#
# Copyright (c) 2007-2016 T. v.Dein <tlinden |AT| cpan.org>.
# All Rights Reserved. Std. disclaimer applies.
# Artistic License, same as perl itself. Have fun.
#
# namespace
package Data::Validate::Struct;

use strict;
use warnings;
use English '-no_match_vars';
use Carp;
use Exporter;
use Encode qw{ encode };
use Regexp::Common::URI::RFC2396 qw /$host $port/;
use Regexp::Common qw /URI net delimited/;

use File::Spec::Functions qw/file_name_is_absolute/;
use File::stat;

use Data::Validate qw(:math is_printable);
use Data::Validate::IP qw(is_ipv4 is_ipv6);

our $VERSION = 0.11;

use vars qw(@ISA);

use vars qw(@ISA @EXPORT @EXPORT_OK %__ValidatorTypes);
require Exporter;
@ISA = qw(Exporter);
@EXPORT = qw(%__ValidatorTypes);
@EXPORT_OK = qw(add_validators);

%__ValidatorTypes = (
                     # primitives
                     int            => sub { return defined(is_integer($_[0])); },
                     hex            => sub { return defined(is_hex($_[0])); },
                     oct            => sub { return defined(is_oct($_[0])); },
                     number         => sub { return defined(is_numeric($_[0])); },

                     word           => qr(^[\w_\-]+$),
                     line           => qr/^[^\n]+$/s,

                     text           => sub { return defined(is_printable($_[0])); },

                     regex          => sub {
                       my $r = ref $_[0];
                       return 1 if $r eq 'Regexp';
                       if ($r eq '') {
                         # this is a bit loosy but should match most regular expressions
                         # using the qr() operator, but it doesn't check if the expression
                         # is valid. we could do this by compiling it, but this would lead
                         # to exploitation possiblities to programs using the module.
                         return $_[0] =~ qr/^qr ( (.).*\1 | \(.*\) | \{.*\} ) $/x;
                       }
                       return 0;
                     },

                     # via imported regexes
                     uri            => qr(^$RE{URI}$),
                     cidrv4         => sub {
                       my ($p, $l) = split(/\//, $_[0]);
                       return defined(is_ipv4($p)) && defined(is_between($l, 0, 32));
                     },
                     ipv4           => sub { defined(is_ipv4($_[0])) },
                     quoted         => qr/^$RE{delimited}{ -delim => qr(\') }$/,
                     hostname       => qr(^$host$),

                     ipv6           => sub { defined(is_ipv6($_[0])) },
                     cidrv6         => sub {
                       my ($p, $l) = split('/', $_[0]);
                       return defined(is_ipv6($p)) && defined(is_between($l, 0, 128));
                     },

                     # matches perl style scalar variables
                     # possible matches: $var ${var} $(var)
                     vars           => qr/(?<!\\) ( \$\w+ | \$\{[^\}]+\} | \$\([^\)]+\) )/x,

                     # closures

                     # this one doesn't do a stat() syscall, so keep cool
                     path           => sub { return file_name_is_absolute($_[0]); },

                     # though this one does it - it stat()s if the file exists
                     fileexists     => sub { return stat($_[0]); },

                     # do a dns lookup on given value, this also fails if
                     # no dns is available - so be careful with this
                     resolvablehost => sub { return gethostbyname($_[0]); },

                     # looks if the given value is an existing user on the host system
                     user           => sub { return (getpwnam($_[0]))[0]; },

                     # same with group
                     group          => sub { return getgrnam($_[0]); },

                     # int between 0 - 65535
                     port           => sub {
                       if ( $_[0] =~ /^$port$/ && ($_[0] > 0 && $_[0] < 65535) )
                         { return 1; } else { return 0; } },

                     # variable integer range, use: range(N1 - N2)
                     range          => sub {
                       if ( defined(is_integer($_[0])) && ($_[0] >= $_[2] && $_[0] <= $_[3]) )
                         { return 1; } else { return 0; } },

                     # just a place holder at make the key exist
                     optional       => 1,
                    );

sub add_validators {
  # class method, add validators globally, not per object
  my(%v) = @_;
  foreach my $type (keys %v) { 
    $__ValidatorTypes{$type} = $v{$type};
  }
}

sub new {
  my ($class, $structure) = @_;
  $class = ref($class) || $class;

  my $self = bless {}, $class;

  $self->{structure} = $structure;

