Passing procedures to procedures

You do not have to define your procedure pointer f to point to a procedure. If the procedure pointer is not referenced, you can pass in any placeholder. However, you cannot define an empty matrix in a procedure call like this: myproc({}, x, a, b, ind).

Here is one way to do it:

b = 1.5;
x = 2;
a = rndn(100, 1).^2;

z = myproc(&logAproc, x, a, b, 0);
z = myproc("NULL", x, a, b, 1);


proc myproc(&f, x, a, b, ind);
    local f:proc, z;
    if ind eq 1;
        z = a*x+b;
    else;
        z=f(x,a,b);
    endif;
    retp(z);
endp;

proc (1) = logAproc(x, a, b);
    local z;
    z = a * ln(x) + b;
    retp(z);
endp;

There is nothing special about the string "NULL". You could replace "NULL" with a scalar 0 or a scalar error code. The code will even work if you replace "NULL" with an integer value. I would not recommend that, however, since procedure pointers are scalar integer pointers that provide the location of the procedure they point at. If you made a mistake in the code, you could end up with a much more confusing bug that way.

Since GAUSS does not treat the procedure pointer as a procedure pointer until after local f:proc, you could use the procedure pointer value as the indicator instead of ind. Below I use "NULL" as the indicator to perform the linear calculation, but you could also use a scalar 0 or a scalar error code as well.

b = 1.5;
x = 2;
a = rndn(100, 1).^2;

z = myproc(&logAproc, x, a, b);
z = myproc("NULL", x, a, b);


proc myproc(&f, x, a, b);
    local z;
    if f == "NULL";
        z = a*x+b;
    else;
        local f:proc;
        z=f(x,a,b);
    endif;
    retp(z);
endp;

proc (1) = logAproc(x, a, b);
    local z;
    z = a * ln(x) + b;
    retp(z);
endp;

While this is maybe a little more concise, there is no real significant advantage. I would recommend you choose the solution that seems most clear to you.

Great answer, thanks!

I like more your second solution, it seems more elegant and parsimonious.

Just one quick question. In the second version you write:

if f "NULL";

Should it not be this?

if f eq "NULL";

If I use, say, a scalar zero instead of the string, I assume it should be

if f eq 0;

Is this correct?

A final question, unrelated to this. How do you create these nice codes with indentations, colours, etc? I can't find these features in the forum editor, but perhaps you have a more fancy version:)

1. Your are correct. I made a mistake. It should read either if f eq "NULL" or if f == "NULL". I corrected the earlier post for future reference.
2. Yes, if you want to use a scalar zero instead of "NULL, then you would change if f eq "NULL" to if f eq 0
3. If you surround code snippets with pre tags like this:

   <pre>
   //insert code here
   </pre>
   

   then your code will be: inside the light grey box, printed with a monospace font and any indentation will be preserved. As to color: for now you have to use span tags with one of the following clases: gkeyword, gcomment or gstring. Adding something like Markdown to do this portion automatically is on our backlog, but I do not have an estimate of when it will be completed.

Thanks for the answer!