# -*- mode: cperl -*-

#
# Author: Slaven Rezic
#
# Copyright (C) 2006,2007,2008,2009,2010,2015 Slaven Rezic. All rights reserved.
# This package is free software; you can redistribute it and/or
# modify it under the same terms as Perl itself.
#
# Mail: srezic@cpan.org
# WWW:  http://www.rezic.de/eserte/
#

package Kwalify;

use strict;
use warnings;

use base qw(Exporter);
use vars qw(@EXPORT_OK $VERSION);
@EXPORT_OK = qw(validate);

$VERSION = '1.22_90';

sub validate ($$) {
    my($schema, $data) = @_;
    my $self = Kwalify::Validator->new;
    $self->validate($schema, $data, "/");
    if (@{$self->{errors}}) {
	die join("\n", map { "  - $_" } @{$self->{errors}}) . "\n";
    } else {
	1;
    }
}

package Kwalify::Validator;

use overload ();

sub new {
    my($class) = @_;
    bless { errors => [] }, $class;
}

sub validate {
    my($self, $schema, $data, $path, $args) = @_;
    $self->{done} = {};
    $self->_validate($schema, $data, $path, $args);
}

sub _validate {
    my($self, $schema, $data, $path, $args) = @_;
    $self->{path} = $path;

    if (!UNIVERSAL::isa($schema, "HASH")) {
	$self->_die("Schema structure must be a hash reference");
    }

    my $type = $schema->{type};
    if (!defined $type) {
	$type = 'str'; # default type;
    }
    my $type_check_method = "validate_" . $type;
    if (!$self->can($type_check_method)) {
	$self->_die("Invalid or unimplemented type '$type'");
    }

    $self->$type_check_method($schema, $data, $path, $args);
}

sub _additional_rules {
    my($self, $schema, $data, $path) = @_;
    no warnings 'uninitialized'; # legal undef values may happen everywhere
    for my $schema_key (keys %$schema) {
	if (defined $schema->{$schema_key}) {
	    if ($schema_key eq 'pattern') {
		(my $pattern = $schema->{pattern}) =~ s{^/(.*)/$}{$1};
		if ($data !~ qr{$pattern}) {
		    $self->_error("Non-valid data '$data' does not match /$pattern/");
		}
	    } elsif ($schema_key eq 'length') {
		if (!UNIVERSAL::isa($schema->{'length'}, "HASH")) {
		    $self->_die("'length' must be a hash with keys max and/or min");
		}
		my $length = length($data);
		for my $sub_schema_key (keys %{ $schema->{'length'} }) {
		    if ($sub_schema_key eq 'min') {
			my $min = $schema->{'length'}->{min};
			if ($length < $min) {
			    $self->_error("'$data' is too short (length $length < min $min)");
			}
		    } elsif ($sub_schema_key eq 'min-ex') {
			my $min = $schema->{'length'}->{'min-ex'};
			if ($length <= $min) {
			    $self->_error("'$data' is too short (length $length <= min $min)");
			}
		    } elsif ($sub_schema_key eq 'max') {
			my $max = $schema->{'length'}->{max};
			if ($length > $max) {
			    $self->_error("'$data' is too long (length $length > max $max)");
			}
		    } elsif ($sub_schema_key eq 'max-ex') {
			my $max = $schema->{'length'}->{'max-ex'};
			if ($length >= $max) {
			    $self->_error("'$data' is too long (length $length >= max $max)");
			}
		    } else {
			$self->_die("Unexpected key '$sub_schema_key' in length specification, expected min, max, min-ex and/or max-ex");
		    }
		}
	    } elsif ($schema_key eq 'enum') {
		if (!UNIVERSAL::isa($schema->{enum}, 'ARRAY')) {
		    $self->_die("'enum' must be an array");
		}
		my %valid = map { ($_,1) } @{ $schema->{enum} };
		if (!exists $valid{$data}) {
		    $self->_error("'$data': invalid " . _base_path($path) . " value");
		}
	    } elsif ($schema_key eq 'range') {
		if (!UNIVERSAL::isa($schema->{range}, "HASH")) {
		    $self->_die("'range' must be a hash with keys max and/or min");
		}
		my($lt, $le, $gt, $ge);
		## yes? no?
		# 	if (eval { require Scalar::Util; defined &Scalar::Util::looks_like_number }) {
		# 	    if (Scalar::Util::looks_like_number($data)) {
		# 		$lt = sub { $_[0] < $_[1] };
		# 		$gt = sub { $_[0] > $_[1] };
		# 	    } else {
		# 		$lt = sub { $_[0] lt $_[1] };
		# 		$gt = sub { $_[0] gt $_[1] };
		# 	    }
		# 	} else {
		#	    warn "Cannot determine whether $data is a number, assume so..."; # XXX show only once
		no warnings 'numeric';
		$lt = sub { $_[0] < $_[1] };
		$gt = sub { $_[0] > $_[1] };
		$le = sub { $_[0] <= $_[1] };
		$ge = sub { $_[0] >= $_[1] };
		#	}

