def test_correct_data(self, filename):
file1 = tf.get_abs_path('chimera_1event_2levels.log')
file2 = tf.get_abs_path('chimera_1event.log')
concat_files([file1, file2], output_filename=filename)
reader1 = get_reader_from_filename(file1)
reader2 = get_reader_from_filename(file2)
reader_out = get_reader_from_filename(filename)
sample_rate1 = reader1.get_sample_rate()
sample_rate2 = reader2.get_sample_rate()
sample_rate_out = reader_out.get_sample_rate()
self.assertEqual(sample_rate_out, sample_rate1,
"Unexpected sample rate. Should be {0}, was {1}.".format(sample_rate1, sample_rate_out))
self.assertEqual(sample_rate_out, sample_rate2,
"Unexpected sample rate. Should be {0}, was {1}.".format(sample_rate2, sample_rate_out))
data1 = reader1.get_all_data()[0]
data2 = reader2.get_all_data()[0]
data_out_should_be = np.zeros(data1.size + data2.size)
data_out_should_be[:data1.size] = data1[:]
data_out_should_be[data1.size:] = data2[:]
data_out = reader_out.get_all_data()[0]
np.testing.assert_array_equal(data_out, data_out_should_be)
reader1.close()
reader2.close()
reader_out.close()
def test_filtered_baseline(self, filename):
"""
Tests that the filtered baseline is the same as the unfiltered.
"""
data_filename = tf.get_abs_path('chimera_1event.log')
reader = get_reader_from_filename(data_filename)
data_all = reader.get_all_data()
data = data_all[0]
reader.close()
baseline = np.mean(data[:150])
# Check at different filter frequencies and re-sample rates
for rates in ([1.e4, 1.e6], [1.e4, 0], [5.e5, 4.e6], [7.7e4, 1.e6]):
filter_freq = rates[0]
re_sample_rate = rates[1]
out_filename = filter_file(data_filename, filter_frequency=filter_freq, out_sample_rate=re_sample_rate,
output_filename=filename)
reader = get_reader_from_filename(out_filename)
data2 = reader.get_all_data()[0]
reader.close()
# Note we re-sampled, which is why only take 30 data points.
baseline2 = np.mean(data2[:20])
ratio = abs((baseline - baseline2) / baseline)
print "ratio:", ratio
self.assertLessEqual(ratio, 0.05, "Filtered baseline different from original. "
"Should be {0}, got {1}.".format(baseline, baseline2))
os.remove(out_filename)
def test_noise(self, filename):
"""
Tests that the noise is added correctly(its mean and std_dev are correct).
"""
seconds = 1.
sample_rate = 1.e6
baseline = 1.
noise_loc = 1.
noise_scale = 0.6
noise = stats.norm(loc=noise_loc, scale=noise_scale)
create_random_data(filename, seconds=seconds, sample_rate=sample_rate, baseline=baseline,
noise=noise)
reader = get_reader_from_filename(filename)
data_all = reader.get_all_data()
data = data_all[0]
mean = np.mean(data)
mean_should_be = baseline + noise_loc
self.assertAlmostEqual(mean, mean_should_be, 1,
"Unexpected mean. Wanted {0}, got {1}.".format(mean_should_be, mean))
std_dev = np.std(data)
self.assertAlmostEqual(noise_scale, std_dev, 1, "Unexpected standard deviation. "
"Wanted {0}, got {1}".format(noise_scale, std_dev))
reader.close()
def test_convert_chimera_file_equality(self, filename):
"""
Test that the original/converted matrices and sample rates are the same for one-channel data.
