[GiNaC-devel] [PATCH] Function and ncmul

Vladimir Kisil kisilv at maths.leeds.ac.uk
Sat Apr 26 20:54:45 CEST 2008


		Dear All,

		Consider the following program:

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

int main(){

	realsymbol a("a");
	
	varidx mu(symbol("mu", "\\mu"), 1);
	ex e = clifford_unit(mu, diag_matrix(lst(-1))),
		e0=e.subs(mu==0);

	cout << e0*e0*exp(a*e0) << endl;
	// -> exp(e0*a)
	cout << exp(a*e0)*e0*e0 << endl;
	// -> exp(e0*a)*e0*e0 
}

  The expression in the second output is not simplified for the
  following reason. When the method ncmul::eval() collects non-commuting
  terms, the function exp(e0*a) gets the type of its first noncommuting
  argument, i.e. e0 (this is defined by function::return_type()). So all
  three factors are of the same type.

  However, later on, when ncmul::eval() calls the method eval_ncmul()
  for the first term in the product, e.i.  exp(e0*a), it falls back to
  basic::eval_ncmul() not clifford::eval_ncmul() as one may expect.

  I am providing below a small patch which makes the behaviour of class
  function more consistent in this respect.

  Best wishes,
  Vladimir
-- 
Vladimir V. Kisil     email: kisilv at maths.leeds.ac.uk
--                      www: http://maths.leeds.ac.uk/~kisilv/

Parent: 6850c16435de33396ccd505ff26b4d4574fa74cf (Fixed bug in mLi summation causing premature drop-out and made Nielsen polylog)
Branch: master
Follows: release_1-4-0
Precedes: 

    Method function::eval_ncmul() is defined to match the behaviour of function::return_type()

------------------------------ ginac/function.pl ------------------------------
index e91278c..3b22143 100644
@@ -507,6 +507,7 @@ public:
 	ex expand(unsigned options=0) const;
 	ex eval(int level=0) const;
 	ex evalf(int level=0) const;
+	ex eval_ncmul(const exvector & v) const;
 	unsigned calchash() const;
 	ex series(const relational & r, int order, unsigned options = 0) const;
 	ex thiscontainer(const exvector & v) const;
@@ -1053,6 +1054,16 @@ ${evalf_switch_statement}
 	throw(std::logic_error("function::evalf(): invalid nparams"));
 }
 
+/*
+* This method is defined to be in line with behaviour of function::return_type()
+*/
+ex function::eval_ncmul(const exvector & v) const
+{
+	// If this function is called then the list of arguments is non-empty
+	// and the first argument is non-commutative, see  function::return_type()
+	return seq.begin()->eval_ncmul(v);
+}
+
 unsigned function::calchash() const
 {
 	unsigned v = golden_ratio_hash(golden_ratio_hash((p_int)tinfo()) ^ serial);



More information about the GiNaC-devel mailing list