Thread: Possible to create methods for data in structs?

Hi there,
I am rather new to C, and I am having a bit of trouble finding information on this thing that I want to do (damn google is good - but it's a shame that the creator of C didn't know about it when he named the language after a letter of the alphabet!)

Basically, I have a struct that looks like:

Code:

typedef struct {
        constpart * electrons;
        double energy;
        double weight ;
        double potential;
        double kinetic;
        char dead;
} diffuser;

and I want to define a method to call on the diffuser object (sorry about this object oriented lingo, I realise that a struct isn't really an object...) that might look like:

Code:

void zeroEnergy(void)
{
        potential = 0.0;
        kinetic = 0.0;
        energy = 0.0;
}

is this possible?
I have been searching around for info, and it does seem like it's do-able. Unfortunately my C book doesn't cover unions, and all the websites I have encountered do lots of things with pre-processor directives (and if anyone could point me to a good book on that topic, and on properly structuring programs by properly allocating the various bits of code between header files and program files then I'd be most grateful) - but the websites give examples that are a little hard to follow and one even included a header that I don't have! (oocp.h or something)...

Any ideas?

many thanks in advance
James


*edit*
oh I should probably clarify - RE the title.
I called this post what I did because I thought it would mean people might understand what I wanted to achieve functionality-wise... i know that structs are not objects and so cannot have 'methods' (i guess)... but I figured people might see the functionality I wanted and tell me the C way of doing it... thanks!

Code:

void zeroEnergy(diffuser *d)
{
        d->potential = 0.0;
        d->kinetic = 0.0;
        d->energy = 0.0;
}

???

[edit]unedit

well of course I could do that... but... I had hoped that there might be a way to package the function up into the struct as part of its definition, so it could be called like:

Code:

diffuser walker;
walker.zeroEnergy();

or is this just one of those 'well why don't you just go LIVE in java!' questions?

Code:

typedef struct tag_diffuser
{
   //constpart * electrons;
   double energy;
   double weight ;
   double potential;
   double kinetic;
   char dead;
   void (*foo)(struct tag_diffuser *this);
} diffuser;

void zeroEnergy(struct tag_diffuser *this)
{
   this->potential = 0.0;
   this->kinetic = 0.0;
   this->energy = 0.0;
}

int main(void)
{
   diffuser diff;
   diff.foo = zeroEnergy;
   diff.foo(&diff);
   return 0;
}

???

ooh thankyou I didn't expect a reply this fast!

to turn your example around a bit, would this be possible?:

Code:

typedef struct tag_diffuser
{
   //constpart * electrons;
   double energy;
   double weight ;
   double potential;
   double kinetic;
   char dead;
   void (*zeroEnergy)(struct tag_diffuser *this);
} diffuser;

void ze(struct tag_diffuser *this)
{
   this->potential = 0.0;
   this->kinetic = 0.0;
   this->energy = 0.0;
}
diff.zeroEnergy = ze;

int main(void)
{
   diffuser diff;
   diff.zeroEnergy(&diff);
   return 0;
}

???

and do i need to explicity pass a reference to the object or is there another way?
(I realise that's how you've defined the function, but is there a way to do it that circumvents that requirement?)

C doesn't support that sort of construct (but C++ does )

ok in that case my next question must be (and as you may have gathered from some of the variable names that this is a scientific computing application, and one that does not scale well with system size (ie number of particles in the molecular system)): are there any performance hits i'd take if i converted it to a C++ application? and would this be the way to go? or is C smaller, faster and sweeter?

as i understand it, C++ is a superset of C, so would I just be able to use the OOP style stuff where I wanted it but keep the rest of it pretty much just C?

are there arguments against this type of thing given the nature of the program?

Code:

typedef struct tag_diffuser
{
   void (*foo)(struct tag_diffuser *this);
   //constpart * electrons;
   double energy;
   double weight ;
   double potential;
   double kinetic;
   char dead;
} diffuser;

void zeroEnergy(struct tag_diffuser *this)
{
   this->potential = 0.0;
   this->kinetic = 0.0;
   this->energy = 0.0;
}

int main(void)
{
   diffuser diff = {zeroEnergy};
   diff.foo(&diff);
   return 0;
}

???

[edit]

>and do i need to explicity pass a reference?

Yes.

C doesn't support that sort of construct (but C++ does ). just use:

Code:

void zeroEnergy(diffuser * this)
{
        this->potential = 0.0;
        this->kinetic = 0.0;
        this->energy = 0.0;
}

[edit]
beat me to it
[/edit]

There are some things that are allowed in C that don't work in C++. Functions with an implicit return type, for example, or passing any amount of parameters to a function with an empty argument list. Old-style parameter lists are also disallowed in C++. (Not that you often see them - the only working piece of code I know that uses them is the zlib library.)

so do you have any thoughts then on the relative virtues of C/C++ in scientific computing applications?

Often, C will be a little faster. Not always. C++ usually allows better modeling of the problem, thus making implementation simpler. In general, C++ requires a little less code to do the same thing.