CODEWITHSUNDEEP
pythonpandas
Frank AK
Frank AK

Reputation: 1781

how to add new row to pandas dataframe based on other columns?

Let's say, I have the below data frame as sample.

    name  age  status  price
0  frank   12       1    100
1   jack   33       0    190
2    joe   22       1    200
******************************

Desired output:

    name  age  status  price
0  frank   12       1    100
1   jack   33       0    190
2   jack   NaN      0    -190
3    joe   22       1    200
******************************

I also posted the sample data frame below, so you can test it easily.

df1 = pd.DataFrame({
        "name":["frank", "jack", "joe"],
        "age": [12, 33, 22],
        "status": [1,0, 1],
        "price": [100,190, 200]
})

As you can see, I want to insert a new row based on the above row, which status is 0, it's meant to be a transaction failure. to statistics more easily, I want to generate a new row below that one. I also want its price to be negative number. and since I don't care about whole column of the new row, so I want the other column be NaN, just like 'age' here in my desired output.

What I've tried so far.

import pandas as pd
import numpy as np

df1 = pd.DataFrame({
        "name":["frank", "jack", "joe"],
        "age": [12, 33, 22],
        "status": [1,0, 1],
        "price": [100,190, 200]
})

df2_list = []

for i, row in df1.iterrows():
    if row["status"] == 0:
        origin_row = row.to_dict()
        new_row = ({
                "name": origin_row.get("name"),
                #"age": origin_row.get("age"),
                "age": np.NaN,
                "status": origin_row.get("status"),
                "price": -origin_row.get("price"),

            })
        df2_list.append(new_row)
df2 = pd.DataFrame(df2_list)
# concat df1 and df2 and sort it .
df3 = pd.concat([df1, df2], ignore_index=True)
df4 = df3.sort_values(['name', 'price'], ascending=[True, False])
print(df4)

I have a loop, and check if it hit my condition status==0, and append it on my tmp list, and ...but it's too many code. I want to know is there any good way, I mean more Pythonic code or pandas has already got some function can it ?

Upvotes: 4

Views: 2725

Answers (1)

cs95
cs95

Reputation: 403128

Use numpy.repeat to add rows, and Series.duplicated to set the price.

df2 = pd.DataFrame(df.values.repeat(df.status.eq(0)+1, axis=0), columns=df.columns)
df2.loc[df2.name.duplicated(), 'price'] *= -1
df2

    name age status price
0  frank  12      1   100
1   jack  33      0   190
2   jack  33      0  -190
3    joe  22      1   200

If you need to mask NaNs in the age column as well, you can do that with Series.mask.

df2.age.mask(df2.name.duplicated())

0     12
1     33
2    NaN
3     22
Name: age, dtype: object

Full code.

df2 = pd.DataFrame(df.values.repeat(df.status.eq(0)+1, axis=0), columns=df.columns)
isdup = df2.name.duplicated()
df2.loc[isdup, 'price'] *= -1
df2['age'] = df2['age'].mask(isdup)

df2
    name  age status price
0  frank   12      1   100
1   jack   33      0   190
2   jack  NaN      0  -190
3    joe   22      1   200

Upvotes: 3

Related Questions