Re: Why are hex numbers inside strings extrapolating to numbers (incorrectly)?

That is a really good question. I did some experimenting and found it was the "0x<number>" construct that is causing it (the leading 0x is required). However, various experiments haven't revealed a pattern that I have noticed:

print "0x1" + 0 = 1
print "0x2" + 0 = 1
print "0x3" + 0 = 1.5
print "0x4" + 0 = 1
print "0x5" + 0 = 1.25
print "0x6" + 0 = 1.5
print "0x7" + 0 = 1.75
print "0x8" + 0 = 1
print "0x9" + 0 = 1.125
print "0x10" + 0 = 1
[download]

And first glance, it would seem some sort of floating point conversion is going on, but exactly what, I don't know except that 9/8 = 1.125 (but why eight?) Actually, the following pattern exists:

print "0x1" + 0 = 1          1/1     = 1
print "0x2" + 0 = 1          2/1     = 2
print "0x3" + 0 = 1.5        3/1.5   = 4
print "0x4" + 0 = 1          4/1     = 4
print "0x5" + 0 = 1.25       5/1.25  = 4
print "0x6" + 0 = 1.5        6/1.5   = 4
print "0x7" + 0 = 1.75       7/1.75  = 4 
print "0x8" + 0 = 1          8/1     = 8
print "0x9" + 0 = 1.125      9/1.125 = 8
print "0x10" + 0 = 1         10/1    = 10
[download]

0x01 is probably being interpreted as just a "0x1", whatever that is. Since all the results are on powers of 2 boundaries, I would presume some bit-oriented operation is happening.

Good question!

Comment on Re: Why are hex numbers inside strings extrapolating to numbers (incorrectly)? Select or Download Code

Replies are listed 'Best First'.
Re: Re: Why are hex numbers inside strings extrapolating to numbers (incorrectly)? by MrNobo1024 (Hermit) on Feb 26, 2001 at 21:02 UTC
0x10 is really 16.	[reply]
Re: Re: Re: Why are hex numbers inside strings extrapolating to numbers (incorrectly)? by mr.nick (Chaplain) on Feb 26, 2001 at 21:06 UTC
No, I think in THIS sitation (which ISN'T taking 0x## as hex numbers) 0x10 is really the same as 0x1, 0x100, 0x183872. Tye believes its a problem with the stdlib call atof() and I tend to agree (the same version of Perl produce different results on different OS's). addendum: it definately is the atof() function call that is the issue. From my Linux box with glibc 2.1.3 (that caused the strange float point results with Perl): `nicholas@neko/3:(pts-0.neko) ~/tmp > cat test.c #include <stdlib.h> int main(void) { printf("%f\n",atof("0x9")); } nicholas@neko/3:(pts-0.neko) ~/tmp > cc test.c -o test ; ./test 1.125000` [download] and from my Solaris 2.6 box (which produced "0" with Perl): `nicholas@peacock2/5 ~ > cat >test.c #include <stdlib.h> int main(void) { printf("%f\n",atof("0x9")); } nicholas@peacock2/5 ~ > cc test.c -o test; ./test 0.000000` [download]	[reply] [d/l] [select]