Everybody interested in learning more about probabilistic data structures and algorithms could be referred to our recently published book:

Probabilistic Data Structures and Algorithms for Big Data Applications by Andrii Gakhov

2019, ISBN: 978-3748190486 (paperback) ASIN: B07MYKTY8W (e-book)

Probabilistic data structures is a common name of data structures based on different hashing techniques.

Unlike regular (or deterministic) data structures, they always provide approximated answers, but usually with reliable ways to estimate the error probability.

The potential losses or errors are fully compensated by extremely low memory requirements, constant query time and scaling.

GitHub repository: https://github.com/gakhov/pdsa

• Python 3.5+ (http://python.org/download/)
• Cython 0.28+ (http://cython.org/#download)

The latest documentation can be found at http://pdsa.readthedocs.io/en/latest/

Membership problem

• Bloom Filter
• Counting Bloom Filter

Cardinality problem

• Linear counter
• Probabilistic counter (Flajolet–Martin algorithm)
• HyperLogLog

Frequency problem

• Count Sketch
• Count-Min Sketch

Rank problem

• Random Sampling
• q-digest

MIT License

• https://github.com/gakhov/pdsa/

• Maintainer: Andrii Gakhov <[email protected]>

Installation requires a working build environment.

Using pip, PDSA releases are currently only available as source packages.

$ pip3 install -U pdsa

When using pip it is generally recommended to install packages in a virtualenv to avoid modifying system state:

$ virtualenv .env -p python3 --no-site-packages
$ source .env/bin/activate
$ pip3 install -U cython
$ pip3 install -U pdsa

The other way to install PDSA is to clone its GitHub repository and build it from source.

$ git clone https://github.com/gakhov/pdsa.git
$ cd pdsa

$ make install

$ bin/pip3 install -r requirements-dev.txt
$ make test