This is the declaration of the function and the compiler gives: "error: array type has incomplete element type" in both the declaration and definition. I don't see the error.
void calculate(float matrM[][], int origMatr[][]);
4 Answers
C doesn't have very powerful arrays, especially not when calling functions. Basically, the only data that gets sent to the function at run-time is the address of the first element. This means that the dimensions must be known in all directions except the topmost, in order for the compiler to be able to generate indexing code. Thus, you can't have two unknown dimensions, that makes indexing impossible and thus isn't allowed.
You can't declare an "any size" array like that, you must specify all dimensions except one, otherwise there's no way to compute the address of a random element in the array.
It's often better to break it down as:
void calculate(float *out, const float *in, size_t width, size_t height);
which will let you treat any memory block as float arrays, but requires you to specify the size, and the write the indexing manually:
out[y * width + x] = in[y * width + x];
4 Comments
Arlind
Is this only when passing two arrays? Because 30 mins ago I passed an array without any parameters and 2 integers and it didn't give me an error
user2384250
well, it depends, I'm not using C that much but I remember a "struct hack" that is usually implemented to avoid this kind of allocation and to get some kind of "automatic and auto-sized" allocation, but it's not really a defined-behaviour, it's more like an undefined behaviour, so it's not really safe to adopt it.
unwind
@Arlind This is when passing arrays. Basically, the only information that gets sent to the function at run-time is the address of the first element. All other information must be present in the declaration, otherwise indexing becomes impossible.
Jens Gustedt
@unwind, in C99 and with VLA parameters, indexing of multidimensional arrays as function parameters is easily doable. please see my answer.
Your "inner" dimensions must have a size, e.g.:
void calculate(float matrM[][5], int origMatr[][7]);
Otherwise, the compiler wouldn't know how to generate indexing code to access a particular element. Note also that these sizes must be compile-time constant expressions.
For more details, see How do I write functions which accept two-dimensional arrays when the width is not known at compile time? from the C FAQ.
answered May 30, 2013 at 15:02
Oliver Charlesworth
274k34 gold badges591 silver badges687 bronze badges
Comments
If you have a modern C compiler (C99 would do) you can have the "extra" dimensions as expressions of other function parameters that precede
void calculate(size_t n, float matrM[][n], int origMatr[][n]);
you'd just have to be careful that you'd have equivalent parameters in your declaration (the one I gave) and the definition (the one that provides the implementation of the function).
Comments
You only get one "free" dimension (the first one) when you're passing an array to a function. The correct declaration will look more like:
void calculate(float matrM[][N], int origMatr[][M]);
1 Comment
Eric Postpischil
N and M do not need to be constants, as of C 1999.
(float matrM[3][], int origMatr[3][])