  # if types will be implemented in Data::Validate, remove our own
  # types from here and use Data::Validate's methods as subroutine
  # checks, which we already support.
  $self->{types} = \%__ValidatorTypes;
  $self->{debug} = 0;
  $self->{errors} = [];

  foreach my $type (keys %{$self->{types}}) {
    # add negative match types
    $self->{types}->{'no' . $type} = $self->{types}->{$type};
  }

  return $self;
}


sub debug {
  shift->{debug} = 1;
}


sub errors {
  my $self = shift;
  return $self->{errors};
}


sub errstr {
  my $self = shift;
  return $self->{errors} ? $self->{errors}->[0] : '';
}


sub type {
  my $self = shift;
  return unless @_;

  my $param = @_ > 1 ? {@_} : {%{$_[0]}};

  foreach my $type (keys %$param) {
    $self->{types}->{$type} = $param->{$type};
    # add negative match types
    $self->{types}->{'no' . $type} = $param->{$type};
  }
}


sub validate {
  my ($self, $config) = @_;

  # reset errors in case it's a repeated run
  $self->{errors} = [];

  $self->_traverse($self->{structure}, $config, ());
  # return TRUE if no errors
  return scalar @{ $self->{errors} } == 0;
}

# Private methods

sub _debug {
  my ($self, $msg) = @_;
  if ($self->{debug}) {
    print STDERR "D::V::S::debug() - $msg\n";
  }
}

sub _traverse {
  my ($self, $reference, $hash, @tree) = @_;

  foreach my $key (keys %{$reference}) {
    if (ref($reference->{$key}) eq 'ARRAY') {
      # just use the 1st one, more elements in array are expected to be the same
      foreach my $item (@{$hash->{$key}}) {
        if (ref($item) eq q(HASH)) {
          # traverse the structure pushing our key to the @tree
          $self->_traverse($reference->{$key}->[0], $item, @tree, $key);
        }
        else {
          # a value, this is tricky
          $self->_traverse(
            { item => $reference->{$key}->[0] },
            { item => $item },
            @tree, $key
          );
        }
      }
    }
    elsif (ref($reference->{$key}) eq 'HASH') {
      $self->_traverse($reference->{$key}, $hash->{$key}, @tree, $key);
    }
    elsif (ref($reference->{$key}) eq '') {
      $self->_debug("Checking $key at " . join(', ', @tree));
      if (my $err = $self->_check_type($key, $reference, $hash)) {
        push @{$self->{errors}}, sprintf(q{%s at '%s'}, $err, join(' => ', @tree));
      }
    }
  }
}

sub _check_type {
  my ($self, $key, $reference, $hash) = @_;

  my (@types, @tmptypes, @tokens);
  @types = @tmptypes = _trim( (split /\|/, $reference->{$key}) );
  # check data types
  if (grep { ! exists $self->{types}->{$_} } map { s/\(.*//; $_ } @tmptypes) {
    return "Invalid data type '$reference->{$key}'";
  }

  # does $key exist in $hash
  unless (exists $hash->{$key}) {
    # is it an optional key?
    if (grep { $_ eq 'optional' } @types) {
      # do nothing
      $self->_debug("$key is optional");
      return;
    }
    else {
      # report error
      return "Required key '$key' is missing";
    }
  }

  # the value in $hash->{$key} (shortcut)
  my $value = $hash->{$key};

  # is the value checkable?
  unless (defined $value) {
    if (grep { $_ eq 'optional' } @types) {
      # do nothing
      $self->_debug("$key is optional");
      return;
    }
    else {
      # report error
      return "value of '$key' is undef";
    }
  }

  # the aggregated match over *all* types
  my $match = 0;
  foreach my $type (@types) {
    # skip optional data type (can't be compared)
    next if $type eq 'optional';

    # tokenize the type into params, only used by coderefs
    # passed to coderef: &code($value, $typename, $unparsed_args, $arg1, $arg2 ...)
    ($type, @tokens) = _tokenize($type);

    # if the type begins with 'no' AND the remainder of the type
    # also exists in the type hash, we are expects something that is
    # FALSE (0), else TRUE (1).
    # we must check for both, if not we will get a false match on a type
    # called 'nothing'.
    my $expects = 1;
    if ($type =~ /^no(.*)/) {
      $expects = 0 if exists $self->{types}->{$1};
    }

    # "Evaluate" this $type. We set $result explicitly to 1 or 0
    # instead of relying the coderef returning a proper value.
    # This makes comparing $expects and $result mush easier, no magic
    # type conversions are needed.
    my $result = ref($self->{types}->{$type}) eq q(CODE)
      # the the type is a code ref, execute the code
      ? &{$self->{types}->{$type}}($value, @tokens)  ? 1 : 0
      # else it's an regexp, check if it's a match
      : $value =~ /$self->{types}->{$type}/ ? 1 : 0;