		for my $sub_schema_key (keys %{ $schema->{range} }) {
		    if ($sub_schema_key eq 'min') {
			my $min = $schema->{range}->{min};
			if ($lt->($data, $min)) {
			    $self->_error("'$data' is too small (< min $min)");
			}
		    } elsif ($sub_schema_key eq 'min-ex') {
			my $min = $schema->{range}->{'min-ex'};
			if ($le->($data, $min)) {
			    $self->_error("'$data' is too small (<= min $min)");
			}
		    } elsif ($sub_schema_key eq 'max') {
			my $max = $schema->{range}->{max};
			if ($gt->($data, $max)) {
			    $self->_error("'$data' is too large (> max $max)");
			}
		    } elsif ($sub_schema_key eq 'max-ex') {
			my $max = $schema->{range}->{'max-ex'};
			if ($ge->($data, $max)) {
			    $self->_error("'$data' is too large (>= max $max)");
			}
		    } else {
			$self->_die("Unexpected key '$sub_schema_key' in range specification, expected min, max, min-ex and/or max-ex");
		    }
		}
	    } elsif ($schema_key eq 'assert') {
		$self->_die("'assert' is not yet implemented");
	    } elsif ($schema_key !~ m{^(type|required|unique|name|classname|class|desc)$}) {
		$self->_die("Unexpected key '$schema_key' in type specification");
	    }
	}
    }
}

sub validate_text {
    my($self, $schema, $data, $path) = @_;
    if (!defined $data || ref $data) {
	return $self->_error("Non-valid data '" . (defined $data ? $data : 'undef') . "', expected text");
    }
    $self->_additional_rules($schema, $data, $path);
}

sub validate_str {
    my($self, $schema, $data, $path) = @_;
    if (!defined $data || ref $data || $data =~ m{^\d+(\.\d+)?$}) {
	return $self->_error("Non-valid data '" . (defined $data ? $data : 'undef') . "', expected a str");
    }
    $self->_additional_rules($schema, $data, $path);
}

sub validate_int {
    my($self, $schema, $data, $path) = @_;
    if ($data !~ m{^[+-]?\d+$}) { # XXX what about scientific notation?
	$self->_error("Non-valid data '" . $data . "', expected an int");
    }
    $self->_additional_rules($schema, $data, $path);
}

sub validate_float {
    my($self, $schema, $data, $path) = @_;
    if ($data !~ m{^[+-]?\d+\.\d+$}) { # XXX other values?
	$self->_error("Non-valid data '" . $data . "', expected a float");
    }
    $self->_additional_rules($schema, $data, $path);
}

sub validate_number {
    my($self, $schema, $data, $path) = @_;
    if ($data !~ m{^[+-]?\d+(\.\d+)?$}) { # XXX combine int+float regexp!
	$self->_error("Non-valid data '" . $data . "', expected a number");
    }
    $self->_additional_rules($schema, $data, $path);
}

sub validate_bool {
    my($self, $schema, $data, $path) = @_;
    if ($data !~ m{^(yes|true|1|no|false|0)$}) { # XXX correct?
	$self->_error("Non-valid data '" . $data . "', expected a boolean");
    }
    $self->_additional_rules($schema, $data, $path);
}

