def test_read_split_file_invalid_input(self):
"""Test parsing invalid input raises an error."""
with self.assertRaises(FeatureMapFileFormatError):
_ = read_split_file(self.split_dup_groups_lines)
with self.assertRaises(FeatureMapFileFormatError):
_ = read_split_file(self.split_dup_objs_lines)
def build_problem_data(group_map_files, mapping_file, prediction_field,
start_level, include_only, negate, n_processes):
#For each scope, build a map from group to object and vice versa
group_to_object = []
object_to_group = []
for map_file in group_map_files:
g_to_o, o_to_g = read_split_file(map_file)
group_to_object.append(g_to_o)
object_to_group.append(o_to_g)
#Find a list of sample names from our group names
#An alternative is 'samplenames = samplemap.keys()', but that may have records without features
samplenames = set()
for grp in group_to_object[start_level]:
l = group_to_object[start_level][grp]
for obj in l:
samplenames.add(parse_object_string_sample(obj))
samplenames = list(samplenames)
#get a map of sample name to it's properties
samplemap = read_mapping_file(mapping_file)
sample_to_response = {}
for samplename in samplenames:
if (include_only is None or
((samplemap[samplename][include_only[0]] in include_only[1]) ^ negate)):
sample_to_response[samplename] = samplemap[samplename][prediction_field]
problem_data = ProblemData(group_to_object, object_to_group, sample_to_response, n_processes)
feature_vector = FeatureVector([FeatureRecord(group, start_level,
len(group_to_object[start_level][group]))
for group in group_to_object[start_level].keys()])
return problem_data, feature_vector
def test_read_split_file(self):
"""Test parsing a feature-to-object map."""
exp = ({'F1': ['A', 'B'], 'F2': ['C'], 'F3': ['D', 'E', 'F']},
{'A': 'F1', 'C': 'F2', 'B': 'F1', 'E': 'F3', 'D': 'F3',
'F': 'F3'})
obs = read_split_file(self.split1_lines)
self.assertEqual(obs, exp)
def build_problem_data(
group_map_files,
mapping_file,
prediction_field,
start_level,
include_only,
negate,
n_processes,
parse_object_string=parse_object_string_sample,
):
simple_var_types = [
("n_processes", types.IntType),
("start_level", types.IntType),
("negate", types.BooleanType),
("prediction_field", types.StringType),
("include_only", (types.NoneType, types.ListType, types.TupleType)),
("group_map_files", types.ListType),
]
for var_name, var_type in simple_var_types:
check_input_type(var_name, locals()[var_name], var_type)
if include_only != None:
if not isinstance(include_only[0], types.StringType):
raise InputTypeError("include_only[0] should be of type string")
if not isinstance(include_only[1], types.ListType) or not all(
[isinstance(value, types.StringType) for value in include_only[1]]
):
raise InputTypeError("include_only[1] should be a list of strings")
if start_level >= len(group_map_files) or start_level < 0:
raise InputTypeError(
"start_level (%s) is not a valid scope index; group_map_files is of length %s"
% (start_level, len(group_map_files))
)
# For each scope, build a map from group to object and vice versa
group_to_object = []
object_to_group = []
for map_file in group_map_files:
g_to_o, o_to_g = read_split_file(map_file)
group_to_object.append(g_to_o)
object_to_group.append(o_to_g)
assert isinstance(g_to_o, types.DictType), "read_split_file did not return a dict type"
assert isinstance(o_to_g, types.DictType), "read_split_file did not return a dict type"
# Find a list of sample names from our group names
# An alternative is 'samplenames = samplemap.keys()', but that may have records without features
samplename_set = set()
for grp in group_to_object[start_level]:
objs = group_to_object[start_level][grp]
for obj in objs:
samplename = parse_object_string(obj)
samplename_set.add(samplename)
samplenames = list(samplename_set)
# get a map of sample name to it's properties
samplemap = read_mapping_file(mapping_file)
def include_samplename(samplename):
if include_only == None:
return True
sample_dict = samplemap[samplename]
try:
if (sample_dict[include_only[0]] in include_only[1]) ^ negate:
return True
return False
except KeyError:
raise KeyError("include_only[0] is not a field in mapping_file")
sample_to_response = {}
for samplename in samplenames:
if include_samplename(samplename):
sample_fields = None
try:
sample_fields = samplemap[samplename]
except KeyError:
raise KeyError(
"A sample name (%s) found in the group files is not a sample in mapping_file." % samplename
)
try:
sample_to_response[samplename] = sample_fields[prediction_field]
except KeyError:
raise KeyError("prediction_field is not a field in mapping_file.")
problem_data = ProblemData(group_to_object, object_to_group, sample_to_response, n_processes, parse_object_string)
return problem_data
