def test_build_already_build(self, calculate_bins_func, add_hist_func):
"""Should not rebuild if Histogram was already build before"""
hist = Histogram()
hist.is_build = True
hist.build()
self.assertFalse(add_hist_func.called)
self.assertFalse(calculate_bins_func.called)
def test_build(self):
"""Should calculate the bin list, and hist values for each column in the Histogram, if the
histogram hasn't been build before"""
hist = Histogram(bins=2)
test_df = self.create_test_df()
column_to_ad = test_df.select(F.col('value'))
column_to_ad_2 = test_df.select(F.col('value2'))
hist.add_column(column_to_ad)
hist.add_column(column_to_ad_2)
hist.build()
self.assertEqual(3, len(hist.bin_list))
self.assertEqual(2, len(hist.hist_dict))
self.assertTrue(hist.is_build)
def test_add_hist_single_value(self):
"""Should set the bin list to n (self.nr_bins) bins (n+1 bin borders) where the min bin border is the
single value -0.5 and the max bin border is the single value +0.5 incase a column is input with only a
single value"""
single_column_value = 1
nr_bins = 5
column_values = [single_column_value] * 100
test_df = self.sqlCtx.createDataFrame(
pd.DataFrame({'foo': column_values}))
hist = Histogram(bins=nr_bins)
hist.add_column(test_df.select(F.col('foo')))
hist.build()
self.assertEqual(6, len(hist.bin_boundaries))
self.assertEqual(single_column_value - 0.5, min(hist.bin_boundaries))
self.assertEqual(single_column_value + 0.5, max(hist.bin_boundaries))
self.assertEqual(len(column_values),
hist.hist_dict['foo'][math.floor(nr_bins / 2)])