|
| 1 | +import datajoint as dj |
| 2 | +import timeit |
| 3 | +import numpy as np |
| 4 | +import uuid |
| 5 | +from . import schema |
| 6 | +from decimal import Decimal |
| 7 | +from datetime import datetime |
| 8 | +from datajoint.blob import pack, unpack |
| 9 | +from numpy.testing import assert_array_equal |
| 10 | +from nose.tools import ( |
| 11 | + assert_equal, |
| 12 | + assert_true, |
| 13 | + assert_false, |
| 14 | + assert_list_equal, |
| 15 | + assert_set_equal, |
| 16 | + assert_tuple_equal, |
| 17 | + assert_dict_equal, |
| 18 | +) |
| 19 | + |
| 20 | + |
| 21 | +def test_pack(): |
| 22 | + for x in ( |
| 23 | + 32, |
| 24 | + -3.7e-2, |
| 25 | + np.float64(3e31), |
| 26 | + -np.inf, |
| 27 | + np.int8(-3), |
| 28 | + np.uint8(-1), |
| 29 | + np.int16(-33), |
| 30 | + np.uint16(-33), |
| 31 | + np.int32(-3), |
| 32 | + np.uint32(-1), |
| 33 | + np.int64(373), |
| 34 | + np.uint64(-3), |
| 35 | + ): |
| 36 | + assert_equal(x, unpack(pack(x)), "Scalars don't match!") |
| 37 | + |
| 38 | + x = np.nan |
| 39 | + assert_true(np.isnan(unpack(pack(x))), "nan scalar did not match!") |
| 40 | + |
| 41 | + x = np.random.randn(8, 10) |
| 42 | + assert_array_equal(x, unpack(pack(x)), "Arrays do not match!") |
| 43 | + |
| 44 | + x = np.random.randn(10) |
| 45 | + assert_array_equal(x, unpack(pack(x)), "Arrays do not match!") |
| 46 | + |
| 47 | + x = 7j |
| 48 | + assert_equal(x, unpack(pack(x)), "Complex scalar does not match") |
| 49 | + |
| 50 | + x = np.float32(np.random.randn(3, 4, 5)) |
| 51 | + assert_array_equal(x, unpack(pack(x)), "Arrays do not match!") |
| 52 | + |
| 53 | + x = np.int16(np.random.randn(1, 2, 3)) |
| 54 | + assert_array_equal(x, unpack(pack(x)), "Arrays do not match!") |
| 55 | + |
| 56 | + x = None |
| 57 | + assert_true(unpack(pack(x)) is None, "None did not match") |
| 58 | + |
| 59 | + x = -255 |
| 60 | + y = unpack(pack(x)) |
| 61 | + assert_true( |
| 62 | + x == y and isinstance(y, int) and not isinstance(y, np.ndarray), |
| 63 | + "Scalar int did not match", |
| 64 | + ) |
| 65 | + |
| 66 | + x = -25523987234234287910987234987098245697129798713407812347 |
| 67 | + y = unpack(pack(x)) |
| 68 | + assert_true( |
| 69 | + x == y and isinstance(y, int) and not isinstance(y, np.ndarray), |
| 70 | + "Unbounded int did not match", |
| 71 | + ) |
| 72 | + |
| 73 | + x = 7.0 |
| 74 | + y = unpack(pack(x)) |
| 75 | + assert_true( |
| 76 | + x == y and isinstance(y, float) and not isinstance(y, np.ndarray), |
| 77 | + "Scalar float did not match", |
| 78 | + ) |
| 79 | + |
| 80 | + x = 7j |
| 81 | + y = unpack(pack(x)) |
| 82 | + assert_true( |
| 83 | + x == y and isinstance(y, complex) and not isinstance(y, np.ndarray), |
| 84 | + "Complex scalar did not match", |
| 85 | + ) |
| 86 | + |
| 87 | + x = True |
| 88 | + assert_true(unpack(pack(x)) is True, "Scalar bool did not match") |
| 89 | + |
| 90 | + x = [None] |
| 91 | + assert_list_equal(x, unpack(pack(x))) |
| 92 | + |
| 93 | + x = { |
| 94 | + "name": "Anonymous", |
| 95 | + "age": 15, |
| 96 | + 99: datetime.now(), |
| 97 | + "range": [110, 190], |
| 98 | + (11, 12): None, |
| 99 | + } |
| 100 | + y = unpack(pack(x)) |
| 101 | + assert_dict_equal(x, y, "Dict do not match!") |
| 102 | + assert_false( |
