problem with anti-symmetric functions and tensors --> solved?

22 Jul 2002

      Hello everybody,

I found that the program

	#include<iostream>
	#include<ginac/ginac.h>
	using namespace std;
	using namespace GiNaC;

	DECLARE_FUNCTION_2P(f)
	ex f_eval(const ex&x,const ex&y)
	{       return f(x,y).hold();
	}
	REGISTER_FUNCTION(f,eval_func(f_eval).set_symmetry(sy_anti(0,1)));

	int main()
	{       cout << f(0,1) << endl;
        	cout << f(1,0) << endl;
        	cout << f(0,0) << endl;
	}

which defines an anti-symmetric function and then tries it with some
values, gives as output

	f(0,1)
	-f(0,1)
	f(0,0)

The last result is not as expected, since an anti-symmetric function is
equal to zero when both arguments are the same. I found that this happens
because in the function *canonicalize* in the file symmetry.cpp the value
of *sign* (which, as expected, turns out to be zero if the two arguments
are the same) is ignored if the variable *something_changed* is *false*
(which is the case in the third example because the two arguments are the
same and thus already sorted in ascending order). The problem is solved by
adding after the line (which appears in the function *canonicalize*)

	sign *= permutation_sign(first, last, sy_is_less(v),
					sy_swap(v,something_changed));

the new line

	if(!sign)return 0;

After this my testprogram gives the result

	f(0,1)
	-f(0,1)
	0

which is what I would have expected in the first place.

Note that this also fixes the same problem for anti-symmetric tensors.

Greetings,
Chris Dams