jsbeckwith / entwiner

WIP: Create and use routable graph databases from a wide range of data sources

Note: this is under very active and early development. Don't use for production

applications yet!

Entwiner

Build, use, and share routable transportation graphs using common geospatial data.

Installation and Usage

Installation

If you use pip version 19 or greater, entwiner can be installed directly:

pip install git+https://github.com/nbolten/entwiner

If you have a version of pip lower than 19, you can install using poetry or render a pip-installable sdist using poetry build, which creates a pip-installable dist/entwiner-*.tar.gz.

Initial usage

Create a database from GeoJSON data using this pattern:

entwiner [INFILES]... OUTFILE.db

where INFILES is any number of linear GeoJSON featurecollections.

You now have an SQLite (technically, Spatialite) database (OUTFILE.db) with two tables that describe a directed graph: nodes and edges.

The transportation network database graph

The nodes table is the list of graph nodes representing the points where your input lines meet one another, and contains only these columns:

• rowid (a unique integer, default in SQLite tables)

• lon (longitude)

• lat (latitude).

The edges table contains these columns:

• rowid (a unique integer, default in SQLite tables)

• u: the "start node" of this edge.

• v: the "end node" of this edge.

• *: all of the "flat" data from the GeoJSON features, encoded as columnar data. The types will be automatically converted to SQLite types, and have null where missing.

Using the database

The database can be used by any applications that consume node/edge information, but entwiner comes with some convenient Python interfaces for doing network analysis and routing.

entwiner.database.DiGraphDB(path): if the db already exists, this creates a (somewhat) networkx-compatible interface to a DiGraph. As of writing, it is compatible with shortest-path algorithms and centrality metrics. Example:

import entwiner as ent
import networkx as nx

G = ent.graphs.digraphdb.DiGraphDB(path='test.db')
path = nx.algorithms.shortest_paths.dijkstra_path(G, '-122.5049849, 48.7798528', '-122.5074134, 48.7798173', 'length')
print(path)

Why SQLite?

• It's everywhere.

• It's just one file!

• It handles locking for you (works great with web apps)

• Flexible persistence: it can live on disk (default), you can set the cache size high so that repeated queries happen in-memory, and you can set the whole thing to live in-memory (fastest)