trait -> rw accessors return $self when used as setter

[forwardever]

is there a trait that makes rw accessors return $self when used as setter, so it is possible to chain

[stevan]

No, but it would be pretty easy to implement one.

sub chained_writer {
    my ($attr) = @_;

    die "rw trait is only valid on attributes"
        unless $attr->isa('mop::attribute');

    my $meta = $attr->associated_meta;
    $meta->add_method(
        $meta->method_class->new(
            name => $attr->key_name,
            body => sub {
                my $self = shift;
                $attr->store_data_in_slot_for($self, shift) if @_;
                $self;
            }
        )
    );
}

[forwardever]

thank you, will try

[forwardever]

okay, get My::Module=SCALAR(0x29f1280) instead of the actual value (getter)

[stevan]

ah, yes, sorry, you will need to do this:

sub chained_writer {
    my ($attr) = @_;

    die "rw trait is only valid on attributes"
        unless $attr->isa('mop::attribute');

    my $meta = $attr->associated_meta;
    $meta->add_method(
        $meta->method_class->new(
            name => $attr->key_name,
            body => sub {
                my $self = shift;
                if (@_) {
                    $attr->store_data_in_slot_for($self, shift);
                    return $self;
                } else {
                    return $attr->fetch_data_in_slot_for($self);
                }
            }
        )
    );
}

[forwardever]

works, thank you, are there docs or tests where I can find out how to get this into my mop implementation (do not want to write it as a sub in my main package)

maybe, this should be implemented in core, the mojolicious guys are making heavy use of it (e.g.)

[stevan]

Sorry, no docs yet.

Basically I suggest making a "MyProject::Traits" package and exporting it from there, this will allow you to control its usage.

I am not sure this is core-worthy, we are trying to keep things to a minimum.

[forwardever]

@stevan why are setters returning the value you set (as in Moose)

it is very hard to find a good reason for this behaviour, returning $self at least allows chaining

most of the time, the value I set is saved in a variable anyway, so return values are useless (I only make use of them in custom setters that do more crazy stuff)

[forwardever]

seems like we discussed this some time ago http://blogs.perl.org/users/forward_ever/2011/10/chaining-in-moose-mouse-moo-and-mo.html

[stevan]

Haha, yes we did :)

[forwardever]

your main argument those days was Moose compatibility

I can also live with personal custom traits (as shown by you above), but still think chaining is the predominant use case (and it would be faster if implemented as default ???)

will leave this issue open in case anybody wants to provide further thoughts, and close later if nobody cares :)

[stevan]

I disagree that chaining is the predominant use case. The standard perl accessor pattern that has been around for many, many years is to return the value that was just set. This is also a common pattern in a number of other languages I have used as well.

The only places where I have seen chaining (on a single instance) be the predominant pattern is jQuery and more recently Java (specifically some of the *Builder APIs that have come out recently).

[forwardever]

okay, I mean use cases in applications, not in the languages themselves, probably have seen too few apps that make good use of accessors returning the value that was just set

(with accessor, I mean simple ones, not the "crazy" ones you can create e.g. with moose)

[doy]

I don't think we're going to do this in core.