On Fri, 2014-08-29 at 10:00 -0700, Karen Etheridge wrote: > On Fri, Aug 29, 2014 at 09:34:41AM -0400, David Mertens wrote: > > We can all agree that Perl likes to give lots of rope to its programmers. > > Calling UNIVERSAL::isa prevents mocking, but it also prevents nefarious > > classes from lying about what they do. Apparently, allowing mocking is *so* > > much more useful than safe-guarding against lying that the porters decided > > to (1) remove the documentation of this behavior (though anybody who > > understands Perl's object system could figure it out) > > Yes, a lot of things are tailored towards the newbie programmer, and > rightly so, because it's an army of well-intentioned but ignorant newbies > that can code us into a deep dark hole from which it takes a long time to > get us out. > > (And, as such, I tailor my answers on publicly archived and searchable > mailing lists accordingly.) > > I'm also saying all this just after fighting with fixing UNIVERSAL::isa's > code (the CPAN module, not the built-in version) and shaking my head at all > the wrongness that I found on the CPAN. > > > and (2) remove it > > from the Exporter list so that it is difficult to bring this function into > > your current package. > > I don't think that's the reason why it was removed. There were technical > issues involving interactions between UNIVERSAL and Export that were the > primary motivators here, instead. The p5p list contains all the gory > details. > > > If you think UNIVERSAL::isa is the > > right way to go, then use it. But you really should know what you're > > doing---and preventing---before using it this way. Chances are quite good > > that the OP had code written by somebody who didn't understand what they > > were doing. > > Agreed on all counts. And, FWIW, all the instances of 'UNIVERSAL::isa' > that I found on my searches on grep.cpan.me were of the > newbie-doing-it-wrong form, rather than the > yes-I-really-want-the-true-answer-not-the-mocked-one-and-I-understand-the-difference > form. > >
This thread has me curious. I wrote Sys::SigAction more than 10 years ago, and I have made minor improvements and bug fixes over the years since. I believe it is still deemed useful... useful enough for the Debian packagers to ask me to change the license to a pure perl license several years ago. Sys::SigAction has one instance of a call to UNIVERSAL::isa (line 135 of SigAction.pm) in a function context: die '$action is not a POSIX::SigAction' if not UNIVERSAL::isa( $action ,'POSIX::SigAction' ); This call is in set_sigaction() which is in a private function and it is essentially just defensive programming. I don't think I have had a problem with this since is was first released. The reason UNIVERSAL::isa is called as a function here is that that within the scope of this function how might one know if the scaler $action is blessed reference, just a reference, or something else. UNIVERSAL::isa() answers all of these questions in one statement. If UNIVERSAL::isa() is now deemed "incorrect" in a context like this, I'm curious, given the requirements this line of code meets, what those on this list think this line of code should be converted to.
