perf(planner): the snow bake skips HR-WSI files that miss a chunk

[timohueser]

Summary

The planner/snow step bakes one zoom-9 chunk at a time. Each chunk opened every HR-WSI file of the region and called transform_bounds before it tested the overlap. transform_bounds sets up a new transform on each call and holds the GIL, so the 4 worker threads did not overlap.

Now laea_footprint gives the EPSG:3035 bounds of each file once per bake. A chunk skips a file that its extent does not touch, before it opens the file. The exact window test for the remaining files does not change.

Measurement

Build: f0d982ee2, on the DACH hr-wsi snapshot 2026-10-16 (2499 files, seasons 2016–2024), one thread, with another bake on the same machine.

Chunk (z9)	Before	After, first chunk	After, cached	Planes equal
268/178	73.0 s	52.5 s	11.6 s	yes
272/182	73.1 s	9.0 s	8.9 s	yes

transform_bounds costs 26.5 ms per call with 1 thread and 26.8 ms with 4 threads. The running DACH bake does about 43 chunks per hour.

Checks

• python.repository-tools (331 tests): pass.
• check_data_sources.py, check_fixture_policy.py, prose.py --check, check_dependencies.py: pass.
• Not run locally: obc-data-steps nextest, obc check fmt and the iOS host check. They do not cover this Python change, and a DACH bake used the shared cargo target. CI runs them.

Public docs: not changed.

Requirements: none

🤖 Generated with Claude Code

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

• Configuration used: defaults
• Review profile: CHILL
• Plan: Advanced
• Run ID: 983db8ac-8ac8-42e3-b67c-8661b9958c84

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}