|
| 1 | +# -------------------------------------------------------------------------------------------- |
| 2 | +# Copyright (c) Microsoft Corporation. All rights reserved. |
| 3 | +# Licensed under the MIT License. |
| 4 | +# -------------------------------------------------------------------------------------------- |
| 5 | +import os |
| 6 | +import sys |
| 7 | +import unittest |
| 8 | +import tempfile |
| 9 | + |
| 10 | +import numpy as np |
| 11 | +import pandas as pd |
| 12 | +from nimbusml import Pipeline, DprepDataStream |
| 13 | +from nimbusml.preprocessing.missing_values import Handler |
| 14 | + |
| 15 | + |
| 16 | +def get_temp_file(suffix=None): |
| 17 | + fd, file_name = tempfile.mkstemp(suffix=suffix) |
| 18 | + fl = os.fdopen(fd, 'w') |
| 19 | + fl.close() |
| 20 | + return file_name |
| 21 | + |
| 22 | + |
| 23 | +class TestDateTimeDataType(unittest.TestCase): |
| 24 | + def test_negative_values(self): |
| 25 | + milliseconds_in_year = 365*24*60*60*1000 |
| 26 | + data = [i * milliseconds_in_year for i in [-1, -2, -3, -3.3]] |
| 27 | + |
| 28 | + df = pd.DataFrame({'c1': data, 'c2': [3,4,5,6]}) |
| 29 | + df = df.astype({'c1': np.dtype('datetime64[ms]')}) |
| 30 | + |
| 31 | + pipeline = Pipeline(steps=[Handler(columns={'c2': 'c2'})]) |
| 32 | + result = pipeline.fit_transform(df) |
| 33 | + |
| 34 | + self.assertTrue(result.loc[:, 'c1'].equals(df.loc[:, 'c1'])) |
| 35 | + self.assertEqual(result.loc[:, 'c1'].dtype, np.dtype('datetime64[ns]')) |
| 36 | + |
| 37 | + self.assertEqual(result.loc[0, 'c1'].year, 1969) |
| 38 | + self.assertEqual(result.loc[0, 'c1'].hour, 0) |
| 39 | + self.assertEqual(result.loc[0, 'c1'].minute, 0) |
| 40 | + self.assertEqual(result.loc[0, 'c1'].second, 0) |
| 41 | + |
| 42 | + self.assertEqual(result.loc[3, 'c1'].year, 1966) |
| 43 | + |
| 44 | + def test_timestamp_boundaries(self): |
| 45 | + # Here are the current min and max for a Pandas Timestamp |
| 46 | + # 1677-09-21 00:12:43.145225 |
| 47 | + # 2262-04-11 23:47:16.854775807 |
| 48 | + |
| 49 | + data = [pd.Timestamp(1677, 9, 22, 1), pd.Timestamp.max] |
| 50 | + df = pd.DataFrame({'c1': data, 'c2': [3,4]}) |
| 51 | + df = df.astype({'c1': np.dtype('datetime64[ms]')}) |
| 52 | + |
| 53 | + pipeline = Pipeline(steps=[Handler(columns={'c2': 'c2'})]) |
| 54 | + result = pipeline.fit_transform(df) |
| 55 | + |
| 56 | + self.assertTrue(result.loc[:, 'c1'].equals(df.loc[:, 'c1'])) |
| 57 | + self.assertEqual(result.dtypes[0], np.dtype('datetime64[ns]')) |
| 58 | + |
| 59 | + self.assertEqual(result.loc[0, 'c1'].year, 1677) |
| 60 | + self.assertEqual(result.loc[0, 'c1'].month, 9) |
| 61 | + self.assertEqual(result.loc[0, 'c1'].day, 22) |
| 62 | + |
| 63 | + self.assertEqual(result.loc[1, 'c1'].year, 2262) |
| 64 | + self.assertEqual(result.loc[1, 'c1'].month, 4) |
| 65 | + self.assertEqual(result.loc[1, 'c1'].day, 11) |
| 66 | + |
| 67 | + def test_datetime_column_parsed_from_string(self): |
| 68 | + dates = ["2018-01-02", "2018-02-01"] |
| 69 | + df = pd.DataFrame({'c1': dates, 'c2': [3,4]}) |
| 70 | + |
| 71 | + file_name = get_temp_file('.csv') |
