Skip to main content

Optimus is the missing framework for cleaning and pre-processing data in a distributed fashion with pyspark.

Project description

Logo Optimus

PyPI version Build Status Documentation Status
built_by iron Updates
GitHub release Codacy Badge StackShare
Platforms Code Health Coverage Status Mentioned in Awesome Data Science
Join the chat at

Optimus is the missing framework to profile, clean, process and do ML in a distributed fashion using Apache Spark(PySpark).

Installation (pip):

In your terminal just type pip install optimuspyspark


  • Apache Spark>= 2.3.0
  • Python>=3.6


After installation go to the examples folder to found notebooks about data cleaning, data munging, and how to create ML and DL models.




Feedback is what drive Optimus future, so please take a couple of minutes to help shape the Optimus' Roadmap:

Also If you want to know what features are the most requested or have and idea you want to see in Optimus let us know at

And if you want to see some cool information and tutorials about Optimus check out our blog

Loading data

Now Optimus can load data in csv, json, parquet and avro from a local file or an URL.

df = op.load.csv("data/foo.csv")

df = op.load.json("data/foo.json")

# parquet
df = op.load.parquet("data/foo.parquet")

# avro
df = op.load.avro("data/foo.avro").table(5)

If you want to load from a URL you just need to use load.url() with the path and the type file (csv, json, parquet, avro)

df = op.load.url("", "json")

Cleaning and Processing

Optimus V2 was created to make data cleaning a breeze. The API was designed to be super easy to newcomers and very familiar for people that comes from Pandas. Optimus expand the Spark DataFrame functionality adding .rows and .cols attributes.

For example you can load data from a url, transform and apply some predefined cleaning functions:

from optimus import Optimus
op = Optimus()

# This is a custom function
def func(value, arg):
    return "this was a number"

df =op.load.url("")

    .cols.date_transform("birth", "new_date", "yyyy/MM/dd", "dd-MM-YYYY")\
    .cols.years_between("birth", "years_between", "yyyy/MM/dd")\
    .cols.apply_by_dtypes("product",func,"string", data_type="integer")\

You transform this:

| id|           firstName|            lastName|billingId|   product|price|     birth|dummyCol|
|  1|                Luis|         Alvarez$$%!|      123|      Cake|   10|1980/07/07|   never|
|  2|               André|              Ampère|      423|      piza|    8|1950/07/08|   gonna|
|  3|               NiELS|          Böhr//((%%|      551|     pizza|    8|1990/07/09|    give|
|  4|                PAUL|              dirac$|      521|     pizza|    8|1954/07/10|     you|
|  5|              Albert|            Einstein|      634|     pizza|    8|1990/07/11|      up|
|  6|             Galileo|             GALiLEI|      672|     arepa|    5|1930/08/12|   never|
|  7|                CaRL|            Ga%%%uss|      323|      taco|    3|1970/07/13|   gonna|
|  8|               David|          H$$$ilbert|      624|  taaaccoo|    3|1950/07/14|     let|
|  9|            Johannes|              KEPLER|      735|      taco|    3|1920/04/22|     you|
| 10|               JaMES|         M$$ax%%well|      875|      taco|    3|1923/03/12|    down|
| 11|               Isaac|              Newton|      992|     pasta|    9|1999/02/15|  never |
| 12|              Emmy%%|            Nöether$|      234|     pasta|    9|1993/12/08|   gonna|
| 13|              Max!!!|           Planck!!!|      111|hamburguer|    4|1994/01/04|    run |
| 14|                Fred|            Hoy&&&le|      553|    pizzza|    8|1997/06/27|  around|
| 15|(((   Heinrich )))))|               Hertz|      116|     pizza|    8|1956/11/30|     and|
| 16|             William|          Gilbert###|      886|      BEER|    2|1958/03/26|  desert|
| 17|               Marie|               CURIE|      912|      Rice|    1|2000/03/22|     you|
| 18|              Arthur|          COM%%%pton|      812|    110790|    5|1899/01/01|       #|
| 19|               JAMES|            Chadwick|      467|      null|   10|1921/05/03|       #|

into this:

| id|firstname|lastname|billingid|          product|price|     birth|  new_date|years_between|
| 10|    james| maxwell|      875|             taco|    3|1923/03/12|12-03-1923|      95.4355|
| 11|    isaac|  newton|      992|            pasta|    9|1999/02/15|15-02-1999|      19.5108|
| 12|     emmy| noether|      234|            pasta|    9|1993/12/08|08-12-1993|      24.6962|
| 13|      max|  planck|      111|       hamburguer|    4|1994/01/04|04-01-1994|      24.6237|
| 14|     fred|   hoyle|      553|            pizza|    8|1997/06/27|27-06-1997|      21.1452|
| 15| heinrich|   hertz|      116|            pizza|    8|1956/11/30|30-11-1956|      61.7204|
| 16|  william| gilbert|      886|             BEER|    2|1958/03/26|26-03-1958|      60.3978|
| 17|    marie|   curie|      912|             Rice|    1|2000/03/22|22-03-2000|      18.4086|
| 18|   arthur| compton|      812|this was a number|    5|1899/01/01|01-01-1899|     119.6317|
| 19|    james|chadwick|      467|             null|   10|1921/05/03|03-05-1921|       97.293|
|  7|     carl|   gauss|      323|             taco|    3|1970/07/13|13-07-1970|      48.0995|
|  8|    david| hilbert|      624|             taco|    3|1950/07/14|14-07-1950|      68.0968|
|  9| johannes|  kepler|      735|             taco|    3|1920/04/22|22-04-1920|      98.3253|

