameter that is undefined is not treated as a scalar. If you
want to allow undefined values, you will have to specify SCALAR |
UNDEF.
[1, 2, 3] or \@foo.
{ a => 1, b => 2 } or \%bar.
\&foo_sub or sub { print "hello" }.
This one is a bit tricky. A glob would be something like *FOO, but
not \*FOO, which is a glob reference. It should be noted that this
trick:
my $fh = do { local *FH; };
makes $fh a glob, not a glob reference. On the other hand, the
return value from Symbol::gensym is a glob reference. Either can
be used as a file or directory handle.
\*FOO. See the GLOB entry above
for more details.
\$x.
UNDEF | SCALAR.
GLOB |
GLOBREF. However, it seems likely that most people interested in
either globs or glob references are likely to really be interested in
whether the parameter in question could be a valid file or directory
handle.
To specify that a parameter must be of a given type when using named parameters, do this:
validate( @_, { foo => { type => SCALAR },
bar => { type => HASHREF } } );
If a parameter can be of more than one type, just use the bitwise or
(|) operator to combine them.
validate( @_, { foo => { type => GLOB | GLOBREF } );
For positional parameters, this can be specified as follows:
validate_pos( @_, { type => SCALAR | ARRAYREF }, { type => CODEREF } );
To specify that a parameter is expected to have a certain set of methods, we can do the following:
validate( @_,
{ foo =>
# just has to be able to ->bar
{ can => 'bar' } } );
... or ...
validate( @_,
{ foo =>
# must be able to ->bar and ->print
{ can => [ qw( bar print ) ] } } );
A word of warning. When constructing your external interfaces, it is probably better to specify what methods you expect an object to have rather than what class it should be of (or a child of). This will make your API much more flexible.
With that said, if you want to validate that an incoming parameter belongs to a class (or child class) or classes, do:
validate( @_,
{ foo =>
{ isa => 'My::Frobnicator' } } );
... or ...
validate( @_,
{ foo =>
{ isa => [ qw( My::Frobnicator IO::Handle ) ] } } );
# must be both, not either!
If you want to specify that a given parameter must match a specific regular expression, this can be done with "regex" spec key. For example:
validate( @_,
{ foo =>
{ regex => qr/^\d+$/ } } );
The value of the "regex" key may be either a string or a pre-compiled
regex created via qr.
If the value being checked against a regex is undefined, the regex is explicitly checked against the empty string ('') instead, in order to avoid "Use of uninitialized value" warnings.
The Regexp::Common module on CPAN is an excellent source of regular
expressions suitable for validating input.
If none of the above are enough, it is possible to pass in one or more callbacks to validate the parameter. The callback will be given the value of the parameter as its first argument. Its second argument will be all the parameters, as a reference to either a hash or array. Callbacks are specified as hash reference. The key is an id for the callback (used in error messages) and the value is a subroutine reference, such as:
validate( @_,
{ foo =>
{ callbacks =>
{ 'smaller than a breadbox' => sub { shift() < $breadbox },
'green or blue' =>
sub { $_[0] eq 'green' || $_[0] eq 'blue' } } } );
validate( @_,
{ foo =>
{ callbacks =>
{ 'bigger than baz' => sub { $_[0] > $_[1]->{baz} } } } } );
If you want values untainted, set the "untaint" key in a spec hashref to a true value, like this:
my %p =
validate( @_, { foo =>
{ type => SCALAR, untaint => 1 },
bar =>
{ type => ARRAYREF } } );
This will untaint the "foo" parameter if the parameters are valid.
Note that untainting is only done if all parameters are valid. Also, only the return values are untainted, not the original values passed into the validation function.
Asking for untainting of a reference value will not do anything, as
Params::Validate will only attempt to untaint the reference itself.
If you want to specify something such as type or interface, plus the fact that a parameter can be optional, do this:
validate( @_, { foo =>
{ type => SCALAR },
bar =>
{ type => ARRAYREF, optional => 1 } } );
or this for positional parameters:
validate_pos( @_, { type => SCALAR }, { type => ARRAYREF, optional => 1 } );
By default, parameters are assumed to be mandatory unless specified as optional.
It also possible to specify that a given optional parameter depends on the presence of one or more other optional parameters.
validate( @_, { cc_number =>
{ type => SCALAR, optional => 1,
depends => [ 'cc_expiration', 'cc_holder_name' ],
},
cc_expiration
{ type => SCALAR, optional => 1 },
cc_holder_name
{ type => SCALAR, optional => 1 },
} );
In this case, "cc_number", "cc_expiration", and "cc_holder_name" are all optional. However, if "cc_number" is provided, then "cc_expiration" and "cc_holder_name" must be provided as well.
This allows you to group together sets of parameters that all must be provided together.
The validate_pos() version of dependencies is slightly different,
in that you can only depend on one other parameter. Also, if for
example, the second parameter 2 depends on the fourth parameter, then
it implies a dependency on the third parameter as well. This is
because if the fourth parameter is required, then the user must also
provide a third parameter so that there can be four parameters in
total.
Params::Validate will die if you try to depend on a parameter not
declared as part of your parameter specification.
If the validate() or validate_pos() functions are called in a
list context, they will return an array or hash containing the
original parameters plus defaults as indicated by the validation spec.
If the function is not called in a list context, providing a default in the validation spec still indicates that the parameter is optional.
The hash or array returned from the function will always be a copy of
the original parameters, in order to leave @_ untouched for the
calling function.