    $self->_debug(sprintf(
      '%s = %s, value %s %s',
      $key,
      encode('UTF-8', $value),
      $result ? 'is' : 'is not',
      $type
    ));
    $match ||= ($expects == $result);
  }

  return if $match;

  return sprintf q{'%s' doesn't match '%s'},
    encode('UTF-8', $value), $reference->{$key};
}


sub _trim {
  my @a = @_;
  foreach (@a) {
    s/^\s+|\s+$//g;
  }
  return wantarray ? @a : $a[0];
}

sub _tokenize {
  my $type = shift;

  if ($type =~ /(.+?)\((.+?)\)/) {
    print "func pattern\n";
    # type matches a function like pattern eg highport(1-1023)
    my $name = $1;
    my $args = $2;
    $args =~ s/\s//g;
    my @params = split /[\,\-]/, $args;
    return ($name, $args, @params);
  }

  # default, just return the name as it is
  return ($type);
}

1;


__END__

=pod

=head1 NAME

Data::Validate::Struct - Validate recursive Hash Structures

=head1 SYNOPSIS

 use Data::Validate::Struct;
 my $validator = new Data::Validate::Struct($reference);
 if ( $validator->validate($config_hash_reference) ) {
   print "valid\n";
 }
 else {
   print "invalid " . $validator->errstr() . "\n";
 }

=head1 DESCRIPTION

This module validates a config hash reference against a given hash
structure in contrast to L<Data::Validate> in which you have to
check each value separately using certain methods.

This hash could be the result of a config parser or just any
hash structure. Eg. the hash returned by L<XML::Simple> could
be validated using this module. You may also use it to validate
CGI input, just fetch the input data from CGI, L<map> it to a
hash and validate it.

Data::Validate::Struct uses some of the methods exported by L<Data::Validate>,
so you need to install it too.


=head1 PREDEFINED BUILTIN DATA TYPES

=over

=item B<int>

Match a simple integer number.

=item B<range(a-b)>

Match a simple integer number in a range between a and b. Eg:

 { loginport => 'range(22-23)' }

=item B<hex>

Match a hex value.

=item B<oct>

Match an octagonal value.

=item B<number>

Match a decimal number, it may contain , or . and may be signed.

=item B<word>

Match a single word, _ and - are tolerated.

=item B<line>

Match a line of text - no newlines are allowed.

=item B<text>

Match a whole text(blob) including newlines. This expression
is very loosy, consider it as an alias to B<any>.

=item B<regex>

Match a perl regex using the operator qr(). Valid examples include:

  qr/[0-9]+/
  qr([^%]*)
  qr{\w+(\d+?)}

Please note, that this doesn't mean you can provide
here a regex against config options must match.

Instead this means that the config options contains a regex.

eg:

  $cfg = {
    grp  = qr/root|wheel/
  };

B<regex> would match the content of the variable 'grp'
in this example.

To add your own rules for validation, use the B<type()>
method, see below.

=item B<uri>

Match an internet URI.

=item B<ipv4>

Match an IPv4 address.

=item B<cidrv4>

The same as above including cidr netmask (/24), IPv4 only, eg:

  10.2.123.0/23

Note: shortcuts are not supported for the moment, eg:

  10.10/16

will fail while it is still a valid IPv4 cidr notation for
a network address (short for 10.10.0.0/16). Must be fixed
in L<Regex::Common>.

=item B<ipv6>

Match an IPv6 address. Some examples:

  3ffe:1900:4545:3:200:f8ff:fe21:67cf
  fe80:0:0:0:200:f8ff:fe21:67cf
  fe80::200:f8ff:fe21:67cf
  ff02:0:0:0:0:0:0:1
  ff02::1

=item B<cidrv6>

The same as above including cidr netmask (/64), IPv6 only, eg:

  2001:db8:dead:beef::1/64
  2001:db8::/32

=item B<quoted>

Match a text quoted with single quotes, eg:

  'barbara is sexy'

=item B<hostname>

Match a valid hostname, it must qualify to the definitions
in RFC 2396.

=item B<resolvablehost>

Match a hostname resolvable via dns lookup. Will fail if no
dns is available at runtime.

=item B<path>

Match a valid absolute path, it won't do a stat() system call.
This will work on any operating system at runtime. So this one:

  C:\Temp

will return TRUE if running on WIN32, but FALSE on FreeBSD!

=item B<fileexists>

Look if value is a file which exists. Does a stat() system call.

=item B<user>

Looks if the given value is an existent user. Does a getpwnam() system call.

