基于列值的Pandas聚合减法

Question

Suppose I have DataFrame

'name'     'quantity'   'day'
'A'         1           'Monday'
'A'         10          'Sunday'
'A'         5           'Friday'
'B'         2           'Monday'
'B'         30          'Sunday'
'B'         5           'Thursday'

What I need to build is another dataframe where for each name I subtract the quantity of Monday from the quantity of Sunday. So, I guess I need a groupBy on the name and then an agg with a function, but I am not sure how to do the filter so that only those days are considered.

Following the example, the end result I seek is

'name'     'sub_quantity'
'A'         9 
'B'         28 

Answer 1

setup

import pandas as pd
from io import StringIO

txt = """name     quantity   day
A         1           Monday
A         10          Sunday
A         5           Friday
B         2           Monday
B         30          Sunday
B         5           Thursday"""

df = pd.read_csv(StringIO(txt), delim_whitespace=True)

option 1
unstack

d1 = df.set_index(['name', 'day']).quantity.unstack()

d1.Sunday.sub(d1.Monday)

name
A     9.0
B    28.0
dtype: float64

option 2
query

s = df.set_index('name').query('day == "Sunday"').quantity
m = df.set_index('name').query('day == "Monday"').quantity
s - m

name
A     9
B    28
Name: quantity, dtype: int64

option 3
xs

d1 = df.set_index(['day', 'name']).quantity
d1.xs('Sunday') - d1.xs('Monday')

name
A     9
B    28
Name: quantity, dtype: int64

option 4
cute apply

def obnoxious(x):
    s = x.day.eq('Sunday').idxmax()
    m = x.day.eq('Monday').idxmax()
    q = 'quantity'
    return x.get_value(s, q) - x.get_value(m, q)

df.groupby('name').apply(obnoxious)

name
A     9
B    28
dtype: int64

---

---

timing
example data

Answer 2

Solution with pivot and substracting by sub :

df = pd.pivot(index=df.name, columns=df.day, values=df.quantity)
print (df.Sunday.sub(df.Monday).reset_index(name='sub_quantity'))
  name  sub_quantity
0    A           9.0
1    B          28.0