Dan Sugalski <[EMAIL PROTECTED]> wrote: > This is entirely a matter of opinion and data design ...
Yep, that's it. The current behavior additionally is inconsistent. Retrieving a reference (that is Parrot) out of a non-existant hash key gives and unrelated new PerlUndef, when assigning to that, nothing happens. When you get a ref by an existing key, you can happily change data inside the hash. I've refered to the subject of the former thread, there are examples. So what will/should Parrot provide, to solve that? IMHO we are additionally lacking vtable methods to achieve the HLL value assign behavior: We only have references inside our aggregates. That doesn't matter per se, you can always clone the PMCs before storing, or you store only newly created PMCs, that's ok, but in the case of retrieving a reference or a value from an hash with an unexisting key, it would matter, the one does autovifiy the latter doesn't. leo