What is testing?

Why test?

  1. Make sure that your code is giving you the correct output on simple cases
  2. Prevent regressions
  3. Prevent changes when someone else’s code updates
  4. Don’t publish the wrong answer because of a bug -> like this

Writing tests for Nose

Create a file whose name begins with the word test (same directory as code to be tested)

Create new my_math.py file:

def add(num1, num2):
    return num1 + num2

def divide(numerator, denominator):
    return numerator / denominator

Create new test_my_math.py file

Import anything the code needs (including the module to be tested)


import my_math

Write functions beginning with test that assert something:

def test_add_integers():
    assert add(2, 3) == 5

Running tests using nose

When tests fail

def test_div_integers():
    assert divide(1, 4) == 0.25

Run tests, fix integer division


def test_div_integers_noneven():
    assert divide(1, 3) == 0.3333

Run tests, add more 3s


import nose
nose.tools.assert_almost_equals(divide(1,3), 0.3333333)

Testing errors

We often want to know how to make sure that our code is throwing errors when it should be. For example, if something goes wrong with a data file we don’t want the code to execute and potentially produce an incorrect result. However, since all of our testing is based on errors any error in our code will typically cause the test to fail. To address this an other issues nose has lots of useful functions for testing (walk through some examples). The one we are looking for here is nose.tools.assert_raises().

def test_add_strings():
    nose.tools.assert_raises(ValueError, add, 'stuff', 'things')

Alternatively we can use the @raises decorator.

def test_add_lists():
    a = add([1, 2], [3, 4])

I often don’t remember what the different error types are, so I will typically trigger the error, make sure the error type makes sense, and then add it to my test.

Try adding a test for adding a string and a number

Testing multiple values

There are some cases where we will want to test multiple values in a single test rather than writing separate tests for each set of values. We could loop over the values, but this will stop after the first failure. Instead, we can do this:

def test_add_badvalues():
    values = [['teststring', 2], [0, 'teststring'], ['stuff', 'things']]
    for val1, val2 in values:
        yield check_badvalues, val1, val2

def check_badvalues(val1, val2):
    nose.tools.assert_raises(ValueError, add, val1, val2)

We can see from the results that this runs and counts 3 separate tests. If one of these tests fails, e.g.,

def test_add_badvalues():
    values = [['teststring', 2], [2, 1], [0, 'teststring'], ['stuff', 'things']]
    for val1, val2 in values:
        yield check_badvalues, val1, val2

Not only will all of the other tests still run, but nose will tell us the values that caused the test to fail.

###What is yield yield is like return, but the function will return a generator. This means that when you call the function, the code doesn’t actually execute. Instead, the function returns the generator object. The first time the function is run, it will run until it hits yield, then it will return the first value of the loop. Then, each additional call will run the loop in the function another time, and return the next value, until there is no value to return. So, test_add_badvalues() ends up being run once for every trip through the for loop, which is why it doesn’t stop at the first error.

Source: http://stackoverflow.com/a/231855/133513

Integration/regression testing

Calls higher level functionality and checks the results. E.g., tax_resolve tests