2025 FOSS4G NA | Taming Dependency Hell - Thomas Machler

In this session from FOSS4G NA 2025, Thomas Machler from Planet introduces a powerful solution for "tame dependency hell"—Pixi. As a fusion of UV and Conda, Pixi is designed to handle the complex intersection of Python dependencies and system-level binaries that often plague geospatial developers.

Thomas shares his firsthand experience at Planet, where managing custom GDAL versions and GPU-supported libraries like TensorFlow required hours of manual configuration. He demonstrates how Pixi automates this process by merging PyPI and Conda dependency trees into a single, reliable lock file, ensuring your local environment perfectly matches your CI/CD and production Docker images.

Highlights:

🛰️ What is Pixi? A deep dive into the "fusion of UV and Conda" that supports multiple platforms and languages

🔄 Beyond Python: Managing system-level binaries (like GDAL) that standard tools like UV or Poetry can't handle alone

🧩 Feature Sets: How to test your project against multiple GDAL versions (3.8 through 3.11) simultaneously using Pixi features

⚙️ Eliminating Makefiles: Using Pixi Tasks to centralize environment activation and CI/CD commands in one file

🌐 Docker Optimization: Building sleek, high-performance images by separating build essentials from runtime environments

🏢 Mono-repo Support: Managing conflicting dependencies and different Python versions within the same repository

For more content like this check out www.projectgeospatial.com

#Pixi #FOSS4G #Geospatial #Python #GDAL #Planet #OpenSource #SoftwareEngineering #Docker #Conda #PackageManagement #Rasterio #ProjectGeospatial #GIS #DevOps #DataScience