Re^3: Refresh a Module

Replies are listed 'Best First'.
Re^4: Refresh a Module by Danny (Chaplain) on Sep 26, 2024 at 03:19 UTC
Actually, in my example I forgot to `delete $INC{"My.pm"};` so the second require wasn't redefining My::Run. When I add the delete it works as expected, printing: `Loading... a: Running Loading... b: Running` [download] In the OPs example there is also a `delete $My::{Run}`. So `require My; print "a: "; My::Run(); # a: Running undef &My::Run; delete $INC{"My.pm"}; delete $My::{Run}; require My; print "b: "; My::Run(); # Undefined subroutine &My::Run called at ...` [download] says My::Run() is undefined in the second call. What does `delete $My::{Run}` do?	[reply] [d/l] [select]
Re^5: Refresh a Module by LanX (Saint) on Sep 26, 2024 at 13:54 UTC
> What does delete $My::{Run} do? It deletes the entry for the symbol (typeglob) in the namespaces/symbol table/ STASH `%My::` I can't test now, but my guess is that the compiled code is "linked" to the old typeglob's CODE slot. (IIRC that's what B::Concise shows š) Redefining the sub in a require or eval or whatever will repopulate the slot. But after deleting the entry in the symbol table a new typeglob is created and autovivified into the stash. ° I could create rather complicated tests to prove all of this with all interpretations of what the OP probably wants to achieve. Unfortunately I'm too busy for guesswork, and honorable ikegami or Dave Mitchell should know the implementation better. You might be interested into following introspection tools B::Terse to scan the OP tree ( B::Deparse ) Devel::Peek blessing the coderef into a package with DESTROY Looking up following documentation might help symbol table hash aka STASH typeglobs import/export mechanisms require and eval execution phases `perlmod` garbage collection and refcounts perlguts the Panther book from O'Reilly Cheers Rolf _{(addicted to the Perl Programming Language :) see Wikisyntax for the Monastery} Updates °) Also diving into implementation details is kind of non productive if `undef` and `sub` already do the job. And typeglobs are particularly icky Perl4 stuff which are badly explained and supported by introspection. š) see Re^2: Refresh a Module	[reply] [d/l]
Re^4: Refresh a Module by Timka (Acolyte) on Sep 26, 2024 at 03:35 UTC
Why bother with un defining a sub when you can just delete it? 🤔	[reply]
Re^5: Refresh a Module by LanX (Saint) on Sep 26, 2024 at 09:36 UTC
> Why bother with un defining a sub when you can just delete it? Two very, very different things. The delete removes the typeglob from the stash. This doesn't necessarily destroy the typeglob if it's still referenced elsewhere (like in another stash, compare refcount and garbage collection). Anyway, the sub's ref is in the CODE slot of the typeglob! Same story, destroying the typeglob doesn't mean destroying the sub, if it's referenced elsewhere. (Think of a typeglob like of a hash with only 6 slots, CODE is one of them) I could do a lot of experiments now to determine the internal mechanisms, question is rather if we are even allowed to rely on them. And others here know the C code better. Fact is that we already have clean interfaces for subs undef to destroy redefine to overwrite Those mechanisms are guaranteed to work. Please keep in mind that typeglobs are a Perl4 mechanism, refs came with Perl5. Not sure to which degree refcounts and garbage collection are applied to typeglobs and/or their slots. Cheers Rolf _{(addicted to the Perl Programming Language :) see Wikisyntax for the Monastery}	[reply]

Updates