Skip to main content

A plugin to fake subprocess for pytest

Project description


PyPI version Python versions Documentation Status

Pytest plugin to fake subprocess.


The plugin adds the fake_process fixture (and fp as an alias). It can be used it to register subprocess results so you won’t need to rely on the real processes. The plugin hooks on the subprocess.Popen(), which is the base for other subprocess functions. That makes the,, subprocess.check_call() and subprocess.check_output() methods also functional.


You can install pytest-subprocess via pip from PyPI:

$ pip install pytest-subprocess

Basic usage

The most important method is fp.register() (or register_subprocess if you prefer to be more verbose), which allows defining the fake processes behavior.

def test_echo_null_byte(fp):
    fp.register(["echo", "-ne", "\x00"], stdout=bytes.fromhex("00"))

    process = subprocess.Popen(
        ["echo", "-ne", "\x00"],
    out, _ = process.communicate()

    assert process.returncode == 0
    assert out == b"\x00"

Optionally, the stdout and stderr parameters can be a list (or tuple) of lines to be joined together with a trailing os.linesep on each line.

def test_git(fp):
    fp.register(["git", "branch"], stdout=["* fake_branch", "  master"])

    process = subprocess.Popen(
        ["git", "branch"],
    out, _ = process.communicate()

    assert process.returncode == 0
    assert out == "* fake_branch\n  master\n"

Passing input

By default, if you use input argument to the Popen.communicate() method, it won’t crash, but also won’t do anything useful. By passing a function as stdin_callable argument for the fp.register() method you can specify the behavior based on the input. The function shall accept one argument, which will be the input data. If the function will return a dictionary with stdout or stderr keys, its value will be appended to according stream.

def test_pass_input(fp):
    def stdin_function(input):
        return {
            "stdout": "This input was added: {data}".format(

        stdout=[b"Just stdout"],

    process = subprocess.Popen(
    out, _ = process.communicate(input=b"sample input\n")

    assert out.splitlines() == [
        b"Just stdout",
        b"This input was added: sample input",

Unregistered commands

By default, when the fp fixture is being used, any attempt to run subprocess that has not been registered will raise the ProcessNotRegisteredError exception. To allow it, use fp.allow_unregistered(True), which will execute all unregistered processes with real subprocess, or use fp.pass_command("command") to allow just a single command.

def test_real_process(fp):
    with pytest.raises(fp.exceptions.ProcessNotRegisteredError):
        # this will fail, as "ls" command is not registered"ls")

    # now it should be fine
    assert"ls") == 0

    # allow all commands to be called by real subprocess
    assert["ls", "-l"]) == 0

Differing results

Each register() or pass_command() method call will register only one command execution. You can call those methods multiple times, to change the faked output on each subprocess run. When you call subprocess more will be raised. To prevent that, call fp.keep_last_process(True), which will keep the last registered process forever.

def test_different_output(fp):
    # register process with output changing each execution
    fp.register("test", stdout="first execution")
    # the second execution will return non-zero exit code
    fp.register("test", stdout="second execution", returncode=1)

    assert subprocess.check_output("test") == b"first execution"
    second_process ="test", stdout=subprocess.PIPE)
    assert second_process.stdout == b"second execution"
    assert second_process.returncode == 1

    # 3rd time shall raise an exception
    with pytest.raises(fp.exceptions.ProcessNotRegisteredError):

    # now, register two processes once again,
    # but the last one will be kept forever
    fp.register("test", stdout="first execution")
    fp.register("test", stdout="second execution")

    # now the processes can be called forever
    assert subprocess.check_output("test") == b"first execution"
    assert subprocess.check_output("test") == b"second execution"
    assert subprocess.check_output("test") == b"second execution"
    assert subprocess.check_output("test") == b"second execution"

Using callbacks

You can pass a function as callback argument to the register() method which will be executed instead of the real subprocess. The callback function can raise exceptions which will be interpreted in tests as an exception raised by the subprocess. The fixture will pass FakePopen class instance into the callback function, that can be used to change the return code or modify output streams.

def callback_function(process):
    process.returncode = 1
    raise PermissionError("exception raised by subprocess")

def test_raise_exception(fp):
    fp.register(["test"], callback=callback_function)

    with pytest.raises(
        PermissionError, match="exception raised by subprocess"
        process = subprocess.Popen(["test"])

    assert process.returncode == 1

It is possible to pass additional keyword arguments into callback by using the callback_kwargs argument:

def callback_function_with_kwargs(process, return_code):
    process.returncode = return_code

def test_callback_with_arguments(fp):
    return_code = 127

        callback_kwargs={"return_code": return_code},

    process = subprocess.Popen(["test"])

    assert process.returncode == return_code

As a context manager

The fp fixture provides context() method that allows us to use it as a context manager. It can be used to limit the scope when a certain command is allowed, e.g. to make sure that the code doesn’t want to execute it somewhere else.

