On Sun, Jan 23, 2000 at 12:27:20PM -0500, [EMAIL PROTECTED] wrote:
> JNP> On Sun, Jan 23, 2000 at 12:05:44PM -0500, [EMAIL PROTECTED] wrote:
> JNP> What is wrong with this:
> >>
> JNP> $watcher->{data} = ...; #?
> >>
> >> that breaks the OO rules. one shoudl never assume knowledge about the
> >> internal stucture of an object. there should be a method/attribute for that.
>
> JNP> But you can't access the internal structure of a watcher except via
> JNP> methods. The hash part is entirely empty.
>
> then i am curious, what and how do you store the info for a watcher? if
> not in the hash of $watcher, then what? a pseudohash? or closures?
Magic:
static SV *wrap_watcher(void *ptr, HV *stash, SV *temple) {
SV *ref;
MAGIC **mgp;
MAGIC *mg;
assert(ptr);
assert(stash);
if (!temple)
temple = (SV*)newHV();
else
SvREFCNT_inc(temple);
if (SvOBJECT(temple))
croak("Can't attach to blessed reference");
assert(!SvROK(temple));
assert(mg_find(temple, '~') == 0); /* multiplicity disallowed! */
ref = newRV_noinc(temple);
sv_bless(ref, stash);
mgp = &SvMAGIC(temple);
while ((mg = *mgp))
mgp = &mg->mg_moremagic;
New(0, mg, 1, MAGIC);
Zero(mg, 1, MAGIC);
mg->mg_type = '~';
mg->mg_obj = (SV*) ptr; /* NOT refcnt'd */
*mgp = mg;
return ref;
}
The real data is not accessible from perl, except by methods.
> JNP> You can also add methods with a prefix. For example,
>
> JNP> package Event::Watcher;
> JNP> sub uri_data { ... }
>
> is that documented?
Not yet.
> what prefix is allowed (any string)? i assume the
> prefix is to avoid method name collisions. this is not cool OO coding
> either IMO. i will write damian conway to see what he thinks about that
> idea.
I think it's cool OO once it's documented. Perl's OO is pretty strange.
--
"Never ascribe to malice that which can be explained by stupidity."
via, but not speaking for Deutsche Bank
