Bedrock, or b6, is Diagonal's geospatial analysis engine. It reads a structured representation of the world into memory, and makes it available for analysis, for example from a Python script or iPython notebook. It also provides a rudimentary web interface for exploring the available data. Communication between Python and b6 happens over GRPC.
We use b6 for the analysis behind our work for clients. Typically, we generate results for different scenarios as JSON, and load them into bespoke visualisations built with d3. We also directly embed the Go library into custom binaries to build tools that support dynamic analysis.
Working in the urban environment, the impact of the projects we support on the communities around them outlasts our own involvement. We have a duty to allow these communities to understand and build on our work beyond our time with the project. We open sourced b6 to enable this - and to comply with our charter, which requires transparency.
We built b6 for ourselves, and don't expect it to be useful for a wide range of people. We've also yet to write as much public documentation as we'd like. However, if you do find it useful, or interesting, we'd be excited to hear about what you're up to.
We provide a docker package for the b6 backend. Running:
docker run -p 8001:8001 -p 8002:8002 europe-docker.pkg.dev/diagonal-public/b6/b6
Will start an instance of the backend, with a web interface on port 8001, and a GRPC interface for analysis from Python on port 8002, hosting a small amount of data from OpenStreetMap for the area of London around Diagonal's spiritual home. Viewing localhost:8001 should show you a map.
To try out analysis, you'll need to install the Python client library, via:
python -m pip install diagonal_b6
Right now, the best (and only) documentation for analysis functionality are the unit tests.
In the web interface, a left click will recenter the
map. Pressing backtick will drop down a terminal window. Shift click will
tell you about the geographic feature at that location. Entering a lat, lng
in the terminal will jump you to that location. Entering
find [#amenity=cafe] | take 10 | show will highlight 10 cafe, as the terminal
supports all the functionality available through Python. We've yet to properly
document it publicly, however.
If you have a small amount of data you'd like to work with, in OSM PBF format,
you can read it directly by putting in a directory by itself and replacing the
/world directory in the image:
docker run -v /path/with/data:/world -p 8001:8001 -p 8002:8002 europe-docker.pkg.dev/diagonal-public/b6/b6
For larger datasets, or datasets in formats other than OSM PBF, you'll need to
use one of the ingestion tools from either the
docker image. These tools convert source data into a compact representation for
efficient reading by the backend, that we call an index. b6-ingest-osm produces
an index for OpenStreetMap data, while b6-ingest-gdal will produce an index for
anything the GDAL library can ready. Our ingest tools aren't documented
publicly yet.