=item B<group>

Looks if the given value is an existent group. Does a getgrnam() system call.

=item B<port>

Match a valid tcp/udp port. Must be a digit between 0 and 65535.

=item B<vars>

Matches a string of text containing variables (perl style variables though)
eg:

  $user is $attribute
  I am $(years) old
  Missing ${points} points to succeed

=back


=head1 MIXED TYPES

If there is an element which could match more than one type, this
can be matched by using the pipe sign C<|> to separate the types.

  { name => 'int | number' }

There is no limit on the number of types that can be checked for, and the
check is done in the sequence written (first the type 'int', and then
'number' in the example above).


=head1 OPTIONAL ITEMS

If there is an element which is optional in the hash, you can use
the type 'optional' in the type. The 'optional' type can also be mixed
with ordinary types, like:

  { name => 'text | optional' }

The type 'optional' can be placed anywhere in the type string.


=head1 NEGATIVE MATCHING

In some rare situations you might require a negative match. So
a test shall return TRUE if a particular value does NOT match the
given type. This might be useful to prevent certain things.

To achieve this, you just have to prepend one of the below mentioned
types with the keyword B<no>.

Example:

 $ref = { path => 'novars' }

This returns TRUE if the value of the given config hash does NOT
contain ANY variables.


=head1 VALIDATOR STRUCTURE

The expected structure must be a standard perl hash reference.
This hash may look like the config you are validating but
instead of real-live values it contains B<types> that define
of what type a given value has to be.

In addition the hash may be deeply nested. In this case the
validated config must be nested the same way as the reference
hash.

Example:

  $reference = { user => 'word', uid => 'int' };

The following config would be validated successful:

  $config = { user => 'HansDampf',  uid => 92 };

this one not:

  $config = { user => 'Hans Dampf', uid => 'nine' };
                           ^                ^^^^
                           |                |
                           |                +----- is not a number
                           +---------------------- space not allowed

For easier writing of references you yould use a configuration
file parser like Config::General or Config::Any, just write the
definition using the syntax of such a module, get the hash of it
and use this hash as validation reference.

=head1 NESTED HASH STRUCTURES

You can also match against nested structures. B<Data::Validate::Struct> iterates
into the given config hash the same way as the reference hash looks like.

If the config hash doesn't match the reference structure, perl will
throw an error, which B<Data::Validate::Struct> catches and returns FALSE.

Given the following reference hash:

  $ref = {
      'b1' => {
          'b2' => {
              'b3' => {
                  'item' => 'int'
              }
          }
      }
  }

Now if you validate it against the following config hash it
will return TRUE:

  $cfg = {
      'b1' => {
          'b2' => {
              'b3' => {
                  'item' => '100'
              }
          }
      }
  }

If you validate it for example against this hash, it will
return FALSE:

  $cfg = {
      'b1' => {
          'b2' => {
              'item' => '100'
          }
      }
  }

=head1 SUBROUTINES/METHODS

=over

=item B<validate($config)>

$config must be a hash reference you'd like to validate.

It returns a true value if the given structure looks valid.

If the return value is false (0), then the error message will
be written to the variable $!.

=item B<type(%types)>

You can enhance the validator by adding your own rules. Just
add one or more new types using a simple hash using the B<type()>
method. Values in this hash can be regexes or anonymous subs.

C<type> does accept either a hash (C<%hash>), a hash ref (C<%$hash>) or a
list of key/values (C<< key => value >>) as input.

For details see L<CUSTOM VALIDATORS>.

=item B<debug()>

Enables debug output which gets printed to STDERR.

=item B<errors>

Returns an array ref with the errors found when validating the hash.
Each error is on the format '<value> doesn't match <types> at <ref>',
where <ref> is a comma separated tree view depicting where in the
the error occurred.

=item B<errstr()>

Returns the last error, which is useful to notify the user
about what happened. The format is like in L</errors>.

=back

=head1 EXPORTED FUNCTIONS

=head2 add_validators

This is a class function which adds types not per object
but globally for each instance of Data::Validate::Struct.

 use Data::Validate::Struct qw(add_validators);
 add_validators( name => .. );
 my $v = Data::Validate::Struct->new(..);

Parameters to B<add_validators> are the same as of the
B<type> method.

For details see L<CUSTOM VALIDATORS>.

=head1 CUSTOM VALIDATORS

You can add your own validators, which maybe regular expressions
or anonymous subs. Validators can be added using the B<type()>
method or globally using the B<add_validators()> function.

=head2 CUSTOM REGEX VALIDATORS

If you add a validator which is just a regular expressions,
it will evaluated as is. This is the most simplest way to
customize validation.

