Complex Style

[Fumeur]

According to my understanding of the standard, both of

_Complex double c1; (1)
double _Complex c2; (2)

should be syntactically correct and equivalent. Now, there's
been a recent thread here according to which (1) is not
accepted by at least one major compiler. On the other hand,
in existing code I'm working on I have almost exclusively
seen (1) and very rarely (2); gcc's documentation also uses
(1) and never (2).

Is my assumption above correct that both styles are legal
in strictly conforming code?

If yes, which style is preferable, seeing that not all
compilers implement both styles?

[Flash Gordon]

Fumeur wrote, On 15/06/08 23:46:

According to my understanding of the standard, both of

_Complex double c1; (1)
double _Complex c2; (2)

should be syntactically correct and equivalent.

They are.

Now, there's
been a recent thread here according to which (1) is not
accepted by at least one major compiler.

Define major. I had never heard of the compiler in question (if it is
the one I suspect) until I joined this group. MSVC and gcc, on the other
hand, I was aware of.

On the other hand,
in existing code I'm working on I have almost exclusively
seen (1) and very rarely (2); gcc's documentation also uses
(1) and never (2).

So? Neither gcc documentation (specifically its example) nor examples in
the standard are proof of what is legal.

Is my assumption above correct that both styles are legal
in strictly conforming code?

They are both acceptable in a strictly conforming C99 program.

If yes, which style is preferable, seeing that not all
compilers implement both styles?

Personally I would use "_Complex double" (or maybe just "_Complex" if
that is legal) because to me it "sounds" better. If, however, programs
using _Complex tend to do it differently then I would switch to however
the majority did it for consistency.
--
Flash Gordon

[Keith Thompson]

Flash Gordon <sp**@flash-gordon.me.ukwrites:

Fumeur wrote, On 15/06/08 23:46:

>According to my understanding of the standard, both of
_Complex double c1; (1)
double _Complex c2; (2)
should be syntactically correct and equivalent.

They are.

> Now, there's been a recent thread here according to which (1) is
not accepted by at least one major compiler.

Define major. I had never heard of the compiler in question (if it is
the one I suspect) until I joined this group. MSVC and gcc, on the
other hand, I was aware of.

> On the other hand, in existing code I'm working on I have almost
exclusively seen (1) and very rarely (2); gcc's documentation also
uses (1) and never (2).

So? Neither gcc documentation (specifically its example) nor examples
in the standard are proof of what is legal.

The standard refers to "float _Complex", "double _Complex", and "long
double _Complex" in both 6.2.5 (types) and 6.7.2 (type specifiers).
Of course other orders are legal (including "double _Complex long if
you want to be especially perverse, but please don't).

>Is my assumption above correct that both styles are legal
in strictly conforming code?

They are both acceptable in a strictly conforming C99 program.

>If yes, which style is preferable, seeing that not all compilers
implement both styles?

Personally I would use "_Complex double" (or maybe just "_Complex" if
that is legal) because to me it "sounds" better. If, however, programs
using _Complex tend to do it differently then I would switch to
however the majority did it for consistency.

No, just "_Complex" is not legal. (An example in C99 6.7.8p24 has
complex c = 5 + 3 * I;
This was corrected in one of the Technical Corrigenda.)

I'd follow the standard's usage (and I don't know why gcc's
documentation doesn't) unless there was some particular reason not to,
such as consistency with existing code.

I'd also add "#include <complex.h>" and use "complex" rather than
"_Complex".

--
Keith Thompson (The_Other_Keith) ks***@mib.org <http://www.ghoti.net/~kst>
Nokia
"We must do something. This is something. Therefore, we must do this."
-- Antony Jay and Jonathan Lynn, "Yes Minister"

[Gestorm]

Fumeur wrote:

According to my understanding of the standard, both of

_Complex double c1; (1)
double _Complex c2; (2)

should be syntactically correct and equivalent. Now, there's
been a recent thread here according to which (1) is not
accepted by at least one major compiler.

Hey, I test the following code:
/*file name is complex.c*/
#include<stdio.h>
int main()
{
double _Complex a = 1.5;
_Complex double b = 1.6;
printf("%f\n", a);
printf("%f\n", b);
}
and compile it buy command
$gcc complex.c
Then execute ./a.out
the result is:
1.500000
1.600000
No error occurs.
^_^

[Keith Thompson]

Gestorm <zh********@126.comwrites:

Fumeur wrote:

>According to my understanding of the standard, both of

_Complex double c1; (1)
double _Complex c2; (2)

should be syntactically correct and equivalent. Now, there's
been a recent thread here according to which (1) is not
accepted by at least one major compiler.

