|
20 | 20 | import itertools
|
21 | 21 |
|
22 | 22 | from absl.testing import absltest
|
23 |
| -from absl.testing import parameterized |
24 | 23 | import numpy as np
|
25 | 24 | import six
|
26 | 25 | from tensorflow_data_validation import types
|
27 | 26 | from tensorflow_data_validation.arrow import arrow_util
|
28 | 27 | from tensorflow_data_validation.pyarrow_tf import pyarrow as pa
|
29 | 28 |
|
30 | 29 |
|
31 |
| -class ArrowUtilTest(absltest.TestCase): |
32 |
| - |
33 |
| - def test_invalid_input_type(self): |
34 |
| - |
35 |
| - functions_expecting_list_array = [ |
36 |
| - arrow_util.ListLengthsFromListArray, |
37 |
| - arrow_util.GetFlattenedArrayParentIndices, |
38 |
| - ] |
39 |
| - functions_expecting_array = [arrow_util.GetArrayNullBitmapAsByteArray] |
40 |
| - functions_expecting_binary_array = [arrow_util.GetBinaryArrayTotalByteSize] |
41 |
| - for f in itertools.chain(functions_expecting_list_array, |
42 |
| - functions_expecting_array, |
43 |
| - functions_expecting_binary_array): |
44 |
| - with self.assertRaisesRegex(RuntimeError, "Could not unwrap Array"): |
45 |
| - f(1) |
46 |
| - |
47 |
| - for f in functions_expecting_list_array: |
48 |
| - with self.assertRaisesRegex(RuntimeError, "Expected ListArray but got"): |
49 |
| - f(pa.array([1, 2, 3])) |
50 |
| - |
51 |
| - for f in functions_expecting_binary_array: |
52 |
| - with self.assertRaisesRegex(RuntimeError, "Expected BinaryArray"): |
53 |
| - f(pa.array([[1, 2, 3]])) |
54 |
| - |
55 |
| - def test_list_lengths(self): |
56 |
| - list_lengths = arrow_util.ListLengthsFromListArray( |
57 |
| - pa.array([], type=pa.list_(pa.int64()))) |
58 |
| - self.assertTrue(list_lengths.equals(pa.array([], type=pa.int32()))) |
59 |
| - list_lengths = arrow_util.ListLengthsFromListArray( |
60 |
| - pa.array([[1., 2.], [], [3.]])) |
61 |
| - self.assertTrue(list_lengths.equals(pa.array([2, 0, 1], type=pa.int32()))) |
62 |
| - list_lengths = arrow_util.ListLengthsFromListArray( |
63 |
| - pa.array([[1., 2.], None, [3.]])) |
64 |
| - self.assertTrue(list_lengths.equals(pa.array([2, 0, 1], type=pa.int32()))) |
65 |
| - |
66 |
| - def test_get_array_null_bitmap_as_byte_array(self): |
67 |
| - array = pa.array([], type=pa.int32()) |
68 |
| - null_masks = arrow_util.GetArrayNullBitmapAsByteArray(array) |
69 |
| - self.assertTrue(null_masks.equals(pa.array([], type=pa.uint8()))) |
70 |
| - |
71 |
| - array = pa.array([1, 2, None, 3, None], type=pa.int32()) |
72 |
| - null_masks = arrow_util.GetArrayNullBitmapAsByteArray(array) |
73 |
| - self.assertTrue( |
74 |
| - null_masks.equals(pa.array([0, 0, 1, 0, 1], type=pa.uint8()))) |
75 |
| - |
76 |
| - array = pa.array([1, 2, 3]) |
77 |
| - null_masks = arrow_util.GetArrayNullBitmapAsByteArray(array) |
78 |
| - self.assertTrue(null_masks.equals(pa.array([0, 0, 0], type=pa.uint8()))) |
79 |
| - |
80 |
| - array = pa.array([None, None, None], type=pa.int32()) |
81 |
| - null_masks = arrow_util.GetArrayNullBitmapAsByteArray(array) |
82 |
| - self.assertTrue(null_masks.equals(pa.array([1, 1, 1], type=pa.uint8()))) |
83 |
| - # Demonstrate that the returned array can be converted to a numpy boolean |
84 |
| - # array w/o copying |
85 |
| - np.testing.assert_equal( |
86 |
| - np.array([True, True, True]), null_masks.to_numpy().view(np.bool)) |
87 |
| - |
88 |
| - def test_get_flattened_array_parent_indices(self): |
