Building a new pd.dataframe with statistics from own functions

I am trying to creat some summary statistics for text data that I am working with, namely the average length of text columns in my dataFrame.

I am working with two columns: short and long

import pandas as pd

data = {
    'short': ['Fruit', 'Vehicle', 'Animal', 'City'],
    'long': ['An edible object that is usually sweet and grows on trees or plants.', 
             'A mode of transportation that is used to move people or goods from one place to another.', 
             'A living organism that typically feeds on organic matter and has the ability to move.', 
             'A large and populous settlement, usually the seat of government or important cultural institutions.']
}

df = pd.DataFrame(data)

For that I have written a function that calculates that information for a single column:

def mean_chars_in_col(data: pd.DataFrame, 
                      col: str):
    return data[col].str.len().mean().round(2)

This gives me the needed information for a single column:

mean_chars_in_col(df, "short")

## 5.5

How can I extend this functionality so that I get this information for every column in the dataFrame, ideally with a new dataframe as output where I have two columns: colname and mean_chars, denoting each observed column in the original dataframe as a row in colname and each result as a row in mean_chars?

Would it also be possible to add onto that with new functions, that then write their output to their own column for each row?

#expected output, ideally a new dataframe
colname, mean_chars, mean_length
short, 5.5, 1
long, 85, 15.25

Thank you.

You can use DataFrame.apply with list of functions, then transpose and set new columns names by list:

def mean_chars_in_col(data: pd.Series):
    return data.str.len().mean()

def mean_length_in_col(data: pd.Series):
    return data.str.split().str.len().mean()


df = (df.apply([mean_chars_in_col, mean_length_in_col])
        .T
        .set_axis(['mean_chars','mean_length'], axis=1))
print (df)
       mean_chars  mean_length
short         5.5         1.00
long         85.0        14.75

Or change function and return Series:

def mean_chars_in_col(data: pd.Series):
    a = data.str.len().mean()
    b = data.str.split().str.len().mean()
    
    return pd.Series([a, b], index=['mean_chars','mean_length'])

df = df.apply(mean_chars_in_col).T
print (df)
       mean_chars  mean_length
short         5.5         1.00
long         85.0        14.75

def mean_chars_in_col(data: pd.Series):
    a = data.str.len().mean()
    b = data.str.split().str.len().mean()
    
    return pd.Series([a, b], index=['mean_chars','mean_length'])

out = df.apply(mean_chars_in_col).T.rename_axis('colname').reset_index()
print (out)
  colname  mean_chars  mean_length
0   short         5.5         1.00
1    long        85.0        14.75

You can use pd.apply, but to do that you will have to modify your function so that takes directly a Series as an argumetn:

data = {
    'short': ['Fruit', 'Vehicle', 'Animal', 'City'],
    'long': ['An edible object that is usually sweet and grows on trees or plants.', 
             'A mode of transportation that is used to move people or goods from one place to another.', 
             'A living organism that typically feeds on organic matter and has the ability to move.', 
             'A large and populous settlement, usually the seat of government or important cultural institutions.']
}

df = pd.DataFrame(data)

# Notice that the function now only takes one argument, 
# which will be a pd.Series

def mean_chars_in_col(col):
    return col.str.len().mean().round(2)

# then use apply to iterate through each column and pass each column into the funcition 

>>> df.apply(mean_chars_in_col)
short     5.5
long     85.0
dtype: float64


</details>