# extract
parse a `.osm.pbf` file and populate the database layer by layer
## blob_chunks
1. scans the `.osm.pbf` file sequentially, reading 4-byte length-prefixed blob headers
2. records each blob's byte range (`first_byte`, `chunk_size`, `data_first_byte`, `data_size`) and type (header / data)
3. bulk-inserts all chunks into `osm_pbf_blob_chunks`
## header
1. reads the header blob byte range from `osm_pbf_blob_chunks`
2. decodes the OSM header (bbox, required/optional features, writingprogram, osmosis replication metadata)
3. updates the corresponding row in `osm_pbf_files`
## osm_data
1. loads all data blob byte ranges from `osm_pbf_blob_chunks`
2. decompresses each blob (raw or zlib) in parallel threads via a bounded queue
3. decodes protobuf primitive blocks → nodes (dense), ways, relations
4. bulk-inserts into `osm_nodes`, `osm_ways`, `osm_relations`
## admin_levels
1. selects candidate ids per level: relations from `osm_relations` by `admin_level` tag (levels 1–9, and 10), or ways from `osm_ways` by `place`/name (levels 10 and 12) — all requiring a non-null `name`
2. assembles the member way segments into rings, building a polygon when the ring closes, otherwise a line (streets at level 12 are always lines)
3. inserts each feature into `admin_levels` (name, country/post codes, and a spatialite WKB geometry blob)
## house_numbers
1. queries house number nodes from `osm_nodes` (those with `addr:housenumber`)
2. assigns each node to a street (`admin_levels` row) using proximity or name matching
3. inserts into `house_numbers` with geometry and the strategy used