89 |
| - indices = arrow_util.GetFlattenedArrayParentIndices( |
90 |
| - pa.array([], type=pa.list_(pa.int32()))) |
91 |
| - self.assertTrue(indices.equals(pa.array([], type=pa.int32()))) |
92 |
| - |
93 |
| - indices = arrow_util.GetFlattenedArrayParentIndices( |
94 |
| - pa.array([[1.], [2.], [], [3.]])) |
95 |
| - self.assertTrue(indices.equals(pa.array([0, 1, 3], type=pa.int32()))) |
96 |
| - |
97 |
| - def test_get_binary_array_total_byte_size(self): |
98 |
| - binary_array = pa.array([b"abc", None, b"def", b"", b"ghi"]) |
99 |
| - self.assertEqual(9, arrow_util.GetBinaryArrayTotalByteSize(binary_array)) |
100 |
| - sliced_1_2 = binary_array.slice(1, 2) |
101 |
| - self.assertEqual(3, arrow_util.GetBinaryArrayTotalByteSize(sliced_1_2)) |
102 |
| - sliced_2 = binary_array.slice(2) |
103 |
| - self.assertEqual(6, arrow_util.GetBinaryArrayTotalByteSize(sliced_2)) |
104 |
| - |
105 |
| - unicode_array = pa.array([u"abc"]) |
106 |
| - self.assertEqual(3, arrow_util.GetBinaryArrayTotalByteSize(unicode_array)) |
107 |
| - |
108 |
| - empty_array = pa.array([], type=pa.binary()) |
109 |
| - self.assertEqual(0, arrow_util.GetBinaryArrayTotalByteSize(empty_array)) |
110 |
| - |
111 |
| - def _value_counts_struct_array_to_dict(self, value_counts): |
112 |
| - result = {} |
113 |
| - for value_count in value_counts: |
114 |
| - value_count = value_count.as_py() |
115 |
| - result[value_count["values"]] = value_count["counts"] |
116 |
| - return result |
117 |
| - |
118 |
| - def test_value_counts_binary(self): |
119 |
| - binary_array = pa.array([b"abc", b"ghi", b"def", b"ghi", b"ghi", b"def"]) |
120 |
| - expected_result = {b"abc": 1, b"ghi": 3, b"def": 2} |
121 |
| - self.assertDictEqual(self._value_counts_struct_array_to_dict( |
122 |
| - arrow_util.ValueCounts(binary_array)), expected_result) |
123 |
| - |
124 |
| - def test_value_counts_integer(self): |
125 |
| - int_array = pa.array([1, 4, 1, 3, 1, 4]) |
126 |
| - expected_result = {1: 3, 4: 2, 3: 1} |
127 |
| - self.assertDictEqual(self._value_counts_struct_array_to_dict( |
128 |
| - arrow_util.ValueCounts(int_array)), expected_result) |
129 |
| - |
130 |
| - def test_value_counts_empty(self): |
131 |
| - empty_array = pa.array([]) |
132 |
| - expected_result = {} |
133 |
| - self.assertDictEqual(self._value_counts_struct_array_to_dict( |
134 |
| - arrow_util.ValueCounts(empty_array)), expected_result) |
135 |
| - |
136 |
| -_MAKE_LIST_ARRAY_INVALID_INPUT_TEST_CASES = [ |
137 |
| - dict( |
138 |
| - testcase_name="invalid_parent_index", |
139 |
| - num_parents=None, |
140 |
| - parent_indices=np.array([0], dtype=np.int64), |
141 |
| - values=pa.array([1]), |
142 |
| - expected_error=RuntimeError, |
143 |
| - expected_error_regexp="Expected integer"), |
144 |
| - dict( |
145 |
| - testcase_name="parent_indices_not_np", |
146 |
| - num_parents=1, |
147 |
| - parent_indices=[0], |
148 |
| - values=pa.array([1]), |
149 |
| - expected_error=TypeError, |
150 |
| - expected_error_regexp="to be a numpy array" |
151 |
| - ), |
152 |
| - dict( |
153 |
| - testcase_name="parent_indices_not_1d", |
154 |
| - num_parents=1, |
155 |
| - parent_indices=np.array([[0]], dtype=np.int64), |
156 |
| - values=pa.array([1]), |
157 |
| - expected_error=TypeError, |
158 |
| - expected_error_regexp="to be a 1-D int64 numpy array" |
159 |
| - ), |
160 |
