def test_required_package_installation():
for package_specs in REQUIRED_MODULE_METADATA:
package = package_specs[0]
min_version = package_specs[1]['min_version']
imported_package = __import__(package)
installed_version = imported_package.__version__
assert _compare_version(installed_version, '>=', min_version)
print(package, 'min:', min_version, 'installed:', installed_version)
def test_deprecation_warning_compute_multi_gray_matter_mask():
imgs = [
Nifti1Image(np.ones((9, 9, 9)), np.eye(4)),
Nifti1Image(np.ones((9, 9, 9)), np.eye(4))
]
if _compare_version(sklearn.__version__, '<', '0.22'):
with pytest.deprecated_call():
masking.compute_multi_gray_matter_mask(imgs)
else:
with pytest.warns(FutureWarning,
match="renamed to 'compute_multi_brain_mask'"):
masking.compute_multi_gray_matter_mask(imgs)
def pytest_collection_modifyitems(items):
# numpy changed the str/repr formatting of numpy arrays in 1.14.
# We want to run doctests only for numpy >= 1.14.Adapted from scikit-learn
if _compare_version(np.__version__, '<', '1.14'):
reason = 'doctests are only run for numpy >= 1.14'
skip_doctests = True
else:
skip_doctests = False
if skip_doctests:
skip_marker = pytest.mark.skip(reason=reason)
for item in items:
if isinstance(item, DoctestItem):
item.add_marker(skip_marker)
def plot_brain_schematics(ax, direction, **kwargs):
"""Creates matplotlib patches from a json custom format and plots them
on a matplotlib Axes.
Parameters
----------
ax : A MPL axes instance
The axes in which the plots will be drawn.
direction : {'x', 'y', 'z', 'l', 'r'}
The directions of the view.
**kwargs :
Passed to the matplotlib patches constructor.
Returns
-------
object_bounds : (xmin, xmax, ymin, ymax) tuple
Useful for the caller to be able to set axes limits.
"""
if _compare_version(matplotlib.__version__, '>=', "2.0"):
get_axis_bg_color = ax.get_facecolor()
else:
get_axis_bg_color = ax.get_axis_bgcolor()
black_bg = colors.colorConverter.to_rgba(get_axis_bg_color) \
== colors.colorConverter.to_rgba('k')
json_filename, transform = _get_json_and_transform(direction)
with open(json_filename) as json_file:
json_content = json.loads(json_file.read())
mpl_patches = _get_mpl_patches(json_content,
transform=transform + ax.transData,
invert_color=black_bg,
**kwargs)
for mpl_patch in mpl_patches:
ax.add_patch(mpl_patch)
object_bounds = _get_object_bounds(json_content, transform)
return object_bounds
# For "manual" documents, if this is true, then toplevel headings are parts,
# not chapters.
#latex_use_parts = False
# Additional stuff for the LaTeX preamble.
# Documents to append as an appendix to all manuals.
#latex_appendices = []
latex_elements = {
'classoptions': ',oneside',
'babel': '\\usepackage[english]{babel}',
# Get completely rid of index
'printindex': '',
}
if _compare_version(sphinx.__version__, '<', '1.5'):
latex_preamble = r"""
\usepackage{amsmath}\usepackage{amsfonts}\usepackage{bm}\usepackage{morefloats}
\let\oldfootnote\footnote
\def\footnote#1{\oldfootnote{\small #1}}
"""
else:
latex_elements['preamble'] = r"""
\usepackage{amsmath}\usepackage{amsfonts}\usepackage{bm}\usepackage{morefloats}
\let\oldfootnote\footnote
\def\footnote#1{\oldfootnote{\small #1}}
"""
# If false, no module index is generated.
if _compare_version(sphinx.__version__, '<', '1.5'):
latex_use_modindex = False
def _safe_cache(memory, func, **kwargs):
"""A wrapper for mem.cache that flushes the cache if the version
number of nibabel has changed.
"""
''' Workaround for
https://github.com/scikit-learn-contrib/imbalanced-learn/issues/482
joblib throws a spurious warning with newer scikit-learn.
This code uses the recommended method first and the deprecated one
if that fails, ensuring th warning is not generated in any case.