"""
data_filename = tf.get_abs_path('chimera_1event.log')
output_filename = convert_file(data_filename, output_filename=filename)
orig_reader = get_reader_from_filename(data_filename)
orig_data_all = orig_reader.get_all_data()
orig_sample_rate = orig_reader.get_sample_rate()
self.assertEqual(len(orig_data_all), 1)
out_reader = DataFileReader(output_filename)
out_data_all = out_reader.get_all_data()
out_sample_rate = out_reader.get_sample_rate()
self.assertEqual(len(out_data_all), 1)
orig_data = orig_data_all[0]
out_data = out_data_all[0]
# assert sample rates are equal
self.assertAlmostEqual(1.0 * orig_sample_rate / out_sample_rate, 1, 4)
# assert the two arrays are equal
np.testing.assert_array_equal(orig_data, out_data)
orig_reader.close()
out_reader.close()
def open_data_files(self, file_names=None):
"""
Analyzes the files for correctness, then adds them to the list widget.
:param ListType<StringType> file_names: The file names to be included in the list widget. If not included,
this function will use a QtGui.QFileDialog.getOpenFileNames to open
files.
:returns: BooleanType -- **True** if files were opened, **False** otherwise.
"""
if file_names is None:
file_names = QtGui.QFileDialog.getOpenFileNames(
self, 'Open data file', '.',
"All types(*.h5 *.hkd *.log *.mat);;"
"Pypore data files *.h5(*.h5);;"
"Heka files *.hkd(*.hkd);;"
"Chimera files *.log(*.log);;Gabys files *.mat(*.mat)")[0]
if len(file_names) > 0:
self.file_list_widget.clear()
else:
return
are_files_opened = False
open_dir = None
for w in file_names:
reader = get_reader_from_filename(w)
# if 'error' in params:
# TODO implement error handling in readers
# pass
# else:
reader.close()
are_files_opened = True
item = FileListItem(w)
open_dir = item.get_directory()
self.file_list_widget.addItem(item)
return are_files_opened
def test_noise(self, filename):
"""
Tests that the noise is added correctly(its mean and std_dev are correct).
"""
seconds = 1.
sample_rate = 1.e6
baseline = 1.
noise_loc = 1.
noise_scale = 0.6
noise = stats.norm(loc=noise_loc, scale=noise_scale)
create_random_data(filename,
seconds=seconds,
sample_rate=sample_rate,
baseline=baseline,
noise=noise)
reader = get_reader_from_filename(filename)
data_all = reader.get_all_data()
data = data_all[0]
mean = np.mean(data)
mean_should_be = baseline + noise_loc
self.assertAlmostEqual(
mean, mean_should_be, 1,
"Unexpected mean. Wanted {0}, got {1}.".format(
mean_should_be, mean))
std_dev = np.std(data)
self.assertAlmostEqual(
noise_scale, std_dev, 1, "Unexpected standard deviation. "
"Wanted {0}, got {1}".format(noise_scale, std_dev))
reader.close()
def convert_file(filename, output_filename=None):
"""
Convert a file to the pypore .h5 file format. Returns the new file's name.
"""
reader = get_reader_from_filename(filename)
sample_rate = reader.get_sample_rate()
n_points = reader.get_points_per_channel_total()
if output_filename is None:
output_filename = filename.split('.')[0] + '.h5'
save_file = data_file.open_file(output_filename, mode='w', sample_rate=sample_rate, n_points=n_points)
blocks_to_get = 1
data = reader.get_next_blocks(blocks_to_get)[0]
n = data.size
i = 0
while n > 0:
save_file.root.data[i:n + i] = data[:]
i += n
data = reader.get_next_blocks(blocks_to_get)[0]
n = data.size
reader.close()
save_file.flush()
save_file.close()
return output_filename
def test_noise_no_events(self, filename):
"""
Tests that noise is added correctly.
