def load(cls, input_file, md5sum=None, sha256sum=None, force=False):
"""Loads a classifier from a file"""
logger = logging.getLogger(__name__)
filename = try_and_get_filename(input_file)
if (md5sum, sha256sum, force) == (None, None, False):
message = "Loading unknown pickled objects is dangerous. Either \
provide an md5 or sha256 for '%s' or explicitly 'force' if you want to \
live dangerously" % filename
logger.error(message)
raise ValueError(message)
(actual_md5sum, actual_sha256sum) = cls._calculate_hashes(input_file)
if not md5sum is None:
if actual_md5sum != md5sum:
message = "Expected md5sum of '%s' to be '%s' but got '%s'" % (
filename, md5sum, actual_md5sum)
if force:
logger.warn("FORCED so ignoring: %s" % message)
else:
logger.error(message)
raise ValueError(message)
if not sha256sum is None:
if actual_sha256sum != sha256sum:
message = "Expected sha256sum of '%s' to be '%s' but got '%s'" % (
filename, sha256sum, actual_sha256sum)
if force:
logger.warn("FORCED so ignoring: %s" % message)
else:
logger.error(message)
raise ValueError(message)
input_file.seek(0)
classifier, feature_labels = pickle.load(input_file)
return SampleClassifier(classifier, feature_labels)
def load(cls, input_file, md5sum=None, sha256sum=None, force=False):
"""Loads a classifier from a file"""
logger = logging.getLogger(__name__)
filename = try_and_get_filename(input_file)
if (md5sum, sha256sum, force) == (None, None, False):
message = "Loading unknown pickled objects is dangerous. Either \
provide an md5 or sha256 for '%s' or explicitly 'force' if you want to \
live dangerously" % filename
logger.error(message)
raise ValueError(message)
(actual_md5sum, actual_sha256sum) = cls._calculate_hashes(input_file)
if not md5sum is None:
if actual_md5sum != md5sum:
message = "Expected md5sum of '%s' to be '%s' but got '%s'" % (filename,
md5sum,
actual_md5sum)
if force:
logger.warn("FORCED so ignoring: %s" % message)
else:
logger.error(message)
raise ValueError(message)
if not sha256sum is None:
if actual_sha256sum != sha256sum:
message = "Expected sha256sum of '%s' to be '%s' but got '%s'" % (filename,
sha256sum,
actual_sha256sum)
if force:
logger.warn("FORCED so ignoring: %s" % message)
else:
logger.error(message)
raise ValueError(message)
input_file.seek(0)
classifier, feature_labels = pickle.load(input_file)
return SampleClassifier(classifier, feature_labels)
def _extract_features(self, features_file):
"""Parses a feature file and returns the sample_names and features"""
features_file.seek(0)
filename = try_and_get_filename(features_file)
feature_file_lines = [line for line in csv.reader(features_file)]
data = np.array(feature_file_lines)
try:
number_of_rows, number_of_columns = data.shape
# shape returns a 2 element tuple if there are a consistent number of
# columns. Otherwise it returns a 1 element tuple which would trigger a
# ValueError.
except ValueError as e:
raise ValueError(
"Issue parsing '%s', some samples have more features than others"
% filename)
feature_labels = data[0, 1:]
sample_names = data[1:, 0]
str_to_int = np.vectorize(int)
try:
features = str_to_int(data[1:, 1:])
except ValueError:
raise ValueError(
"Issue parsing '%s', expected features to be 0 or 1" %
filename)
if not self._check_features_binary(features):
raise ValueError(
"Issue parsing '%s', expected features to be 0 or 1" %
filename)
return sample_names, features
def write_vcf_file(self, vcf_output_file):
reader = self._get_records_from_vcf()
reader.add_GT_format_header(
) # This might not be true but it doesn't hurt
writer = vcf.Writer(vcf_output_file, reader)
filename = try_and_get_filename(vcf_output_file)
self.logger.info("Writing VCF records to '%s'" % filename)
for record in reader:
writer.write_record(record)
def export(self, output_file):
"""Writes the classifier to a file"""
filename = try_and_get_filename(output_file)
self.logger.info("Writing classifier to '%s'" % filename)
pickle.dump((self.classifier, self.feature_labels), output_file)
output_file.flush()
with open(filename, 'rb') as hash_check_file:
(md5sum, sha256sum) = self._calculate_hashes(hash_check_file)
self.logger.info("Wrote classifier with md5sum '%s' to '%s'" % (md5sum,
filename))
self.logger.info("Wrote classifier with sha256sum '%s' to '%s'" % (sha256sum,
filename))
def export(self, output_file):
"""Writes the classifier to a file"""
filename = try_and_get_filename(output_file)
self.logger.info("Writing classifier to '%s'" % filename)
pickle.dump((self.classifier, self.feature_labels), output_file)
output_file.flush()
with open(filename, 'rb') as hash_check_file:
(md5sum, sha256sum) = self._calculate_hashes(hash_check_file)
self.logger.info("Wrote classifier with md5sum '%s' to '%s'" %
(md5sum, filename))
self.logger.info("Wrote classifier with sha256sum '%s' to '%s'" %
(sha256sum, filename))
def get_feature_labels_from_file(self, feature_file):
"""Parses a file of features and returns the feature labels
Some files have more features others don't. This is a useful
pre-computation step which extracts the feature labels used in one file
which can be used to filter the features in another"""
feature_file.seek(0)
features_line = csv.reader(feature_file).next()
if features_line[0] != 'Features':
filename = try_and_get_filename(feature_file)
raise ValueError("Issue parsing '%s', expected first element to be 'Features'" % filename)
feature_labels = features_line[1:]
return np.array(feature_labels)
def get_feature_labels_from_file(self, feature_file):
"""Parses a file of features and returns the feature labels
Some files have more features others don't. This is a useful
pre-computation step which extracts the feature labels used in one file
which can be used to filter the features in another"""
feature_file.seek(0)
features_line = csv.reader(feature_file).next()
if features_line[0] != 'Features':
filename = try_and_get_filename(feature_file)
raise ValueError(
"Issue parsing '%s', expected first element to be 'Features'" %
filename)
feature_labels = features_line[1:]
return np.array(feature_labels)
def _extract_features(self, features_file):
"""Parses a feature file and returns the sample_names and features"""
features_file.seek(0)
filename = try_and_get_filename(features_file)
feature_file_lines = [line for line in csv.reader(features_file)]
data = np.array(feature_file_lines)
try:
number_of_rows, number_of_columns = data.shape
# shape returns a 2 element tuple if there are a consistent number of
# columns. Otherwise it returns a 1 element tuple which would trigger a
# ValueError.
except ValueError as e:
raise ValueError("Issue parsing '%s', some samples have more features than others" % filename)
feature_labels = data[0,1:]
sample_names = data[1:,0]
str_to_int = np.vectorize(int)
try:
features = str_to_int(data[1:,1:])
except ValueError:
raise ValueError("Issue parsing '%s', expected features to be 0 or 1" % filename)
if not self._check_features_binary(features):
raise ValueError("Issue parsing '%s', expected features to be 0 or 1" % filename)
return sample_names, features
