# Real-world fixtures
These checked-in fixtures are the fastest local proof path for the repo's schema-first workflow.
They back the end-to-end conversion/query tests in `src/engine.rs`, and they are the safest first
stop before claiming that a new doc flow or object-storage workflow works remotely.
## How agents should use these fixtures
When following the tracked `AGENTS.md` guidance or `.codex/skills/object-storage-ops/SKILL.md`:
1. convert one of these fixtures to Parquet locally
2. inspect schema first
3. run a direct CLI query
4. only then mirror the same flow against S3/MinIO if remote access is part of the task
## Our World in Data COVID-19
### CSV fixture
- File: `owid-covid-latest.csv`
- Source: Our World in Data COVID-19 latest dataset
- Direct download URL: `https://raw.githubusercontent.com/owid/covid-19-data/master/public/data/latest/owid-covid-latest.csv`
- Project docs: `https://docs.owid.io/projects/covid/en/latest/dataset.html`
- Retrieved: 2026-04-14
- Covered by: `real_world_owid_csv_converts_and_queries`
### JSON fixture
- File: `owid-covid-latest.json`
- Source: local OWID-derived object-map fixture for JSON ingestion coverage
- Derived from: `owid-covid-latest.csv`
- Retrieved: 2026-04-14
- Covered by: `real_world_owid_json_object_map_converts_and_queries`
## Government of Canada wastewater CSV
- File: `canada-wastewater-aggregate.csv`
- Source: Government of Canada / Public Health Infobase wastewater dashboard
- Direct download URL: `https://health-infobase.canada.ca/src/data/wastewater/wastewater_aggregate.csv`
- Source page: `https://health-infobase.canada.ca/covid-19/wastewater/`
- Retrieved: 2026-04-14
- Covered by: `canada_government_csv_converts_and_queries`
## Spreadsheet fixture
- File: `keyfoods_0708.xlsx`
- Source: checked-in spreadsheet fixture for XLSX ingestion coverage
- Retrieved: 2026-04-14
- Covered by: `real_world_usda_xlsx_converts`
## Notes
- Synthetic benchmark parquet files are generated under `benchmark-data/` or `tmp/`; they are not checked in here.
- When refreshing a fixture, keep the retrieval date and covering test name in sync with the code.