Simplify code for AtomicStructureAdapter indexing.

Remove unnecessary branching in the get_slice function.

Author: pavoljuhas

srrealmodule/wrap_AtomicStructureAdapter.cpp

@@ -374,25 +374,22 @@ class atomadapter_indexing : public vector_indexing_suite<
         static object
         get_slice(Container& container, index_type from, index_type to)
         {
+            // make sure slice is of a correct type and has a copy
+            // of any additional structure data.
             StructureAdapterPtr rv = container.clone();
             AtomicStructureAdapterPtr rva;
             rva = boost::static_pointer_cast<AtomicStructureAdapter>(rv);
-            // short-circuit for long slices
-            if ((to - from) > rva->countSites() / 2)
+            // handle index ranges for a valid and empty slice
+            if (from <= to)
             {
                 rva->erase(rva->begin() + to, rva->end());
                 rva->erase(rva->begin(), rva->begin() + from);
-                return object(rv);
-            }
-            // otherwise save memory for short slices
-            rva->clear();
-            rv = rva->clone();
-            rva = boost::static_pointer_cast<AtomicStructureAdapter>(rv);
-            if (from <= to)
-            {
-                rva->assign(container.begin() + from, container.begin() + to);
             }
-            return object(rv);
+            else  rva->clear();
+            // save memory by making a new copy for short slices
+            const bool longslice = ((to - from) > rva->countSites() / 2);
+            object pyrv(longslice ? rv : rv->clone());
+            return pyrv;
         }