Supplement for the paper - Efficient and Portable Einstein Summation in SQL

You will find information about our group and other projects at ti2.uni-jena.de.

If you are not familiar with einstein summation, good entrypoints are the article Einstein Summation in NumPy or Einsum is all you Need - Einstein Summation in Deep Learning.

Structure of the Repository

.
├── case_study                 # Stand alone files for some case studies of the paper
│   ├── discussion             # Stand alone files for the discussion section
│   ├── graphical_models       # Stand alone files for the graphical model experiments
│   └── quantum_circuits       # Stand alone files for the quantum circuits experiments
├── experiments                # Experimental files to reproduce the results of the paper
└── generate_sql_code          # Stand alone file for generating sql-code

See each folder for more information.

Requirements

For a docker setup see below.

Python and Anaconda

All files were tested using Python version 3.8 and Anaconda version 4.9.2

Further we require the following Python packages.

• numpy (version 1.21.1)
• pandas (version 1.4.2)
• tableauhyperapi (version 0.0.13287)
• psycopg2 (version 2.9.3)
• opt_einsum (version 3.3.0)
• rdflib (version 6.2.0)
• python-sat (version 0.1.7.dev19)

All packages can be installed using Anaconda and the yaml file with

conda env create -f environment.yml

after this the environment is activated with

conda activate sql-einstein

and can be deleted with

conda env remove -n sql-einstein.

Postgres

The experiments further need a postgres DBMS installation. We used psql version 12.7. You can install postgres on linux using

sudo apt-get install postgresql-12.11 postgresql-contrib-12.11

The postgres installation should have a database with the following configuaration:

• name: 'postgres',
• user: 'postgres',
• password: 'password',
• host: 'localhost'.

Docker Setup

We also support a docker setup. A Dockerfile is available and can be build with

docker build -t sql-einstein .

After building you can run the docker container with

docker run --rm -it sql-einstein

This command removes (--rm) the container after execution. Also it runs the script startup.sh as an entrypoint. This script does start the postgres server.

After running the docker run ... command you will be prompted with an interactive shell where you can run experiments as described in each subfolder. For a quickstart you can run

cd experiments
python run_experiments

to start all experiments or

cd case_study/quantum_circuits
python qc_sqlite.py

to start the quantum circuit sqlite experiment.