# XXX is this correct?
sub validate_scalar {
    shift->validate_text(@_);
}

sub validate_date {
    my($self, $schema, $data, $path) = @_;
    if ($data !~ m{^\d{4}-\d{2}-\d{2}$}) {
	$self->_error("Non-valid data '" . $data . "', expected a date (YYYY-MM-DD)");
    }
    $self->_additional_rules($schema, $data, $path);
}

sub validate_time {
    my($self, $schema, $data, $path) = @_;
    if ($data !~ m{^\d{2}:\d{2}:\d{2}$}) {
	$self->_error("Non-valid data '" . $data . "', expected a time (HH:MM:SS)");
    }
    $self->_additional_rules($schema, $data, $path);
}

sub validate_timestamp {
    my($self) = @_;
    $self->_error("timestamp validation NYI"); # XXX
}

sub validate_any {
    my($self, $schema, $data, $path) = @_;
    $self->_additional_rules($schema, $data, $path);
}

sub validate_seq {
    my($self, $schema, $data, $path) = @_;
    if (!exists $schema->{sequence}) {
	$self->_die("'sequence' missing with 'seq' type");
    }
    my $sequence = $schema->{sequence};
    if (!UNIVERSAL::isa($sequence, 'ARRAY')) {
	$self->_die("Expected array in 'sequence'");
    }
    if (@$sequence != 1) {
	$self->_die("Expect exactly one element in sequence");
    }
    if (!UNIVERSAL::isa($data, 'ARRAY')) {
	$self->_error("Non-valid data " . $data . ", expected sequence");
	return;
    }

    return if ($self->{done}{overload::StrVal($data)}{overload::StrVal($schema)});
    $self->{done}{overload::StrVal($data)}{overload::StrVal($schema)} = 1;

    my $subschema = $sequence->[0];
    my $unique = _get_boolean($subschema->{unique});
    my %unique_val;
    my %unique_mapping_val;
    my $index = 0;
    for my $elem (@$data) {
	my $subpath = _append_path($path, $index);
	$self->_validate($subschema, $elem, $subpath, { unique_mapping_val => \%unique_mapping_val});
	if ($unique) {
	    if (exists $unique_val{$elem}) {
		$self->_error("'$elem' is already used at '$unique_val{$elem}'");
	    } else {
		$unique_val{$elem} = $subpath;
	    }
	}
	$index++;
    }
}

sub validate_map {
    my($self, $schema, $data, $path, $args) = @_;
    my $unique_mapping_val;
    if ($args && $args->{unique_mapping_val}) {
	$unique_mapping_val = $args->{unique_mapping_val};
    }
    if (!exists $schema->{mapping}) {
	$self->_die("'mapping' missing with 'map' type");
    }
    my $mapping = $schema->{mapping};
    if (!UNIVERSAL::isa($mapping, 'HASH')) {
	$self->_die("Expected hash in 'mapping'");
    }
    if (!defined $data) {
	$self->_error("Undefined data, expected mapping");
	return;
    }
    if (!UNIVERSAL::isa($data, 'HASH')) {
	$self->_error("Non-valid data " . $data . ", expected mapping");
	return;
    }

    return if ($self->{done}{overload::StrVal($data)}{overload::StrVal($schema)});
    $self->{done}{overload::StrVal($data)}{overload::StrVal($schema)} = 1;

    my %seen_key;
    my $default_key_schema;

    ## Originally this was an each-loop, but this could lead into
    ## endless recursions, because mapping may be reused in Kwalify,
    ## thus the each iterator was shared between recursion levels.
    # while(my($key,$subschema) = each %$mapping) {
    for my $key (keys %$mapping) {
	my $subschema = $mapping->{$key};
	if ($key eq '=') { # the "default" key
	    $default_key_schema = $subschema;
	    next;
	}
	my $subpath = _append_path($path, $key);
	$self->{path} = $subpath;
	if (!UNIVERSAL::isa($subschema, 'HASH')) {
	    $self->_die("Expected subschema (a hash)");
	}
	my $required = _get_boolean($subschema->{required});
	if (!defined $data->{$key}) {
	    if ($required) {
		$self->{path} = $path;
		$self->_error("Expected required key '$key'");
		next;
	    } else {
		$seen_key{$key}++;
		next;
	    }
	}
	my $unique = _get_boolean($subschema->{unique});
	if ($unique) {
	    if (defined $unique_mapping_val->{$data->{$key}}->{val} 
		&& $unique_mapping_val->{$data->{$key}}->{val} eq $data->{$key}) {
		$self->_error("'$data->{$key}' is already used at '$unique_mapping_val->{$data->{$key}}->{path}'");
	    } else {
		$unique_mapping_val->{$data->{$key}} = { val  => $data->{$key},
							 path => $subpath,
						       };
	    }
	}