def test_context_manager(fp):
    with pytest.raises(fp.exceptions.ProcessNotRegisteredError):
        # command not registered, so will raise an exception

    with fp.context() as nested_process:
        nested_process.register("test", occurrences=3)
        # now, we can call the command 3 times without error
        assert subprocess.check_call("test") == 0
        assert subprocess.check_call("test") == 0

    # the command was called 2 times, so one occurrence left, but since the
    # context manager has been left, it is not registered anymore
    with pytest.raises(fp.exceptions.ProcessNotRegisteredError):

Non-exact command matching

If you need to catch a command with some non-predictable elements, like a path to a randomly-generated file name, you can use fake_subprocess.any() for that purpose. The number of arguments that should be matched can be controlled by min and max arguments. To use fake_subprocess.any() you need to define the command as a tuple or list. The matching will work even if the subprocess command will be called with a string argument.

def test_non_exact_matching(fp):
    # define a command that will take any number of arguments
    fp.register(["ls", fp.any()])
    assert subprocess.check_call("ls -lah") == 0

    # `fake_subprocess.any()` is OK even with no arguments
    fp.register(["ls", fp.any()])
    assert subprocess.check_call("ls") == 0

    # but it can force a minimum amount of arguments
    fp.register(["cp", fp.any(min=2)])

    with pytest.raises(fp.exceptions.ProcessNotRegisteredError):
        # only one argument is used, so registered command won't match
        subprocess.check_call("cp /source/dir")
    # but two arguments will be fine
    assert subprocess.check_call("cp /source/dir /tmp/random-dir") == 0

    # the `max` argument can be used to limit maximum amount of arguments
    fp.register(["cd", fp.any(max=1)])

    with pytest.raises(fp.exceptions.ProcessNotRegisteredError):
        # cd with two arguments won't match with max=1
        subprocess.check_call("cd ~/ /tmp")
    # but any single argument is fine
    assert subprocess.check_call("cd ~/") == 0

    # `min` and `max` can be used together
    fp.register(["my_app", fp.any(min=1, max=2)])
    assert subprocess.check_call(["my_app", "--help"]) == 0

You can also specify just the command name, and have it match any command with the same name, regardless of the location. This is accomplished with fake_subprocess.program("name").

def test_any_matching_program(fp):
    # define a command that can come from anywhere
    assert subprocess.check_call("/bin/ls") == 0

Check if process was called

You may want to simply check if a certain command was called, you can do this by accessing fp.calls, where all commands are stored as-called. You can also use a utility function fp.call_count() to see how many a command has been called. The latter supports fp.any().

def test_check_if_called(fp):
    # any command can be called

    subprocess.check_call(["cp", "/tmp/source", "/source"])
    subprocess.check_call(["cp", "/source", "/destination"])
    subprocess.check_call(["cp", "/source", "/other/destination"])

    # you can check if command is in ``fp.calls``
    assert ["cp", "/tmp/source", "/source"] in fp.calls
    assert ["cp", "/source", "/destination"] in fp.calls
    assert ["cp", "/source", "/other/destination"] in fp.calls

    # or check how many it was called, possibly with wildcard arguments
    assert fp.call_count(["cp", "/source", "/destination"]) == 1

    # with ``call_count()`` you don't need to use the same type as
    # the subprocess was called
    assert fp.call_count("cp /tmp/source /source") == 1

    # can be used with ``fp.any()`` to match more calls
    assert fp.call_count(["cp", fp.any()]) == 3

Handling signals

You can use standard kill(), terminate() or send_signal() methods in Popen instances. There is an additional received_signals() method to get a tuple of all signals received by the process. It is also possible to set up an optional callback function for signals.

import signal

def test_signal_callback(fp):
    """Test that signal callbacks work."""

    def callback(process, sig):
        if sig == signal.SIGTERM:
            process.returncode = -1

    # the `register()` method returns a ProgressRecorder object, where
    # all future matching `Popen()` instances will be appended
    process_recorder = fp.register("test", signal_callback=callback)

    process = subprocess.Popen("test")

    assert process.returncode == -1
    assert process.received_signals() == (signal.SIGTERM,)

    # the instance appended to `register()` output is the `Popen` instance
    # created later
    assert process_recorder.first_call is process

Asyncio support

The plugin now supports asyncio and works for asyncio.create_subprocess_shell and asyncio.create_subprocess_exec:

async def test_basic_usage(
        ["some-command-that-is-definitely-unavailable"], returncode=500

    process = await asyncio.create_subprocess_shell(
    returncode = await process.wait()

    assert process.returncode == returncode
    assert process.returncode == 500


For full documentation, including API reference, please see


Contributions are very welcome. Tests can be run with tox, please ensure the coverage at least stays the same before you submit a pull request.


