Unexpected truncation to integer when using evalf

[knuesel]

Calling Sum(1.9**k/factorial(k), (k, 0, oo)).evalf() returns the value for e^1 instead of e^1.9. Calling Sum(x**k/factorial(k), (k, 0, oo)).doit().subs(x,1.9) however gives the expected result.

[jksuom]

SymPy evaluates sums of hypergeometric series by using integer arithmetic. All numbers are truncated to integers by MPZ (hypsum). That will also affect factors like 1.9:

>>> from mpmath.libmp.backend import MPZ
>>> MPZ(1.9)
mpz(1)

To make use of non-integer factors the truncation should occur after the operation, not before:

        term = MPZ(term*func1(k - 1))
        term = MPZ(term/func2(k - 1))

[smichr]

Possible solution:

diff --git a/sympy/core/evalf.py b/sympy/core/evalf.py
index 0ce7def..5815c61 100644
--- a/sympy/core/evalf.py
+++ b/sympy/core/evalf.py
@@ -1127,11 +1127,10 @@ def hypsum(expr, n, start, prec):
         s = term
         k = 1
         while abs(term) > 5:
-            term *= MPZ(func1(k - 1))
-            term //= MPZ(func2(k - 1))
+            term *= func1(k - 1)/func2(k - 1)
             s += term
             k += 1
-        return from_man_exp(s, -prec)
+        return from_man_exp(MPZ(s), -prec)
     else:
         alt = g < 0
         if abs(g) < 1:
diff --git a/sympy/core/tests/test_evalf.py b/sympy/core/tests/test_evalf.py
index 27daed9..46de791 100644
--- a/sympy/core/tests/test_evalf.py
+++ b/sympy/core/tests/test_evalf.py
@@ -525,3 +525,9 @@ def test_issue_13098():
     assert ceiling(log(S('9.'+'9'*20), 10)) == 1
     assert floor(log(20 - S('9.'+'9'*20), 10)) == 1
     assert ceiling(log(20 - S('9.'+'9'*20), 10)) == 2
+
+
+def test_issue_13365():
+    s = Sum(1.9**x/factorial(x), (x, 0, oo))
+    a = S.Exp1**1.9
+    assert (s - a).evalf() < 1e-12