"""
sample_rate = 1.e5
baseline = 1.0
n = 100000
noise_scale = 1.
create_specified_data(filename=filename, n=n, sample_rate=sample_rate, baseline=baseline,
noise_scale=noise_scale, events=[])
reader = get_reader_from_filename(filename)
data_all = reader.get_all_data()
data = data_all[0]
reader.close()
self.assertEqual(n, data.size, "Unexpected array size. Wanted {0}, got {1}.".format(n, data.size))
decimal = 1
mean = np.mean(data)
self.assertAlmostEqual(baseline, mean, decimal,
"Unexpected baseline. Wanted {0}, got {1}. Tested to {2} decimals.".format(baseline,
mean,
decimal))
decimal = 1
std_dev = np.std(data)
self.assertAlmostEqual(noise_scale, std_dev, decimal,
msg="Unexpected stddev. Wanted{0}, got {1}. Tested to {2} decimals.".format(noise_scale,
std_dev,
decimal))
def filter_file(filename,
filter_frequency,
out_sample_rate,
output_filename=None):
"""
Reads data from the filename file and uses a Butterworth low-pass filter with cutoff at filter_frequency. Outputs
the filtered waveform to a new :py:class:`pypore.filetypes.data_file.DataFile`.
:param StringType filename: Filename containing data to be filtered.
:param DoubleType filter_frequency: Cutoff frequency for the low-pass Butterworth filter.
:param DoubleType out_sample_rate: After the data is filtered, it can be resampled to roughly out_sample_rate. If \
out_sample_rate <= 0, the data will not be resampled.
:param StringType output_filename: (Optional) Filename for the filtered data. If not specified,
for an example filename='test.mat', the default output_filename would be 'test_filtered.h5'
:returns: StringType -- The output filename of the filtered data.
Usage:
>>> import pypore.file_converter as fc
>>> fc.filter_file("filename", 1.e4, 1.e5, "output.h5") # filter at 10kHz, resample at 100kHz
"""
reader = get_reader_from_filename(filename)
data = reader.get_all_data()[0]
sample_rate = reader.get_sample_rate()
final_sample_rate = sample_rate
n_points = len(data)
if output_filename is None:
output_filename = os.path.splitext(filename)[0] + '_filtered.h5'
# wn is a fraction of the Nyquist frequency (half the sampling frequency).
wn = filter_frequency / (0.5 * sample_rate)
b, a = sig.butter(6, wn)
filtered = sig.filtfilt(b, a, data)[:]
# resample the data, if requested.
if 0 < out_sample_rate < sample_rate:
n_out = int(np.ceil(n_points * out_sample_rate / sample_rate))
filtered = sig.resample(filtered, num=n_out)
final_sample_rate = sample_rate * (1.0 * n_out / n_points)
save_file = data_file.open_file(output_filename,
mode='w',
sample_rate=final_sample_rate,
n_points=filtered.size)
save_file.root.data[:] = filtered[:]
save_file.flush()
save_file.close()
return output_filename
def run(self):
if not self.filename == '' or self.plot_options['datadict'] == '':
reader = get_reader_from_filename(self.filename)
self.plot_options['data'] = reader.get_all_data(self.decimate)
if self.sample_rate == 0.0:
self.sample_rate = reader.get_sample_rate()
reader.close()
if self.cancelled:
return
self.plot_options['sample_rate'] = self.sample_rate
self.dataReady.emit({'plot_options': self.plot_options, 'status_text': '', 'thread': self})
def test_original_files_unmodified(self, filename):
file1 = tf.get_abs_path('chimera_1event_2levels.log')
file2 = tf.get_abs_path('chimera_1event.log')
reader1 = get_reader_from_filename(file1)
reader2 = get_reader_from_filename(file2)
sample_rate1_orig = reader1.get_sample_rate()
sample_rate2_orig = reader2.get_sample_rate()
data1_orig = reader1.get_all_data()[0]
data2_orig = reader2.get_all_data()[0]
reader1.close()
reader2.close()
concat_files([file1, file2], output_filename=filename)
reader1 = get_reader_from_filename(file1)
reader2 = get_reader_from_filename(file2)
reader_out = get_reader_from_filename(filename)
sample_rate1_final = reader1.get_sample_rate()
sample_rate2_final = reader2.get_sample_rate()
self.assertEqual(sample_rate1_final, sample_rate1_orig,
"Sample rate changed. Was {0}, now {1}.".format(sample_rate1_orig, sample_rate1_final))
self.assertEqual(sample_rate2_final, sample_rate2_orig,
"Sample rate changed. Was {0}, now {1}.".format(sample_rate2_orig, sample_rate2_final))
data1 = reader1.get_all_data()[0]
data2 = reader2.get_all_data()[0]
np.testing.assert_array_equal(data1, data1_orig)
np.testing.assert_array_equal(data2, data2_orig)
reader1.close()
reader2.close()
reader_out.close()
def test_set_output_sample_rate(self, filename):
"""
Tests that we can successfully set the output sample rate, and the number of data points changes correctly.
