def test_join_features_libsvm_output_argparse():
"""
Make sure that join_features exits when passing in output libsvm files
"""
jf_cmd_args = ['foo.csv', 'bar.csv', 'baz.libsvm']
jf.main(argv=jf_cmd_args)
def test_join_features_unknown_output_format():
"""
Make sure that join_features exits when passing in an unknown output file format
"""
jf_cmd_args = ['foo.csv', 'bar.csv', 'baz.xxx']
jf.main(argv=jf_cmd_args)
def check_join_features_raises_system_exit(cmd_args):
"""
Little helper to make test output cleaner for tests that check that
join_features exits with the specified arguments.
"""
jf.main(cmd_args)
def check_join_features_argparse(extension, label_col='y', id_col='id',
quiet=False):
"""
A utility function to check that we are setting up argument parsing
correctly for join_features for ALL file types. We are not checking
whether the results are correct because we have separate tests for that.
"""
# replace the run_configuration function that's called
# by the main() in filter_feature with a mocked up version
writer_class = EXT_TO_WRITER[extension]
# create some dummy input and output filenames
infile1 = join(_my_dir, 'other', 'test_join_features1{}'.format(extension))
infile2 = join(_my_dir, 'other', 'test_join_features2{}'.format(extension))
outfile = 'bar{}'.format(extension)
# create a simple featureset with actual ids, labels and features
fs1, _ = make_classification_data(num_labels=3,
train_test_ratio=1.0,
random_state=1234)
fs2, _ = make_classification_data(num_labels=3,
train_test_ratio=1.0,
feature_prefix='g',
random_state=5678)
jf_cmd_args = [infile1, infile2, outfile]
if extension in ['.tsv', '.csv', '.arff']:
writer1 = writer_class(infile1, fs1, label_col=label_col,
id_col=id_col)
writer2 = writer_class(infile2, fs2, label_col=label_col,
id_col=id_col)
jf_cmd_args.extend(['-l', label_col])
jf_cmd_args.extend(['--id_col', id_col])
else:
writer1 = writer_class(infile1, fs1)
writer2 = writer_class(infile2, fs2)
writer1.write()
writer2.write()
if quiet:
jf_cmd_args.append('-q')
# Substitute mock methods for the main methods that get called by
# filter_features: FeatureSet.filter() and the __init__() method
# of the appropriate writer. We also need to mock the write()
# method to prevent actual writing.
with patch.object(FeatureSet, '__add__', autospec=True) as add_mock, \
patch.object(writer_class, '__init__', autospec=True,
return_value=None) as write_init_mock, \
patch.object(writer_class, 'write', autospec=True) as write_mock:
jf.main(argv=jf_cmd_args)
# get the various arguments from the three mocked up methods
add_pos_arguments, add_kw_arguments = add_mock.call_args
write_pos_arguments, write_kw_arguments = write_init_mock.call_args
# make sure that the arguments they got were the ones we specified
eq_(write_pos_arguments[1], outfile)
eq_(write_kw_arguments['quiet'], quiet)
# note that we cannot test the label_col column for the writer
# the reason is that is set conditionally and those conditions
# do not execute with mocking
eq_(add_pos_arguments[0], fs1)
eq_(add_pos_arguments[1], fs2)