package Ref::Util;
# ABSTRACT: Utility functions for checking references
$Ref::Util::VERSION = '0.203';
use strict;
use warnings;

use Exporter 5.57 'import';

{
    my $impl = $ENV{PERL_REF_UTIL_IMPLEMENTATION}
        || our $IMPLEMENTATION
        || 'XS';
    if ($impl ne 'PP' && eval { require Ref::Util::XS; 1 }) {
        _install_aliases('Ref::Util::XS');
    }
    else {
        require Ref::Util::PP;
        _install_aliases('Ref::Util::PP');
    }
}

sub _install_aliases {
    my ($package) = @_;
    no warnings 'once';
    no strict 'refs';
    our %EXPORT_TAGS = %{"${package}::EXPORT_TAGS"};
    our @EXPORT_OK   = @{"${package}::EXPORT_OK"};
    *$_ = \&{"${package}::$_"} for '_using_custom_ops', @EXPORT_OK;
}

1;

__END__

=pod

=encoding UTF-8

=head1 NAME

Ref::Util - Utility functions for checking references

=head1 VERSION

version 0.203

=head1 SYNOPSIS

    use Ref::Util qw( is_plain_arrayref is_plain_hashref );

    if ( is_plain_arrayref( $something ) ) {
        print for @{ $something };
    } elsif ( is_plain_hashref( $something ) ) {
        print for sort values %{ $something };
    }

=head1 DESCRIPTION

Ref::Util introduces several functions to help identify references in a
B<smarter> (and usually faster) way. In short:

    # conventional approach             # with Ref::Util

    ref( $foo ) eq 'ARRAY'              is_plain_arrayref( $foo )

    use Scalar::Util qw( reftype );
    reftype( $foo ) eq 'ARRAY'          is_arrayref( $foo )

The difference:

=over 4

=item * No comparison against a string constant

When you call C<ref>, you stringify the reference and then compare it
to some string constant (like C<ARRAY> or C<HASH>). Not just awkward,
it's brittle since you can mispell the string.

If you use L<Scalar::Util>'s C<reftype>, you still compare it as a
string:

    if ( reftype($foo) eq 'ARRAY' ) { ... }

=item * Supports blessed variables

B<Note:> In future versions, the idea is to make the default functions
use the B<plain> variation, which means explicitly non-blessed references.

If you want to explicitly check for B<blessed> references, you should use
the C<is_blessed_*> functions. There will be an C<is_any_*> variation
which will act like the current main functions - not caring whether it's
blessed or not.

When calling C<ref>, you receive either the reference type (B<SCALAR>,
B<ARRAY>, B<HASH>, etc.) or the package it's blessed into.

When calling C<is_arrayref> (et. al.), you check the variable flags,
so even if it's blessed, you know what type of variable is blessed.

    my $foo = bless {}, 'PKG';
    ref($foo) eq 'HASH'; # fails

    use Ref::Util 'is_hashref';
    my $foo = bless {}, 'PKG';
    is_hashref($foo); # works

On the other hand, in some situations it might be better to specifically
exclude blessed references. The rationale for that might be that merely
because some object happens to be implemented using a hash doesn't mean it's
necessarily correct to treat it as a hash. For these situations, you can use
C<is_plain_hashref> and friends, which have the same performance benefits as
C<is_hashref>.

There is also a family of functions with names like C<is_blessed_hashref>;
these return true for blessed object instances that are implemented using
the relevant underlying type.

=item * Supports tied variables and magic

Tied variables (used in L<Readonly>, for example) are supported.

    use Ref::Util qw<is_plain_hashref>;
    use Readonly;

    Readonly::Scalar my $rh2 => { a => { b => 2 } };
    is_plain_hashref($rh2); # success

L<Ref::Util> added support for this in 0.100. Prior to this version
the test would fail.

=item * Ignores overloading

These functions ignore overloaded operators and simply check the
variable type. Overloading will likely not ever be supported, since I
deem it problematic and confusing.

Overloading makes your variables opaque containers and hides away
B<what> they are and instead require you to figure out B<how> to use
them. This leads to code that has to test different abilities (in
C<eval>, so it doesn't crash) and to interfaces that get around what
a person thought you would do with a variable. This would have been
alright, except there is no clear way of introspecting it.

