matbon01 / joinbench

Imperial 4th year Mathematics and Computer Science MEng project

JOINBENCH: Designing Tools for the Evaluation of Efficient Equijoins in Haskell

In order to evaluate alternative equijoin implementation in Haskell, this paper introduced the JOINBENCH relation and surrounding methodology and tooling. The JOINBENCH relation is a scheme designed for the synthesis of data optimised to help evaluate the performance of equijoins in a variety of scenarios. Furthermore, this paper presents a low-level library that helps users customise and define their own synthetic data sources for future benchmarking purposes.

In their distinguished paper "Relational Algebra by way of Adjunctions" it was noted that the monadic structure of bulk types can help explain most of relational algebra. Using this structure, the authors designed a new method to facilitate the use of monad comprehensions in an efficient implementation of equijoins of relational databases. This project presents an implementation of such a system and an evaluation of the performance such query optimisations carry using the bespoke tooling described above.

Repository outline

• Database system library see: src/, test/ (requires environment as described by joinbench.cabal).
• Database benchmarking client code: app/, test/ (requires environment as described by joinbench.cabal).
• Synthetic database generation modules: database_generation/.
• Data visualisation and analysis modules: analysis/ (requires environment set up using conda-env.txt).
• Report: report/.
• Various automations: see Makefiles in directories and .github/ for CI/CD pipeline.

joinbench's Issues

Add custom words to spelling dictionary

Proof read and fix adjunction definition draft

Report spelling errors to the status in a more useful way

Fix need for "def" after definitions I want to turn into commands

Add software pipeline

Spelling check takes certain latex variables as words

Rename linter to report linter

Reporting linting issues cannot deal with errors that include "

I am not sure if it is about uneven number of "s, but it breaks the pipeline

Add page count to status

Deal with linting issues in a better way

Currently the status of the commit ignores any linting issues

Add word count to status