SUMIF 等效于 Python 中的唯一日期范围（如果日期落在变量创建的不同日期范围内则求和）

Question

我希望创建基于不同功能/类别特有的日期范围求和的变量，以在 Python 中自动执行当前的 Excel 任务。它类似于 Excel 中的 SUMIF，但不同变量的日期范围不同。我会尝试重新创建一个类似的情况，因为我无法共享确切的数据。目前，我有一个销售数据框，其中包含每周按地区划分的销售额，如下所示：

Week	Area	Sales
08/02/2019	London	200
08/02/2019	Scotland	150
15/02/2019	London	100
15/02/2019	Scotland	120
22/02/2019	London	50
22/02/2019	Scotland	20

我想合并日期是否属于产品的销售期，所以说我有另一个这样的数据框：

Product	Sale Start Week	Sale End Week
Boots	08/02/2019	15/02/2019
Accessories	15/02/2019	22/02/2019

如果日期在为每个产品指定的日期范围内，我想创建一些总和。例如，对于下面的 Boots，如果 Sales 中的周数在 Sales Periods 日期范围内，则对 Sales 求和：

Area	Boots	Accessories
London	300	150
Scotland	270	140

我试过 groupby 和数据透视表，但我不确定如何将销售日期过滤器合并到其中。目前，销售期数据框和销售数据框是分开的。

这就是我所拥有的几乎就在那里的数据透视代码：

test = pd.pivot_table(df,index=['Area','Week'],columns=sales_period_df['Product'],values=['Sales'],aggfunc=np.sum)

但这不包括销售期间的过滤，我不确定如何合并它。非常感谢您的建议，在此先感谢！

Answer 1

# DF: sales (top DF in question)
# DF2: sales period (second DF in question)

# format the date into datetime
df['Week'] = pd.to_datetime(df['Week'], dayfirst=True)
df2[['Sale Start Week','Sale End Week']]=df2[['Sale Start Week','Sale End Week']].apply(pd.to_datetime, dayfirst=True)
df2

# merge using merge_asof 
df3=pd.merge_asof( df.sort_values('Week'),
                  df2.sort_values('Sale Start Week'),
                  left_on = 'Week',
                  right_on='Sale Start Week')

# including only when week falls within end week
df3=df3.loc[df3['Week'] <= df3['Sale End Week']]
df3

# cross tab for resultset
out= (pd.crosstab(index=df3['Area'], 
                  columns=df3['Product'], 
                  values=df3['Sales'], 
                  aggfunc='sum')
      .reset_index()
      .rename_axis(columns=None))

out

    Area    Accessories     Boots
0   London      150         200
1   Scotland    140         150

Answer 2

由于周期重叠，我们不能在这种情况下使用经典的旋转（除非我们为每个周期复制重叠的销售记录，这看起来太多了）。所以我们必须手动创建这个表。

首先，让我们准备一些要使用的数据：

import pandas as pd
from io import StringIO

data = '''Week,Area,Sales
08/02/2019,London,200
08/02/2019,Scotland,150
15/02/2019,London,100
15/02/2019,Scotland,120
22/02/2019,London,50
22/02/2019,Scotland,20'''

df = pd.read_csv(StringIO(data), index_col=0, parse_dates=True, dayfirst=True).sort_index()

data = '''Product,Sale Start Week,Sale End Week
Boots,08/02/2019,15/02/2019
Accessories,15/02/2019,22/02/2019
Something,08/02/2019,22/02/2019'''

sales_period_df = pd.read_csv(StringIO(data), index_col=0, parse_dates=[1, 2], dayfirst=True)

df 和sales_period_df 的结构略有修改，Week 和Product 现在是索引。

接下来，我们准备输出帧和支持数据：

import pandas.IndexSlice as idx

# create slices from sales_period_df
# which can be used to locate data in df
periods = sales_period_df.agg(lambda row: idx[row['Sale Start Week']:row['Sale End Week']], axis=1)

# separate sales by area
sales_by_area = df.groupby('Area')['Sales']

# create the output DataFrame with unique areas as indexes 
# and products as columns
output = pd.DataFrame(index=df['Area'].unique(), columns=sales_period_df.index)

要填写数据，我们可以像这样使用 eather apply 或 agg：

for product in output.columns:
    output[product] = sales_by_area.agg(lambda sales: sales.loc[periods[product]].sum())

让我们组装代码：

import pandas as pd
from pandas import IndexSlice as idx
from io import StringIO

data = '''Week,Area,Sales
08/02/2019,London,200
08/02/2019,Scotland,150
15/02/2019,London,100
15/02/2019,Scotland,120
22/02/2019,London,50
22/02/2019,Scotland,20'''

df = pd.read_csv(StringIO(data), index_col=0, parse_dates=True, dayfirst=True).sort_index()

data = '''Product,Sale Start Week,Sale End Week
Boots,08/02/2019,15/02/2019
Accessories,15/02/2019,22/02/2019
Something,08/02/2019,22/02/2019'''

sales_period_df = pd.read_csv(StringIO(data), index_col=0, parse_dates=[1, 2], dayfirst=True)

periods = sales_period_df.agg(lambda row: idx[row['Sale Start Week']:row['Sale End Week']], axis=1)
output = pd.DataFrame(index=df['Area'].unique(), columns=sales_period_df.index)
sales_by_area = df.groupby('Area')['Sales']

for product in output.columns:
    output[product] = sales_by_area.agg(lambda sales: sales.loc[periods[product]].sum())

print(output)

输出：

Product   Boots  Accessories  Something
London      300          150        350
Scotland    270          140        290

Answer 3

一种选择是使用 conditional_join 计算非等值连接以获取匹配项，最后进行 groupby 和求和：

# pip install pyjanitor
import pandas as pd
import janitor

(dates
.conditional_join(
    products, 
    ('Week', 'Sale Start Week', '>='), 
    ('Week', 'Sale End Week', '<='),
    # for larger data, numba may offer 
    # better performance
    use_numba = False, 
    df_columns=['Area','Sales'], 
    right_columns='Product')
.pivot_table(
    index='Area',
    columns='Product',
    values='Sales',
    aggfunc='sum')
)
Product   Accessories  Boots
Area                        
London            150    300
Scotland          140    270