Writing pytest tests against tools written with argparse#

I usually build command-line tools using Click (and my click-app cookiecutter template), which includes a really nice set of tools for writing tests.

Today I decided to try building a tool called stream-delay using argparse from the Python standard library, since it didn’t need any other dependencies.

The one challenge I had was how to write the tests. I used pytest as a test-only dependency.

Here’s the pattern I came up with, using the capsys pytest fixture to capture standard output from my tool.

1
from stream_delay import main
2
import pytest
3

4
@pytest.mark.parametrize("option", ("-h", "--help"))
5
def test_help(capsys, option):
6
    try:
7
        main([option])
8
    except SystemExit:
9
        pass
10
    output = capsys.readouterr().out
11
    assert "Stream one or more files with a delay" in output

My main() function starts like this:

1
import argparse, sys
2

3
parser = argparse.ArgumentParser(
4
    description="Stream one or more files with a delay between each line"
5
)
6
parser.add_argument("files", type=argparse.FileType("r"), nargs="*", default=["-"])
7
parser.add_argument("-d", "--delay-in-ms", type=int, default=100)
8

9

10
def main(args=None):
11
    parsed_args = parser.parse_args(args)
12
    delay_in_s = float(parsed_args.delay_in_ms) / 1000
13
    # ...

As you can see, main() takes an optional list of arguments. The default for that is None which will cause argparse to read sys.argv - but I can inject arguments to the function from my tests if I need to.

I’m catching the SystemExit exception because this will be raised by default if you use -h or --help - but I still want to finish my test execution so I can inspect the captured output.

Complete code: