Skip to content

Latest commit

 

History

History
53 lines (39 loc) · 2.27 KB

pytest-subprocess.md

File metadata and controls

53 lines (39 loc) · 2.27 KB

Mocking subprocess with pytest-subprocess

For apple-notes-to-sqlite I needed to write some tests that simulated executing the osascript command using the Python subprocess module.

I wanted my tests to run on Linux CI machines, where that command would not exist.

After failing to use unittest.mock.patch to solve this, I went looking for alternatives. I found pytest-subprocess.

Here's the relevant section of the test I wrote:

from apple_notes_to_sqlite.cli import cli, COUNT_SCRIPT

FAKE_OUTPUT = b"""
The stuff I would expect to be returned by log lines in
my osascript script.
"""

def test_apple_notes_to_sqlite_dump(fp):
    fp.register_subprocess(["osascript", "-e", COUNT_SCRIPT], stdout=b"2")
    fp.register_subprocess(["osascript", "-e", fp.any()], stdout=FAKE_OUTPUT)
    runner = CliRunner()
    with runner.isolated_filesystem():
        result = runner.invoke(cli, ["--dump"])
        # ...

fp is the fixture provided by the package (you need to pip install pytest-subprocess for this to work).

COUNT_SCRIPT here is the first of my osascript constants. It looks like this (in cli.py):

COUNT_SCRIPT = """
tell application "Notes"
    set noteCount to count of notes
end tell
log noteCount
"""

That first fixture line says that any time my program calls osascript -e that-count-script the return value sent to standard output should be a binary string 2.

fp.register_subprocess(["osascript", "-e", COUNT_SCRIPT], stdout=b"2")

The second call to subprocess made by my script is more complicated - it involves a script that is dynamically generated.

fp.register_subprocess(["osascript", "-e", fp.any()], stdout=FAKE_OUTPUT)

I eventually figured that using fp.any() was easier than specifying the exact script. This is a wildcard value which matches any string. It returns the full FAKE_OUTPUT variable as the simulated standard out.

What's useful about pytest-subprocess is that it works for both subprocess.check_output() and more complex subprocess.Popen() calls - both of which I was using in this script.