Dr. Strangedata or: How I Learned To Stop Worrying and Love the_geom
summary
An application that intersects user areas with reference layers repeatedly fails because worst‑case spatial relations, not data size, drive cost; buffering 5000 drill points creates millions of vertices and causes the system to blow up. Empirical specifics: intersection cost grows with input size and K (edge intersections) — the Martinez R8 clipping complexity appears — so mitigation is to use coarse coverage estimates, simplified reference layers, async execution policies, and realistic test sets via procedural generation rather than purely random geometries.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:18Phases of Spatial Data
- 2:13Something Special About Spatial Data
- 2:26Performance Issues
- 4:06Reactive Manifesto
- 5:09Guarantees vs Technology
- 6:24Basic Spatial Operation
- 7:04Open Source Software
- 13:18Strategies
- 14:06Avoid needlessly complex data
- 15:09Geometry generation tools
related talks
45mBreakout Session
44mHow to use Geospatial Data in SQL Server
19mThe Power of GeoDjango
47mPostGIS and pgRouting: Extensions for spatial data in PostgreSQL
05mWhy Your Data Sucks
46mGetting Down To Earth: Geospatial Analysis With Rails
30mGeospatial analysis made easy with PostGIS and Geoalchemy
54mGeoDjango