"""
data_file_names = [tf.get_abs_path('chimera_1event.log'),
tf.get_abs_path('chimera_1event_2levels.log')]
for data_filename in data_file_names:
# Open a reader and read the original sample rate
orig_reader = get_reader_from_filename(data_filename)
orig_sample_rate = orig_reader.get_sample_rate()
orig_data = orig_reader.get_all_data()
n_orig = orig_data[0].size
orig_reader.close()
for out_sample_rate in (100.e4, 1.e6, 5.e5):
out_filename = filter_file(data_filename, 10.e4, out_sample_rate, output_filename=filename)
# The output number of data points should be int(np.ceil(n_orig * out_sample_rate / orig_sample_rate))
n_out = int(np.ceil(n_orig * out_sample_rate / orig_sample_rate))
# The output sample rate should be set by n_out
out_sample_rate = orig_sample_rate * (1.0 * n_out) / n_orig
# Get the params from the output file
out_reader = get_reader_from_filename(out_filename)
o_s_r = out_reader.get_sample_rate()
out_data = out_reader.get_all_data()
n_o = out_data[0].size
out_reader.close()
self.assertEqual(n_out, n_o,
"Number of re-sample points not correct. "
"Original data {0}, output {1}, should be {2}.".format(n_orig, n_o, n_out))
self.assertAlmostEqual(o_s_r, out_sample_rate, 2,
"Sample rate not set correctly. Was {0}, should be {1}".format(o_s_r,
out_sample_rate))
os.remove(out_filename)
def _test_params_equality(self, filename, data_should_be, sample_rate):
"""
Tests that the data and sample_rate in filename equal data_should_be and sample_rate, respectively.
"""
reader = get_reader_from_filename(filename)
data_all = reader.get_all_data()
out_data = data_all[0]
out_sample_rate = reader.get_sample_rate()
reader.close()
self.assertEqual(sample_rate, out_sample_rate,
"Sample rates not equal. Wanted {0}, got {1}.".format(sample_rate, out_sample_rate))
np.testing.assert_array_equal(out_data, data_should_be)
def run(self):
if not self.filename == '' or self.plot_options['datadict'] == '':
reader = get_reader_from_filename(self.filename)
self.plot_options['data'] = reader.get_all_data(self.decimate)
if self.sample_rate == 0.0:
self.sample_rate = reader.get_sample_rate()
reader.close()
if self.cancelled:
return
self.plot_options['sample_rate'] = self.sample_rate
self.dataReady.emit({
'plot_options': self.plot_options,
'status_text': '',
'thread': self
})
def filter_file(filename, filter_frequency, out_sample_rate, output_filename=None):
"""
Reads data from the filename file and uses a Butterworth low-pass filter with cutoff at filter_frequency. Outputs
the filtered waveform to a new :py:class:`pypore.filetypes.data_file.DataFile`.
:param StringType filename: Filename containing data to be filtered.
:param DoubleType filter_frequency: Cutoff frequency for the low-pass Butterworth filter.
:param DoubleType out_sample_rate: After the data is filtered, it can be resampled to roughly out_sample_rate. If \
out_sample_rate <= 0, the data will not be resampled.
:param StringType output_filename: (Optional) Filename for the filtered data. If not specified,
for an example filename='test.mat', the default output_filename would be 'test_filtered.h5'
:returns: StringType -- The output filename of the filtered data.
