view release on metacpan or  search on metacpan

lib/Tupelo/Munge.pm  view on Meta::CPAN

Other than writing to a single slot and replacement of the complete
sequence of slot values, tuples do not naturally support operations to
edit the slot value sequence, such as adding an object reference onto
the end of a tuple.  Such an operation can be constructed via complete
replacement of a tuple's slot value sequence, but if the desire to do
that arises then it suggests that an array would be a more appropriate
data structure.

A tuple can be identified as such through the
L<C<builtin::reftype>|builtin/reftype> function, which will return
C<OBJECT> when given a reference to one.  Correspondingly, the
L<C<ref>|perlfunc/ref> function will return C<OBJECT> if given a reference
to an unblessed tuple, and the default stringification of a reference to a
tuple will include C<OBJECT>.  Beware that this usage of the word "object"
is confusing: it was already a somewhat overloaded term before the tuple
data type existed, so it makes a poor way to identify the tuple data type.

Via this module, a tuple can be constructed from a list of slot values,
and the slot values in a tuple can be read and written.  This module
represents each slot value in the form of a reference value (for a slot
referencing an object) or an undefined value (for an empty slot).

=head2 Core class system

The L<Perl core class system|perlclass> (introduced alongside tuples in
Perl 5.37.9) uses tuples as the representation format for its classful
objects.  This class system is not the subject of this module, but it
has a special relationship to the tuple data type, which is worthy of
comment.  Beware that the core documentation overloads the term "class",
using it both to refer to Perl classes in general (i.e., packages into
which objects get blessed) and to refer specifically to the classes of
this class system.

A blessed object constructed through the automatically-generated
constructor for a core class system class is always a tuple object,
and its slots reference the field variables for that class and its
superclasses.  The field variables can only be scalars, arrays, and
hashes, so those are the only types of objects that will be referenced
by such tuples.  Once the object is fully constructed and visible as
C<$self> to method code, none of the tuple slots are empty, and the
sequence of object references never subsequently changes.

Beware that mutating a tuple used as the representation of one of these
blessed objects can easily cause malfunction of its class.  Of course,
the same goes for mutating the innards of any data structure representing
a classful object.

=head2 Experimental status

The tuple data type as supplied by the Perl core is experimental.
This means that it could change significantly, or be removed entirely,
in a future version of Perl.

Furthermore, the manner in which this module uses the tuple data type
is somewhat speculative.  The Perl core does not document precisely what
kinds of operations are intended to be possible on tuples, and there is no
established common practice, so the semantics offered by this module are
in part guesses.  For example, rewriting a tuple in a way that changes
the number of slots it has is something that's naturally possible to do
with the data structure, but isn't ever performed by core code, and it
might be decided in the future that it should never happen to a tuple
beyond its initial construction.

The mutability semantics offered by this module are particularly
speculative.  As of Perl 5.37.10, the Perl core neither sets nor honours
the read-only flag that this module uses.  It is not possible to discern
in what situations the core intends tuples to be mutable, because the core
code is not self-consistent on this point: it assumes that slot values
will not change in situations in which it can actually change them.
When some coherent semantics are decided upon for the core, it should
be possible to achieve cooperation between this module and the core,
and this module might also change behaviour on the buggy Perl versions
to get closer to the consensus semantics.

=cut

package Tupelo::Munge;

{ use 5.032001; }
use warnings;
use strict;

use XSLoader;

our $VERSION = "0.000";

use parent "Exporter";
our @EXPORT_OK = qw(
	pure_tuple constant_tuple variable_tuple
	tuple_mutable tuple_length tuple_slot tuple_slots
	tuple_set_slot tuple_set_slots tuple_seal
);

XSLoader::load(__PACKAGE__, $VERSION);

=head1 FUNCTIONS

=head2 Construction

=over

=item pure_tuple(REF ...)

Returns a reference to an immutable tuple.  The tuple is composed of
the slot values supplied by the I<REF>s, each of which must be either
a reference or undefined.  The tuple is not necessarily fresh.

=item constant_tuple(REF ...)

Creates a fresh immutable tuple and returns a reference to it.  The tuple
is composed of the slot values supplied by the I<REF>s, each of which
must be either a reference or undefined.

=item variable_tuple(REF ...)

Creates a fresh mutable tuple and returns a reference to it.  The tuple
is initialised to contain the slot values supplied by the I<REF>s,
each of which must be either a reference or undefined.

=back

=head2 Examination

=over

=item tuple_mutable(TUPLE)

I<TUPLE> must be a reference to a tuple.  Returns a truth value indicating
whether the tuple is mutable.