Handle numpy array scalars in pdfRepr

lib/matplotlib/backends/backend_pdf.py

@@ -139,7 +139,7 @@ def pdfRepr(obj):
     # Floats. PDF does not have exponential notation (1.0e-10) so we
     # need to use %f with some precision.  Perhaps the precision
     # should adapt to the magnitude of the number?
-    elif isinstance(obj, float):
+    elif isinstance(obj, (float, np.floating)):
         if not np.isfinite(obj):
             raise ValueError("Can only output finite numbers in PDF")
         r = ("%.10f" % obj).encode('ascii')
@@ -151,7 +151,7 @@ def pdfRepr(obj):
         return [b'false', b'true'][obj]
 
     # Integers are written as such.
-    elif isinstance(obj, (int, long)):
+    elif isinstance(obj, (int, long, np.integer)):
         return ("%d" % obj).encode('ascii')
 
     # Unicode strings are encoded in UTF-16BE with byte-order mark.