openpoiservice 0.1.7

POI service consuming OpenStreetMap data

Openpoiservice

Openpoiservice (ops) is a flask application which hosts a highly customizable points of interest database derived from OpenStreetMap.org data and thereby exploits it's notion of tags...

OpenStreetMap tags consisting of a key and value describe specific features of map elements (nodes, ways, or relations) or changesets. Both items are free format text fields, but often represent numeric or other structured items.

This service consumes OSM tags on nodes, ways and relations by grouping them into predefined categories. If it picks up an OSM object tagged with one of the osm keys defined in categories.yml it will import this point of interest with specific additional tags which may be defined in ops_settings.yml. Any additional tag, for instance wheelchair or smoking may then be used to query the service via the API after import.

For instance, if you want to request all pois accessible by wheelchair within a geometry, you could add then add wheelchair: ['yes', 'dedicated] in filters within the body of your HTTP POST request.

You may pass 3 different types of geometry within the request to the database. Currently "Point" and "LineString" with a corresponding and buffer are supported as well as a polygon. Points of interest will be returned within the given geometry.

You can control the maximum size of geometries and further restrictions in the settings file of this service.

Import Process

The osm file(s) to be imported are parsed several times to extract points of interest from relations (osm_type 3), ways (osm_type 2) and nodes (osm_type 1) in order. Which type the specific point of interest originated from will be returned in the response - this will help you find the object directly on OpenStreetMap.org.

Installation

You can either run openpoiservice on your host machine in a virtual environment or simply with docker. The Dockerfile provided installs a WSGI server (gunicorn) which starts the flask service on port 5000.

Technical specs for storing and importing OSM files

Python version

As this service makes use of the python collections library, in particular the notion of deque's and its functions it only supports python 3.5 and greater.

Database

This application uses a psql/postgis setup for storing the points of interest. We highly recommend using this docker container.

Importer

Please consider the following technical requirements for parsing & importing osm files.

Region	Memory
Germany	8 GB
Europe	32 GB
Planet	128 GB

Note: Openpoiservice will import any osm pbf file located in the osm folder or subdirectory within. This way you can split the planet file into smaller regions (e.g. download from Geofabrik, scraper script for the download links to be found in the osm folder) and use a smaller instance to import the global data set (as long as the OSM files don't exceed 5 GB of disk space, 16 GB of memory will suffice to import the entire planet).

Run as Docker Container (Flask + Gunicorn)

Make your necessary changes to the settings in the file ops_settings_docker.yml. This file will be copied to the docker container. If you are planning to import a different osm file, please download it to the osm folder (any folder within will be scanned for osm files) as this will be a shared volume.

Afterwards run:

$ docker-compose up -d -f /path/to/docker-compose.yml

Once the container is built you can either, create the empty database:

$ docker exec -it container_name /ops_venv/bin/python manage.py create_db

Delete the database:

$ docker exec -it container_name /ops_venv/bin/python manage.py drop_db

Or import the OSM data:

$ docker exec -it container_name /ops_venv/bin/python manage.py import_data

Run in a Virtual Environment

1. Create and activate a virtualenv
2. This repository uses imposm.parser to parse the OpenStreetMap data. To this end, make sure google's protobuf is installed on your system:

• Ubuntu (16.04 and earlier, supported on 17.10): most likely you will have to install protobuf from source if https://imposm.org/docs/imposm.parser/latest/install.html#requirements doesn't do the job.
• OS X Using homebrewon OS Xbrew install protobuf` will suffice.

3. Afterwards you can install the necessary requirements via pipwith pip install -r requirements.txt

Prepare settings.yml

Update openpoiservice/server/ops_settings.yml with your necessary settings and then run one of the following commands.

[

$ export APP_SETTINGS="openpoiservice.server.config.ProductionConfig|DevelopmentConfig"

]

Create the POI DB

$ python manage.py create_db

Drop the POI DB

$ python manage.py drop_db

Parse and import OSM data

$ python manage.py import_data

Run the Application with Flask-Werkzeug

$ python manage.py run

Per default you can access the application at the address http://localhost:5000/

Want to specify a different port?