Usage:
>>> import pypore.file_converter as fC
>>> fC.filter_file("filename", 1.e4, 1.e5, "output.h5") // filter at 10kHz, resample at 100kHz
"""
reader = get_reader_from_filename(filename)
data = reader.get_all_data()[0]
sample_rate = reader.get_sample_rate()
final_sample_rate = sample_rate
n_points = len(data)
if output_filename is None:
output_filename = filename.split('.')[0] + '_filtered.h5'
# wn is a fraction of the Nyquist frequency (half the sampling frequency).
wn = filter_frequency / (0.5 * sample_rate)
b, a = sig.butter(6, wn)
filtered = sig.filtfilt(b, a, data)[:]
# resample the data, if requested.
if 0 < out_sample_rate < sample_rate:
n_out = int(np.ceil(n_points * out_sample_rate / sample_rate))
filtered = sig.resample(filtered, num=n_out)
final_sample_rate = sample_rate * (1.0*n_out/n_points)
save_file = data_file.open_file(output_filename, mode='w', sample_rate=final_sample_rate, n_points=filtered.size)
save_file.root.data[:] = filtered[:]
save_file.flush()
save_file.close()
return output_filename
def _test_out_sample_rate_data_len_equality(self, orig_data, orig_sample_rate, out_filename, sample_rate):
out_reader = get_reader_from_filename(out_filename)
out_sample_rate = out_reader.get_sample_rate()
out_data = out_reader.get_all_data()
out_reader.close()
self.assertAlmostEqual(orig_sample_rate, out_sample_rate, 2,
msg="Sampling rate changed during filter_file, "
"when it should not. Was {0}, wanted {1}, got {2}".format(
orig_sample_rate, sample_rate, out_sample_rate))
self.assertEqual(len(orig_data), len(orig_data),
msg="Filtering changed the number of channels. Was {0}, output {1}".format(len(orig_data),
len(out_data)))
self.assertEqual(orig_data[0].size, out_data[0].size,
msg="Output data size doesn't match the original. Was {0}, output {1}.".format(
orig_data[0].size, out_data[0].size))
return out_data
def _test_params_equality(self, filename, data_should_be, sample_rate):
"""
Tests that the data and sample_rate in filename equal data_should_be and sample_rate, respectively.
"""
reader = get_reader_from_filename(filename)
data_all = reader.get_all_data()
out_data = data_all[0]
out_sample_rate = reader.get_sample_rate()
reader.close()
self.assertEqual(
sample_rate, out_sample_rate,
"Sample rates not equal. Wanted {0}, got {1}.".format(
sample_rate, out_sample_rate))
np.testing.assert_array_equal(out_data, data_should_be)
def test_same_sample_rate_no_change(self, filename):
"""
Tests that if we set the output sample rate to < 0, the sampling doesn't change, but the file is filtered.
"""
data_filename = tf.get_abs_path('chimera_1event.log')
# Open a reader and read the original sample rate
orig_reader = get_reader_from_filename(data_filename)
orig_sample_rate = orig_reader.get_sample_rate()
orig_data = orig_reader.get_all_data()
orig_reader.close()
for sample_rate in (-1, 0., orig_sample_rate, orig_sample_rate + 100.):
# filter the file, with negative sample rate
out_filename = filter_file(data_filename, 10.e4, sample_rate, output_filename=filename)
self._test_out_sample_rate_data_len_equality(orig_data, orig_sample_rate, out_filename, sample_rate)
os.remove(out_filename)
def test_noise_no_events(self, filename):
"""
Tests that noise is added correctly.
"""
sample_rate = 1.e5
baseline = 1.0
n = 100000
noise_scale = 1.
create_specified_data(filename=filename,
n=n,
sample_rate=sample_rate,
baseline=baseline,
noise_scale=noise_scale,
events=[])
reader = get_reader_from_filename(filename)
data_all = reader.get_all_data()
data = data_all[0]
reader.close()
self.assertEqual(
n, data.size,
"Unexpected array size. Wanted {0}, got {1}.".format(n, data.size))
decimal = 1
mean = np.mean(data)
self.assertAlmostEqual(
baseline, mean, decimal,
"Unexpected baseline. Wanted {0}, got {1}. Tested to {2} decimals."
