Hello Andrey,

I was just playing with this feature in v0.89 in order to make some Wiki pages about it. Here is the example to the other users, just in case to clarify things.

Here is the function definition of "f" with two arguments- one of the argument is variable and another is function "F" (argumet function) - with single argument
f(x,F(1))←F(x+2)
On the LHS of the definition F(1) means that argument function F has one argument (I prefer this way) although it could be defined as F(dummy) where "dummy" might be whatever we like - it does not matter.
Let's put some variables and functions here
a←6
x←mat(1,3,2,1)
fun(y)←y
fun1(z)←z+a
Now, few calls of function "f"
f(x,fun(xx))=mat(3,5,2,1)
f(x,fun1(xx))=mat(9,11,2,1)
It must be taken care that arguments of argument function "fun" and "fun1" have the following restrictions:
1. constants not allowed (like fun(5))
2. must be a variable different that the name of arguments in the function "f" (fun(x) not allowed here - x is the argument of "f" )
3. also the name of previously defined variable is not allowed (f(a) not allowed - variable "a" is defined above)

I hope this is Ok.

Regards
Radovan

P.S. I think this is not possible (it would be quite to much at the moment )
F(x,y,U(x,f(1)))←U(x,f(x))+2
Could you confirm this, please. This would be another restriction, but might be logical from the SMath implementation - not sure?
4. Argument function "U" in the argument list of "F" can not have another argument function in the argument list - "f(1)"

Edited by user 31 July 2010 11:22:35(UTC)  | Reason: Not specified

Hi,

I can't manage to use functions as arguments in following scenario:

f(g(1);x):=g(x)
g(x):=eval(x+x)
f(g(_);2)=Error: "_ Not Defined".

eval here disables using function as argument (yet another eval() deficiency! ). This is highly unfortunate, as my functions have to use evals to avoid overflow.

Another problem:

f(g(2);x;y):=g(x;y)
g(x;y):=x^y
f(g(_;_);1;2)=4
f(g(_1;_2);1;2)=1

(_, _1 and _2 are all uninitialized)

i.e. dummy argument names to function as argument affect the function expansion, in this specific case making it use y for both x and y! Of course, using different names makes things OK, but this is still a bug, because in the function definition the call is g(x,y), not g(y,y). I believe that the names of dummy arguments must be completely ignored (only tested for not being initialized), and repeating them should make no difference.

Edit: second problem is discussed in clause 3 of previous post #22. Still, I suppose that it should be fixed.

Edited by user 22 April 2014 05:35:08(UTC)  | Reason: Not specified

If eval() would return itself without evaluation if the numeric evaluation fails, this particular problem should be solved. Filed as proposal SS-123.