QLever Petrimaps

Interactive visualization of SPARQL query results with geospatial information on a map, either as individual objects or as a heatmap. Petrimaps can display hundreds of millions of objects interactively, for example, a heatmap of all the streets in a given country, or even in the whole world. Petrimaps is implemented as a middleware. In the QLever UI, whenever a SPARQL result has WKT literals in its last column, a "Map view" button is displayed, which leads to an instance of Petrimaps.

Here are two example queries: All railway lines in OSM, All streets in Germany in OSM.

Requirements

• gcc > 5.0 || clang > 3.9
• xxd
• libcurl
• libpng (for PNG rendering)
• Java Runtime Environment (for compiling the JS of the web frontend)

Optional Requirements

• zlib (for gzip compression)
• OpenMP

Installation

Compile yourself:

$ git clone --recurse-submodules https://github.com/ad-freiburg/qlever-petrimaps
$ cd qlever-petrimaps
$ mkdir -p build && cd build
$ cmake ..
$ make

via Docker:

$ docker build -t petrimaps .

Usage

To start:

$ petrimaps [-p <port=9090>] [-m <memory limit] [-c <cache dir>]

Requests can be send via the ?query get parameter. The QLever backend to use must be specified via the ?backend get parameter.

IMPORTANT: the tool currently expects the geometry to be the last selected column.

PREFIX geo: <http://www.opengis.net/ont/geosparql#>
PREFIX osmkey: <https://www.openstreetmap.org/wiki/Key:>
SELECT ?osm_id ?geometry WHERE {
  ?osm_id osmkey:building ?building .
  ?osm_id geo:hasGeometry ?geometry .
}

Cache + Memory Management

The tool caches query results and memory usage will thus slowly build up. There is a primitive memory limit which can be set via the -m parameter (in GB). By default, 90% of the available system memory are used.

If a query runs out of memory, you can clear all existing caches by requesting

/clearsession

/clearsessions will also work. Optionally, you can specify the session id via `?id='.

Disk Cache

If -c specifies a serialization cache directory, the complete geometries downloaded from a QLever backend will be serialized to disk and re-used on later startups. This significantly speeds up the loading times.