bugfix: fix nan check to account for values castable to float (VIO-1616)

synse_server/cmd/read.py

@@ -36,13 +36,11 @@ def reading_to_dict(reading: api.V3Reading) -> Dict[str, Any]:
         # Ensure the value is not NaN, as NaN is not a part of the
         # JSON spec and could cause clients to error.
         try:
-            float(value)
-        except ValueError:
+            if math.isnan(float(value)):
+                value = None
+        except (ValueError, TypeError):
             # Not a real number, nothing to do.
             pass
-        else:
-            if math.isnan(value):
-                value = None
 
     if not reading.unit or (reading.unit.symbol == '' and reading.unit.name == ''):
         unit = None

tests/unit/cmd/test_read.py

@@ -704,3 +704,45 @@ def test_reading_to_dict_4_nan(state_reading):
         'unit': None,
         'context': {},
     }
+
+
+@pytest.mark.parametrize(
+    'reading_value', [
+        "000000",
+        "0",
+        "1",
+        "-1",
+        "ff0033",
+        "fab",
+        "FFAADD",
+        "319a81",
+    ]
+)
+def test_reading_to_dict_5_float_not_nan(state_reading, reading_value: str) -> None:
+    """Convert a reading to a dictionary. Here, the string value is castable to
+    a float, and thus will go through the NaN check, but should fail the check since
+    the float values do not resolve to NaN.
+
+    Regression for: https://vaporio.atlassian.net/browse/VIO-1616
+    """
+
+    msg = api.V3Reading(
+        id='ddd',
+        timestamp='2019-04-22T13:30:00Z',
+        type='color',
+        deviceType='led',
+        deviceInfo='Example LED Device',
+        string_value=reading_value,
+    )
+
+    actual = reading_to_dict(msg)
+    assert actual == {
+        'device': 'ddd',
+        'timestamp': '2019-04-22T13:30:00Z',
+        'type': 'color',
+        'device_type': 'led',
+        'device_info': 'Example LED Device',
+        'value': reading_value,
+        'unit': None,
+        'context': {},
+    }