Testing for KeyError

Question

I'm trying to write a unit test that verifies a KeyError is created when a bad key is passed to a dictionary.

The code that raises the exception:

connections = SettingsManager().get_connections()
try:
    connection = connections[self.conn_name]
except Exception:
    self.log.error("Connection %s does not exist, exiting." % conn_name)
    self.log.error(sys.exc_info()[0])
    raise

I've looked and found KeyError tests using lambda, but I've not had much luck. Here is the test I have thus far, but it errors with the actual KeyError.

def test_bad_connection(self):
    #Testing to see if a non existant connection will fail gracefully.
    records = [1, 2]
    given_result = DataConnector("BadConnection").generate_data(self.table, records)
    expected_result = "BadConnection"

    self.assertRaises(KeyError, given_result[:1])

Martijn Pieters · Accepted Answer · 2014-06-20T15:36:13.087

15

assertRaises() will call the function for you, and assert that that call raises the exception:

records = [1, 2]
connector = DataConnector("BadConnection")

self.assertRaises(KeyError, connector.generate_data, self.table, records)

Alternatively, use assertRaises() as a context manager:

with self.assertRaises(KeyError) as raises:
    DataConnector("BadConnection").generate_data(self.table, records)

which has the added advantage that the context manager then lets you access the raised exception:

self.assertEqual(raises.exception.message, "BadConnection")

edited Jun 20 '14 at 15:36

answered Jun 20 '14 at 15:29

Martijn Pieters

1,048,767
296
4,058
3,343

Thank you - that explains it perfectly! For some reason, I was unable to use the 'raises.exception.msg', rather I used 'raises.exception[0]' and that gave me the expected result. – OpenDataAlex Jun 20 '14 at 15:35
1

@OpenDataAlex: sorry; the `exception.args[0]` value is also exposed as `exception.message`; I just misspelled it. – Martijn Pieters Jun 20 '14 at 15:37

score 14 · Answer 2 · answered Jul 06 '16 at 07:08

14

self.assertRaise() only takes a callable, so while

self.assertRaises(KeyError, given_result[:1])
would give you an actual KeyError while testing,

self.assertRaises(KeyError, lambda: given_result[:1])
should work.

In general:
Does not work: self.assertRaises(KeyError, mydict[mykey]) #KeyError in tests
Does work: self.assertRaises(KeyError, lambda: mydict[mykey])
Does work: self.assertRaises(KeyError, mydict.__getitem__, mykey) #but is requires an actual dict, instead of a function

answered Jul 06 '16 at 07:08

Roman

8,826
10
63
103

2

This answer was easier to understand and reason with than the accepted answer, because it provides a more general solution and not a custom solution. Thanks! – Vadorequest Nov 25 '18 at 12:50
1

`self.assertRaises(KeyError, mydict.get(mykey))` also works. – Jonatan Feb 02 '19 at 10:54

Testing for KeyError

2 Answers2