Sample:

 use Data::Validate::Struct qw(add_validators);
 add_validators(address => qr(^\w+\s\s*\d+$));
 my $v = Data::Validate::Struct->new({place => 'address'});
 $v->validate({place => 'Livermore 19'});

Regexes will be executed exactly as given. No flags or ^ or $
will be used by the module. Eg. if you want to match the whole
value from beginning to the end, add ^ and $, like you can see
in our 'address' example above.

=head2 CUSTOM VALIDATOR FUNCTIONS

If the validator is a coderef, it will be executed as a sub.

Example:

 use Data::Validate::Struct qw(add_validators);
 add_validators(
    list => sub {
      my $list = shift;
      my @list = split /\s*,\s*/, $list;
      return scalar @list > 1;
    },
 );

In this example we add a new type 'list', which
is really simple. 'list' is a subroutine which gets called
during evaluation for each option which you define as type 'list'.

Such a subroutine must return a true value in order to produce a match.
It receives the following arguments:

=over

=item *

value to be evaluated

=item *

unparsed arguments, if defined in the reference

=item *

array of parsed arguments, tokenized by , and -

=back

That way you may define a type which accepts an arbitrary number
of arguments, which makes the type customizable. Sample:

 # new validator
 $v4 = Data::Validate::Struct->new({ list => nwords(4) });
 
 # define type 'nwords' with support for 1 argument
 $v4->type(
   nwords => sub {
     my($val, $ignore, $count) = @_;
     return (scalar(split /\s+/, $val) == $count) ? 1 : 0;
   },
 );
 
 # validate
 $v4->validate({ list => 'these are four words' });
 

=head2 CUSTOM VALIDATORS USING A GRAMMAR

Sometimes you want to be more flexible, in such cases you may
use a parser generator to validate input. This is no feature
of Data::Validate::Struct, you will just write a custom code
ref validator, which then uses the grammar.

Here's a complete example using L<Parse::RecDescent>:

 use Parse::RecDescent;
 use Data::Validate::Struct qw(add_validators);
 
 my $grammar = q{
    line: expr(s)
    expr: number operator number
    number: int | float
    int: /\d+/
    float: /\d*\\.\d+/
    operator: '+' | '-' | '*' | '/'
 };
 
 my $parse = Parse::RecDescent->new($grammar);
 
 add_validators(calc => sub { defined $parse->line($_[0]) ? 1 : 0; });
 
 my $val = Data::Validate::Struct->new({line => 'calc'});
 
 if ($val->validate({line => "@ARGV"})) {
   my $r;
   eval "\$r = @ARGV";
   print "$r\n";
 }
 else {
   print "syntax error\n";
 }

Now you can use it as follows:

 ./mycalc 54 + 100 - .1
 153.9
 
 ./mycalc 8^2
 syntax error

=head2 NEGATED VALIDATOR

A negative/reverse match is automatically added as well, see
L</NEGATIVE MATCHING>.

=head1 EXAMPLES

Take a look to F<t/run.t> for lots of examples.

=head1 CONFIGURATION AND ENVIRONMENT

No environment variables will be used.

=head1 SEE ALSO

I recommend you to read the following documentations, which are supplied with perl:

L<perlreftut> Perl references short introduction.

L<perlref> Perl references, the rest of the story.

L<perldsc> Perl data structures intro.

L<perllol> Perl data structures: arrays of arrays.

L<Data::Validate> common data validation methods.

L<Data::Validate::IP> common data validation methods for IP-addresses.

=head1 LICENSE AND COPYRIGHT

Copyright (c) 2007-2015 T. v.Dein

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

=head1 BUGS AND LIMITATIONS

Some implementation details as well as the API may change
in the future. This will no more happen if entering a stable
release (starting with 1.00).

To submit use L<http://rt.cpan.org>.

=head1 INCOMPATIBILITIES

None known.

=head1 DIAGNOSTICS

To debug Data::Validate::Struct use B<debug()> or the perl debugger, see L<perldebug>.

For example to debug the regex matching during processing try this:

 perl -Mre=debug yourscript.pl

=head1 DEPENDENCIES

Data::Validate::Struct depends on the module L<Data::Validate>,
L<Data::Validate:IP>, L<Regexp::Common>, L<File::Spec> and L<File::stat>.

=head1 AUTHORS

T. v.Dein <tlinden |AT| cpan.org>

Per Carlson <pelle |AT| cpan.org>

Thanks to David Cantrell for his helpful hints.

=head1 VERSION

0.11

=cut