.format(baseline, mean, decimal))
decimal = 1
std_dev = np.std(data)
self.assertAlmostEqual(
noise_scale,
std_dev,
decimal,
msg="Unexpected stddev. Wanted{0}, got {1}. Tested to {2} decimals."
.format(noise_scale, std_dev, decimal))
def open_data_files(self, file_names=None):
"""
Analyzes the files for correctness, then adds them to the list widget.
:param ListType<StringType> file_names: The file names to be included in the list widget. If not included,
this function will use a QtGui.QFileDialog.getOpenFileNames to open
files.
:returns: BooleanType -- **True** if files were opened, **False** otherwise.
"""
if file_names is None:
file_names = QtGui.QFileDialog.getOpenFileNames(self,
'Open data file',
'.',
"All types(*.h5 *.hkd *.log *.mat);;"
"Pypore data files *.h5(*.h5);;"
"Heka files *.hkd(*.hkd);;"
"Chimera files *.log(*.log);;Gabys files *.mat(*.mat)")[0]
if len(file_names) > 0:
self.file_list_widget.clear()
else:
return
are_files_opened = False
open_dir = None
for w in file_names:
reader = get_reader_from_filename(w)
# if 'error' in params:
# TODO implement error handling in readers
# pass
# else:
reader.close()
are_files_opened = True
item = FileListItem(w)
open_dir = item.get_directory()
self.file_list_widget.addItem(item)
return are_files_opened
def concat_files(files, output_filename=None):
"""
This function concatenates multiple files into one data file. All of the sampling rates of the original files
must be the same.
:param list files: List of string file names OR
:py:class:`Readers <pypore.i_o.abstract_reader.AbstractReader>`.
:param output_filename: Optional file name for the resulting file.
:raises: :py:exc:`ValueError` -- if the length of the files list is < 2.
:raises: :py:exc:`SamplingRatesMismatchError <pypore.file_converter.SamplingRatesMismatchError>` -- if the sampling
rates do not match in all of the files.
>>> from pypore.i_o.data_file_reader import DataFileReader
>>> concat_files(['file1.log', DataFileReader('dataFile.h5')], output_filename='concatenated.h5') # can pass strings or Readers
"""
if len(files) < 2:
raise ValueError("Minimum length of files list is 2.")
# Get the first sample rate
should_close_reader = False
reader = files[0]
if not isinstance(reader, AbstractReader):
reader = get_reader_from_filename(reader)
should_close_reader = True
sample_rate = reader.get_sample_rate()
if output_filename is None:
basename = os.path.basename(reader.get_filename())
output_filename = basename.split(
'.')[0] + '_concatenated_' + datetime.datetime.now().strftime(
"%Y%m%d_%H%M%S") + '.h5'
if should_close_reader:
reader.close()
n = 0
# Get the total number of data points, and check that the sampling rates are equal.
for i, reader in enumerate(files):
should_close_reader = False
# If it's not already a reader
if not isinstance(reader, AbstractReader):
reader = get_reader_from_filename(reader)
should_close_reader = True
curr_sample_rate = reader.get_sample_rate()
if curr_sample_rate != sample_rate:
raise SamplingRatesMismatchError(
"Sampling rates differ in files. Found {0} and {1}.".format(
curr_sample_rate, sample_rate))
n += reader.get_all_data()[0].size
if should_close_reader:
reader.close()
# Open a new data file.
new_data_file = df.open_file(output_filename,
mode='w',
n_points=n,
sample_rate=sample_rate)
curr_i = 0
for i, reader in enumerate(files):
should_close_reader = False
# If it's not already a reader
if not isinstance(reader, AbstractReader):
reader = get_reader_from_filename(reader)
should_close_reader = True
n_i = reader.get_points_per_channel_total()
new_data_file.root.data[curr_i:curr_i + n_i] = reader.get_all_data()[0]
curr_i += n_i
if should_close_reader:
reader.close()
new_data_file.close()
return output_filename
