Sigh...
And well you might. You're obviously finding this quite hard. Read back. No one suggested that the compiler should suddenly stop catering for all cases did they!
Prior to this recent change, both lvalue and rvalue uses--and yes, even your obscured, aliased lvalue assignment--worked just fine. And they didn't double or treble the size of small anonymous arrays in the process. But subsequent to this change, any reference to $#--lvalue or rvalue--has the immediate affect of adding magic to the array, regardless whether it is actually used!
My suggestion was that the application of magic be deferred until it is actually required--when it is used as an lvalue--rather than when it is first referenced. As ikegami has demonstrated, that is perfectly feasible to do.
The whole point about knowing which are the common cases, and which are the obscure--may only happen in golf and pointless demonstrations--cases, is that you optimise for the former, not the latter.
In reply to Re^11: Does @{ } copy arrays?
by BrowserUk
in thread Does @{ } copy arrays?
by tford
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |