06 — Project Archive

Notebook: examples/06_project_archive.py

Full-project archival demonstration using archive_project() and inspect_project() on the BaldEagleCrkMulti2D example project.

What it demonstrates

1. Extract BaldEagleCrkMulti2D via RasExamples
2. inspect_project() — discover geometry/plan files without extracting anything
3. archive_project() — export all geometry layers to a consolidated GeoParquet archive
4. Read manifest.json and inspect layer inventory
5. Filter by layer column to get homogeneous geometry types
6. Query with DuckDB using WHERE layer = 'mesh_cells'

Run it

marimo edit examples/06_project_archive.py
# or
python examples/06_project_archive.py

Expected output structure

out/06_project_archive/
├── manifest.json
├── BaldEagleCrkMulti2D.parquet       # Project metadata (RasPrj dataframes)
└── BaldEagleCrkMulti2D.g01.parquet   # All geometry layers in one file

Each consolidated GeoParquet contains a layer column for filtering:

-- Query specific layers with DuckDB
SELECT * FROM 'Model.g01.parquet' WHERE layer = 'mesh_cells'
SELECT * FROM 'Model.g01.parquet' WHERE layer = 'bc_lines'

-- List available layers
SELECT DISTINCT layer FROM 'Model.g01.parquet'

-- Project metadata
SELECT * FROM 'Model.parquet' WHERE _table = 'plan_df'

Validate with geoparquet-io (optional)

# Best practice: validate GeoParquet spec compliance
pipx install geoparquet-io
gpio check all output/Model.g01.parquet
gpio inspect output/Model.g01.parquet

Notes

• BaldEagleCrkMulti2D has no plan HDF results (model not pre-run), so --results would yield no output
• Terrain COG conversion is skipped in this notebook (set include_terrain=True to enable)
• All layers that are absent from this particular model are silently skipped
• Hilbert spatial sorting is applied within each layer by default for optimal query performance
• Per-row bounding box columns (bbox_xmin, bbox_ymin, bbox_xmax, bbox_ymax) enable spatial predicate pushdown in DuckDB
• ZSTD compression is used instead of snappy for better compression ratios