NOTE this project is not production ready at all, maybe never will, but may be useful nonetheless.
This is a collection of scripts that can process data extracts from OpenStreetMap (PBF files) allowing the creation of static sites (that is, a bunch of file hosted without any backend processing) to do a few useful operations:
- display an interactive map (zoom, pan, scroll) based on vector tiles
- locate addresses and places by name
- TODO: nearest neighbor
- TODO: routing (road network can be extracted, but it's not reasonably usable by the frontend)
Activate the virtual environment and then:
pip install static_osm_indexer
(Or clone this repo and use make install)
this will make a set of CLI utilities, all prefixed with soi_, available. The library can also be used programmatically.
This command will generate a fully static map from a PBF extract
soi_generate_full_map italy-latest.osm.pbf 3.8,36.2,20.2,47.8 test_italy_all
it needs the name of the folder where to store it and the bounding box.
Use --help to see all the options and their usage
Use the --publish-address to specify the URL of your map, including protocol and port. By default is http://127.0.0.1:9000, be aware that mapbox needs absolute addresses, so it must match the protocol, host and port of your deployment. If missing, can be trivially changed afterwards on the generated files.
Run soi_list_named_locations, this will generate a file in which every line is a JSON with a name field and lat, lon coordinates (as EPSG:4326).
By default it will extract the name tag which generally corresponds to the local name, but you can add specific locales, for example with --tags 'name,name:it' you will get local names and Italian names when available. Duplicates are ignored.
Once you have a file with the list of locations you need to index it using soi_index_location_names. This command requires an output folder for the output, and will rearrange the large location names file into smaller files quicker to retrieve on the fly. In this folder a file called index_metadata.json contains the metadata needed by the frontend to use the index.
The default config is fine for most cases, but there are two improvements you may want to look into:
-
--stopwordsallows you to ignore words that are very common in addresses, for example the word for street in your language. Using it you can create a more balanced index. -
--token_lengthis the amount of characters to be retrieved before fetching a file. By default 3, if you are processing Chinese or Japanese you should set it to 1 given the different statistical distribution of ideograms.
Use soi_extract_road_network to extract the road network graph into a SQLLite database. Use --help for further instructions, it has flags to filter for the walking, bicycle and car network. The --collapse-distance flag allows to aggregate nodes that are close together to greatly reduce the complexity of the output.
The soi_road_network_to_geojson command can produce a geoJSON representation of a road network. It will generate a file with the edges and another with only the nodes, for inspection. Notice that unless the area is very small or you used the --collapse-distance flag these files are going to be quite large.
This project is under MIT license. Using the data from OSM is fine including commercial projects, but there are some rules and you must credit OpenStreetMap.