CubedPandas

Multi-dimensional data analysis for Pandas dataframes.

---

Note: CubedPandas is at an early stage of its development. Features may change. Ideas, Issues and Feedback are very welcome!

CubedPandas provides an easy, fast & fun approach to perform multi-dimensional data analysis on Pandas dataframes. CubedPandas wraps almost any dataframe into a multi-dimensional cube, which can be analyzed, aggregated, sliced, filtered, viewed and much more.

CubedPandas is inspired by OLAP databases (online analytical processing), which are typically used for reporting, business intelligence, data warehousing and financial analysis purposes. Check out the sample code below and give it a try.

CubedPandas is licensed under the BSD 3-Clause License and is available on GitHub and PyPi.

If you have fallen in love with CubedPandas or find it otherwise valuable, please consider to sponsor the project on GitHub Sponsors, so many cool ideas/features to come.

Getting started

After installing CubedPandas...

pip install cubedpandas

...you are ready to go. "Cubing" a Pandas DataFrame is as simple as this:

import pandas as pd
from common import cubed

df = pd.DataFrame({"product":  ["Apple",  "Pear",   "Banana", "Apple",  "Pear",   "Banana"],
                   "channel":  ["Online", "Online", "Online", "Retail", "Retail", "Retail"],
                   "customer": ["Peter",  "Peter",  "Paul",   "Paul",   "Mary",   "Mary"  ],
                   "mailing":  [True,     False,    True,     False,    True,     False   ],
                   "revenue":  [100,      150,      300,      200,      250,      350     ],
                   "cost":     [50,       90,       150,      100,      150,      175     ]})

cdf = cubed(df)  # That's it! 'cdf' is now a CubedDataFrame

Multi-Dimensional Cubes?

CubedPandas automatically infers a multi-dimensional schema from the dataframe. By default, numeric columns are considered as measures - the values to analyse & aggregate - all other columns are considered as the dimensions to filter, navigate and view the data. The values in the dimensions are called members.

But you can also define your own schema. Schemas are quite powerful and flexible, as they will allow you to define not only your dimensions and measures, but also custom aggregations, some business logic, sorting, number formating etc. Note: This feature not yet available, planned for release 0.2.0.

Multi-Dimensional Cells - Numbers Please...

One key feature of CubePandas is its easy & intuitive access to individual data cells in the cube. You define a multi-dimensional address and CubedPandas will evaluate and return the corresponding value.

Cells behave like numbers (float, int), so you can use them in arithmetic operations. In the following examples, all addresses will refer to the same data and return the same value of 100.

# Using a proper, fully qualified and unambiguous address, order doesn't matter
a = cdf["product:Apple", "channel:Online", "customer:Peter", "revenue"]

# If there are no ambiguities in your data, you can use a shorthand address
b = cdf["Online", "Apple", "Peter", "revenue"]

# And if member values are compliant with Python naming conventions, you can use
c = cdf.Online.Apple.Peter.revenue

assert a == b == c  == 100

It's All About Slicing & Aggregating

CubedPandas allows you to slice & dice your data. You can filter by dimensions, aggregate by measures in a very convenient way. The first measure in the dataframe (left to right) is used as the default measure, here revenue. So, if no measure is specified, the default measure is used, hence cdf["Apple", "Online"] is equivalent to cdf["Apple", "Online", "revenue"].

a = cdf["Online"]              # 550 = 100 + 150 + 300
b = cdf["product:Banana"]      # 650 = 300 + 350
c = cdf["Apple", "cost"]       # 100 = 100 -> explicit sum
d = cdf["Apple", "cost"].avg   # 75 = (50 + 100) / 2
e = cdf["*"]                   # 1350 -> '*' means 'all' 
f = cdf.count                  # 6 -> number of affected records
g = cdf["customer:P*"]         # 750 = 100 + 150 + 300 + 200  -> wildcard search
h = cdf.Peter + cdf.Mary       # 850 = (100 + 150) + (250 + 350)
i = cdf.cost                   # 715 -> sum of all records for the measure 'cost'

Seeing Is Believing

Note: This feature not yet available, planned for release 0.2.0

Pandas printing capabilities are tied to the tabular structure of a dataframe. CubedPandas unlocks the full potential of dataframes by providing multi-dimensional views. These are called Slice and look like an Excel Pivot-Table.

The following code will show a slice with product and customer nested on the rows, channel on the columns, and mailing as a filter. The default measure revenue is used.

cdf.slice(("product", "customer"), "channel", "mailing").show()

Your feedback is very welcome!

Please help improve and extend CubedPandas with your feedback & ideas and use the CubedPandas GitHub Issues to request new features and report bugs. For general questions, discussions and feedback, please use the CubedPandas GitHub Discussions.

Enjoy ... and happy cubing!