series expansion once again...

[Sheplyakov Alexei]

Hello!


> Series expansion now works predictably. All terms with the exponent
> of the expansion variable smaller than the given order are 
> calculated exactly. If the series is not terminating, the Order 
> function is (at least) of the given order.

( GiNaC 1.1.7 changelog )

That sounds nice, but let's consider a simple program.

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

int main(int argc, char** argv) {

	try { 
		symbol eps("epsilon");

		ex part1 = tgamma(2-eps)*pow(eps, -1)*pow(tgamma(3-eps), -1);
		ex part2 = tgamma(2-eps)*pow(eps, 2)*pow(tgamma(3-eps), -1);
		ex test = part1 + part2;

		cout << test << " = " <<  test.series(eps, 1) << endl;
		cout << part1 << " = " <<  part1.series(eps, 1) << endl;
		cout << part2 << " = " <<  part2.series(eps, 1) << endl;

		return 0;
	}
	catch ( exception & ee) {
		cerr << ee.what() << endl;
		return 1;
	}
}

My naive expectations are:
	1) first and second lines should be the same
	2) third line should be Order(epsilon)

But the output is:

tgamma(2-epsilon)*tgamma(3-epsilon)^(-1)*epsilon^2+tgamma(2-epsilon)*tgamma(3-epsilon)^(-1)*epsilon^(-1) = Order(epsilon^(-1))
tgamma(2-epsilon)*tgamma(3-epsilon)^(-1)*epsilon^(-1) = 1/2*epsilon^(-1)+1/4+Order(epsilon)
tgamma(2-epsilon)*tgamma(3-epsilon)^(-1)*epsilon^2 = Order(epsilon^(-1))

	
Is this a bug or a feature?

Thanks,
	Alexei.


P.S.

$ ginac-config --version
1.1.7

$ g++ --version | head -n 1
g++ (GCC) 3.3.3 (Debian 20040401)