Matt S Trout (dbix-...@trout.me.uk)

Hey, Dave, where would it be good to put this in the docs so Ovid knows
which file to patch?

I figure we want to use this as a specific example of a general thing to
know in roles, right?

Looking at perldoc Memoized, it can take a subref instead, and returns
the wrapped routine, so:

__PACKAGE__->meta->add_method('foo' => memoize(sub { __PACKAGE__ }));

should work fine. So should

*foo = Sub::Name::subname 'My::Role::foo' => memoize('foo');
sub foo { __PACKAGE__ }

and so should pretty much any variant thereon.

Maybe this wants documenting somewhere in Moose for people who don't read the
Memoize docs?

On Thu, Nov 06, 2008 at 04:59:41AM -0800, Michael G Schwern wrote:
> I still say subroutine attributes are the way. That's the syntax for
> attaching things to a subroutine in Perl 5. I'd make a decorate() attribute
> and pass it the decorators.

That's not what attributes do though, they call a class method at a random
point when the sub doesn't have a name yet and then you do a crazy dance to
try and do something useful with them.

We hates attributes, we do.

> sub incr : decorate(memoize, log($file)) {
>     $_[0] + 1;
> }

Meh. How about

memoized logged (to => $file) sub incr (Int $i) {
  $i + 1;
}

?

We can certainly decorate 'method' like that. Not sure if it's possible
with sub; when rafl took over trying to make sub signatures work I kind of
forgot most of what little I'd worked out about how 'sub' is parsed in
statement form.

Better still, how about we shift off to the perl5-syntax list and stop
bothering p5p until we have something worth shoving into core? :)

Other than the bigass diff to sqlite.sql, is there anything left to do here?

If not, lob me an htpasswd line and I'll set you up to commit this.

> The recreation of the test-schema generates many changes - many more
> then my :-(

Diffs against sqlite.sql are irrelevant - the only reason it's in svn at all
is for the (now vanishingly small) number of people patching DBIC who don't
have SQL::Translator.

Might actually just have Makefile.PL create it and drop the damn thing from
svn entirely - thoughts people?

DBIC_TRACE=1 perl myscript.pl

and see what queries you're running.

Also please note it's DBIx::Class or DBIC, DBIx:: contains lots of other
projects.