| 72 | + df.to_csv(file_name) |
| 73 | + df = pd.read_csv(file_name, parse_dates=['c1'], index_col=0) |
| 74 | + |
| 75 | + self.assertEqual(df.dtypes[0], np.dtype('datetime64[ns]')) |
| 76 | + |
| 77 | + pipeline = Pipeline(steps=[Handler(columns={'c2': 'c2'})]) |
| 78 | + result = pipeline.fit_transform(df) |
| 79 | + |
| 80 | + self.assertEqual(result.loc[0, 'c1'].year, 2018) |
| 81 | + self.assertEqual(result.loc[0, 'c1'].month, 1) |
| 82 | + self.assertEqual(result.loc[0, 'c1'].day, 2) |
| 83 | + self.assertEqual(result.loc[0, 'c1'].hour, 0) |
| 84 | + self.assertEqual(result.loc[0, 'c1'].minute, 0) |
| 85 | + self.assertEqual(result.loc[0, 'c1'].second, 0) |
| 86 | + |
| 87 | + self.assertEqual(result.loc[1, 'c1'].year, 2018) |
| 88 | + self.assertEqual(result.loc[1, 'c1'].month, 2) |
| 89 | + self.assertEqual(result.loc[1, 'c1'].day, 1) |
| 90 | + self.assertEqual(result.loc[1, 'c1'].hour, 0) |
| 91 | + self.assertEqual(result.loc[1, 'c1'].minute, 0) |
| 92 | + self.assertEqual(result.loc[1, 'c1'].second, 0) |
| 93 | + |
| 94 | + self.assertEqual(len(result), 2) |
| 95 | + self.assertEqual(result.dtypes[0], np.dtype('datetime64[ns]')) |
| 96 | + |
| 97 | + os.remove(file_name) |
| 98 | + |
| 99 | + @unittest.skipIf(sys.version_info[:2] == (2, 7), "azureml-dataprep is not installed.") |
| 100 | + def test_dprep_datastream(self): |
| 101 | + import azureml.dataprep as dprep |
| 102 | + |
| 103 | + dates = ["2018-01-02 00:00:00", "2018-02-01 10:00:00"] |
| 104 | + col2 = ['0', '1'] |
| 105 | + label_array = np.repeat([0], 2) |
| 106 | + train_df = pd.DataFrame({'col1': dates, 'col2': col2, 'label': label_array}) |
| 107 | + |
| 108 | + pipeline = Pipeline(steps=[ |
| 109 | + Handler(columns={'2': 'col2'}, concat=False, impute_by_slot=True, replace_with='Mean') |
| 110 | + ]) |
| 111 | + |
| 112 | + file_name = get_temp_file('.csv') |
| 113 | + train_df.to_csv(file_name) |
| 114 | + |
| 115 | + dataflow = dprep.read_csv(file_name, infer_column_types=True) |
| 116 | + dprepDataStream = DprepDataStream(dataflow) |
| 117 | + |
| 118 | + result = pipeline.fit_transform(dprepDataStream) |
| 119 | + |
| 120 | + self.assertEqual(result.loc[:, 'col1'].dtype, np.dtype('datetime64[ns]')) |
| 121 | + |
| 122 | + self.assertEqual(result.loc[0, 'col1'].year, 2018) |
| 123 | + self.assertEqual(result.loc[0, 'col1'].month, 1) |
| 124 | + self.assertEqual(result.loc[0, 'col1'].day, 2) |
| 125 | + self.assertEqual(result.loc[0, 'col1'].hour, 0) |
| 126 | + self.assertEqual(result.loc[0, 'col1'].minute, 0) |
| 127 | + self.assertEqual(result.loc[0, 'col1'].second, 0) |
| 128 | + |
| 129 | + self.assertEqual(result.loc[1, 'col1'].year, 2018) |
| 130 | + self.assertEqual(result.loc[1, 'col1'].month, 2) |
| 131 | + self.assertEqual(result.loc[1, 'col1'].day, 1) |
| 132 | + self.assertEqual(result.loc[1, 'col1'].hour, 10) |
| 133 | + self.assertEqual(result.loc[1, 'col1'].minute, 0) |
| 134 | + self.assertEqual(result.loc[1, 'col1'].second, 0) |
| 135 | + |
| 136 | + os.remove(file_name) |
| 137 | + |
| 138 | + |
| 139 | +if __name__ == '__main__': |
| 140 | + unittest.main() |
0 commit comments