That would not serve my purposes, as any modules in question will not yet be loaded when the subref receives it. | [reply] |
| [reply] |
I think the OP's example is misleading you somewhat. When /dev/urandom says "That would not serve my purposes, as any modules in question will not yet be loaded when the subref receives it." I take that to mean that what /dev/urandom will do with the repurposed require function is to actually manipulate the module the instant it gets loaded, before anyone else can do anything with it. The original post said that it illustrated the issue, but not the why of it. Perhaps a better illustration would be:
*CORE::GLOBAL::require = sub {
CORE::require($_[0]);
if ($_[0] eq 'Net/Ping.pm') {
$Net::Ping::TIMEOUT = 5000; # milliseconds
}
};
I believe the OP means that they want to do something like that (I don't think that the code there would work even if everything else works - I don't think Net::Ping has a $TIMEOUT available). Perhaps check the $VERSION, or whatever. Modifying the @INC to include a code ref wouldn't allow for this - you merely return the code (via a filehandle), but at this point, it hasn't been compiled, so you can't interact with it yet.
I suspect it'd be better to derive from Net::Ping, and change the code to require that module instead... but we really don't have enough details to be definitive about this. | [reply] [d/l] |
BEGIN { unshift @INC, sub { ... } }
use lib 'somewhere/special';
require Net::Ping; # Resolves to "somewhere/special/Net/Ping.pm".
as lib also unshifts @INC. This accident may not be obvious if it's a module that uses lib.
For many purposes it works, but one should be aware of that it's not bullet proof against accidents.
lodin | [reply] [d/l] [select] |