| Re: - unsigned variables [message #41051] |
Thu, 23 September 2004 17:53 |
JD Smith
Messages: 850
Registered: December 1999
|
Senior Member |
|
|
On Thu, 23 Sep 2004 15:56:16 -0700, Holger Fleckenstein wrote:
> A strange behavior in IDL occured to me.
>
> In C++ if I do:
> unsigned short x=1;
> printf("%d",-x);
> I get
> -1
> like I would expect.
>
> In IDL however:
> x=1U
> print, -x
> gives
> 65535
> So it basically treats it like I had done:
> print, uint(-1)
>
> Does anybody have an explanation for this? Is this, because of a typecast
> before executing the print? (Can creat bugs, which are hard to localize.)
Nope, it's because your print format is treating it as a signed long
integer. Try:
unsigned short x=1;
printf("%hu",-x);
which gives:
65535
The difference is, IDL *knows* your integer is an unsigned short. C
doesn't know or care, and so is happy to print it however you like.
You can always change IDL's mind by explicitly casting it:
IDL> print,fix(-1U)
-1
Note that short -1 and 65535 are actually represented by the exact
same bit pattern, namely:
1111111111111111
JD
|
|
|
|
| Re: - unsigned variables [message #41053 is a reply to message #41051] |
Thu, 23 September 2004 17:44  |
Longtime Lurker
Messages: 1
Registered: September 2004
|
Junior Member |
|
|
Holger Fleckenstein wrote:
> A strange behavior in IDL occured to me.
>
> In C++ if I do:
> unsigned short x=1;
> printf("%d",-x);
> I get
> -1
> like I would expect.
Whereas:
printf("%u",-x);
4294967295
printf("%hu", -x);
65535
like I would expect
%d simply tells C to output the bit pattern stored in the argument(s) as if
it were a *signed* integer
Similarly
printf("%f", -x);
gives
2.102785
on my little-endian machine. The bit pattern is identical in all four cases
all that changes is the way it is interpreted.
>
> In IDL however:
> x=1U
> print, -x
> gives
> 65535
> So it basically treats it like I had done:
> print, uint(-1)
But -x is a UINT just like x
help, -x
<Expression> UINT = 65535
So IDL is doing exactly what you ask it to
Consider
print, -x, format = '(F)'
65535.0000000000000000
>
> Does anybody have an explanation for this?
> Is this, because of a typecast before executing the print?
> (Can creat bugs, which are hard to localize.)
IDL outputs the true value of -x taking its type into account - any
formatting is applied to this value. C[++] outputs the value of the bits
interpreted according to the rules of the supplied conversion specifier.
The C behaviour has caught me out with code like
long long x = 1, y = 1;
printf("%d %d\n", x, y);
1 0
When I should have had
printf("%lld %lld\n", x, y);
1 1
IDL's behaviour seems preferable to me...
Paul
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