=item * Ignores subtle types:

The following types, provided by L<Scalar::Util>'s C<reftype>, are
not supported:

=over 4

=item * C<VSTRING>

This is a C<PVMG> ("normal" variable) with a flag set for VSTRINGs.
Since this is not a reference, it is not supported.

=item * C<LVALUE>

A variable that delegates to another scalar. Since this is not a
reference, it is not supported.

=item * C<INVLIST>

I couldn't find documentation for this type.

=back

Support might be added, if a good reason arises.

=item * Usually fast

When possible, Ref::Util uses L<Ref::Util::XS> as its implementation. (If
you don't have a C compiler available, it uses a pure Perl fallback that has
all the other advantages of Ref::Util, but isn't as fast.)

In fact, Ref::Util::XS has two alternative implementations available
internally, depending on the features supported by the version of Perl
you're using. For Perls that supports custom OPs, we actually add an OP
(which is faster); for other Perls, the implementation that simply calls an
XS function (which is still faster than the pure-Perl equivalent).

See below for L<benchmark results|/"BENCHMARKS">.

=back

=head1 EXPORT

Nothing is exported by default. You can ask for specific subroutines
(described below) or ask for all subroutines at once:

    use Ref::Util qw<is_scalarref is_arrayref is_hashref ...>;

    # or

    use Ref::Util ':all';

=head1 SUBROUTINES

=head2 is_ref($ref)

Check for a reference to anything.

    is_ref([]);

=head2 is_scalarref($ref)

Check for a scalar reference.

    is_scalarref(\"hello");
    is_scalarref(\30);
    is_scalarref(\$value);

Note that, even though a reference is itself a type of scalar value, a
reference to another reference is not treated as a scalar reference:

    !is_scalarref(\\1);

The rationale for this is two-fold. First, callers that want to decide how
to handle inputs based on their reference type will usually want to treat a
ref-ref and a scalar-ref differently. Secondly, this more closely matches
the behavior of the C<ref> built-in and of L<Scalar::Util/reftype>, which
report a ref-ref as C<REF> rather than C<SCALAR>.

=head2 is_arrayref($ref)

Check for an array reference.

    is_arrayref([]);

=head2 is_hashref($ref)

Check for a hash reference.

    is_hashref({});

=head2 is_coderef($ref)

Check for a code reference.

    is_coderef( sub {} );

=head2 is_regexpref($ref)

Check for a regular expression (regex, regexp) reference.

    is_regexpref( qr// );

=head2 is_globref($ref)

Check for a glob reference.

    is_globref( \*STDIN );

=head2 is_formatref($ref)

Check for a format reference.

    # set up format in STDOUT
    format STDOUT =
    .

    # now we can test it
    is_formatref( *main::STDOUT{'FORMAT'} );

This function is not available in Perl 5.6 and will trigger a
C<croak()>.

=head2 is_ioref($ref)

Check for an IO reference.

    is_ioref( *STDOUT{IO} );

=head2 is_refref($ref)

Check for a reference to a reference.

    is_refref( \[] ); # reference to array reference

=head2 is_plain_scalarref($ref)

Check for an unblessed scalar reference.

    is_plain_scalarref(\"hello");
    is_plain_scalarref(\30);
    is_plain_scalarref(\$value);

=head2 is_plain_ref($ref)

Check for an unblessed reference to anything.

    is_plain_ref([]);

=head2 is_plain_arrayref($ref)

Check for an unblessed array reference.

    is_plain_arrayref([]);

=head2 is_plain_hashref($ref)

Check for an unblessed hash reference.

    is_plain_hashref({});

=head2 is_plain_coderef($ref)

Check for an unblessed code reference.

    is_plain_coderef( sub {} );

=head2 is_plain_globref($ref)

Check for an unblessed glob reference.

    is_plain_globref( \*STDIN );

=head2 is_plain_formatref($ref)

Check for an unblessed format reference.

