omarkohl / pytest-datafiles Goto Github PK

It seems that if relative paths are given in the decorator, they're assumed to be relative to the directory where you ran pytest from. To me, it makes more sense to have the paths relative to the file that defines the tests that are running. Breaking backwards compatibility is problematic, but what would you think of having a flag which can be used to make the paths relative to the test itself?

I see a similar issue with an earlier version of pytest (4.3.0) which appears to have been fixed with pytest v4.5.0. But I'm using pytest v6.2.4 - is there something I'm missing?

The sdist (tarball) at PyPI doesnt contain the test files. i.e. https://files.pythonhosted.org/packages/26/60/4e49c18caf0ba7b7222da47ea6143f4f1ea3993f043be7bdf4ae409e544d/pytest-datafiles-2.0.1.tar.gz

Not including the tests makes it harder to package this for distros.

I notice that #10 is currently in flight, so it may be better to do that first, as poetry has its own way of recording which .py and data files need to be included in the sdist.

for reference, poetry needs something like the following in pyproject.toml:

packages = [
    { include = "pytest_datafiles.py"},  # not sure about this line
    { include = "tests", format = "sdist" },  # this is the voodoo, but maybe something extra is needed to ensure the _fixture_files is also included
]

It seems that the datafiles fixture is a py.path (well, a LocalPath) - does this have any functionality that pathlib.Path does not? If not, I think using pathlib's Path would be nice. The only function I've missed so far is is_file, but I haven't tested this much. Additionally, the py website says

Note: The ‘py’ library is in “maintenance mode” and so is not recommended for new projects. Please check out pathlib or pathlib2 for path operations.

I want to test two files like so:

@pytest.mark.datafiles(os.path.join(FIXTURE_DIR, '2019_DPD.txt'),
                       os.path.join(FIXTURE_DIR, 'expected_DPD.csv'))
def test_dpd_to_dataframe(datafiles):
    path_to_data_file = str(datafiles.listdir()[0])
    # Do stuff

    expected_dpd_df = pd.read_csv(str(datafiles.listdir()[1]))
    # Do other stuff

    assert dpd_df.equals(expected_dpd_df)

and this will work only if file1 and file2 are stored in that order (The functions I am testing with are different and the files, hence, are different)

Is there any way to make sure that datafiles.listdir() returns the files in the exact same order and not switch it. (Like it is happening now)

Hi,

I've just seen this warning when upgrading to pytest-3.8.0:

pytest_datafiles.py:71: RemovedInPytest4Warning: MarkInfo objects are deprecated as they contain merged marks which are hard to deal with correctly.
Please use node.get_closest_marker(name) or node.iter_markers(name).
Docs: https://docs.pytest.org/en/latest/mark.html#updating-code
  content = request.keywords.get('datafiles').args

pytest_datafiles.py:76: RemovedInPytest4Warning: MarkInfo objects are deprecated as they contain merged marks which are hard to deal with correctly.
Please use node.get_closest_marker(name) or node.iter_markers(name).
Docs: https://docs.pytest.org/en/latest/mark.html#updating-code
  options.update(request.keywords.get('datafiles').kwargs)

I may try to fix this in the next few days (it should be easy enough), but just wanted to let you know now.

Thanks for the useful module!

Tom

I'm trying to write a fixture that returns safely a data file for testing.
Here's an example:

import os
import pytest
import json
from pathlib import Path

# from real_estate_tools.utils import get_property_detail, fetch_properties
FIXTURE_DIR = Path(__file__).parent.resolve() / "test_files"


# Works as expected
@pytest.mark.datafiles(FIXTURE_DIR / "complete_result.json")
def test_load_file(datafiles) -> dict:
    with open(datafiles / os.listdir(datafiles)[0], "r") as f:
        assert json.load(f)


# Turn to fixture
@pytest.mark.datafiles(FIXTURE_DIR / "complete_result.json")
@pytest.fixture
def load_file(datafiles) -> dict:
    with open(datafiles / os.listdir(datafiles)[0], "r") as f:
        return json.load(f)


# Doesn't work
# In the fixture, the part os.listdir(datafiles)
# has zero length, i.e. nothing is copied to the
# target temp dir.
def test_using_fixture(load_file):
    print(load_file)
    assert True

My motivation is that I want to execute multiple different tests using complete_result.json as an input. Therefore, I'm trying to write the "loader" once as a fixture but it doesn't work.

What am I doing wrong?

I'd like to use datafiles in a parameterized fixture. In other words, I want my test function to be automatically invoked with different data files, as with a normal fixture with the params argument. How do I achieve this? I tried the code below, but my datafiles are not copied to the tmpdir, and the test collection fails, because the test_files() fixture does not yield any output. I'm quite new to pytest, so possibly I don't fully understand how it works.

@pytest.fixture(params = [1,2])
@pytest.mark.datafiles('file1.txt','file1.txt')
def test_files(request,datafiles):
    for testfile in datafiles.listdir():
        yield testfile

@pytest.fixture(params = ['expected_output1','expected_output2'])
def expected_output(request):
    return request.param

def my_test_function(test_files,expected_output):
    assert myFcn(test_files) == expected_output

Mode bits are not copied by default (https://py.readthedocs.io/en/latest/path.html#py._path.local.LocalPath.copy). This results in unexpected modes in testing. It would be useful if there were some way to use the datafiles fixture and copy modes.

With pytest 4.3.0 I am constantly seeing the following warning when executing tests using datafiles:

.tox/test-py37/lib/python3.7/site-packages/_pytest/mark/structures.py:324
  xxxx/.tox/test-py37/lib/python3.7/site-packages/_pytest/mark/structures.py:324: PytestUnknownMarkWarning: Unknown pytest.mark.datafiles - is this a typo?  You can register custom marks to avoid this warning - for details, see https://docs.pytest.org/en/latest/mark.html
    PytestUnknownMarkWarning,

The plugin works as expected though

I am running into an issue where I get a "ScopeMismatch" error if I try to use datafiles to create another fixture with a different scope than "function".

I want the scope to be "module" based because I would like to keep the database open for all tests since loading the database is time consuming.

Here is a simplified example of what I am trying to do:

@pytest.fixture(scope="module")
@pytest.mark.datafiles(TEST_DATA_DIR / "image.png")
def image_db(datafiles):
    """Loads image.png into database and returns database entry"""
   with database():
      database.add(datafiles / "image.png")
      for db_entry in database.query_images():
         yield db_entry
         break

test setup failed
ScopeMismatch: You tried to access the 'function' scoped fixture 'datafiles' with a 'module' scoped request object, involved factories

Any help is a appreciated.

@pytest.fixture
@pytest.mark.datafiles('/tmp/big_files/film1.mp4')
def lots_of_files(datafiles):
    pass

currently not providing the files I'm looking for

I'm using pytest-datafiles to test BuildStream, which is a build tool and as such does a lot of filesystem manipulation (naturally, we also have a lot of test cases which need to manipulate files mirrored outside of the git directory).

For us, the difference between a symlink to a file, and a copy of the said file, is an important difference; we need to test build scenarios where there are symlinks.

Currently our problem is that, if I commit symlinks to my test data dirs in git, and then use @pytest.mark.datafiles() to test with a directory that contains symlinks, the symlinks just dont appear in the test environment. Instead copies of the symlink targets are created.

Currently I have some ugly workarounds in place which create the required symlinks in the tmpdir manually in advance of testing.