osmo

osmo copied to clipboard

osmo: Populate a MongoDB collection with OpenStreetMap data

Features

• osmo populates a MongoDB database collection with all nodes, ways and relations

• Currently, only OSM XML (.osm) is supported

• It's just slightly slower than Osmosis (with Postgres)

• Prerequisites: see DEPENDENCY

• Building:

  make

  (A priority is it to make this work in both Mac OS and Linux!)

• Synopsis:

  bin/osmo DATAFILE DATABASE.COLLECTION

• To add an index on the primary key:

  mongo DATABASE --eval
  "printjson(db.COLLECTION.ensureIndex({ p: 1, id: 1 }, { unique: true, dropDups: true, background: true }))"

Data model

The data model supported by osmo is focussed on minimal data size and unification. Some principles:

• nodes, ways and relations are transformed to GeoObject instances to unify the concept and to simplify queries on only 1 collection

• Every geoObject has a key named "p" (prefix). It can have the values, "n" (node), "w" (way) or "r" (relation).

• keys for regular fields (user, timestamp, ...) are abbreviated to save space

• The geoObject primary key is [ 'p', 'id' ]

• Example for a GeoObject instance:

  { "_id" : ObjectId("4e3213dbe9b522a655e3f5ff"), "c" : 4977984, "id" : 318214, "p" : "r", "parts" : { "list" : [ { "p" : "n", "id" : 319621296 }, { "p" : "n", "id" : 560090559 }, { "p" : "n", "id" : 292635574 }, { "p" : "n", "id" : 292635576 }, { "p" : "n", "id" : 245431191 }, { "p" : "w", "id" : 48289915 }, { "p" : "w", "id" : 49366289 }, ... ], "info" : [ { "role" : "forward" }, { "role" : "forward" }, { "role" : "" }, { "role" : "forward" }, { "role" : "" }, { "role" : "forward" }, { "role" : "forward" }, ... ] }, "t" : 1276427410, "tags" : { "network" : "VRT", "operator" : "SWT", "ref" : "4", "route" : "bus", "type" : "route" }, "u" : { "id" : 109925, "name" : "WanMil" }, "v" : 6 }

• "parts" is an ObjectView object that has always two properties: "list" (a list of GeoObject references) and "info" (that optionally holds information about the referenced objects in the same order). GeoObjects with p=n do not have parts, with p=w, parts are always references to nodes, with p=r parts can reference any type (node, way, relation).

Contribute

Take a look at TODO to see a list of priorized features that are still lacking.