Tests in software are like rumble strips on the highway - they provide feedback to the driver as early as possible when they depart their normal lane and gives them time to correct the situation comfortably before they depart the road and have an accident.
The Four-Phase Test Pattern is described in Gerard Meszaros’ book “xUnit Testing Patterns” and is a nice clear way to structure your tests. It allows the reader to easily see what is being testing and greatly increases the documentation value of the test.
PyTest will automatically look for files named test_*.py or *_test.py and
from them run functions whose names begin with test and/or methods that
begin with test contained in classes that begin with Test and have no
__init__. There are some more custom configurations you can provide and the
full test discovery process is detailed
here
in the documentation.
pytest and let the suite run.All tests need to use a form of the assert statement which will pass if the
conditional is True and fail if it is False. This can be as simple as
assert a==b and we should be striving for simplicity!
Comparing strings is done with a simple equality check.
def test_title_case():
# Setup
input = 'this is a test string'
desired = 'This Is A Test String'
# Exercise
actual = input.title()
# Verify
assert actual == desired
# Cleanup - none necessary
def test_build_asos_request_url_single_digit_datetimes():
"""
Test building URL with single digit month and day.
"""
# Setup
start = datetime.datetime(2018, 1, 5, 1)
end = datetime.datetime(2018, 1, 9, 1)
station = 'FSD'
# Exercise
result_url = meteogram.build_asos_request_url(station, start, end)
# Verify
truth_url = ('https://mesonet.agron.iastate.edu/request/asos/1min_dl.php?'
'station%5B%5D=FSD&tz=UTC&year1=2018&month1=01&day1=05&hour1=01'
'&minute1=00&year2=2018&month2=01&day2=09&hour2=01&minute2=00&'
'vars%5B%5D=tmpf&vars%5B%5D=dwpf&vars%5B%5D=sknt&vars%5B%5D=drct&'
'sample=1min&what=view&delim=comma&gis=yes')
assert result_url == truth_url
# Cleanup - none necessary
def test_build_asos_request_url_double_digit_datetimes():
"""
Test building URL with double digit month and day.
"""
# Setup
start = datetime.datetime(2018, 10, 11, 12)
end = datetime.datetime(2018, 10, 16, 15)
station = 'MLI'
# Exercise
result_url = meteogram.build_asos_request_url(station, start, end)
# Verify
truth_url = ('https://mesonet.agron.iastate.edu/request/asos/1min_dl.php?'
'station%5B%5D=MLI&tz=UTC&year1=2018&month1=10&day1=11&hour1=12'
'&minute1=00&year2=2018&month2=10&day2=16&hour2=15&minute2=00&'
'vars%5B%5D=tmpf&vars%5B%5D=dwpf&vars%5B%5D=sknt&vars%5B%5D=drct&'
'sample=1min&what=view&delim=comma&gis=yes')
assert result_url == truth_url
# Cleanup - none necessary
Numerical comparisons are a common place that new testers start getting into trouble. Testing integers is straightforward:
def test_does_three_equal_three():
assert 3 == 3
Trouble begins immediately when we leave the integer world though! Floating point comparisons are fraught with peril. If you really want to get into why there is a lot of light bedtime reading about the IEEE 754 specification and how machine precision results in small rounding errors. For our purposes we will simply say that this is NOT a safe thing to do:
def test_floating_subtraction():
# Setup
desired = 0.293
# Exercise
actual = 1 - 0.707
# Verify
assert actual == desired
# Cleanup
The best way to deal with this is to add a tolerance check, something like:
def test_floating_subtraction():
# Setup
desired = 0.293
# Exercise
actual = 1 - 0.707
# Verify
assert(abs(actual-desired) < 0.00001)
# Cleanup
That would result in too much logic and duplicated code in our tests and seems
like a solved problem right? Numpy has a testing helper for exactly this:
numpy.testing.assert_almost_equal.
Now our test can simplify to:
from numpy.testing import assert_almost_equal
def test_floating_subtraction():
# Setup
desired = 1.5
# Exercise
actual = 3 - 1.5
# Verify
assert_almost_equal(actual, desired)
# Cleanup - none necessary
def wind_components(speed, direction):
"""
Calculate the U, V wind vector components from the speed and direction.
Parameters
----------
speed : array_like
The wind speed (magnitude)
wdir : array_like
The wind direction, specified as the direction from which the wind is
blowing (0-360 degrees), with 360 degrees being North.
Returns
-------
u, v : tuple of array_like
The wind components in the X (East-West) and Y (North-South)
directions, respectively.
"""
direction = np.radians(direction)
u = -speed * np.sin(direction)
v = -speed * np.cos(direction)
return u, v
def test_wind_components_north():
# Setup
speed = 10
direction = 0
# Exercise
u, v = meteogram.wind_components(speed, direction)
# Verify
true_u = 0
true_v = -10
assert_almost_equal(u, true_u)
assert_almost_equal(v, true_v)
def test_wind_components_northeast():
# Setup
speed = 10
direction = 45
# Exercise
u, v = meteogram.wind_components(speed, direction)
# Verify
true_u = -7.0710
true_v = -7.0710
assert_almost_equal(u, true_u, 3)
assert_almost_equal(v, true_v, 3)
def test_wind_components_threesixty():
# Setup
speed = 10
direction = 360
# Exercise
u, v = meteogram.wind_components(speed, direction)
# Verify
true_u = 0
true_v = -10
assert_almost_equal(u, true_u)
assert_almost_equal(v, true_v)
def test_wind_components_zero_wind():
# Setup
speed = 0
direction = 45
# Exercise
u, v = meteogram.wind_components(speed, direction)
# Verify
true_u = 0
true_v = 0
assert_almost_equal(u, true_u)
assert_almost_equal(v, true_v)
# Cleanup
Array comparisons are much like integer and floating point comparisons.
In this case we’re going to use np.testing.assert_array_almost_equal to
cleanup our tests we just wrote.
import numpy as np
from numpy.testing import assert_almost_equal, assert_array_almost_equal
def test_wind_components():
# Setup
speed = np.array([10, 10, 10, 0])
direction = np.array([0, 45, 360, 45])
# Exercise
u, v = meteogram.wind_components(speed, direction)
# Verify
true_u = np.array([0, -7.0710, 0, 0])
true_v = np.array([-10, -7.0710, -10, 0])
assert_array_almost_equal(u, true_u, 3)
assert_array_almost_equal(v, true_v, 3)
# Cleanup