test_common.py

# License: GNU AGPLv3
from functools import partial

import pytest
from sklearn.utils.estimator_checks import parametrize_with_checks

from gtda.images.preprocessing import Binarizer, Inverter


# mark checks to skip
SKIP_TESTS = {
    'Binarizer':  [],
    'Inverter':  [],
    }

# mark tests as a known failure
# TODO: these should be addressed later.
# Note with scikit-learn 0.23 these can be moved to estimator tags
XFAIL_TESTS = {
    'Binarizer':  ["check_transformer_data_not_an_array",
                   "check_transformer_general",
                   "check_transformer_general(readonly_memmap=True)", ],
    'Inverter':  ["check_transformer_data_not_an_array",
                  "check_transformer_general",
                  "check_transformer_general(readonly_memmap=True)", ],
    }


# adapted from sklearn.utils.estimator_check v0.22
def _get_callable_name(obj):
    """Get string representation of a function or a partial function name

    Examples
    --------
    >>> def f(x=2): pass
    >>> _get_callable_name(f)
    'f'
    >>> _get_callable_name(partial(f, x=1))
    'f(x=1)'
    """
    if not isinstance(obj, partial):
        return obj.__name__

    if not obj.keywords:
        return obj.func.__name__

    kwstring = ",".join([f"{k}={v}"
                         for k, v in obj.keywords.items()])
    return f"{obj.func.__name__}({kwstring})"


def _get_estimator_name(estimator):
    """Get string representation for classes and class instances

    Examples
    --------
    >>> from sklearn.preprocessing import StandardScaler
    >>> _get_estimator_name(StandardScaler)
    'StandardScaler'
    >>> _get_estimator_name(StandardScaler())
    'StandardScaler'
    """
    if isinstance(estimator, type):
        # this is class
        return estimator.__name__
    else:
        # this an instance
        return estimator.__class__.__name__


@pytest.mark.filterwarnings("ignore:Input of `fit` contains")
@parametrize_with_checks([Binarizer(), Inverter()])
def test_sklearn_api(check, estimator, request):
    estimator_name = _get_estimator_name(estimator)
    check_name = _get_callable_name(check)

    if check_name in SKIP_TESTS[estimator_name]:
        # skip this test
        pytest.skip()

    if check_name in XFAIL_TESTS[estimator_name]:
        # mark tests as a known failure
        request.applymarker(pytest.mark.xfail(
            run=True, reason='known failure'))

    check(estimator)