I was thinking of allocation to be handing over a chunk of raw memory and initialization to be things like setting ref count to 1, [...]
Then allocation and initialization are done by the same function (not opcode). newSV and newAV, for example.
Surely allocation of a new array would only be done on entry to the block at the start of each iteration?
No, it's done on exit.
The directive to which I referred is SAVEt_CLEARSV. It's placed on the stack by SAVECLEARSV. In the code relating to SAVEt_CLEARSV in Perl_leave_scope in scope.c, you'll spot
switch (SvTYPE(sv)) { /* Console ourselves with a new value */ case SVt_PVAV: *(SV**)ptr = (SV*)newAV(); break; case SVt_PVHV: *(SV**)ptr = (SV*)newHV(); break; default: *(SV**)ptr = newSV(0); break; }
If you would care to point me to documentation where I can read some more that would be helpful
I don't know where or if this detail is documented. I discovered it while searching for my's runtime effect.
perl -MO=Concise -e"code" gives the names of the opcodes (e.g padav for my @array), which you'll find in the pp* files.
In reply to Re^10: references--hard vs anonymous operational weirdness
by ikegami
in thread references--hard vs anonymous operational weirdness
by hill
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |