pyjanitor

pyjanitor is a Python implementation of the R package janitor, and provides a clean API for cleaning data.

Quick start

Installation: conda install -c conda-forge pyjanitor. Read more installation instructions here.
Check out the collection of general functions.

Why janitor?

Originally a port of the R package, pyjanitor has evolved from a set of convenient data cleaning routines into an experiment with the method chaining paradigm.

Data preprocessing usually consists of a series of steps that involve transforming raw data into an understandable/usable format. These series of steps need to be run in a certain sequence to achieve success. We take a base data file as the starting point, and perform actions on it, such as removing null/empty rows, replacing them with other values, adding/renaming/removing columns of data, filtering rows and others. More formally, these steps along with their relationships and dependencies are commonly referred to as a Directed Acyclic Graph (DAG).

The pandas API has been invaluable for the Python data science ecosystem, and implements method chaining of a subset of methods as part of the API. For example, resetting indexes (.reset_index()), dropping null values (.dropna()), and more, are accomplished via the appropriate pd.DataFrame method calls.

Inspired by the ease-of-use and expressiveness of the dplyr package of the R statistical language ecosystem, we have evolved pyjanitor into a language for expressing the data processing DAG for pandas users.

To accomplish this, actions for which we would need to invoke imperative-style statements, can be replaced with method chains that allow one to read off the logical order of actions taken. Let us see the annotated example below. First off, here is the textual description of a data cleaning pathway:

Create a DataFrame.
Delete one column.
Drop rows with empty values in two particular columns.
Rename another two columns.
Add a new column.

Let's import some libraries and begin with some sample data for this example:

# Libraries
import numpy as np
import pandas as pd
import janitor

# Sample Data curated for this example
company_sales = {
    'SalesMonth': ['Jan', 'Feb', 'Mar', 'April'],
    'Company1': [150.0, 200.0, 300.0, 400.0],
    'Company2': [180.0, 250.0, np.nan, 500.0],
    'Company3': [400.0, 500.0, 600.0, 675.0]
}

In pandas code, most users might type something like this:

# The Pandas Way

# 1. Create a pandas DataFrame from the company_sales dictionary
df = pd.DataFrame.from_dict(company_sales)

# 2. Delete a column from the DataFrame. Say 'Company1'
del df['Company1']

# 3. Drop rows that have empty values in columns 'Company2' and 'Company3'
df = df.dropna(subset=['Company2', 'Company3'])

# 4. Rename 'Company2' to 'Amazon' and 'Company3' to 'Facebook'
df = df.rename(
    {
        'Company2': 'Amazon',
        'Company3': 'Facebook',
    },
    axis=1,
)

# 5. Let's add some data for another company. Say 'Google'
df['Google'] = [450.0, 550.0, 800.0]

# Output looks like this:
# Out[15]:
#   SalesMonth  Amazon  Facebook  Google
# 0        Jan   180.0     400.0   450.0
# 1        Feb   250.0     500.0   550.0
# 3      April   500.0     675.0   800.0

Slightly more advanced users might take advantage of the functional API:

df = (
    pd.DataFrame(company_sales)
    .drop(columns="Company1")
    .dropna(subset=["Company2", "Company3"])
    .rename(columns={"Company2": "Amazon", "Company3": "Facebook"})
    .assign(Google=[450.0, 550.0, 800.0])
)

# The output is the same as before, and looks like this:
# Out[15]:
#   SalesMonth  Amazon  Facebook  Google
# 0        Jan   180.0     400.0   450.0
# 1        Feb   250.0     500.0   550.0
# 3      April   500.0     675.0   800.0

With pyjanitor, we enable method chaining with method names that are explicitly named verbs, which describe the action taken.

df = (
    pd.DataFrame.from_dict(company_sales)
    .remove_columns(["Company1"])
    .dropna(subset=["Company2", "Company3"])
    .rename_column("Company2", "Amazon")
    .rename_column("Company3", "Facebook")
    .add_column("Google", [450.0, 550.0, 800.0])
)

# Output looks like this:
# Out[15]:
#   SalesMonth  Amazon  Facebook  Google
# 0        Jan   180.0     400.0   450.0
# 1        Feb   250.0     500.0   550.0
# 3      April   500.0     675.0   800.0

As such, pyjanitor's etymology has a two-fold relationship to "cleanliness". Firstly, it's about extending Pandas with convenient data cleaning routines. Secondly, it's about providing a cleaner, method-chaining, verb-based API for common pandas routines.

