On Mon, Feb 21, 2005 at 04:10:29PM -0800, David Wheeler wrote:
> On Feb 21, 2005, at 2:17 PM, Tim Bunce wrote:
>
> >>Which is also pretty easily done, eh? Something like this?
> >
> >Yes, though I wouldn't bother validating the contents of the hash.
> >Just checking you're getting a hash ref is enough I think.
>
> Well, I figured it was faster to do it up-front (during the
> construction of a database handle) than for each and every call to a
> method (where you had C<if ref $cb eq 'CODE'>.
True. The 'if ref' was a little over the top though. If the user
messes with the contents of $h->{Callbacks} then they get what
they've asked for.
Note that the STORE code for "Callbacks" only gets fired
in a few situations. Specifically, doing
$h->{Callbacks}->{foo} = ...;
will try to FETCH Callbacks first and if it doesn't exist perl will
STORE an *empty* hash ref which is then returned by the $h->{Callbacks}
expression. The following ->{foo} = ... doesn't trigger any code.
> >The dispatcher (written in C) will handle the normal pre- and
> >post-method callbacks,
> >so the explicit post-connect callback above wouldn't be needed in perl
> >code.
>
> Oh. Well the Perl stuff is what I know, and I was going for a working
> implementation based on my knowledge, as a proof-of-concept. In fact,
> Perl code is where you had it, too (commented out, for all practical
> purposes).
Yeap. Just me leaving a trail of breadcrumbs to follow later.
> Feel free to re-implement for performance to your heart's content.
:)
> >If any method wants to have extra callbacks then it could define them
> >to have
> >a "method.foo" name like connect_cached.reused here:
> >
> >>+ # If the caller has provided a callback then call it
> >>+ if (my $cb = $dbh->{Callbacks}{"connect_cached.reused"}) {
> >>+ $cb->($dbh, $dsn, $user, $auth, $attr);
> >>+ }
>
> Right. Would these be arbitrary, then, defined on a method-by-method
> basis?
Yeap. Though the standard way of writing it would be
if (my $cb = $dbh->{Callbacks}) {
...
otherwise $dbh->{Callbacks}{"connect_cached.reused"} will
autovivify the Callbacks hash and so cause a (slight) slowdown in
the method dispatcher for all later uses of the handle.
> >p.s. Don't forget tests, and then there's PurePerl support to
> >consider...
> >Feel free to throw it back at me if it looks like too much work, or
> >just
> >checkin bite sized pieces and see how it goes. Probably makes sense for
> >me to do at least the dispatcher changes.
>
> I was going to add a couple of tests once I got it compiling. It's
> compiling now, but I think I've bitten off more than I can chew in
> getting to work. The C is getting over my head. :-(. I've done some
> stuff, blindly, and attached the patch. At least the Pure Perl version
> is working. :-)
That's great. You've got it to the point were I can't ignore it any
more :-) I'll apply it and finish it up.
Many thanks David!
So, do you (or anyone else) fancy doing anything else?...
Tim.
