Re^2: symboltable problem

Ok, call me stupid but I don't get what you are trying to say.

I re-phrase my question:

For package "Hubba", it's symbol table is stored in a hash %Hubba:: that maps symbol-names to typeglobs.

So when I do %Bubba:: = %Hubba:: (I am NOT doing *Bubba::=*Hubba::) the whole symbol-to-typeglob mapping is copied to another hash (which happens to be the symbol-table of another package).

So why does that then not create aliases?

I can do this:

our $var = "zappa";

$Bubba::{abba} = *var;

print $Bubba::abba;
[download]

which shows that I can create variables by manipulating the symbol-table. Why does that not happen in my script?

Can someone explain that in plain english without using the word "magic"?

Comment on Re^2: symboltable problem Download Code

Replies are listed 'Best First'.
Re^3: symboltable problem by Eliya (Vicar) on Jun 20, 2012 at 21:26 UTC
I think the problem is that `%Bubba:: = %Hubba::` is creating a new hash, which destroys the binding of `$Bubba::abba` to the original hash set up at compile time of the latter expression. This reasoning is based on the observation that both of the following work fine: `#!/usr/bin/perl use strict; $Hubba::abba = "zappa\n"; %Bubba:: = %Hubba::; # compile time deferred eval 'print $Bubba::abba';` [download] `#!/usr/bin/perl use strict; $Hubba::abba = "zappa\n"; # copying via hash slice doesn't create new hash @Bubba::{keys %Hubba::} = @Hubba::{keys %Hubba::}; # or: @Bubba::{keys %Hubba::} = values %Hubba::; print $Bubba::abba;` [download]	[reply] [d/l] [select]
Re^4: symboltable problem by choroba (Cardinal) on Jun 20, 2012 at 21:55 UTC
Yes, that seems to eplain it. Another solution then is to create the alias in a BEGIN block.	[reply]
Re^4: symboltable problem by morgon (Priest) on Jun 21, 2012 at 16:00 UTC
Thanks - that is very interesting. I did not expect that... I find this illustration quite amusing: `use strict; our $foo = "old value\n"; %:: = (%::); our $var = "new value\n"; $::{foo} = *var; print $foo; eval q\| print $foo \|;` [download] btw: Is there a nicer way to create a typeglob with a given value in the scalar-slot?	[reply] [d/l]