Calling real c-men...

[Gary Tarolli]

In article <9104082133.AA15472 at karron.med.nyu.edu>, Dan Karron at UCBVAX.BERKELEY.EDU writes:
> 
> Here is a problem at the border of the C-language. It is on the edge of 
> c's value as a terse, compact language, and a slippery mess when it comes to
> arrays and pointers. I would like someone to show me that this is really
> a beautiful feature to be appreciated and loved!
> 
> Here is my problem. I have allocated a tree of structures, and I want to
> pass a pointer to a piece of a structure to a subroutine. Now what I want to
> pass is a Pointer to an 2 dim CONTIGUOUS array of floats (Like a Matrix typedef,
> in that each float is in storage adjacent to each other, not an array of pointers to a list of floats)
> 
> In the receiving subroutine, I would like to be able to reference each
> element in the array pointed to by the same indexes as in the main structure.
> 

I am not sure that this is what you need, but here goes.  If you have an
array of float, eg. mat[4][4] and you pass it to a subroutine, you can
declare the incoming parameter as 

	float (*inmat)[4];	/* inmat is a pointer to 4 floats	*/

You can then index the array as inmat[row][column];  With different 
declarations (or castings) you can treat the array as having different
dimensions, eg.   float (*inmat5)[5]; inmat5 = ((*)[5])inmat; inmat5[1][4];
You can also assign these pointers and increment them:

	float mat[4][4];
	inmat = mat[2];		/* assign it to point to third row */
	inmat++;		/* increment it to fourth row	*/

Although you can do most of this with a paramenter declaration, 
eg. inmat[][4],  you cannot declare extra parameters.  Thus this technique
of using a pointer to a single dimen. array of floats is much more
versatile.

WARNING - although I have done this and I know it works, I have not checked 
	my examples above with the compiler.  So they may be off by a paren 
	or two.

--------------------
	Gary Tarolli