Distributed under the terms of the MIT license, “pytest-subprocess” is free and open source software


If you encounter any problems, please file an issue along with a detailed description.

This pytest plugin was generated with Cookiecutter along with @hackebrot’s cookiecutter-pytest-plugin template.


1.5.0 (2023-01-28)


  • #109: Match also os.PathLike.

  • #105: Add program matcher.

Other changes

  • #110: Produce TypeError on Win Py<3.8 for Path args.

1.4.2 (2022-10-02)


  • #87: Add support for Python 3.11.

  • #80, #86: The register() method returns an auxiliary object that will contain all matching FakePopen instances.

Bug fixes

  • #93: Raise callback exceptions on communicate() calls.

Other changes

  • #97: Fixed warnings in tests, treat warnings as errors.

  • #91: Use sys.executable instead just “python” in tests while invoking python subprocess.

  • #90: Fix documentation build, add CI check for it.

1.4.1 (2022-02-09)

Other changes

  • #74: Add fp alias for the fixture, and register for the regisiter_subprocess.

1.4.0 (2022-01-23)


  • #71: Add support for stdin with asyncio.

Bug fixes

  • #68: Make stdout and stderr an asyncio.StreamReader instance when using asyncio functions.

  • #63, #67: Add missing items to asyncio.subprocess.

Other changes

  • #69: Extracted code into separate files to improve navigation.

1.3.2 (2021-11-07)

Bug fixes

  • #61: Fixed behavior of asyncio.create_subproess_exec().

1.3.1 (2021-11-01)

Bug fixes

  • #58: Correctly handle file stream output.

1.3.0 (2021-10-24)


  • #55: Add support for terminate(), kill(), send_signal().

1.2.0 (2021-10-09)


  • #49, #52: Add support for asyncio.

Other changes

  • #50: Change docs theme.

1.1.2 (2021-07-17)

Bug fixes

  • #47: Prevent allow_unregistered() and keep_last_process() from affecting other tests.

1.1.1 (2021-06-18)

Bug fixes

  • #43: Wait for callback thread to finish when calling communicate().

Other changes

  • #42: Fix type annotations for register_subprocess().

1.1.0 (2021-04-18)

Bug fixes

  • #37: Preserve original command in proc.args to prevent leaking the internal Command type.

Other changes

  • #38: Switched CI from Azure Pipelines to GitHub Actions.

  • #35: Drop support for python 3.4 and 3.5. Move type annotations from .pyi files into sources.

1.0.1 (2021-03-20)

Bug fixes

  • #34: Prevent appending newlines to outputs unless defined as list/tuple.

Other changes

  • #32: Make the Command class iterable.

1.0.0 (2020-08-22)


  • #29: Remember subprocess calls to check if expected commands were executed.

  • #28: Allow to match a command with variable arguments (non-exact matching).

0.1.5 (2020-06-19)

Bug fixes

  • #26: encoding and errors arguments will properly trigger text mode.

0.1.4 (2020-04-28)

Bug fixes

  • #22: The returncode will not be ignored when callback is used.

  • #21: The exception raised from callback will take precedence over those from subprocess.

  • #20: Registering process will be now consistent regardless of the command type.

  • #19: Fixed crash for stderr redirect with an empty stream definition.

0.1.3 (2020-03-04)


  • #13: Allow passing keyword arguments into callbacks.

Bug fixes

  • #12: Properly raise exceptions from callback functions.

Documentation changes

  • #15: Add documentation chapter about the callback functions.

0.1.2 (2020-01-17)


  • #3: Add basic support for process input.

Bug fixes

  • #5: Make wait() method to raise TimeoutError after the desired time will elapse.

Documentation changes

  • #7, #8, #9: Create Sphinx documentation.

Other changes

  • #10: Switch from tox to nox for running tests and tasks.

  • #4: Add classifier for Python 3.9. Update CI config to test also on that interpreter version.

0.1.1 (2019-11-24)

Other changes

  • #1, #2: Enable support for Python 3.4, add CI tests for that version.

0.1.0 (2019-11-23)

Initial release

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pytest-subprocess-1.5.0.tar.gz (24.1 kB view hashes)

Uploaded Source

Built Distribution

pytest_subprocess-1.5.0-py3-none-any.whl (20.5 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page