	$self->_validate($subschema, $data->{$key}, $subpath);
	$seen_key{$key}++;
    }

#    while(my($key,$val) = each %$data) {
    for my $key (keys %$data) {
	my $val = $data->{$key};
	my $subpath = _append_path($path, $key);
	$self->{path} = $subpath;
	if (!$seen_key{$key}) {
	    if ($default_key_schema) {
		$self->_validate($default_key_schema, $val, $subpath);
	    } else {
		$self->_error("Unexpected key '$key'");
	    }
	}
    }
}

sub _die {
    my($self, $msg) = @_;
    $msg = "[$self->{path}] $msg";
    die $msg."\n";
}

sub _error {
    my($self, $msg) = @_;
    $msg = "[$self->{path}] $msg";
    push @{$self->{errors}}, $msg;
    0;
}

# Functions:
sub _append_path {
    my($root, $leaf) = @_;
    $root . ($root !~ m{/$} ? "/" : "") . $leaf;
}

sub _base_path {
    my($path) = @_;
    my($base) = $path =~ m{([^/]+)$};
    $base;
}

sub _get_boolean {
    my($val) = @_;
    defined $val && $val =~ m{^(yes|true|1)$}; # XXX check for all boolean trues
}

1;
__END__

=head1 NAME

Kwalify - Kwalify schema for data structures

=head1 SYNOPSIS

  use Kwalify qw(validate);
  validate($schema, $data);

Typically used together with YAML or JSON:

  use YAML;
  validate(YAML::LoadFile($schema_file), YAML::LoadFile($data_file));

  use JSON;
  validate(decode_json($schema_data), decode_json($data));

=head1 DESCRIPTION

Kwalify is a Perl implementation for validating data structures
against the Kwalify schema. For a schema definition, see
L<http://www.kuwata-lab.com/kwalify/ruby/users-guide.01.html>, but see
also below L</SCHEMA DEFINITION>.

=head2 validate($schema_data, $data)

Validate I<$data> according to Kwalify schema specified in
I<$schema_data>. Dies if the validation fails.

B<validate> may be exported.

=head1 SCHEMA DEFINITION

The original schema definition document is not very specific about
types and behaviour. Here's how B<Kwalify.pm> implements things:

=over

=item pattern

Perl regular expressions are used for patterns. This may or may not be
compatible with other Kwalify validators, so restrict to "simple"
regular expression constructs to be compatible with other validators.

=item type

=over

=item str

Any defined value which is B<not> a number. Most probably you will
want to use B<text> instead of B<str>.

=item int

A possibly signed integer. Note that scientific notation is not
supported, and it is also not clear whether it should be supported.

=item float

A possibly signed floating value with a mandatory decimal point. Note
that scientific notation is also not supported here.

=item bool

The values B<yes>, B<true>, and B<1> for true values and the values
B<no>, B<false>, and B<0> for false values are allowed. The ruby
implementation possibly allows more values, but this is not
documented.

Note that this definition is problematic, because for example the
string B<no> is a true boolean value in Perl. So one should stick to
B<0> and B<1> as data values, and probably define an additional
B<pattern> or B<enum> to ensure this:

    type: bool
    enum: [0, 1]

=item scalar

Currently the same as B<text>, but it's not clear if this is correct.

=item date

A string matching C<< /^\d{4}-\d{2}-\d{2}$/ >> (i.e. YYYY-MM-DD). Note
that no date range checks are done (yet).

=item time

A string matching C<< /^\d{2}:\d{2}:\d{2}$/ >> (i.e. HH:MM:SS). Note
that no time range checks are done (yet).

=item timestamp

Not supported --- it is not clear what this is supposed to be.

=back

=item assert

Currently not supported by the Perl implementation.

=item classname

Previously defined what is now B<class>, see L<http://web.archive.org/web/20071230173101/http://www.kuwata-lab.com/kwalify/users-guide.01.html>.

=item class

Currently not used, as there's no genclass action.

=item default

Currently not used, as there's no genclass action.

=back

=head1 TECHNICAL NOTES

As B<Kwalify.pm> is a pure validator and de-coupled from a parser (in
fact, it does not need to deal with YAML at all, but just with pure
perl data structures), there's no connection to the original validated
document. This means that no line numbers are available to the
validator. In case of validation errors the validator is only able to
show a path-like expression to the data causing the error.

=head1 AUTHOR

Slaven ReziE<x0107>, E<lt>srezic@cpan.orgE<gt>

=head1 COPYRIGHT AND LICENSE

Copyright (C) 2006,2007,2008,2009,2010,2015 by Slaven ReziE<x0107>

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

=head1 SEE ALSO

L<pkwalify>, L<kwalify(1)>.

Other non-XML schema languages: L<http://rx.codesimply.com/>

=cut