Front page | perl.perl5.porters |
Postings from December 2002
Hugo van der Sanden (via RT) writes:
> Tels <perl_dummy@bloodgate.com> wrote:
> :"khkramer@allafrica.com (via RT)" <perlbug@perl.org> wrote:
> :>lead to $foo's reference count being high by one, if $foo is an object
> :>that returns a reference to itself in its copy constructor.
> :
> :Uh, but isn't a copy constructor supposed to copy something, instead of
> :merely returning the same thing? How should that supposed to work with a
> :"defect" copy constructor?
> :
> :*puzzled*
>
> I couldn't see anything in the docs to state or imply that returning the
> thing itself should be illegal, and I can see no reason for it to be so.
> And if it should be illegal, we should make it an error rather than
> silently doing the wrong thing.
>
In addition, if we "force" the copy constructor to actually copy
something, a whole raft of common and useful approaches become
tortuous and inefficient. I'm thinking particularly of
while ( $foo++ ) {} # and
if ( $foo++ ) {}
constructions on iterator-like $foos.
Thanks to Hugo for further stripping down the test-case to demonstrate
that the problem isn't while- and if-centric (and mea culpa for not
doing so myself), but my problems that led to submitting the bug
report were exactly the above. If I have to actually copy my $foos in
order to use the post-increment operator -- which include lots of
stored state and (usually) a live database handle or two -- I'll have
to stop using overloading.
Kwin
Thread Previous
|
Thread Next