[C++] How to tell function which part of struct to use

[meesa]

As an explanation, I'll use a very simple example.

Say I have a structure defined as such:

struct data{
  char c;
  int x;
};

And I want to create a function to print either c or x. However, I don't want to just use a switch or something to figure this out, I want to reference which part I'm outputting.

Now here's the twist

vector<data> dataList;

I need to access element x or c in the vector, but all I pass is the vector and which element, so that I might have something like:

void (vector<data> x, DataType i){
  cout << x[0].i;
}

So now, if I pass either x or c (someway), I can output one of them. Does that make sense, and can it be done?

Thanks.

[N36fSL3A]

I don't understand.

If you wanted to just pass the element of a vector, why not just make a function like this?

vector <Data>myvector;

void yourFunction(Data dat)
{
    // code
}

int main()
{
    yourFunction(myvector[0]);
    
    return 0;
}

So now, if I pass either x or c (someway), I can output one of them. Does that make sense, and can it be done?

I'm trying to figure out if you mean the struct member or the vector

[meesa]

I would give the function the vector, and the function would then loop through the vector (Yes, I know I didn't put a loop) printing element i. However, i could really be either x or c.

So when I say
cout << dat[index].i << endl;
That could be either of the following:

cout << dat[index].x >> endl;
// OR
cout << dat[index].c >> endl;

Which one i points to is based on what is passed to the function.

[N36fSL3A]

Or you could just do it once, then loop through the vector and in the for loop call the function for each element.

[Disch]

I'm pretty sure this cannot be accomplished in C++ or any other strong typed language. I've only ever seen this functionality in scripting languages like Python.

There might be a way to simulate it with templates and some very nasty pointer math... but I'm hesitant to even suggest it because it's so ugly.


EDIT:

Here's my idea, untested. Again I don't recommend this and am just showing it because I find it interesting. I would strongly suggest you reconsider whatever it is you're trying to do and come up with a different way to do it.

#include <cstdlib>  // <- need stdlib for offsetof()

template <typename T>
void yourfunction(const std::vector<data>& d, size_t offset)
{
    std::cout << *static_cast<const T*>(  static_cast<const char*>(&d[0]) + offset ) << std::endl;
}

//.....
// usage:

std::vector<data> v;
// assume 'v' populated here

yourfunction<int>(v,offsetof(data,x));  // to print 'x' as an 'int'
yourfunction<char>(v,offsetof(data,c));  // to print 'c' as a 'char' 

[meesa]

Well, how would you suggest I sort a vector (End goal) based on the values of certain element inside the structure, so that at one point in the program, I might sort by c, and at another point, I might sort by x? Do I just have to create a function for each element? Even though the code will look the exact same except for it being .c and .x?

[Daleth]

You could have the second parameter be an int/unsigned whose value indicates which data member to sort by. (Probably a lame way to do so)

[meesa]

Well, in that case, then I'd still have nearly identical code inside of the if statements, which is what I was trying to prevent. It would be cleaner code to just have separate functions, considering that will be using different data types.

[N36fSL3A]

Wait, I think I'm understanding this. You want to return 2 different types for 2 different functions?

In that case you could just have overloaded functions.

int functionForX(std::vector);
char functionForC(std::vector);

// The compiler will determine which one to use based on the data type. 

[BHX]

Do you mean something like this?

#include <iostream>
#include <vector>

using namespace std;

struct data{
	char c;
	int x;
};

void printfunctx(vector<data> s, int index)
{
	cout << s[index].x << endl;
}

void printfunctc(vector<data> x, int index)
{
	cout << x[index].c << endl;
}

int main()
{
	vector<data> Dat;
	Dat.push_back(data()); 
	Dat.push_back(data()); 
	
	Dat[0].c = 'y';
	Dat[0].x = 24;
	
	Dat[1].c = 'n';
	Dat[1].x = 2;
	
	printfunctx(Dat, 0);
	printfunctc(Dat, 0);
	printfunctx(Dat, 1);
	printfunctc(Dat, 1);
	
	return 0;
}

[Disch]

Well, how would you suggest I sort a vector

Oh jeez, this is for sorting?

You can specify your own routine to sort by. If you want to sort by different values, just give a different routine.

http://www.cplusplus.com/reference/algorithm/sort/

See the "custom" version.

EDIT:

example (with C++11 lambdas):

// sort by 'c':
std::sort(  yourvector.begin(), yourvector.end(),
    [](const data& a, const data& b) { return a.c < b.c; } );

// sort by 'x':
std::sort(  yourvector.begin(), yourvector.end(),
    [](const data& a, const data& b) { return a.x < b.x; } );

[BHX]

Oops, missed the sort part. Thought he was just trying to figure how to use structs with vectors and how to print each one of the members.

[meesa]

Disch, I know that exists. Essentially, I'm trying to not have neatly identical code. BHX just gave a perfect example of what I was hoping to avoid. The functions are identical, save a few letters. But if that's what I have to do, I guess I have to. I was just hoping there was some way to not do it that way.

[BHX]

Yeah, I'm not sure if there is a way to do it, unless you add a third parameter to use an if statement to compare.

void printfunct(vector<data> s, int index, int check)
{
        if(check == 0)
             cout << s[index].c << endl;
        if(check == 1)
             cout << s[index].x << endl;
}

printfunct(Dat, 0, 0);  
printfunct(Dat, 0, 1);

[Disch]

Disch, I know that exists. Essentially, I'm trying to not have neatly identical code

I'm not sure I see the reason for wanting to avoid this. In my example, the sort routine is literally a single comparison. It's no more redundant to write the sorting lambda than it is to specify which element you want to sort by. It's just more verbose.

[meesa]

Less code = easier to change if necessary.

The idea was that maybe I could pass something that would say which element i pointed to. It would seem not, and now I know. I appreciated everybody's help and input.

[N36fSL3A]

Sorry, I couldn't clearly understand the question.

[Daleth]

//Assuming this is a struct only for holding multiple data types
struct data{
   char c;
   int I;
   char operator char() ()const{return c;}
   int operator int() ()const{return I;}
};

template<typename T>
void sortbytype(vector<data>& tosort, T dummy){
   sort(tosort.begin(), tosort.end(),
      [](const data& d1, const data& d2){
         return
            static_cast<decltype(dummy)>(d1)
            < static_cast<decltype(dummy)>(d2)
         ;
      }
   );
}

Pretty silly, and I don't even know if it would work.

[meesa]

I think I'll keep it clean, and just use functions. At least until I can fully understand that. :)

[Lachlan Easton]

@Daleth:
I may be wrong but do you really need the dummy? Couldn't you just pass the type as a template parameter? Also I think it would be safer to have those operators private and either friend sortbytype or have it as a member.