    # set up format in STDOUT
    format STDOUT =
    .

    # now we can test it
    is_plain_formatref(bless *main::STDOUT{'FORMAT'} );

=head2 is_plain_refref($ref)

Check for an unblessed reference to a reference.

    is_plain_refref( \[] ); # reference to array reference

=head2 is_blessed_scalarref($ref)

Check for a blessed scalar reference.

    is_blessed_scalarref(bless \$value);

=head2 is_blessed_ref($ref)

Check for a blessed reference to anything.

    is_blessed_ref(bless [], $class);

=head2 is_blessed_arrayref($ref)

Check for a blessed array reference.

    is_blessed_arrayref(bless [], $class);

=head2 is_blessed_hashref($ref)

Check for a blessed hash reference.

    is_blessed_hashref(bless {}, $class);

=head2 is_blessed_coderef($ref)

Check for a blessed code reference.

    is_blessed_coderef( bless sub {}, $class );

=head2 is_blessed_globref($ref)

Check for a blessed glob reference.

    is_blessed_globref( bless \*STDIN, $class );

=head2 is_blessed_formatref($ref)

Check for a blessed format reference.

    # set up format for FH
    format FH =
    .

    # now we can test it
    is_blessed_formatref(bless *FH{'FORMAT'}, $class );

=head2 is_blessed_refref($ref)

Check for a blessed reference to a reference.

    is_blessed_refref( bless \[], $class ); # reference to array reference

=head1 BENCHMARKS

Here is a benchmark comparing similar checks.

    my $bench = Dumbbench->new(
        target_rel_precision => 0.005,
        initial_runs         => 20,
    );

    my $amount = 1e7;
    my $ref    = [];
    $bench->add_instances(
        Dumbbench::Instance::PerlSub->new(
            name => 'Ref::Util::is_plain_arrayref (CustomOP)',
            code => sub {
                Ref::Util::is_plain_arrayref($ref) for ( 1 .. $amount )
            },
        ),

        Dumbbench::Instance::PerlSub->new(
            name => 'ref(), reftype(), !blessed()',
            code => sub {
                ref $ref
                    && Scalar::Util::reftype($ref) eq 'ARRAY'
                    && !Scalar::Util::blessed($ref)
                    for ( 1 .. $amount );
            },
        ),

        Dumbbench::Instance::PerlSub->new(
            name => 'ref()',
            code => sub { ref($ref) eq 'ARRAY' for ( 1 .. $amount ) },
        ),

        Dumbbench::Instance::PerlSub->new(
            name => 'Data::Util::is_array_ref',
            code => sub { is_array_ref($ref) for ( 1 .. $amount ) },
        ),

    );

The results:

    ref():                                   5.335e+00 +/- 1.8e-02 (0.3%)
    ref(), reftype(), !blessed():            1.5545e+01 +/- 3.1e-02 (0.2%)
    Ref::Util::is_plain_arrayref (CustomOP): 2.7951e+00 +/- 6.2e-03 (0.2%)
    Data::Util::is_array_ref:                5.9074e+00 +/- 7.5e-03 (0.1%)

(Rounded run time per iteration)

A benchmark against L<Data::Util>:

    Ref::Util::is_plain_arrayref: 3.47157e-01 +/- 6.8e-05 (0.0%)
    Data::Util::is_array_ref:     6.7562e-01 +/- 7.5e-04 (0.1%)

=head1 SEE ALSO

=over 4

=item * L<Params::Classify>

=item * L<Scalar::Util>

=item * L<Data::Util>

=back

=head1 THANKS

The following people have been invaluable in their feedback and support.

=over 4

=item * Yves Orton

=item * Steffen Müller

=item * Jarkko Hietaniemi

=item * Mattia Barbon

=item * Zefram

=item * Tony Cook

=item * Sergey Aleynikov

=back

=head1 AUTHORS

=over 4

=item * Aaron Crane

=item * Vikentiy Fesunov

=item * Sawyer X

=item * Gonzalo Diethelm

=item * p5pclub

=back

=head1 LICENSE

This software is made available under the MIT Licence as stated in the
accompanying LICENSE file.

=head1 AUTHORS

=over 4

=item *

Sawyer X <xsawyerx@cpan.org>

=item *

Aaron Crane <arc@cpan.org>

=item *

Vikenty Fesunov <vyf@cpan.org>

=item *

Gonzalo Diethelm <gonzus@cpan.org>

=item *

Karen Etheridge <ether@cpan.org>

=back

=head1 COPYRIGHT AND LICENSE

This software is Copyright (c) 2017 by Sawyer X.

This is free software, licensed under:

  The MIT (X11) License

=cut