def test_yes_tau(self):
# should usually drop some rows
reader = CSVReader(schema, df)
private_reader = PrivateQuery(reader, schema, 0.01)
lengths = []
for i in range(10):
rs = private_reader.execute_typed("SELECT COUNT(*) AS c FROM PUMS.PUMS WHERE age > 90 GROUP BY educ")
lengths.append(len(rs['c']))
l = lengths[0]
print(lengths)
assert(any([l != ll for ll in lengths]))
def run_agg_query(self, df, metadata_path, query, confidence):
metadata = MetadataLoader(metadata_path).read_schema()
reader = CSVReader(metadata, df)
private_reader = PrivateQuery(reader, metadata, self.epsilon)
query_ast = private_reader.parse_query_string(query)
subquery, query, syms, types, sens, srs_orig = private_reader._preprocess(
query_ast)
#exact_values = private_reader.execute_ast(query)
#bounds_centered_zero = list(private_reader._apply_noise(*exact_values, confidence)[1].values())[1]
#actual_value = exact_values[1:][0][1]
#bounds = np.array([bounds_centered_zero[0] + actual_value, bounds_centered_zero[1] + actual_value])
noisy_values = []
for idx in range(self.repeat_count):
srs = TypedRowset(srs_orig.rows(), types, sens)
noisy_values.append(
private_reader._postprocess(subquery, query, syms, types, sens,
srs).rows()[1:][0][0])
return np.array(noisy_values) #, bounds
def test_count_exact(self):
reader = CSVReader(schema, df)
rs = reader.execute("SELECT COUNT(*) AS c FROM PUMS.PUMS")
assert (rs[1][0] == 1000)
def _load_reader(dataset_document):
return CSVReader(DataverseAdapter.load_metadata(dataset_document),
DataverseAdapter.load_df(dataset_document))
def test_sum_no_rows_exact_typed(self):
reader = CSVReader(schema, df)
query = QueryParser(schema).queries("SELECT SUM(age) as c FROM PUMS.PUMS WHERE age > 100")[0]
trs = reader.execute_typed(query)
assert(trs['c'][0] == None)
def test_group_by_noisy_typed_order_desc(self):
reader = CSVReader(schema, df)
private_reader = PrivateQuery(reader, schema, 1.0)
rs = private_reader.execute_typed("SELECT COUNT(*) AS c, married AS m FROM PUMS.PUMS GROUP BY married ORDER BY c DESC")
assert(rs['c'][0] > rs['c'][1])
def test_group_by_noisy_order(self):
reader = CSVReader(schema, df)
private_reader = PrivateQuery(reader, schema, 1.0)
rs = private_reader.execute("SELECT COUNT(*) AS c, married AS m FROM PUMS.PUMS GROUP BY married ORDER BY c")
assert(rs[1][0] < rs[2][0])
def test_group_by_exact_order_expr_desc(self):
reader = CSVReader(schema, df)
rs = reader.execute("SELECT COUNT(*) * 5 AS c, married AS m FROM PUMS.PUMS GROUP BY married ORDER BY c DESC")
assert(rs[1][0] == 549 * 5)
assert(rs[2][0] == 451 * 5)
def test_empty_result_typed(self):
reader = CSVReader(schema, df)
rs = reader.execute("SELECT age as a FROM PUMS.PUMS WHERE age > 100")
trs = TypedRowset(rs, ['int'], [None])
assert(len(trs) == 0)
def test_empty_result(self):
reader = CSVReader(schema, df)
rs = reader.execute("SELECT age as a FROM PUMS.PUMS WHERE age > 100")
print(rs)
assert(len(rs) == 1)
def _load_reader(dataset_document):
return CSVReader(LocalCSVAdapter.load_metadata(dataset_document),
LocalCSVAdapter.load_df(dataset_document))