Hello, dear esteemed monks!

tl;dr: I'm thinking of a module that (1) compiles a set of validation rules once and (2) for each hashref given to it later, returns another object containing valid data, errors, AND initial input for processing and/or resubmission.

Is there such a module already? If not, should I roll my own? If yes, does the API described below seem sensible?

Now the whole story. Some time ago there was a discussion here at Perlmonks pointing out that it would be nice to make create button act as preview if post content was edited. Can't find proofs now, but the idea impressed me much.

Recently I started looking for a form validator. My initial idea was to build a permanent validator object once and keep throwing inputs at it later. And I found a module with almost the interface I imagined - Validator::LIVR.

However, after trying to actually implement form validation & resubmission, I found myself juggling 3 hashrefs (valid data, errors, raw user input for reentry). This was not very convenient, so I decided to pack them into one object, adding is_valid() method on top. And I found a module on CPAN with almost the interface I imagined (but without the "compile once" part) - Data::CGIForm.

So I crossed these two and got roughly the following API:

# initialization
# the hash describes regexps, requiredness, and other checks
# per input key
my $validator = My::Class->new ( { ... } );

# later when processing request
my $form = $validator->validate( { get => "params" } );
if ($form->is_valid) {
    do_something( $form->data );
    redirect( "/somewhere" );
} else {
    show_form_again( display_errors => $form->error, input_defaults =>
+ $form->raw );
};
[download]

Also error content (and thus is_valid return value) can be modified, just like in Data::CGIForm:

if (!load_user($form->data->{user})) {
    $form->error( user => "No such user in database" );
};
[download]

Unfortunately, the Data::CGIForm has a fatal flaw: error() without arguments acts like my is_valid(), while separate errors() method returns all errors as hash. Error/errors is prone to typoes in my opinion, but maybe I should shut up my ego and stay compatible to an existing API instead?

Thank you