I'm trying to interface with C code, where I have a function
definition that takes two const char[]'s:

PetscErrorCode PetscInitialize(int*, char***, const char[], const
char[]);

However, the typical way that you pass "Null" values instead of
the last two arguments is "PETSC_NULL"

The problem is that PETSC_NULL is a macro defined to be 0.  I
can't do the normal "int PETSC_NULL = 0" thing, because it
doesn't follow the type signature of PetscInitialize.

How to I pass 0 in place of the last two arguments?  cast(const
char[])(0) doesn't work, and just an empty string ("\0") doesn't
work either.

-Andrew

On Tuesday, 14 August 2012 at 19:27:26 UTC, Andrew wrote:
> I'm trying to interface with C code, where I have a function
> definition that takes two const char[]'s:
>
> PetscErrorCode PetscInitialize(int*, char***, const char[], const
> char[]);
>
> However, the typical way that you pass "Null" values instead of
> the last two arguments is "PETSC_NULL"
>
> The problem is that PETSC_NULL is a macro defined to be 0.  I
> can't do the normal "int PETSC_NULL = 0" thing, because it
> doesn't follow the type signature of PetscInitialize.
>
> How to I pass 0 in place of the last two arguments?  cast(const
> char[])(0) doesn't work, and just an empty string ("\0") doesn't
> work either.
>
> -Andrew

How about cast(const(char[]))([0]) ?

On Tue, 14 Aug 2012 21:27:25 +0200, Andrew <andrew.spott@gmail.com> wrote:

> I'm trying to interface with C code, where I have a function
> definition that takes two const char[]'s:
>
> PetscErrorCode PetscInitialize(int*, char***, const char[], const
> char[]);
>
> However, the typical way that you pass "Null" values instead of
> the last two arguments is "PETSC_NULL"
>
> The problem is that PETSC_NULL is a macro defined to be 0.  I
> can't do the normal "int PETSC_NULL = 0" thing, because it
> doesn't follow the type signature of PetscInitialize.
>
> How to I pass 0 in place of the last two arguments?  cast(const
> char[])(0) doesn't work, and just an empty string ("\0") doesn't
> work either.
>
> -Andrew

I'm a tad confused. Why can't you just pass null? In D, the literal
0 is not implicitly convertible to void*, char***, or any other
pointer or reference type, so instead you should use the literal
null, which has no nameable type, but is implicitly converted to
any pointer or reference type you may wish.

If you really want to, you can have

enum PETSC_NULL = null;

so that you can write PetscInitialize( &i, PETSC_NULL, PETSC_NULL ),
but I'd recommend just using the null we have.

-- 
Simen

> How about cast(const(char[]))([0]) ?

Although I think what you actually want is probably just null, since a C function expecting an array is expecting a pointer, and passing cast(const(char[]))([0]) satisfies the signature but gives you a non-null array. I would guess what you really want to alias PETSC_NULL to null and pass that...

On Tuesday, 14 August 2012 at 20:10:13 UTC, Simen Kjaeraas wrote:
> On Tue, 14 Aug 2012 21:27:25 +0200, Andrew <andrew.spott@gmail.com> wrote:
>
>> I'm trying to interface with C code, where I have a function
>> definition that takes two const char[]'s:
>>
>> PetscErrorCode PetscInitialize(int*, char***, const char[], const
>> char[]);
>>
>> However, the typical way that you pass "Null" values instead of
>> the last two arguments is "PETSC_NULL"
>>
>> The problem is that PETSC_NULL is a macro defined to be 0.  I
>> can't do the normal "int PETSC_NULL = 0" thing, because it
>> doesn't follow the type signature of PetscInitialize.
>>
>> How to I pass 0 in place of the last two arguments?  cast(const
>> char[])(0) doesn't work, and just an empty string ("\0") doesn't
>> work either.
>>
>> -Andrew
>
> I'm a tad confused. Why can't you just pass null? In D, the literal
> 0 is not implicitly convertible to void*, char***, or any other
> pointer or reference type, so instead you should use the literal
> null, which has no nameable type, but is implicitly converted to
> any pointer or reference type you may wish.
>
> If you really want to, you can have
>
> enum PETSC_NULL = null;
>
> so that you can write PetscInitialize( &i, PETSC_NULL, PETSC_NULL ),
> but I'd recommend just using the null we have.

This appears to be the correct answer.  I was mostly just
confused.  something #defined is just replaced by the C
preprocessor (or so I thought), which means that in C, you would
be passing 0 to PetscInitialize... which seems weird.

passing null seems to work though.

-Andrew

On Tuesday, August 14, 2012 23:28:08 Andrew Spott wrote:
> This appears to be the correct answer. I was mostly just confused. something #defined is just replaced by the C preprocessor (or so I thought), which means that in C, you would be passing 0 to PetscInitialize... which seems weird.

0 _is_ the null value in C. What's weird is that they had a non-standard macro for it (NULL is frequently used, but PETSC_NULL is not standard at all).

- Jonathan M Davis

On Tuesday, 14 August 2012 at 21:33:34 UTC, Jonathan M Davis
wrote:
> On Tuesday, August 14, 2012 23:28:08 Andrew Spott wrote:
>> This appears to be the correct answer. I was mostly just
>> confused. something #defined is just replaced by the C
>> preprocessor (or so I thought), which means that in C, you would
>> be passing 0 to PetscInitialize... which seems weird.
>
> 0 _is_ the null value in C. What's weird is that they had a non-standard macro
> for it (NULL is frequently used, but PETSC_NULL is not standard at all).
>
> - Jonathan M Davis


There is some explanation in the header.  PETSC tries to be as
portable as possible, apparently there is some weird cases where
0 *isn't* the default value for NULL.