There is nothing wrong with returning a pointer to a static variable. It's just that the return type must be declared properly. It kind of makes sense if you try to reproduce what the declarations mean, and what the compiler accordingly tries to do. Consider the declaration static string flow[23][2];
. It declares 23 rows of strings, each with 2 elements. It helps if you look at it as a one-dimensional array. It just so happens that the array elements are arrays, but that's not so important right now (but we'll come back to it). From this perspective the array has just 23 elements, and each element has the size of 2 strings. Like with all arrays, the elements (here: arrys of 2 strings) are simply lined up in memory.
Like any array, flow
will in most contexts decay to a pointer to its first element. Incrementing that pointer will point to the next element, i.e the second row. Numerically the compiler must add 2*sizeof(string) to the address of flow in order to compute the address of flow's next element, which would be flow[1]. (It comes directly behind flow[0]. No magic here.)
Now if you declare string **flowpp
, flowpp is a pointer already, no need to decay. If we think it is pointing to the first element in an array, what type would the elements have? Sure enough: plain pointers. Incrementing flowpp would let it point to the next element. My pointers are 4 bytes large, so that numerically adding just 4 to flowpp would be enough to access flowpp's next element. Compared to what needs to be added to flow (remember, 2*sizeof(string)), that's completely different. The compiler computes the offsets of elements depending of what the pointers point to! Which is very different in the two cases.
So what can your function return? What does flow decay to when you return it? It decays to a pointer to its first element. The elements are arrays of two strings. It must be string xxx[2]
, with xxx being a pointer: hence string (*p)[2]
. If the pointer is actually returned by a function, we have a function call instead of plain p, so it's (*f())[2]
.
Here is a complete example:
#include<iostream>
using namespace std;
const int numFlowElems = 3, numArrElems = 2;
/** @return a pointer to the first element of a static array
of string[numArrElems]s.
*/
string (*flowFile())[numArrElems]
{ // init so that we see something below.
static string flow[numFlowElems][numArrElems]
= {{"1","2"},
{"3","4"},
{"5","6"}
};
// your function code ...
return flow;
}
int main()
{
// array decays to ptr, like usual. Ptr elems are string[numArrElems].
// ptrToArr is a pointer to arrays of two strings.
string (*ptrToArr)[numArrElems] = flowFile();
for( int flowInd= 0; flowInd<numFlowElems; ++flowInd )
{
for(int strInd = 0; strInd<numArrElems; ++strInd)
{
cout << ptrToArr[flowInd][strInd] << ' ';
}
cout << endl;
}
return 0;
}
How do you parse string (*flowFile())[numArrElems]
? I needed two attempts to get the declaration right, if that's any consolation. The key is that in C and C++ (not in C#, mind you!) a declaration has the shape of an expression.
You can do it from the inside to the outside: flowFile() is a function. The result is dereferenced because the function call has higher precedence than the star: *flowFile() is the dereferenced result. Apparently that result is an array of size numArrElems, with elements which are strings.
You can do it outside in: The result of (*flowFile())[numArrElems]
is declared as a string. (*flowFile())
is an array of strings with numArrElems elements. Apparently flowFile()
must be dereferenced to obtain that array so that flowfile is a function which returns a pointer to an array of numArrElems strings. That's true! It returns the first element of flow, which is exactly an array of strings.
Vectors of vectors might indeed be easier; if you want to retain the semantics you should pass references, as others mentioned: After all, all functions in your original program will operate on the same static array. If you pass vectors by value that will not be the case any longer. But then, that may actually be beneficial.