[Kirby Urner]
>  ...
>  Question:
> 
>  Why does float -> long work like this:
> 
>      >>> long(1.0826786300000142e+045)
>      1082678630000014218353234260713996413124476928L
> 
>  and not like this?
> 
>      >>> long(1.0826786300000142e+045)
>      1082678630000014200000000000000000000000000000L
> 
>  Kirby

Because your machine floating-point isn't decimal, it's binary:

> >> x = 2.**100
> >> print x
1.26765060023e+030
> >> print long(x)
1267650600228229401496703205376
> >> print 2L ** 100
1267650600228229401496703205376
> >>

So making long(2.**100)

1267650600230000000000000000000L

instead would be plain wrong.  The same thing happens in your example, but
is harder to picture because, offhand, I bet you don't know the exact
IEEE-754 bit pattern corresponding to 1.0826786300000142e+045 <wink> .

> >> x = 1.0826786300000142e+045
> >> import math
> >> mantissa, exponent = math.frexp(x)
> >> print mantissa
0.758577950788
> >> print exponent
150
> >> print math.ldexp(mantissa, 53)
6.832662753e+015
> >> mantissa_as_long = long(math.ldexp(mantissa, 53))
> >> mantissa_as_long
6832662753002049L
> >> print mantissa_as_long << (exponent-53)
1082678630000014218353234260713996413124476928
> >> print long(x)
1082678630000014218353234260713996413124476928
> >>

IOW, long(some_huge_integer_in_float_format) *does* have a large number of
trailing zeroes, but in base 2, not necessarily in base 10.

> >> print hex(long(x))
0x308C8A88868482000000000000000000000000L
> >>


_______________________________________________
Edu-sig mailing list
Edu-sig@[...].org
http://mail.python.org/mailman/listinfo/edu-sig