I probably now have to implement all other arithmetic operators used in that JS code as above.
I can't come up with a better solution ... but if you were to use a perl whose ivsize was 4, then I think (not rigorously tested) you could take care of it with a simple use integer; at the top of the script.
On perl-5.38.0, built with just that configuration:
D:\>perl -V:ivsize
ivsize='4';
D:\>perl -le "print 1169367104<<5;"
3060008960
D:\>perl -Minteger -le "print 1169367104<<5;"
-1234958336
Cheers,
Rob | [reply]
[d/l]
[select] |
I believe you can do the complete calculation without conversion and just at the end downgrade to 32 bit and convert to signed. The following should work:
sub calc_32bit_signed( $value ) {
return unpack( "l", pack "L", ($value & 0xffffffff)))
}
...
say calc_32bit_signed( $n << 6 );
say calc_32bit_signed( $n + 2 );
| [reply]
[d/l] |
return unpack( "l", pack "L", ($value & 0xffffffff)))
For which finite numeric values do any of the following 3 renditions differ:
say unpack( "l", pack "L", ($value & 0xffffffff));
say unpack( "l", pack "L", ($value));
say unpack( "l", pack "l", ($value));
Sorry ... I'm not trying to be critical ... not even nitpicking ... but I haven't found any such values, and I'm genuinely curious as to whether they exist.)
Cheers,
Rob | [reply]
[d/l]
[select] |
| [reply]
[d/l]
[select] |
I don't think that will work, for example consider something like ($value << 6) + length($value). So if the first part overflows in JS then it is necessary to overflow (the first part) in Perl too. Simulating the overflow for the sum would not do. At the moment I am running JS code via shelling out to node.js and that works just fine.
I was hoping that there would be a module like Math::BigInt where you specify the number of bits or something like use smallint; But now, come to think of it, perhaps a new module could consist of only overloading arithmetic operators. Let me see if I can whip up something based on your pack receipe.
Edit: see Re: Trying to translate overflowing JS code to Perl with working code for this.
| [reply]
[d/l]
[select] |
package Math32 {
use Inline 'C', <<~'END';
#include <stdint.h>
long calc(long input) {
return (int32_t)((int32_t)input << 5);
}
END
}
use v5.36;
say Math32::calc(1169367104);
-1234958336
Then all you need to do is write all the JS code in C, using int32_t variables and typecasts. If you want to declare the function parameter and return type as int32_t, you'd need to add it to the typemap, which is a hassle. | [reply]
[d/l]
[select] |