Installation

pyjanitor is currently installable from PyPI:

pip install pyjanitor

pyjanitor also can be installed by the conda package manager:

conda install pyjanitor -c conda-forge

pyjanitor can be installed by the pipenv environment manager too. This requires enabling prerelease dependencies:

pipenv install --pre pyjanitor

pyjanitor requires Python 3.6+.

Functionality

Current functionality includes:

Cleaning columns name (multi-indexes are possible!)
Removing empty rows and columns
Identifying duplicate entries
Encoding columns as categorical
Splitting your data into features and targets (for machine learning)
Adding, removing, and renaming columns
Coalesce multiple columns into a single column
Date conversions (from matlab, excel, unix) to Python datetime format
Expand a single column that has delimited, categorical values into dummy-encoded variables
Concatenating and deconcatenating columns, based on a delimiter
Syntactic sugar for filtering the dataframe based on queries on a column
Experimental submodules for finance, biology, chemistry, engineering, and pyspark

API

The idea behind the API is two-fold:

Copy the R package function names, but enable Pythonic use with method chaining or pandas piping.
Add other utility functions that make it easy to do data cleaning/preprocessing in pandas.

Continuing with the company_sales dataframe previously used:

import pandas as pd
import numpy as np
company_sales = {
    'SalesMonth': ['Jan', 'Feb', 'Mar', 'April'],
    'Company1': [150.0, 200.0, 300.0, 400.0],
    'Company2': [180.0, 250.0, np.nan, 500.0],
    'Company3': [400.0, 500.0, 600.0, 675.0]
}

As such, there are three ways to use the API. The first, and most strongly recommended one, is to use pyjanitor's functions as if they were native to pandas.

import janitor  # upon import, functions are registered as part of pandas.

# This cleans the column names as well as removes any duplicate rows
df = pd.DataFrame.from_dict(company_sales).clean_names().remove_empty()

The second is the functional API.

from janitor import clean_names, remove_empty

df = pd.DataFrame.from_dict(company_sales)
df = clean_names(df)
df = remove_empty(df)

The final way is to use the pipe() method:

from janitor import clean_names, remove_empty
df = (
    pd.DataFrame.from_dict(company_sales)
    .pipe(clean_names)
    .pipe(remove_empty)
)

Contributing

Follow the development guide for a full description of the process of contributing to pyjanitor.

Adding new functionality

Keeping in mind the etymology of pyjanitor, contributing a new function to pyjanitor is a task that is not difficult at all.

Define a function

First off, you will need to define the function that expresses the data processing/cleaning routine, such that it accepts a dataframe as the first argument, and returns a modified dataframe:

import pandas_flavor as pf

@pf.register_dataframe_method
def my_data_cleaning_function(df, arg1, arg2, ...):
    # Put data processing function here.
    return df

We use pandas_flavor to register the function natively on a pandas.DataFrame.

Add a test case

Secondly, we ask that you contribute a test case, to ensure that the function works as intended. Follow the contribution docs for further details.

Feature requests

If you have a feature request, please post it as an issue on the GitHub repository issue tracker. Even better, put in a PR for it! We are more than happy to guide you through the codebase so that you can put in a contribution to the codebase.

Because pyjanitor is currently maintained by volunteers and has no fiscal support, any feature requests will be prioritized according to what maintainers encounter as a need in our day-to-day jobs. Please temper expectations accordingly.

API Policy

pyjanitor only extends or aliases the pandas API (and other dataframe APIs), but will never fix or replace them.

Undesirable pandas behaviour should be reported upstream in the pandas issue tracker. We explicitly do not fix the pandas API. If at some point the pandas devs decide to take something from pyjanitor and internalize it as part of the official pandas API, then we will deprecate it from pyjanitor, while acknowledging the original contributors' contribution as part of the official deprecation record.

Credits

Test data for chemistry submodule can be found at Predictive Toxicology.

pyjanitor-devs / pyjanitor

readme