| - dict( |
161 |
| - testcase_name="parent_indices_not_int64", |
162 |
| - num_parents=1, |
163 |
| - parent_indices=np.array([0], dtype=np.int32), |
164 |
| - values=pa.array([1]), |
165 |
| - expected_error=TypeError, |
166 |
| - expected_error_regexp="to be a 1-D int64 numpy array" |
167 |
| - ), |
168 |
| - dict( |
169 |
| - testcase_name="parent_indices_length_not_equal_to_values_length", |
170 |
| - num_parents=1, |
171 |
| - parent_indices=np.array([0], dtype=np.int64), |
172 |
| - values=pa.array([1, 2]), |
173 |
| - expected_error=RuntimeError, |
174 |
| - expected_error_regexp="values array and parent indices array must be of the same length" |
175 |
| - ), |
176 |
| - dict( |
177 |
| - testcase_name="num_parents_too_small", |
178 |
| - num_parents=1, |
179 |
| - parent_indices=np.array([1], dtype=np.int64), |
180 |
| - values=pa.array([1]), |
181 |
| - expected_error=RuntimeError, |
182 |
| - expected_error_regexp="Found a parent index 1 while num_parents was 1" |
183 |
| - ) |
184 |
| -] |
185 |
| - |
186 |
| - |
187 |
| -_MAKE_LIST_ARRAY_TEST_CASES = [ |
188 |
| - dict( |
189 |
| - testcase_name="parents_are_all_empty", |
190 |
| - num_parents=5, |
191 |
| - parent_indices=np.array([], dtype=np.int64), |
192 |
| - values=pa.array([], type=pa.int64()), |
193 |
| - expected=pa.array([None, None, None, None, None], |
194 |
| - type=pa.list_(pa.int64()))), |
195 |
| - dict( |
196 |
| - testcase_name="long_num_parent", |
197 |
| - num_parents=(long(1) if six.PY2 else 1), |
198 |
| - parent_indices=np.array([0], dtype=np.int64), |
199 |
| - values=pa.array([1]), |
200 |
| - expected=pa.array([[1]]) |
201 |
| - ), |
202 |
| - dict( |
203 |
| - testcase_name="leading nones", |
204 |
| - num_parents=3, |
205 |
| - parent_indices=np.array([2], dtype=np.int64), |
206 |
| - values=pa.array([1]), |
207 |
| - expected=pa.array([None, None, [1]]), |
208 |
| - ), |
209 |
| - dict( |
210 |
| - testcase_name="same_parent_and_holes", |
211 |
| - num_parents=4, |
212 |
| - parent_indices=np.array([0, 0, 0, 3, 3], dtype=np.int64), |
213 |
| - values=pa.array(["a", "b", "c", "d", "e"]), |
214 |
| - expected=pa.array([["a", "b", "c"], None, None, ["d", "e"]]) |
215 |
| - ) |
216 |
| -] |
217 |
| - |
218 |
| - |
219 |
| -class MakeListArrayFromParentIndicesAndValuesTest(parameterized.TestCase): |
220 |
| - |
221 |
| - @parameterized.named_parameters(*_MAKE_LIST_ARRAY_INVALID_INPUT_TEST_CASES) |
222 |
| - def testInvalidInput(self, num_parents, parent_indices, values, |
223 |
| - expected_error, expected_error_regexp): |
224 |
| - with self.assertRaisesRegex(expected_error, expected_error_regexp): |
225 |
| - arrow_util.MakeListArrayFromParentIndicesAndValues( |
226 |
| - num_parents, parent_indices, values) |
227 |
| - |
228 |
| - @parameterized.named_parameters(*_MAKE_LIST_ARRAY_TEST_CASES) |
229 |
| - def testMakeListArray(self, num_parents, parent_indices, values, expected): |
230 |
| - actual = arrow_util.MakeListArrayFromParentIndicesAndValues( |
231 |
| - num_parents, parent_indices, values) |
232 |
| - self.assertTrue( |
233 |
| - actual.equals(expected), |
234 |
| - "actual: {}, expected: {}".format(actual, expected)) |
235 |
| - |
236 |
| - |
237 | 30 | class EnumerateArraysTest(absltest.TestCase):
|
238 | 31 |
|
239 | 32 | def testInvalidWeightColumnMissingValue(self):
|
|
0 commit comments