# index
> build lookup structures over extracted admin_levels data
## coordinates
1. drops and recreates `admin_levels_rtree`
2. loads all `admin_levels` rows that have geometry, in batches
3. computes the bounding box of each geometry and inserts into the rtree
## hierarchy
1. loads all non-street admin boundaries into memory with their polygon rings and centroids
2. builds an in-memory rtree for fast spatial candidate lookup
3. for each boundary (level ASC), finds the smallest enclosing parent via point-in-polygon
4. writes ancestor id chain and user-friendly name (`"street, city, country"`) to `admin_levels_hierarchy`
5. processes streets in batches against the same in-memory tree
## user_friendly_name
1. builds (or rebuilds) the tantivy full-text index (`admin_levels_hierarchy_tantivy`) at the index path (e.g. `database.tantivy`), one document per `admin_levels_hierarchy` row
2. indexes each row's own name (plus post code, both original and digits-only) and its concatenated ancestor names (`hier`), each in folded / strict / lower field variants used for ranking
3. expands abbreviations bidirectionally (e.g. `rua` ↔ `r.`) so a query matches either form