3d array allocation

Would this work?

Code:

---

int (*p)[x][y] = malloc(z * sizeof(*p));

---

If I understand correctly, p is a pointer to a 2D array and sizeof(*p) will be sizeof(int)*x*y?

You are correct... but you tell me if z is appropriately sized?

Wait... I just noticed your title. Let me ask you this, is this a 3d array?

I suppose. I mean since you have a pointer and the appropriate space you could do:

Code:

---

for (int i=0; i<z; ++i)
  for (int j=0; j<y; ++j)
      for (int k=0; k<x; ++k)
          printf("%d", p[k][y][i]; //or p[i][y][k]?

---

The loop limits would be z, x, y
The subscript would be [i][j][k]

Quote:

---

Originally Posted by Salem

The loop limits would be z, x, y
The subscript would be [i][j][k]

---

Yeah, typed it wrong. But to get this straight.... would my first code malloc() a 3rd array?

Not on it's own, no.

It allocates ONE array element of z dimension. you need x * y of them, each one installed in the respective [x][y] pointer.

--
Mats

Code:

---

int (*p)[x][y] = malloc(z * sizeof(*p));

---

So how much space does this allocates?
Since p is a pointer to a 2D array doesn't that mean that sizeof(*p) is the size of the 2D array, thus x*y*sizeof(int) ?
If that is true, then the above will allocate z*y*x*sizeof(int) space, thus space for a 3D array.

I could do of course:

Code:

---

int ***p = malloc(z * sizeof(*p));
for (...) {
  p[...] = malloc(...)
  for (...)
      p[...][...] = malloc(...)
}

---

and allocate a 3d array, but I am just wondering if the one line code would work, since it is simpler and guarantees a contiguous space of memory.

If it won't work, I would like to know what would happen. The pointer to a 2D array syntax is not that clear in C

A 3D that uses dynamic allocation will need to allocate one chunk of memory for each [x][y] location, so even if you avoid doing the first two levels of allocation, you still need to use a loop to fill in the bottom layer of the pointers.

--
Mats

Lets say you allocate one array. You would do:

Code:

---

int *p = malloc(z*sizeof(int));

---

So you have this:
p--> M1
The pointer points to one int, so you just allocate z int and have an array.
p--> M1 ..... Mz

If you had though (*p)[3][3] wouldn't that meant that you have this:
p--> M1 M2 M3 M4 M5 M6 M7 M8 M9
So if you do what I posted you should have z times that.

Maybe I am not understanding the (*)[][] syntax correctly?

So you are saying that my line would be the first line of code to allocate and I would need a double loop to fill the array?

It's not about the amount of data you allocate, but rather how it is distributed in the array.

To make things simpler, let's just use a 2D array:

Code:

---

int (*p)[3];

p[0] = malloc(3 * sizeof(int));

---

Now, you have a 2D array, but only one set of second level data, p[0][0..2] is available.

To make that a proper 2D array, you would need to do:

Code:

---

for(i = 0; i < 3; i++)
  p[i] = malloc(3 * sizeof(int));

---

Likewise, you need to allocate x * y blocks of z elements to make a 3D array, since each pointer p[x][y] is an individual pointer, and thus needs to point to the z elements it represents.

--
Mats

Yes, but if you do this:

Code:

---

int (*p)[3];
p = malloc(3 * sizeof(int[3]);

---

which I assume is equivalent with this:

Code:

---

int (*p)[3];
p = malloc(3 * sizeof(*p);

---

The thing is that I can understand how pointers work. But int*** p is no the same as int p[][x][y]. How the second is implemented is not really clear. But I understand that if you do this:

Code:

---

int a[2][2][2];
int (*p)[2][2] = a;
p[0][0][1] = 2;

---

then p points to a and you can use the indexes correctly.
So you point p to a chunk of memory that has 8*sizeof(int).
Instead of that, can't you dynamically allocate that piece of memory?

I should use this example which is more to my point, for a 2D array:

Code:

---

char (*p)[10] = malloc(100);

---

So you have a pointer that points to a chunk of 100byts. Shouldn't the indexes work as you expect?

Am I just missing an index here? I mean would this work?

Code:

---

char (*p)[10][10] = malloc(100);

---

And use it like:

Code:

---

p[0][0] = '1'; p[1][0] = '2'; etc etc

---

> It allocates ONE array element of z dimension. you need x * y of them, each one installed in the respective [x][y] pointer.
But that's what the sizeof(*p) in the original post tells you surely.
It should give you the same result as sizeof( int[x][y] ).

IMO, int (*p)[x][y] = malloc(z * sizeof(*p)); does what you want, so long as x and y are compile-time constants.

> It allocates ONE array element of z dimension. you need x * y of them, each one installed in the respective [x][y] pointer.
But that's what the sizeof(*p) in the original post tells you surely.
It should give you the same result as sizeof( int[x][y] ).

IMO, int (*p)[x][y] = malloc(z * sizeof(*p)); does what you want, so long as x and y are compile-time constants.

Tested it. Seems to work fine. Just looking at the net I get more complicated results on how to allocate a 3D array. So why not just use a single line??

It only works if all the minor dimensions are compile time constants, and thus the compiler can work out the size from the declaration of the pointer.

Why? This works fine:

Code:

---

#include <stdlib.h>
#include <stdio.h>

int main()
{
        int x,y,z;
        int t;
        int maxx, maxy, maxz;
        scanf("%d %d %d", &maxx, &maxy, &maxz);
        int (*p)[maxy][maxx] = malloc(maxz * sizeof(*p));
        for (x=0; x<maxx; ++x)
                for (y=0; y<maxy; ++y)
                        for (z=0; z<maxz; ++z)
                                t = p[x][y][z];
}

---

When I put a large number for maxx, maxxy it crushes though. But it seems to allocate a 3D array with maxx, maxy, maxz not being at constant-time valued.

You know Ptolemy's model of planet paths worked to accurately map where celestial bodies should be. Though completely wrong, it did work just fine too.