What is the distinction between IV and NV variables?

[Kc12349]

Hello All,

Why is a NV and IV created in this case as opposed to just an IV? If instead $scalar2 is set to '0', only an IV is created.

Additionally, why is NV used to track numeric value for $scalar1 when it is modified and not IV? If $scalar2 is instead set to '0' initially, numeric value will be stored in IV.

Thanks for any enlightenment you can offer.

use Devel::Peek;

my $scalar1 = '';
Dump($scalar1);
my $temp1 = $scalar1+1;
Dump($scalar1);
$scalar1 = $scalar1+1;
Dump($scalar1);

SV = PV(0x36974) at 0x43ef74
  REFCNT = 1
  FLAGS = (PADMY,POK,pPOK)
  PV = 0x2b23ad4 ""\0
  CUR = 0
  LEN = 4
Argument "" isn't numeric in addition (+) at C:\Projects\test.pl line 
+17.
SV = PVNV(0x2bbc254) at 0x43ef74
  REFCNT = 1
  FLAGS = (PADMY,POK,pIOK,pNOK,pPOK)
  IV = 0
  NV = 0
  PV = 0x2b23ad4 ""\0
  CUR = 0
  LEN = 4
SV = PVNV(0x2bbc254) at 0x43ef74
  REFCNT = 1
  FLAGS = (PADMY,NOK,pNOK)
  IV = 0
  NV = 1
  PV = 0x2b23ad4 ""\0
  CUR = 0
  LEN = 4

###

my $scalar2 = '0';
Dump($scalar2);
my $temp2 = $scalar2+1;
Dump($scalar2);
$scalar2 = $scalar2+1;
Dump($scalar2);

SV = PV(0xa26974) at 0x2965d3c
  REFCNT = 1
  FLAGS = (PADMY,POK,pPOK)
  PV = 0x2aa9cd4 "0"\0
  CUR = 1
  LEN = 4
SV = PVIV(0x43297c) at 0x2965d3c
  REFCNT = 1
  FLAGS = (PADMY,IOK,POK,pIOK,pPOK)
  IV = 0
  PV = 0x2aa9cd4 "0"\0
  CUR = 1
  LEN = 4
SV = PVIV(0x43297c) at 0x2965d3c
  REFCNT = 1
  FLAGS = (PADMY,IOK,pIOK)
  IV = 1
  PV = 0x2aa9cd4 "0"\0
  CUR = 1
  LEN = 4

[derby]

Perl tries to do the right thing and if it thinks the integer addition may cause an overflow, it will create the NV slots and use fp arithmetic. You can follow through the perl source (sv.c and/or pp_hot.c) but I believe what you're seeing is perl dropping the string '0' into the case of not causing an overflow and '' into the default possible overflow case.

[ikegami]

Why is a NV and IV created in this case as opposed to just an IV?

PVNV scalars always have a field for an IV/UV/RV. The PVNV in question doesn't have an IV, though. (IOK is false.)

You actually want to know why a floating point is used (converting the PV scalar to a PVNV scalar) instead of an integer (which would convert the PV scalar to a PVIV scalar).

Don't know.