I'm attempting to write a unit test using PyTest and Mock that tests that the correct arguments are being passed to json.dump

The function that houses the json.dump accepts 1 argument that is a list of ids that is used to make queries to the db. For each id 2 calls are made to retrieve tables that contain that same id. Then that information is formatted into a dictionary. At the end of the list, I pass the final dictionary to json.dump like so:

    with open(
        "./src/bloomberg/ecdpmm/bmpmsvc/read/serialize_change_set/serialize_change_set.json",
        "w",
        encoding="utf-8",
    ) as serialized_file:

        json.dump(response_json, serialized_file, default=object_handler, sort_keys=True, indent=4)

What confuses me is how, in another file, I can write a test cases asserts the correct arguments are being passed. Like I've seen code snippets that do something along the lines of:

json = Mock()

json.loads('{"key": "value"}')
>> <Mock name='mock.loads()' id='4550144184'>

json.loads.assert_called_with('{"key": "value"}')

When doing the assertion are the arguments (that I am mocking) being compared to the arguments being passed in my main file ? Are the functions above the correct approach to what I am trying to do ?

Any guidance/tips would be greatly appreciated!

Here is the function I am trying to test:

def serialize_change_set(change_set_ids: list[int] | None = None) -> None:
    """
    Get a list of altered tables in the change set and serialize them
    """
    logger.info("serialize all the tables in the change set ")

    response_json = {}
    for cs_id in change_set_ids:
        response_json[cs_id] = {"Changed_Tables": {}}

        description = db.get_change_set_description(cs_id)
        response_json[cs_id]["Description"] = description

        altered_tables = db.get_change_set(cs_id)
        for table_name, table_rows in altered_tables.items():
            formatted_rows = [dict(row) for row in table_rows]
            response_json[cs_id]["Changed_Tables"][table_name] = formatted_rows

    with open(
        "serialize_change_set.json",
        "w",
        encoding="utf-8",
    ) as serialized_file:

        json.dump(response_json, serialized_file, default=object_handler, sort_keys=True, indent=4)

CodePudding user response：

If you have a module (say, my_module.py) that does this:

import json


def my_function(response_json):
    with open(
        "serialize_change_set.json",
        "w",
        encoding="utf-8",
    ) as serialized_file:
        json.dump(
            response_json,
            serialized_file,
            sort_keys=True,
            indent=4,
        )

Then you can write a test like this that mocks out json.dump and then later tests to see how it was called. In test_my_module.py we have:

import my_module
from unittest import mock


def test_my_function():
    test_data = {"some": "value"}

    with mock.patch("my_module.json"), mock.patch(
        "my_module.open", mock.mock_open()
    ) as _open:
        my_module.my_function(test_data)

        my_module.json.dump.assert_called_with(
            test_data, _open(), sort_keys=True, indent=4
        )

We have to mock out both json.dump and the builtin open function (because (a) we don't want our unit tests to create files, and (b) we want to be able to match the value returned by open in our call to assert_called_with).

Our test passing:

$ pytest -v
========================================== test session starts ==========================================
platform linux -- Python 3.10.4, pytest-6.2.4, py-1.11.0, pluggy-0.13.1 -- /usr/bin/python3
cachedir: .pytest_cache
rootdir: /home/lars/tmp/python
plugins: anyio-3.5.0
collected 1 item

test_my_module.py::test_my_function PASSED                                                        [100%]

=========================================== 1 passed in 0.02s ===========================================

CodePudding user response：

ISTM that the best possible test for arguments to json.dump() is to run them through _ = json.dump(*args, **kwargs) to see if it raises an exception.

Since the two functions have the same internal logic, the dumps() function can serve as a validator for dump without having to write to disk.

Put that is a validation function and use Mock to attach the call.