TST Fixes failing test on main #19639

thomasjpfan · 2021-03-07T20:32:06Z

Fixes failing tests on main.

Here are the failing runs: https://dev.azure.com/scikit-learn/scikit-learn/_build/results?buildId=27158&view=logs&jobId=dfe99b15-50db-5d7b-b1e9-4105c42527cf&j=dfe99b15-50db-5d7b-b1e9-4105c42527cf&t=ef785ae2-496b-5b02-9f0e-07a6c3ab3081

jnothman · 2021-03-08T00:49:31Z

This is one of the failing runs? https://github.com/scikit-learn/scikit-learn/runs/2048815281

jnothman

there are some unrelated changes in here, aren't there?

thomasjpfan

@jnothman I left comments connecting each diff and the corresponding test that it fixes

thomasjpfan · 2021-03-08T14:17:39Z

sklearn/datasets/tests/test_openml.py

+        feature = data_by_id.data[:, feature_idx]
+        if np.all(np.isnan(feature)):
+            # no need to check when all values are nan
+            continue
+        values = np.unique(feature)


To resolve test_fetch_openml_anneal and other test in test_openml

This avoids possible DeprecationWarning.

This is now fixed in #20327, right?

thomasjpfan · 2021-03-08T14:18:49Z

sklearn/ensemble/tests/test_weight_boosting.py

@@ -33,10 +33,10 @@

 # Toy sample
 X = [[-2, -1], [-1, -1], [-1, -2], [1, 1], [1, 2], [2, 1]]
-y_class = ["foo", "foo", "foo", 1, 1, 1]    # test string class labels
+y_class = ["foo", "foo", "foo", "bar", "bar", "bar"]


To resolve test_classification_toy. Numpy now warns when running np.array(["foo", 1]).

thomasjpfan · 2021-03-08T14:19:04Z

sklearn/linear_model/_sgd_fast.pyx

@@ -70,13 +70,37 @@ cdef class LossFunction:
        """Python version of `dloss` for testing.

        Pytest needs a python function and can't use cdef functions.
+
+        Parameters


To resolve test_docstring_parameters

thomasjpfan · 2021-03-08T14:19:31Z

sklearn/neighbors/_binary_tree.pxi

@@ -1620,11 +1620,12 @@ cdef class BinaryTree:
            - 'linear'
            - 'cosine'
            Default is kernel = 'gaussian'
-        atol, rtol : float, default=0, 1e-8
+        atol : float, default=0


To resolve test_docstring_parameters. Building with the newest version of Cython now exposes these docstrings correctly.

thomasjpfan · 2021-03-08T14:19:56Z

sklearn/neighbors/_classification.py

@@ -475,7 +475,18 @@ def fit(self, X, y):
                    raise TypeError("The outlier_label of classes {} is "
                                    "supposed to be a scalar, got "
                                    "{}.".format(classes, label))
-                if np.append(classes, label).dtype != classes.dtype:
+                # Starting in numpy 1.21 np.append admits a warning when


To resolve test_radius_neighbors_classifier_outlier_labeling

thomasjpfan · 2021-03-08T14:20:21Z

sklearn/preprocessing/_encoders.py

-                                 force_all_finite=force_all_finite)
+
+            # starting in numpy 1.21 mixed typed `X` will raise a FutureWarning
+            with warnings.catch_warnings():


To resolve test_one_hot_encoder_warning

thomasjpfan · 2021-03-08T14:20:46Z

sklearn/utils/class_weight.py

@@ -63,7 +63,11 @@ def compute_class_weight(class_weight, *, classes, y):
            raise ValueError("class_weight must be dict, 'balanced', or None,"
                             " got: %r" % class_weight)
        for c in class_weight:
-            i = np.searchsorted(classes, c)
+            try:


To resolve test_compute_class_weight_not_present

thomasjpfan · 2021-03-08T14:21:00Z

sklearn/utils/tests/test_encode.py

@@ -48,10 +49,19 @@ def test_encode_with_check_unknown():

 def _assert_check_unknown(values, uniques, expected_diff, expected_mask):
    diff = _check_unknown(values, uniques)
-    assert_array_equal(diff, expected_diff)
+    if is_scalar_nan(expected_diff[-1]):


To resolve test_check_unknown_missing_values failing

thomasjpfan · 2021-03-08T14:21:42Z

sklearn/utils/multiclass.py

@@ -303,7 +310,15 @@ def type_of_target(y):
        _assert_all_finite(y)
        return 'continuous' + suffix

-    if (len(np.unique(y)) > 2) or (y.ndim >= 2 and len(y[0]) > 1):
+    try:
+        unique_values = np.unique(y)


Resolves test_unique_labels_mixed_types

thomasjpfan · 2021-03-08T14:25:13Z

sklearn/utils/multiclass.py

@@ -160,7 +160,14 @@ def is_multilabel(y):
                (y.dtype.kind in 'biu' or  # bool, int, uint
                 _is_integral_float(np.unique(y.data))))
    else:
-        labels = np.unique(y)
+        try:
+            labels = np.unique(y)


Resolves test_unique_labels_mixed_types:

with pytest.raises(ValueError): unique_labels([["1", 2], [1, 3]])

lorentzenchr · 2021-06-23T16:30:30Z

I get DeprecationWarning: elementwise comparison failed; this will raise an error in the future. in #20250 test_fetch_openml_anneal[False]. It stems from calling np.unique on nan values.

thomasjpfan · 2021-06-23T17:13:16Z

@lorentzenchr For now, we are ignoring that issue here: #20327 since it is a bug in numpy which has a fix here: numpy/numpy#19301

jjerphan · 2021-07-07T06:52:19Z

Have the tests targeted by this PR been fixed?

thomasjpfan · 2022-02-09T14:59:38Z

Closing, since main looks okay now.

TST Fixes failing test on main

52f2265

thomasjpfan added the No Changelog Needed label Mar 7, 2021

thomasjpfan marked this pull request as ready for review March 7, 2021 22:05

ENH Capture all exception for future releases of numpy

705b998

jnothman reviewed Mar 8, 2021

View reviewed changes

thomasjpfan commented Mar 8, 2021

View reviewed changes

thomasjpfan added the cython label Apr 13, 2021

thomasjpfan closed this Feb 9, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

TST Fixes failing test on main #19639

TST Fixes failing test on main #19639

thomasjpfan commented Mar 7, 2021 •

edited

jnothman commented Mar 8, 2021

jnothman left a comment

thomasjpfan left a comment

thomasjpfan Mar 8, 2021

lorentzenchr Jun 23, 2021

lorentzenchr Jun 23, 2021

thomasjpfan Mar 8, 2021

thomasjpfan Mar 8, 2021

thomasjpfan Mar 8, 2021

thomasjpfan Mar 8, 2021

thomasjpfan Mar 8, 2021

thomasjpfan Mar 8, 2021

thomasjpfan Mar 8, 2021

thomasjpfan Mar 8, 2021

thomasjpfan Mar 8, 2021

lorentzenchr commented Jun 23, 2021

thomasjpfan commented Jun 23, 2021

jjerphan commented Jul 7, 2021

thomasjpfan commented Feb 9, 2022

TST Fixes failing test on main #19639

TST Fixes failing test on main #19639

Conversation

thomasjpfan commented Mar 7, 2021 • edited

jnothman commented Mar 8, 2021

jnothman left a comment

Choose a reason for hiding this comment

thomasjpfan left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

lorentzenchr commented Jun 23, 2021

thomasjpfan commented Jun 23, 2021

jjerphan commented Jul 7, 2021

thomasjpfan commented Feb 9, 2022

thomasjpfan commented Mar 7, 2021 •

edited