PostGIS schema

MapLark stores OpenStreetMap in three geometry tables in schema osm. Use this page to choose type, way_shape, and tags so a query hits the right table and index. You query through HTTP or the SDK. There is no SQL endpoint and no database URL for API keys.

Three tables

Geometry is a single geom column in EPSG:4326. Remaining OSM tags stay in tags JSONB. A few high-traffic keys are also copied onto typed text columns so common filters stay index-friendly.

Table	What it holds	Typical query
`osm.osm_point`	Tagged nodes. Always a Point. POIs, peaks, bus stops, benches.	`type=node`
`osm.osm_line`	Linear ways and some relations (routes, boundaries). Roads, paths, rivers, roundabouts. Closed highways stay lines, not polygons.	`type=way&way_shape=line`
`osm.osm_polygon`	Area ways and multipolygon or boundary relations. Buildings, parks, lakes, landuse. Closed ways with area tags land here.	`type=way&way_shape=polygon`

How to pick a table

type is the OSM element kind: node, way, or relation. way_shape is the geometry class for ways and relations only. Nodes always go to osm_point.

Nodes (POIs, peaks, benches): type=node.
Roads, paths, rivers: type=way or type=relation, and way_shape=line.
Buildings, parks, landuse: way_shape=polygon.
POIs that may be a node or a building polygon: type=node,way&way_shape=polygon.
Unknown mix: omit both. The API unions all three tables.

Reference mapping:

Params	Reads
`type=node`	`osm.osm_point`
`type=way&way_shape=line` or `type=relation&way_shape=line`	`osm.osm_line`
`type=way&way_shape=polygon`	`osm.osm_polygon`
`type=node,way&way_shape=polygon`	points plus polygons
omit both	points, lines, and polygons

limit applies per table or union arm. Omitting way_shape for ways can return up to two pages of that size (line plus polygon). Set way_shape when the geometry class is known. Parameter reference: OSM Features API.

Columns

Every table has osm_id, osm_type (N / W / R), tags JSONB, and geom. GeoJSON id looks like way/123. Properties keep the full tag map. Import drops noisy prefixes such as source, tiger, and created_by.

Table	Extra columns	Typed tag columns (common filters)
`osm_point`		`place`, `natural`, `aeroway`, `amenity`, `shop`, `tourism`, `public_transport`, `opening_hours`
`osm_line`	`length_m`	`highway`, `railway`, `waterway`, `aerialway`, `boundary`, `admin_level`, `tunnel`, `bridge`, `oneway`, `layer`, `ref`
`osm_polygon`	`area_m2`	`building`, `landuse`, `natural`, `leisure`, `amenity`, `shop`, `aeroway`, `waterway`, `water`, `boundary`, `admin_level`, `opening_hours`

Line size filters: min_length_m, max_length_m.
Polygon size filters: min_area_m2, max_area_m2.
Any other OSM key still filters through tags as key, key=value, or a numeric compare such as ele>500.

Indexes

Plan queries so they can use these families. Always include a spatial anchor: bbox, location plus radius, within, or osm_ids. Tag-only scans are rejected.

Index	On	What it speeds up
GiST on `geom`	all three	Bounding box and around envelope (`geom &&` an envelope). Radius queries then refine with distance.
GIN on `tags`	all three	Arbitrary OSM keys that are not promoted to a typed column.
btree on `(osm_type, osm_id)` and `osm_id`	all three	`osm_ids` lookup, keyset pagination, and `within` container load.
btree on name	all three	Name filters.
btree on `length_m` / `area_m2`	line / polygon	Size filters.
Partial btree on typed tag columns	per table	Hot keys such as `amenity`, `highway`, `building`, `shop`, `natural`, `landuse`. Prefer those tags when they match the ask.

Places search uses the point-plus-polygon union (POIs that are nodes or building polygons).
Routing reads highway-tagged rows from osm.osm_line.
Stats histograms use the same tables and filters as GeoJSON, with larger spatial caps.

How to plan a query