$ python manage.py run -h 0.0.0.0 -p 8080

Tests

$ export TESTING="True" && python manage.py test

Category IDs and their configuration

openpoiservice/server/categories/categories.yml is a list of (note: not all!) OpenStreetMap tags with arbitrary category IDs. If you keep the structure as follows, you can manipulate this list as you wish.

transport:
   id: 580
   children:
       aeroway:
           aerodrome: 581        
           aeroport: 582 
           helipad: 598         
           heliport: 599 
       amenity:
           bicycle_parking: 583  

sustenance:
   id: 560             
   children:
       amenity:
           bar: 561             
           bbq: 562   
...

Openpoiservice uses this mapping while it imports pois from the OpenStreetMap data and assigns the custom category IDs accordingly.

column_mappings in openpoiservice/server/ops_settings.yml controls which OSM information will be considered in the database and also if these may be queried by the user via the API , e.g.

wheelchair:

smoking:

fees:

For instance means that the OpenStreetMap tag wheelchair will be considered during import and save to the database. A user may then add a list of common values in the filters object wheelchair: ['yes', 'dedicated', ...] which correspond to the OSM common values of the tag itself, e.g. https://wiki.openstreetmap.org/wiki/Key:wheelchair.

API Documentation

The documentation for this flask service is provided via flasgger and can be accessed via http://localhost:5000/apidocs/.

Generally you have three different request types pois, stats and list.

Using request=pois in the POST body will return a GeoJSON FeatureCollection in your specified bounding box or geometry.

Using request=stats will do the same but group by the categories, ultimately returning a JSON object with the absolute numbers of pois of a certain group.

Finally, request=list will return a JSON object generated from openpoiservice/server/categories/categories.yml.

Examples

POIS around a buffered point

curl -X POST \
  http://localhost:5000/pois \
  -H 'Content-Type: application/json' \
  -d '{
  "request": "pois",
  "geometry": {
    "bbox": [
      [8.8034, 53.0756],
      [8.7834, 53.0456]
    ],
    "geojson": {
      "type": "Point",
      "coordinates": [8.8034, 53.0756]
    },
    "buffer": 250  
  }
}'

POIs of given categories

curl -X POST \
  http://localhost:5000/pois \
  -H 'Content-Type: application/json' \
  -d '{
  "request": "pois",
  "geometry": {
    "bbox": [
      [8.8034, 53.0756],
      [8.7834, 53.0456]
    ],
    "geojson": {
      "type": "Point",
      "coordinates": [8.8034, 53.0756]
    },
    "buffer": 100  
  },
  "limit": 200,
  "filters": {
    "category_ids": [180, 245]
  } 
}'

POIs of given category groups

curl -X POST \
  http://localhost:5000/pois \
  -H 'Content-Type: application/json' \
  -d '{
  "request": "pois",
  "geometry": {
    "bbox": [
      [8.8034, 53.0756],
      [8.7834, 53.0456]
    ],
    "geojson": {
      "type": "Point",
      "coordinates": [8.8034, 53.0756]
    },
    "buffer": 100  
  },
  "limit": 200,
  "filters": {
    "category_group_ids": [160]
  } 
}'

POI Statistics

curl -X POST \
  http://129.206.7.157:5005/pois \
  -H 'Content-Type: application/json' \
  -d '{
  "request": "stats",
  "geometry": {
    "bbox": [
      [8.8034, 53.0756],
      [8.7834, 53.0456]
    ],
    "geojson": {
      "type": "Point",
      "coordinates": [8.8034, 53.0756]
    },
    "buffer": 100  
  }
}'

POI Categories as a list

curl -X POST \
  http://127.0.0.1:5000/pois \
  -H 'content-type: application/json' \
  -d '{
	"request": "list"
}'