| 103 | + isinstance(["range"][0], np.ndarray), "Scalar int was coerced into array." |
| 104 | + ) |
| 105 | + |
| 106 | + x = uuid.uuid4() |
| 107 | + assert_equal(x, unpack(pack(x)), "UUID did not match") |
| 108 | + |
| 109 | + x = Decimal("-112122121.000003000") |
| 110 | + assert_equal(x, unpack(pack(x)), "Decimal did not pack/unpack correctly") |
| 111 | + |
| 112 | + x = [1, datetime.now(), {1: "one", "two": 2}, (1, 2)] |
| 113 | + assert_list_equal(x, unpack(pack(x)), "List did not pack/unpack correctly") |
| 114 | + |
| 115 | + x = (1, datetime.now(), {1: "one", "two": 2}, (uuid.uuid4(), 2)) |
| 116 | + assert_tuple_equal(x, unpack(pack(x)), "Tuple did not pack/unpack correctly") |
| 117 | + |
| 118 | + x = ( |
| 119 | + 1, |
| 120 | + {datetime.now().date(): "today", "now": datetime.now().date()}, |
| 121 | + {"yes!": [1, 2, np.array((3, 4))]}, |
| 122 | + ) |
| 123 | + y = unpack(pack(x)) |
| 124 | + assert_dict_equal(x[1], y[1]) |
| 125 | + assert_array_equal(x[2]["yes!"][2], y[2]["yes!"][2]) |
| 126 | + |
| 127 | + x = {"elephant"} |
| 128 | + assert_set_equal(x, unpack(pack(x)), "Set did not pack/unpack correctly") |
| 129 | + |
| 130 | + x = tuple(range(10)) |
| 131 | + assert_tuple_equal( |
| 132 | + x, unpack(pack(range(10))), "Iterator did not pack/unpack correctly" |
| 133 | + ) |
| 134 | + |
| 135 | + x = Decimal("1.24") |
| 136 | + assert_true(x == unpack(pack(x)), "Decimal object did not pack/unpack correctly") |
| 137 | + |
| 138 | + x = datetime.now() |
| 139 | + assert_true(x == unpack(pack(x)), "Datetime object did not pack/unpack correctly") |
| 140 | + |
| 141 | + x = np.bool_(True) |
| 142 | + assert_true(x == unpack(pack(x)), "Numpy bool object did not pack/unpack correctly") |
| 143 | + |
| 144 | + x = "test" |
| 145 | + assert_true(x == unpack(pack(x)), "String object did not pack/unpack correctly") |
| 146 | + |
| 147 | + x = np.array(["yes"]) |
| 148 | + assert_true( |
| 149 | + x == unpack(pack(x)), "Numpy string array object did not pack/unpack correctly" |
| 150 | + ) |
| 151 | + |
| 152 | + x = np.datetime64("1998").astype("datetime64[us]") |
| 153 | + assert_true(x == unpack(pack(x))) |
| 154 | + |
| 155 | + |
| 156 | +def test_recarrays(): |
| 157 | + x = np.array([(1.0, 2), (3.0, 4)], dtype=[("x", float), ("y", int)]) |
| 158 | + assert_array_equal(x, unpack(pack(x))) |
| 159 | + |
| 160 | + x = x.view(np.recarray) |
| 161 | + assert_array_equal(x, unpack(pack(x))) |
| 162 | + |
| 163 | + x = np.array([(3, 4)], dtype=[("tmp0", float), ("tmp1", "O")]).view(np.recarray) |
| 164 | + assert_array_equal(x, unpack(pack(x))) |
| 165 | + |
| 166 | + |
| 167 | +def test_object_arrays(): |
| 168 | + x = np.array(((1, 2, 3), True), dtype="object") |
| 169 | + assert_array_equal(x, unpack(pack(x)), "Object array did not serialize correctly") |
| 170 | + |
| 171 | + |
| 172 | +def test_complex(): |
| 173 | + z = np.random.randn(8, 10) + 1j * np.random.randn(8, 10) |
| 174 | + assert_array_equal(z, unpack(pack(z)), "Arrays do not match!") |
| 175 | + |
| 176 | + z = np.random.randn(10) + 1j * np.random.randn(10) |
| 177 | + assert_array_equal(z, unpack(pack(z)), "Arrays do not match!") |
| 178 | + |
| 179 | + x = np.float32(np.random.randn(3, 4, 5)) + 1j * np.float32(np.random.randn(3, 4, 5)) |