Simple examples of defaults would be:
my %p = validate( @_, { foo => 1, bar => { default => 99 } } );
my @p = validate( @_, 1, { default => 99 } );
In scalar context, a hash reference or array reference will be returned, as appropriate.
By default, when validation fails Params::Validate calls
Carp::confess(). This can be overridden by setting the on_fail
option, which is described in the "GLOBAL" OPTIONS
section.
When using this module to validate the parameters passed to a method
call, you will probably want to remove the class/object from the
parameter list before calling validate() or validate_pos().
If your method expects named parameters, then this is necessary for
the validate() function to actually work, otherwise @_ will not
be useable as a hash, because it will first have your object (or
class) followed by a set of keys and values.
Thus the idiomatic usage of validate() in a method call will look
something like this:
sub method
{
my $self = shift;
my %params = validate( @_, { foo => 1, bar => { type => ARRAYREF } } );
}
Because the API for the validate() and validate_pos() functions
does not make it possible to specify any options other than the the
validation spec, it is possible to set some options as
pseudo-'globals'. These allow you to specify such things as whether
or not the validation of named parameters should be case sensitive,
for one example.
These options are called pseudo-'globals' because these settings are only applied to calls originating from the package that set the options.
In other words, if I am in package Foo and I call
validation_options(), those options are only in effect when I call
validate() from package Foo.
While this is quite different from how most other modules operate, I feel that this is necessary in able to make it possible for one module/application to use Params::Validate while still using other modules that also use Params::Validate, perhaps with different options set.
The downside to this is that if you are writing an app with a standard
calling style for all functions, and your app has ten modules, each
module must include a call to validation_options(). You could of
course write a module that all your modules use which uses various
trickery to do this when imported.
This option is only relevant when dealing with named parameters.
This callback will be used to transform the hash keys of both the
parameters and the parameter spec when validate() or
validate_with() are called.
Any alterations made by this callback will be reflected in the parameter hash that is returned by the validation function. For example:
sub foo {
return
validate_with( params => \@_,
spec => { foo => { type => SCALAR } },
normalize_keys =>
sub { my $k = shift; $k =~ s/^-//; return uc $k },
);
}
%p = foo( foo => 20 );
# $p{FOO} is now 20
%p = foo( -fOo => 50 );
# $p{FOO} is now 50
The callback must return a defined value.
If a callback is given than the deprecated "ignore_case" and "strip_leading" options are ignored.
If given, this callback will be called whenever a validation check fails. It will be called with a single parameter, which will be a string describing the failure. This is useful if you wish to have this module throw exceptions as objects rather than as strings, for example.
This callback is expected to die() internally. If it does not, the
validation will proceed onwards, with unpredictable results.
The default is to simply use the Carp module's confess() function.
validate() or
validate_pos(). If this option is set, then the given number of
frames are skipped instead.
DEPRECATED
This is only relevant when dealing with named parameters. If it is true, then the validation code will ignore the case of parameter names. Defaults to false.
DEPRECATED
This too is only relevant when dealing with named parameters. If this
is given then any parameters starting with these characters will be
considered equivalent to parameters without them entirely. For
example, if this is specified as '-', then -foo and foo would be
considered identical.
The validate_with() function can be used to set the options listed
above on a per-invocation basis. For example:
my %p =
validate_with
( params => \@_,
spec => { foo => { type => SCALAR },
bar => { default => 10 } },
allow_extra => 1,
);
In addition to the options listed above, it is also possible to set the option "called", which should be a string. This string will be used in any error messages caused by a failure to meet the validation spec.
This subroutine will validate named parameters as a hash if the "spec" parameter is a hash reference. If it is an array reference, the parameters are assumed to be positional.
my %p =
validate_with
( params => \@_,
spec => { foo => { type => SCALAR },
bar => { default => 10 } },
allow_extra => 1,
called => 'The Quux::Baz class constructor',
);
my @p =
validate_with
( params => \@_,
spec => [ { type => SCALAR },
{ default => 10 } ],
allow_extra => 1,
called => 'The Quux::Baz class constructor',
);
If the environment variable PERL_NO_VALIDATION is set to something
true, then validation is turned off. This may be useful if you only
want to use this module during development but don't want the speed
hit during production.
The only error that will be caught will be when an odd number of parameters are passed into a function/method that expects a hash.
If you want to selectively turn validation on and off at runtime, you
can directly set the $Params::Validate::NO_VALIDATION global
variable. It is strongly recommended that you localize any
changes to this variable, because other modules you are using may
expect validation to be on when they execute. For example:
{
local $Params::Validate::NO_VALIDATION = 1;
# no error
foo( bar => 2 );
}
# error
foo( bar => 2 );
sub foo
{
my %p = validate( @_, { foo => 1 } );
...
}
But if you want to shoot yourself in the foot and just turn it off, go ahead!
Right now there is no way (short of a callback) to specify that something must be of one of a list of classes, or that it must possess one of a list of methods. If this is desired, it can be added in the future.
Ideally, there would be only one validation function. If someone figures out how to do this, please let me know.
Please submit bugs and patches to the CPAN RT system at http://rt.cpan.org/NoAuth/ReportBug.html?Queue=Params%3A%3AValidate or via email at bug-params-validate@rt.cpan.org.
Support questions can be sent to Dave at autarch@urth.org.
The code repository is at https://svn.urth.org/svn/Params-Validate/
Dave Rolsky, <autarch@urth.org> and Ilya Martynov <ilya@martynov.org>
Copyright (c) 2004-2007 David Rolsky. All rights reserved. This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.