diff --git a/_imaging.c b/_imaging.c
index 7d0bcba04..7c55f633b 100644
--- a/_imaging.c
+++ b/_imaging.c
@@ -393,35 +393,28 @@ getlist(PyObject* arg, int* length, const char* wrong_length, int type)
         return NULL;
     }
     
-    switch (type) {
-    case TYPE_UINT8:
-        for (i = 0; i < n; i++) {
-            op = PySequence_Fast_GET_ITEM(seq, i);
+    for (i = 0; i < n; i++) {
+        op = PySequence_Fast_GET_ITEM(seq, i);
+        // DRY, branch prediction is going to work _really_ well 
+        // on this switch. And 3 fewer loops to copy/paste. 
+        switch (type) {
+        case TYPE_UINT8:
             itemp = PyInt_AsLong(op);
             ((UINT8*)list)[i] = CLIP(itemp);
-        }
-        break;
-    case TYPE_INT32:
-        for (i = 0; i < n; i++) {
-            op = PySequence_Fast_GET_ITEM(seq, i);
+            break;
+        case TYPE_INT32:
             itemp = PyInt_AsLong(op);
             ((INT32*)list)[i] = itemp;
-        }
-        break;
-    case TYPE_FLOAT32:
-         for (i = 0; i < n; i++) {
-             op = PySequence_Fast_GET_ITEM(seq, i);
-             dtemp = PyFloat_AsDouble(op);
-             ((FLOAT32*)list)[i] = (FLOAT32) dtemp;
-         }
-        break;
-    case TYPE_DOUBLE:
-        for (i = 0; i < n; i++) {
-            op = PySequence_Fast_GET_ITEM(seq, i);
+            break;
+        case TYPE_FLOAT32:
+            dtemp = PyFloat_AsDouble(op);
+            ((FLOAT32*)list)[i] = (FLOAT32) dtemp;
+            break;
+        case TYPE_DOUBLE:
             dtemp = PyFloat_AsDouble(op);
             ((double*)list)[i] = (double) dtemp;
-         }
-        break;
+            break;
+        }
     }
 
     if (length)