Thread: problem with double array assignment with a single type

i have two structures declared as such:

Code:

struct rtpkt {
	int src;          /* source router                                   */
	int dst;          /* destination router (immediate neighbor)         */
	int mincost[N];   /* minimun cost to nodes 0, 1, 2, and 3            */
};

struct dt
{
  int dist[N][N];     /* dist[i][j] is the distance to node i via node j */
};

And i have a line in the .c file:

Code:

            struct rtpkt send_neighbor;
            send_neighbor.mincost = dt[node].dist[node];

and the error that im getting is (cant assign two different types) not a quote.
What am i doing wrong here?

-thank you

Originally Posted by tabstop

The left side is an array of ints, the right side is one int. Perhaps you want to put the number into a particular slot in your array?

how is the right side one int? in my understanding, (apparently a wrong one)
doublearray[10][10] is an array of double pointers, so doublearray[1] is pointing to a location in memory that is the start of a 10*sizeof(int) block.
so is onearray[10], which onearray is pointing to the same sort of structure.

what is wrong from my understanding?

memcpy works in C...

Hrm, well, let's leave that subject untouched until more information appears.

No, you cannot. Use memcpy.

Or use a more obvious for loop to copy each member of the array.

Why?
1. It's the more natural thing to do.
2. Chances are, you'll screw up the 'size' parameter of memcpy at some point, and either copy too little, or too much data.
3. It might actually be quicker at run-time, depending on your implementation of memcpy.
4. When you decide to use C++, then the assignment approach will seem natural, and applying the technique to complex objects (with an overloaded assignment operator) will continue to "do the right thing". memcpy in this instance would be a disaster.

Originally Posted by Salem

3. It might actually be quicker at run-time, depending on your implementation of memcpy.

Normally I would agree, but I don't think I can do that with this one...
memcpy is an optimized and fast function for copying memory, so it will probably be faster, if anything. Or should be.

Your code seems misleading, since you are actually creating a new pointer and assigning to it instead of assigning to the destination array.
Although it does support assigning a structure to a structure, which pretty much means it is possible to assign a range of values to another range of values, just in a different form. You may think it is strange, but it's just how it is.

Originally Posted by Elysia

Normally I would agree, but I don't think I can do that with this one...
memcpy is an optimized and fast function for copying memory, so it will probably be faster, if anything. Or should be.

But it's probably better at copying 400-40000 bytes than 40 bytes. After all, the compiler may well come up with roughly the same loop as a short memcpy. And there is the overhead of the call or stuffing the right values in the right registers for a rep movs sequence (assuming x86). So I'm with Salem here. Use a loop.

Obviously, if we have 1000 indices at each size, memcpy() will probably be a tad faster.

--
Mats