Dear GiNaCers, Here's a patch. Find out below why you want this patch ;-). The integral class has a .derivative() method, so in principle integrals can be series expanded. However, if somewhere deep down in the integrant a sum with a lot of terms occurs, the same kind of problems that have troubled the series expansion of powers, happen again. Basically we get expression explosion. To solve this I wrote an integral::series() method. While doing this I discovered several bugs/undesirable features, namely (1) When requesting terms of a pseries via pseries::op(), the order term looks like O(1)^n instead of O(x^n). This looks strange and different from the output that one gets from printing the entire series. (2) Since pseries does not have a let_op() method for understandable reasons, eval_integ() does not work for a power series. I have added a new method pseries::eval_integ() to solve this. (3) When trying to find coefficients and exponents of power series, using pseries::op() is not convenient and not efficient. This returns a multiplication of the coefficient with a power that the user then may attempt to extract the coefficient and the exponent from... To make this easier, I added pseries::coeffop() and pseries::exponop() methods. (4) pseries::mul_series() does not know that zero times a power series is zero. (5) If pseries::power_const() discovers that the user is expanding to such a low degree that only the order term needs to be returned, GiNaC sometimes tries to save memory by allocating a negative amount of bytes in order to store a negative number of terms in it. Best, Chris