17 Aug
2004
17 Aug
'04
2:05 p.m.
Trying to be more exact, if you require a main variable as parameter to quo, then you already impose an ordering on the variables, and in this case quo is always well-defined, eg quo(y,y*x,y) would be 1/x. So the restriction to univariate input would be unnecessary if the implementation would follow this logic. Also, if quo would act as described an implementation of floor and frac[tional part] for rational functions would be a natural. However, if the ginac designers don't want this, then quo should throw an exception on *any* multivariate input, IMHO. ralf