Patch by Stefan Weinzierl that fixes a bug in series expansion.

diff --git a/check/exam_pseries.cpp b/check/exam_pseries.cpp
index b0ccac3eba904839a21499b0fcbd39812fa7510e..1c8acd939712ddacb3e8dee733907dc76e027783 100644 (file)
--- a/check/exam_pseries.cpp
+++ b/check/exam_pseries.cpp
@@ -348,6 +348,18 @@ static unsigned exam_series12()
        return result;
 }
 
+// Test of the patch of Stefan Weinzierl that prevents an infinite loop if
+// a factor in a product is a complicated way of writing zero.
+static unsigned exam_series13()
+{
+       unsigned result = 0;
+
+       ex e = pow(2,x)*( 1/x*(-(1+x)/(1-x)) + (1+x)/x/(1-x));
+       ex d = Order(x);
+       result += check_series(e,0,d,1);
+
+       return result;
+}
 
 unsigned exam_pseries()
 {
@@ -368,6 +380,7 @@ unsigned exam_pseries()
        result += exam_series10();  cout << '.' << flush;
        result += exam_series11();  cout << '.' << flush;
        result += exam_series12();  cout << '.' << flush;
+       result += exam_series13();  cout << '.' << flush;
        
        if (!result) {
                cout << " passed " << endl;

diff --git a/ginac/pseries.cpp b/ginac/pseries.cpp
index ec8e8c652560bb60eceadc99dbf44b3e1aedd4f5..0da6ae8d4854908fc1e59e116e036ae56e433e66 100644 (file)
--- a/ginac/pseries.cpp
+++ b/ginac/pseries.cpp
@@ -892,10 +892,13 @@ ex mul::series(const relational & r, int order, unsigned options) const
                
        // holds ldegrees of the series of individual factors
        std::vector<int> ldegrees;
+       std::vector<bool> ldegree_redo;
 
        // find minimal degrees
        const epvector::const_iterator itbeg = seq.begin();
        const epvector::const_iterator itend = seq.end();
+       // first round: obtain a bound up to which minimal degrees have to be
+       // considered
        for (epvector::const_iterator it=itbeg; it!=itend; ++it) {
 
                ex expon = it->coeff;
@@ -909,19 +912,61 @@ ex mul::series(const relational & r, int order, unsigned options) const
                }
 
                int real_ldegree = 0;
+               bool flag_redo = false;
                try {
                        real_ldegree = buf.expand().ldegree(sym-r.rhs());
                } catch (std::runtime_error) {}
 
                if (real_ldegree == 0) {
+                       if ( factor < 0 ) {
+                               // This case must terminate, otherwise we would have division by
+                               // zero.
+                               int orderloop = 0;
+                               do {
+                                       orderloop++;
+                                       real_ldegree = buf.series(r, orderloop, options).ldegree(sym);
+                               } while (real_ldegree == orderloop);
+                       } else {
+                               // Here it is possible that buf does not have a ldegree, therefore
+                               // check only if ldegree is negative, otherwise reconsider the case
+                               // in the second round.
+                               real_ldegree = buf.series(r, 0, options).ldegree(sym);
+                               if (real_ldegree == 0)
+                                       flag_redo = true;
+                       }
+               }
+
+               ldegrees.push_back(factor * real_ldegree);
+               ldegree_redo.push_back(flag_redo);
+       }
+
+       int degbound = order-std::accumulate(ldegrees.begin(), ldegrees.end(), 0);
+       // Second round: determine the remaining positive ldegrees by the series
+       // method.
+       // here we can ignore ldegrees larger than degbound
+       size_t j = 0;
+       for (epvector::const_iterator it=itbeg; it!=itend; ++it) {
+               if ( ldegree_redo[j] ) {
+                       ex expon = it->coeff;
+                       int factor = 1;
+                       ex buf;
+                       if (expon.info(info_flags::integer)) {
+                               buf = it->rest;
+                               factor = ex_to<numeric>(expon).to_int();
+                       } else {
+                               buf = recombine_pair_to_ex(*it);
+                       }
+                       int real_ldegree = 0;
                        int orderloop = 0;
                        do {
                                orderloop++;
                                real_ldegree = buf.series(r, orderloop, options).ldegree(sym);
-                       } while (real_ldegree == orderloop);
+                       } while ((real_ldegree == orderloop)
+                                       && ( factor*real_ldegree < degbound));
+                       ldegrees[j] = factor * real_ldegree;
+                       degbound -= factor * real_ldegree;
                }
-
-               ldegrees.push_back(factor * real_ldegree);
+               j++;
        }
 
        int degsum = std::accumulate(ldegrees.begin(), ldegrees.end(), 0);