# Licensed to Elasticsearch B.V. under one or more contributor # license agreements. See the NOTICE file distributed with # this work for additional information regarding copyright # ownership. Elasticsearch B.V. licenses this file to you under # the Apache License, Version 2.0 (the "License"); you may # not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, # software distributed under the License is distributed on an # "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY # KIND, either express or implied. See the License for the # specific language governing permissions and limitations # under the License. from datetime import datetime # File called _pytest for PyCharm compatability import numpy as np import pytest from eland import Series from tests.common import TestData, assert_pandas_eland_series_equal class TestSeriesArithmetics(TestData): def test_ecommerce_datetime_comparisons(self): pd_df = self.pd_ecommerce() ed_df = self.ed_ecommerce() ops = ["__le__", "__lt__", "__gt__", "__ge__", "__eq__", "__ne__"] # this datetime object is timezone naive datetime_obj = datetime(2016, 12, 18) # FIXME: the following timezone conversions are just a temporary fix # to run the datetime comparison tests # # The problem: # - the datetime objects of the pandas DataFrame are timezone aware and # can't be compared with timezone naive datetime objects # - the datetime objects of the eland DataFrame are timezone naive (which # should be fixed) # - however if the eland DataFrame is converted to a pandas DataFrame # (using the `to_pandas` function) the datetime objects become timezone aware # # This tests converts the datetime objects of the pandas Series to # timezone naive ones and utilizes a class to make the datetime objects of the # eland Series timezone naive before the result of `to_pandas` is returned. # The `to_pandas` function is executed by the `assert_pandas_eland_series_equal` # function, which compares the eland and pandas Series # convert to timezone naive datetime object pd_df["order_date"] = pd_df["order_date"].dt.tz_localize(None) class ModifiedElandSeries(Series): def to_pandas(self): """remove timezone awareness before returning the pandas dataframe""" series = super().to_pandas() series = series.dt.tz_localize(None) return series for op in ops: pd_series = pd_df[getattr(pd_df["order_date"], op)(datetime_obj)][ "order_date" ] ed_series = ed_df[getattr(ed_df["order_date"], op)(datetime_obj)][ "order_date" ] # "type cast" to modified class (inherits from ed.Series) that overrides the `to_pandas` function ed_series.__class__ = ModifiedElandSeries assert_pandas_eland_series_equal( pd_series, ed_series, check_exact=False, rtol=1e-3, # previously known as check_less_precise ) def test_ecommerce_series_invalid_div(self): pd_df = self.pd_ecommerce() ed_df = self.ed_ecommerce() # eland / pandas == error with pytest.raises(TypeError): _ = ed_df["total_quantity"] / pd_df["taxful_total_price"] def test_ecommerce_series_simple_arithmetics(self): pd_df = self.pd_ecommerce().head(100) ed_df = self.ed_ecommerce().head(100) pd_series = ( pd_df["taxful_total_price"] + 5 + pd_df["total_quantity"] / pd_df["taxless_total_price"] - pd_df["total_unique_products"] * 10.0 + pd_df["total_quantity"] ) ed_series = ( ed_df["taxful_total_price"] + 5 + ed_df["total_quantity"] / ed_df["taxless_total_price"] - ed_df["total_unique_products"] * 10.0 + ed_df["total_quantity"] ) assert_pandas_eland_series_equal(pd_series, ed_series, rtol=True) def test_ecommerce_series_simple_integer_addition(self): pd_df = self.pd_ecommerce().head(100) ed_df = self.ed_ecommerce().head(100) pd_series = pd_df["taxful_total_price"] + 5 ed_series = ed_df["taxful_total_price"] + 5 assert_pandas_eland_series_equal(pd_series, ed_series, rtol=True) def test_ecommerce_series_simple_series_addition(self): pd_df = self.pd_ecommerce().head(100) ed_df = self.ed_ecommerce().head(100) pd_series = pd_df["taxful_total_price"] + pd_df["total_quantity"] ed_series = ed_df["taxful_total_price"] + ed_df["total_quantity"] assert_pandas_eland_series_equal(pd_series, ed_series, rtol=True) def test_ecommerce_series_basic_arithmetics(self): pd_df = self.pd_ecommerce().head(100) ed_df = self.ed_ecommerce().head(100) ops = [ "__add__", "__truediv__", "__floordiv__", "__pow__", "__mod__", "__mul__", "__sub__", "add", "truediv", "floordiv", "pow", "mod", "mul", "sub", ] for op in ops: pd_series = getattr(pd_df["taxful_total_price"], op)( pd_df["total_quantity"] ) ed_series = getattr(ed_df["taxful_total_price"], op)( ed_df["total_quantity"] ) assert_pandas_eland_series_equal(pd_series, ed_series, rtol=True) pd_series = getattr(pd_df["taxful_total_price"], op)(10.56) ed_series = getattr(ed_df["taxful_total_price"], op)(10.56) assert_pandas_eland_series_equal(pd_series, ed_series, rtol=True) pd_series = getattr(pd_df["taxful_total_price"], op)(np.float32(1.879)) ed_series = getattr(ed_df["taxful_total_price"], op)(np.float32(1.879)) assert_pandas_eland_series_equal(pd_series, ed_series, rtol=True) pd_series = getattr(pd_df["taxful_total_price"], op)(int(8)) ed_series = getattr(ed_df["taxful_total_price"], op)(int(8)) assert_pandas_eland_series_equal(pd_series, ed_series, rtol=True) def test_supported_series_dtypes_ops(self): pd_df = self.pd_ecommerce().head(100) ed_df = self.ed_ecommerce().head(100) # Test some specific operations that are and aren't supported numeric_ops = [ "__add__", "__truediv__", "__floordiv__", "__pow__", "__mod__", "__mul__", "__sub__", ] non_string_numeric_ops = [ "__add__", "__truediv__", "__floordiv__", "__pow__", "__mod__", "__sub__", ] # __mul__ is supported for int * str in pandas # float op float for op in numeric_ops: pd_series = getattr(pd_df["taxful_total_price"], op)( pd_df["taxless_total_price"] ) ed_series = getattr(ed_df["taxful_total_price"], op)( ed_df["taxless_total_price"] ) assert_pandas_eland_series_equal(pd_series, ed_series, rtol=True) # int op float for op in numeric_ops: pd_series = getattr(pd_df["total_quantity"], op)( pd_df["taxless_total_price"] ) ed_series = getattr(ed_df["total_quantity"], op)( ed_df["taxless_total_price"] ) assert_pandas_eland_series_equal(pd_series, ed_series, rtol=True) # float op int for op in numeric_ops: pd_series = getattr(pd_df["taxful_total_price"], op)( pd_df["total_quantity"] ) ed_series = getattr(ed_df["taxful_total_price"], op)( ed_df["total_quantity"] ) assert_pandas_eland_series_equal(pd_series, ed_series, rtol=True) # str op int (throws) for op in non_string_numeric_ops: with pytest.raises(TypeError): pd_series = getattr(pd_df["currency"], op)(pd_df["total_quantity"]) with pytest.raises(TypeError): ed_series = getattr(ed_df["currency"], op)(ed_df["total_quantity"]) with pytest.raises(TypeError): pd_series = getattr(pd_df["currency"], op)(1) with pytest.raises(TypeError): ed_series = getattr(ed_df["currency"], op)(1) # int op str (throws) for op in non_string_numeric_ops: with pytest.raises(TypeError): pd_series = getattr(pd_df["total_quantity"], op)(pd_df["currency"]) with pytest.raises(TypeError): ed_series = getattr(ed_df["total_quantity"], op)(ed_df["currency"]) def test_ecommerce_series_basic_rarithmetics(self): pd_df = self.pd_ecommerce().head(10) ed_df = self.ed_ecommerce().head(10) ops = [ "__radd__", "__rtruediv__", "__rfloordiv__", "__rpow__", "__rmod__", "__rmul__", "__rsub__", "radd", "rtruediv", "rfloordiv", "rpow", "rmod", "rmul", "rsub", ] for op in ops: pd_series = getattr(pd_df["taxful_total_price"], op)( pd_df["total_quantity"] ) ed_series = getattr(ed_df["taxful_total_price"], op)( ed_df["total_quantity"] ) assert_pandas_eland_series_equal(pd_series, ed_series, rtol=True) pd_series = getattr(pd_df["taxful_total_price"], op)(3.141) ed_series = getattr(ed_df["taxful_total_price"], op)(3.141) assert_pandas_eland_series_equal(pd_series, ed_series, rtol=True) pd_series = getattr(pd_df["taxful_total_price"], op)(np.float32(2.879)) ed_series = getattr(ed_df["taxful_total_price"], op)(np.float32(2.879)) assert_pandas_eland_series_equal(pd_series, ed_series, rtol=True) pd_series = getattr(pd_df["taxful_total_price"], op)(int(6)) ed_series = getattr(ed_df["taxful_total_price"], op)(int(6)) assert_pandas_eland_series_equal(pd_series, ed_series, rtol=True) def test_supported_series_dtypes_rops(self): pd_df = self.pd_ecommerce().head(100) ed_df = self.ed_ecommerce().head(100) # Test some specific operations that are and aren't supported numeric_ops = [ "__radd__", "__rtruediv__", "__rfloordiv__", "__rpow__", "__rmod__", "__rmul__", "__rsub__", ] non_string_numeric_ops = [ "__radd__", "__rtruediv__", "__rfloordiv__", "__rpow__", "__rmod__", "__rsub__", ] # __rmul__ is supported for int * str in pandas # float op float for op in numeric_ops: pd_series = getattr(pd_df["taxful_total_price"], op)( pd_df["taxless_total_price"] ) ed_series = getattr(ed_df["taxful_total_price"], op)( ed_df["taxless_total_price"] ) assert_pandas_eland_series_equal(pd_series, ed_series, rtol=True) # int op float for op in numeric_ops: pd_series = getattr(pd_df["total_quantity"], op)( pd_df["taxless_total_price"] ) ed_series = getattr(ed_df["total_quantity"], op)( ed_df["taxless_total_price"] ) assert_pandas_eland_series_equal(pd_series, ed_series, rtol=True) # float op int for op in numeric_ops: pd_series = getattr(pd_df["taxful_total_price"], op)( pd_df["total_quantity"] ) ed_series = getattr(ed_df["taxful_total_price"], op)( ed_df["total_quantity"] ) assert_pandas_eland_series_equal(pd_series, ed_series, rtol=True) # str op int (throws) for op in non_string_numeric_ops: with pytest.raises(TypeError): pd_series = getattr(pd_df["currency"], op)(pd_df["total_quantity"]) with pytest.raises(TypeError): ed_series = getattr(ed_df["currency"], op)(ed_df["total_quantity"]) with pytest.raises(TypeError): pd_series = getattr(pd_df["currency"], op)(10.0) with pytest.raises(TypeError): ed_series = getattr(ed_df["currency"], op)(10.0) # int op str (throws) for op in non_string_numeric_ops: with pytest.raises(TypeError): pd_series = getattr(pd_df["total_quantity"], op)(pd_df["currency"]) with pytest.raises(TypeError): ed_series = getattr(ed_df["total_quantity"], op)(ed_df["currency"])