Hey, I test the following code:
/*file name is complex.c*/
#include<stdio.h>
int main()
{
double _Complex a = 1.5;
_Complex double b = 1.6;
printf("%f\n", a);
printf("%f\n", b);
}
and compile it buy command
$gcc complex.c
Then execute ./a.out
the result is:
1.500000
1.600000
No error occurs.

gcc was not the "major compiler" in question, so this isn't
surprising.

But you're printing values of type double _Complex using a "%f"
format, which requires an argument of type double.

Here's a corrected version:

#include <stdio.h>
#include <complex.h>

int main(void)
{
double _Complex a = 1.5;
_Complex double b = 1.6;
printf("a = %f + %f*I\n", creal(a), cimag(a));
printf("b = %f + %f*I\n", creal(b), cimag(b));
return 0;
}

The output is:

a = 1.500000 + 0.000000*I
b = 1.600000 + 0.000000*I

--
Keith Thompson (The_Other_Keith) ks***@mib.org <http://www.ghoti.net/~kst>
Nokia
"We must do something. This is something. Therefore, we must do this."
-- Antony Jay and Jonathan Lynn, "Yes Minister"

[Flash Gordon]

Keith Thompson wrote, On 16/06/08 01:47:

Flash Gordon <sp**@flash-gordon.me.ukwrites:

>Fumeur wrote, On 15/06/08 23:46:

<snip>

>Personally I would use "_Complex double" (or maybe just "_Complex" if
that is legal) because to me it "sounds" better. If, however, programs
using _Complex tend to do it differently then I would switch to
however the majority did it for consistency.

No, just "_Complex" is not legal. (An example in C99 6.7.8p24 has
complex c = 5 + 3 * I;
This was corrected in one of the Technical Corrigenda.)

I'd follow the standard's usage (and I don't know why gcc's
documentation doesn't) unless there was some particular reason not to,
such as consistency with existing code.

Ah well, my preference is not strong, so I may well do the same if I
ever end up using the complex types. I can certainly see the advantage
of following the way the standard expresses it.

I'd also add "#include <complex.h>" and use "complex" rather than
"_Complex".

I may well also do that if I ever use the type.
--
Flash Gordon

[Serve Lau]

"Fumeur" <[e-mail] in bericht
news:54*******************@invalid.invalid...

>
According to my understanding of the standard, both of

_Complex double c1; (1)
double _Complex c2; (2)

should be syntactically correct and equivalent. Now, there's
been a recent thread here according to which (1) is not
accepted by at least one major compiler.

The world wants to know which compiler

[Fumeur]

On Mon, 16 Jun 2008 18:41:16 +0200, "Serve Lau" <ni***@qinqin.comwrote:

>
"Fumeur" <[e-mail] in bericht
news:54*******************@invalid.invalid...

>>
According to my understanding of the standard, both of

_Complex double c1; (1)
double _Complex c2; (2)

should be syntactically correct and equivalent. Now, there's
been a recent thread here according to which (1) is not
accepted by at least one major compiler.

The world wants to know which compiler

Do your own research!

[Serve Lau]

"Fumeur" <[e-mail] in bericht
news:26*******************@invalid.invalid...

>
On Mon, 16 Jun 2008 18:41:16 +0200, "Serve Lau" <ni***@qinqin.comwrote:

>>The world wants to know which compiler

Do your own research!

*cry* never mind then, the world doesnt want to know that bad :P

[Fumeur]

On Mon, 16 Jun 2008 23:48:17 +0200, "Serve Lau" <ni***@qinqin.comwrote:

>
"Fumeur" <[e-mail] in bericht
news:26*******************@invalid.invalid...

>>
On Mon, 16 Jun 2008 18:41:16 +0200, "Serve Lau" <ni***@qinqin.comwrote:

>>>The world wants to know which compiler

Do your own research!

*cry* never mind then, the world doesnt want to know that bad :P

After reading back through the archives, I think "the world" really
may not know because the fact first popped up buried deeply inside
one of those homework threads.

The relevant article is "<ud************@localhost.invalid>" wherein
a C.L.C reader posted a piece of code using (among others) these two
type specifications:

_Complex double (1)

and

double _Complex (2)

In a followup article, "<31****************@aioe.org>", another
reader presented the compiler diagnostics he got on the line(s)
containing (1).

(being mainly a gcc user, I was a bit surprised by that...)