Learning FastAPI on hard-mode: getting out lots of the hardest (aw yeah, geo) data fast
summary
Production stack: FastAPI, Pydantic v2, pyproj and shapely serving the G-NAF dataset (about 15 million records split across 203 files) with daily refreshes. Code examples show async endpoints, uvicorn hosting, and using Pydantic models for validation/serialization. GeoJSON geometries include Points, LineStrings, Polygons and quadruply nested MultiPolygons; datum conversions rely on pyproj. Claims include parallel awaiting to reduce combined latency, real responses in the 600–8000 ms range for very large feature collections, and pragmatic tradeoffs around flattened vs authority-file formats.
AI-written summary · about
chapters (22)
- 0:00Introduction
- 1:04What this FastAPI talk covers
- 2:28Australian address data
- 4:13Australian census history
- 5:16National address file
- 6:39FastAPI basics
- 7:01FastAPI code walkthrough
- 8:28Installing FastAPI
- 9:10Live demo: address search
- 9:53Address parsing challenges
- 10:13Async in Python
- 11:40Pydantic V2 overview
- 13:02Pydantic validation
- 16:11GeoJSON standards
- 16:51GeoJSON structure
- 18:13Coordinate systems
- 19:58Pydantic + Shapely geometries
- 21:02GNAF querying
- 21:43Datum conversion
- 22:27Regional data visualization
- 23:10Async parallel fetching
- 24:12Q&A
