CODEWITHSUNDEEP
pythonpython-3.xpandasdataframeseries
singularity2047
singularity2047

Reputation: 1071

How to get maximum length of each column in the data frame using pandas python

I have a data frame where most of the columns are varchar/object type. Length of the column varies a lot and could be anything within the range of 3 - 1000+ . Now, for each column, I want to measure maximum length.

I know how to calculate maximum length for a col. If its varchar then:

max(df.char_col.apply(len))

and if its number (float8 or int64) then:

max(df.num_col.map(str).apply(len))

But my dataframe has hundreds of column and I want to calculate maximum length for all columns at the same time. The problem for that is, there are different data types, and I dont know how to do all at once.

So Question 1: How to get maximum column length for each columns in the data frame

Now I am trying to do that only for varchar/object type columns using following code:

xx = df.select_dtypes(include = ['object'])
for col in [xx.columns.values]:
   maxlength = [max(xx.col.apply(len))]

I selected only object type columns and tried to write a for loop. But its not working. probably using apply() within for loop is not a good idea.

Question 2: How to get maximum length of each column for only object type columns

Sample data frame:

d1 = {'name': ['john', 'tom', 'bob', 'rock', 'jimy'], 'DoB': ['01/02/2010', '01/02/2012', '11/22/2014', '11/22/2014', '09/25/2016'], 'Address': ['NY', 'NJ', 'PA', 'NY', 'CA'], 'comment1': ['Very good performance', 'N/A', 'Need to work hard', 'No Comment', 'Not satisfactory'], 'comment2': ['good', 'Meets Expectation', 'N', 'N/A', 'Incompetence']}
df1 = pd.DataFrame(data = d1)
df1['month'] = pd.DatetimeIndex(df1['DoB']).month
df1['year'] = pd.DatetimeIndex(df1['DoB']).year

Upvotes: 38

Views: 89477

Answers (7)

DSH
DSH

Reputation: 1139

Both answers below using dict comprehension:

So Question 1: How to get maximum column length for each columns in the data frame

max_length_all_cols = {col: df.loc[:, col].astype(str).apply(len).max() for col in df.columns}

Question 2: How to get maximum length of each column for only object type columns

max_length_object_cols = {col: df.loc[:, col].astype(str).apply(len).max() for col in df.select_dtypes([object]).columns}

Upvotes: 1

Hrvoje
Hrvoje

Reputation: 15202

Finding max number of chars for all columns (any type) in dataframe:

for column in df:
    print(column,"->", df[column].astype(str).str.len().max())

works fairly fast, I'm using it on 80 million rows df.

Upvotes: 18

Osmond Bishop
Osmond Bishop

Reputation: 8190

Select only object type columns

df2 = df1[[x for x in df1 if df1[x].dtype == 'object']]

Get the maximum length in each column

max_length_in_each_col = df2.applymap(lambda x: len(x)).max()

Upvotes: 10

Mohamad A Sallal
Mohamad A Sallal

Reputation: 696

You can use min max after using str and len method

df["A"].str.len().max()
df["A"].str.len().min()

df["Column Name"].str.len().max()
df["Column Name"].str.len().min()

Upvotes: 42

Azhar Ansari
Azhar Ansari

Reputation: 136

I tried numpy.vectorize but it gave 'Memory Error' for huge dataframe.

The below code worked perfectly for me. It will give you a list of maximum lengths for each column in an excel spreadsheet (read into a dataframe using pandas)

import pandas as pd

xl = pd.ExcelFile('sample.xlsx')
df = xl.parse('Sheet1')

maxColumnLenghts = []
for col in range(len(df.columns)):
    maxColumnLenghts.append(max(df.iloc[:,col].astype(str).apply(len)))
print('Max Column Lengths ', maxColumnLenghts)

Upvotes: 4

alif
alif

Reputation: 805

Some great answers here and I would like to contribute mine

Solution:

dict([(v, df[v].apply(lambda r: len(str(r)) if r!=None else 0).max())for v in df.columns.values])

Explanation:

#convert tuple to dictionary
dict( 
    [
        #create a tuple such that (column name, max length of values in column)
        (v, df[v].apply(lambda r: len(str(r)) if r!=None else 0).max()) 
            for v in df.columns.values #iterates over all column values
    ])

Sample output

{'name': 4, 'DoB': 10, 'Address': 2, 'comment1': 21, 'comment2': 17}

Upvotes: 11

jpp
jpp

Reputation: 164773

One solution is to use numpy.vectorize. This may be more efficient than pandas-based solutions.

You can use pd.DataFrame.select_dtypes to select object columns.

import pandas as pd
import numpy as np

df = pd.DataFrame({'A': ['abc', 'de', 'abcd'],
                   'B': ['a', 'abcde', 'abc'],
                   'C': [1, 2.5, 1.5]})

measurer = np.vectorize(len)

Max length for all columns

res1 = measurer(df.values.astype(str)).max(axis=0)

array([4, 5, 3])

Max length for object columns

res2 = measurer(df.select_dtypes(include=[object]).values.astype(str)).max(axis=0)

array([4, 5])

Or if you need output as a dictionary:

res1 = dict(zip(df, measurer(df.values.astype(str)).max(axis=0)))

{'A': 4, 'B': 5, 'C': 3}

df_object = df.select_dtypes(include=[object])
res2 = dict(zip(df_object, measurer(df_object.values.astype(str)).max(axis=0)))

{'A': 4, 'B': 5}

Upvotes: 39

Related Questions