| 180 | + assert_array_equal(x, unpack(pack(x)), "Arrays do not match!") |
| 181 | + |
| 182 | + x = np.int16(np.random.randn(1, 2, 3)) + 1j * np.int16(np.random.randn(1, 2, 3)) |
| 183 | + assert_array_equal(x, unpack(pack(x)), "Arrays do not match!") |
| 184 | + |
| 185 | + |
| 186 | +def test_insert_longblob(): |
| 187 | + insert_dj_blob = {"id": 1, "data": [1, 2, 3]} |
| 188 | + schema.Longblob.insert1(insert_dj_blob) |
| 189 | + assert (schema.Longblob & "id=1").fetch1() == insert_dj_blob |
| 190 | + (schema.Longblob & "id=1").delete() |
| 191 | + |
| 192 | + query_mym_blob = {"id": 1, "data": np.array([1, 2, 3])} |
| 193 | + schema.Longblob.insert1(query_mym_blob) |
| 194 | + assert (schema.Longblob & "id=1").fetch1()["data"].all() == query_mym_blob[ |
| 195 | + "data" |
| 196 | + ].all() |
| 197 | + (schema.Longblob & "id=1").delete() |
| 198 | + |
| 199 | + query_32_blob = ( |
| 200 | + "INSERT INTO djtest_test1.longblob (id, data) VALUES (1, " |
| 201 | + "X'6D596D00530200000001000000010000000400000068697473007369646573007461736B73007374" |
| 202 | + "616765004D000000410200000001000000070000000600000000000000000000000000F8FF00000000" |
| 203 | + "0000F03F000000000000F03F0000000000000000000000000000F03F00000000000000000000000000" |
| 204 | + "00F8FF230000004102000000010000000700000004000000000000006C006C006C006C00720072006C" |
| 205 | + "0023000000410200000001000000070000000400000000000000640064006400640064006400640025" |
| 206 | + "00000041020000000100000008000000040000000000000053007400610067006500200031003000')" |
| 207 | + ) |
| 208 | + dj.conn().query(query_32_blob).fetchall() |
| 209 | + dj.blob.use_32bit_dims = True |
| 210 | + assert (schema.Longblob & "id=1").fetch1() == { |
| 211 | + "id": 1, |
| 212 | + "data": np.rec.array( |
| 213 | + [ |
| 214 | + [ |
| 215 | + ( |
| 216 | + np.array([[np.nan, 1.0, 1.0, 0.0, 1.0, 0.0, np.nan]]), |
| 217 | + np.array(["llllrrl"], dtype="<U7"), |
| 218 | + np.array(["ddddddd"], dtype="<U7"), |
| 219 | + np.array(["Stage 10"], dtype="<U8"), |
| 220 | + ) |
| 221 | + ] |
| 222 | + ], |
| 223 | + dtype=[("hits", "O"), ("sides", "O"), ("tasks", "O"), ("stage", "O")], |
| 224 | + ), |
| 225 | + } |
| 226 | + (schema.Longblob & "id=1").delete() |
| 227 | + dj.blob.use_32bit_dims = False |
| 228 | + |
| 229 | + |
| 230 | +def test_datetime_serialization_speed(): |
| 231 | + # If this fails that means for some reason deserializing/serializing |
| 232 | + # np arrays of np.datetime64 types is now slower than regular arrays of datetime |
| 233 | + |
| 234 | + optimized_exe_time = timeit.timeit( |
| 235 | + setup="myarr=pack(np.array([np.datetime64('2022-10-13 03:03:13') for _ in range(0, 10000)]))", |
| 236 | + stmt="unpack(myarr)", |
| 237 | + number=10, |
| 238 | + globals=globals(), |
| 239 | + ) |
| 240 | + print(f"np time {optimized_exe_time}") |
| 241 | + baseline_exe_time = timeit.timeit( |
| 242 | + setup="myarr2=pack(np.array([datetime(2022,10,13,3,3,13) for _ in range (0, 10000)]))", |
| 243 | + stmt="unpack(myarr2)", |
| 244 | + number=10, |
| 245 | + globals=globals(), |
| 246 | + ) |
| 247 | + print(f"python time {baseline_exe_time}") |
| 248 | + |
| 249 | + assert optimized_exe_time * 900 < baseline_exe_time |
0 commit comments