Note that you can use Optimus functions and Spark functions(.WithColumn()) at the same time. To know about all the Optimus functionality please go to this notebooks

Custom functions

Spark have multiple ways to transform your data like rdd, Column Expression ,udf and pandas udf. In Optimus we create the apply() and apply_expr which handle all the implementation complexity.

Here you apply a function to the "billingid" column. Sum 1 and 2 to the current column value. All powered by Pandas UDF

from optimus import Optimus
op = Optimus()

df =op.load.url("")

def func(value, args):
    return value +args[0] + args[1]

df.cols.apply("billingid",func,"int", [1,2]).show()

If you want to apply a Column Expression use apply_expr() like this. In this case we pasa an argument 10 to divide the actual column value

from pyspark.sql import functions as F
from optimus import Optimus
op = Optimus()

df =op.load.url("")

def func(col_name, args):
    return F.col(col_name)/20

df.cols.apply_expr("billingid", func, 20).show()

Data profiling

Optimus comes with a powerful and unique data profiler. Besides basic and advance stats like min, max, kurtosis, mad etc, it also let you know what type of data has every column. For example if a string column have string, integer, float, bool, date Optimus can give you an unique overview about your data. Just run df.profile("*") to profile all the columns. For more info about the profiler please go to this notebook

Machine Learning

Machine Learning is one of the last steps, and the goal for most Data Science WorkFlows.

Apache Spark created a library called MLlib where they coded great algorithms for Machine Learning. Now with the ML library we can take advantage of the Dataframe API and its optimization to create easily Machine Learning Pipelines.

Even though this task is not extremely hard, is not easy. The way most Machine Learning models work on Spark are not straightforward, and they need lots feature engineering to work. That's why we created the feature engineering section inside the Transformer.

To import the Machine Learning Library you just need to say to import Optimus and the ML API:

    from optimus import Optimus

    op = Optimus()

One of the best "tree" models for machine learning is Random Forest. What about creating a RF model with just one line? With Optimus is really easy.

    df_cancer =op.load.url("")
    columns = ['diagnosis', 'radius_mean', 'texture_mean', 'perimeter_mean', 'area_mean', 'smoothness_mean',
           'compactness_mean', 'concavity_mean', 'concave points_mean', 'symmetry_mean',
    df_predict, rf_model =, columns, "diagnosis")

This will create a DataFrame with the predictions of the Random Forest model.

So lets see the prediction compared with the actual label:["label","prediction"]).show()
|  1.0|       1.0|
|  1.0|       1.0|
|  1.0|       1.0|
|  1.0|       1.0|
|  1.0|       1.0|
|  1.0|       1.0|
|  1.0|       1.0|
|  1.0|       1.0|
|  1.0|       1.0|
|  1.0|       1.0|
|  1.0|       1.0|
|  1.0|       1.0|
|  1.0|       1.0|
|  1.0|       1.0|
|  1.0|       1.0|
|  1.0|       1.0|
|  1.0|       0.0|
|  1.0|       1.0|
|  1.0|       1.0|
|  0.0|       0.0|
only showing top 20 rows

The rf_model variable contains the Random Forest model for analysis.

Contributing to Optimus

Contributions go far beyond pull requests and commits. We are very happy to receive any kind of contributions


[Become a backer] and get your image on our README on Github with a link to your site.


[Become a sponsor] and get your image on our README on Github with a link to your site.

Optimus for Spark 1.6.x

Optimus main stable branch will work now for Spark 2.3.1 The 1.6.x version is now under maintenance, the last tag release for this Spark version is the 0.4.0. We strongly suggest that you use the >2.x version of the framework because the new improvements and features will be added now on this version.

Core Team

Argenis Leon and Favio Vazquez


Here is the amazing people that make Optimus possible:



Apache 2.0 © Iron

Logo Iron

Optimus twitter

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Filename, size & hash SHA256 hash help File type Python version Upload date
optimuspyspark-2.1.0-py3-none-any.whl (74.6 kB) Copy SHA256 hash SHA256 Wheel py3 Sep 13, 2018
optimuspyspark-2.1.0.tar.gz (66.8 kB) Copy SHA256 hash SHA256 Source None Sep 13, 2018

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page