Improving Accessibility and Usability of the dingo Python Library for Constraint Sampling

[vfisikop]

Project

GeomScale

Summary

This proposal aims to improve the packaging and usability of GeomScale’s dingo Python library by refining its PyPI deployment and simplifying its interface. These enhancements will support broader adoption of efficient constraint sampling methods across fields such as systems biology, Bayesian inference, and computational statistics.

submitter

Vissarion Fisikopoulos

project lead

vfisikop

Community benefit

This project will lower the barrier to entry for using advanced constraint sampling methods by making the dingo Python library easier to install, use, and integrate into existing scientific workflows. By improving its packaging and interface, dingo will better serve researchers working in systems biology, Bayesian inference, and computational statistics—areas where constraint-aware methods are increasingly important. The improved usability will also enhance interoperability with other NumFOCUS projects such as PyMC and Stan, which lack native support for sampling over general constrained domains, and Equadratures, where dingo can complement numerical integration with efficient sampling under convex constraints. In doing so, this project will extend the reach and impact of GeomScale’s methods across a broader segment of the open-source scientific computing ecosystem.

Amount requested

6000

Execution plan

Execution Plan:

M1: Improve PyPI Packaging and Testing
Refactor dingo to resolve dependency issues and ensure smooth installation from PyPI. Set up automated cross-platform testing to guarantee compatibility and reliability.

M2: Simplify Interface and Documentation
Create a unified, user-friendly interface for sampling from convex domains and metabolic networks. Enhance documentation and provide clear examples to support broader adoption.

Timeline:
Months 1–3: Packaging improvements, testing, and PyPI release.
Months 4–6: Interface redesign, documentation, and user feedback integration.

Who (github usernames)
vfisikop, TolisChal
They are both very experienced with the volesti and dingo codebase (already very active contributors).