'''
try:
location = os.path.join(memory.location, 'joblib')
except AttributeError:
location = memory.location
except TypeError:
location = None
if location is None or location in __CACHE_CHECKED:
return memory.cache(func, **kwargs)
version_file = os.path.join(location, 'module_versions.json')
versions = dict()
if os.path.exists(version_file):
with open(version_file, 'r') as _version_file:
versions = json.load(_version_file)
modules = (nibabel, )
# Keep only the major + minor version numbers
my_versions = dict((m.__name__, m.__version__) for m in modules)
commons = set(versions.keys()).intersection(set(my_versions.keys()))
collisions = [
m for m in commons
if not _compare_version(versions[m], '==', my_versions[m])
]
# Flush cache if version collision
if len(collisions) > 0:
if nilearn.CHECK_CACHE_VERSION:
warnings.warn("Incompatible cache in %s: "
"different version of nibabel. Deleting "
"the cache. Put nilearn.CHECK_CACHE_VERSION "
"to false to avoid this behavior." % location)
try:
tmp_dir = (os.path.split(location)[:-1] +
('old_%i' % os.getpid(), ))
tmp_dir = os.path.join(*tmp_dir)
# We use rename + unlink to be more robust to race
# conditions
os.rename(location, tmp_dir)
shutil.rmtree(tmp_dir)
except OSError:
# Another process could have removed this dir
pass
try:
os.makedirs(location)
except OSError:
# File exists?
pass
else:
warnings.warn("Incompatible cache in %s: "
"old version of nibabel." % location)
# Write json files if configuration is different
if versions != my_versions:
with open(version_file, 'w') as _version_file:
json.dump(my_versions, _version_file)
__CACHE_CHECKED[location] = True
return memory.cache(func, **kwargs)
def test_compare_version(version_a, operator, version_b):
assert _compare_version(version_a, operator, version_b)
def test_compare_version_error():
with pytest.raises(ValueError,
match=("'_compare_version' received an "
"unexpected operator <>.")):
_compare_version('0.1.0', '<>', '1.1.0')
def test_butterworth():
rng = np.random.RandomState(42)
n_features = 20000
n_samples = 100
sampling = 100
low_pass = 30
high_pass = 10
# Compare output for different options.
# single timeseries
data = rng.standard_normal(size=n_samples)
data_original = data.copy()
'''
May be only on py3.5:
Bug in scipy 1.1.0 generates an unavoidable FutureWarning.
(More info: https://github.com/scipy/scipy/issues/9086)
The number of warnings generated is overwhelming TravisCI's log limit,
causing it to fail tests.
This hack prevents that and will be removed in future.
'''
buggy_scipy = (_compare_version(scipy.__version__, '<', '1.2')
and _compare_version(scipy.__version__, '>', '1.0'))
if buggy_scipy:
warnings.simplefilter('ignore')
''' END HACK '''
out_single = nisignal.butterworth(data,
sampling,
low_pass=low_pass,
high_pass=high_pass,
copy=True)
np.testing.assert_almost_equal(data, data_original)
nisignal.butterworth(data,
sampling,
low_pass=low_pass,
high_pass=high_pass,
copy=False)
np.testing.assert_almost_equal(out_single, data)
np.testing.assert_(id(out_single) != id(data))
# multiple timeseries
data = rng.standard_normal(size=(n_samples, n_features))
data[:, 0] = data_original # set first timeseries to previous data
data_original = data.copy()
out1 = nisignal.butterworth(data,
sampling,
low_pass=low_pass,
high_pass=high_pass,
copy=True)
np.testing.assert_almost_equal(data, data_original)
np.testing.assert_(id(out1) != id(data_original))
# check that multiple- and single-timeseries filtering do the same thing.
np.testing.assert_almost_equal(out1[:, 0], out_single)
nisignal.butterworth(data,
sampling,
low_pass=low_pass,
high_pass=high_pass,
copy=False)
np.testing.assert_almost_equal(out1, data)
# Test nyquist frequency clipping, issue #482
out1 = nisignal.butterworth(data, sampling, low_pass=50., copy=True)
out2 = nisignal.butterworth(
data,
sampling,
low_pass=80., # Greater than nyq frequency
copy=True)
np.testing.assert_almost_equal(out1, out2)
np.testing.assert_(id(out1) != id(out2))