TripleTrail is an RDF triple store for spatial networks such as railways, roads and utilities. It answers SPARQL 1.1 and 1.2 with complete spatial and routing support, built in and fast:
- GeoSPARQL 1.1 and the PostGIS
ST_*function set, answered from a spatial index. - pgRouting's
pgr_*functions (shortest paths, isochrones, flows, vehicle routing and more) on any data model, plus routing with constraints, SHACL vehicle rules and contraction hierarchies. - A drop-in replacement for Apache Jena Fuseki: the same URLs, admin API and Graph Store Protocol, with access control, backups, blue-green releases and a web UI.
- Data in, data out: query CSV, JSON, XML, Excel, HTML, Shapefile and more with SPARQL Anything queries (Facade-X) or R2RML/RML mappings; serve features as OGC API Features and vector tiles; read and write GeoParquet.
- Validation and reasoning: SHACL Core and SHACL-SPARQL, SHACL-AF rules, RDFS and OWL 2 RL entailment, kept up to date as data changes.
It is built on the Oxigraph crates, with its own storage engine and query planner.
Releases publish binaries for Linux, macOS and Windows and a Docker image,
ghcr.io/matdata-eu/tripletrail. Until the first release, build it from source with
Rust 1.95 or later:
git clone https://github.com/Matdata-eu/tripletrail
cd tripletrail
cargo install --locked --path crates/tripletrail-server # installs the `tripletrail` command
or build the Docker image yourself:
docker build -t tripletrail .
docker run -p 7878:7878 -v tripletrail-data:/data tripletrail
For Python, pip install ./python builds the tripletrail package (it needs Rust too); see
python/README.md.
Save a few stations as stations.ttl:
@prefix ex: <http://example.org/> .
@prefix geo: <http://www.opengis.net/ont/geosparql#> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
ex:brussels a ex:Station ; rdfs:label "Brussels Central" ;
geo:hasGeometry [ geo:asWKT "POINT(4.3571 50.8456)"^^geo:wktLiteral ] .
ex:leuven a ex:Station ; rdfs:label "Leuven" ;
geo:hasGeometry [ geo:asWKT "POINT(4.7158 50.8813)"^^geo:wktLiteral ] .
ex:antwerp a ex:Station ; rdfs:label "Antwerp Central" ;
geo:hasGeometry [ geo:asWKT "POINT(4.4211 51.2172)"^^geo:wktLiteral ] .
Start the server with them loaded into an in-memory dataset named ds:
tripletrail serve --load stations.ttl
Open http://localhost:7878/ for the web UI. Query runs SPARQL with table, map and chart views, Map shows the stations on OpenStreetMap, and Workbench is a playground for shapes, rules and pipelines. Try the stations within 50 km of Brussels:
PREFIX geo: <http://www.opengis.net/ont/geosparql#>
PREFIX geof: <http://www.opengis.net/def/function/geosparql/>
PREFIX uom: <http://www.opengis.net/def/uom/OGC/1.0/>
PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>
SELECT ?name ?km WHERE {
?station rdfs:label ?name ; geo:hasGeometry/geo:asWKT ?wkt .
BIND(geof:distance(?wkt, "POINT(4.3571 50.8456)"^^geo:wktLiteral, uom:metre) / 1000 AS ?km)
FILTER(?km < 50)
} ORDER BY ?km
The same works from any SPARQL client, at Fuseki's URLs:
curl http://localhost:7878/ds/query -H 'Accept: text/csv' --data-urlencode query@near.rq
curl http://localhost:7878/ds/data -H 'Content-Type: text/turtle' --data-binary @more.ttl # add data
curl http://localhost:7878/ds/ogc/collections # OGC API Features
To keep the data on disk, give the server a directory. A large file loads faster offline first:
tripletrail load --location ./data --dataset kg big-file.nt # bulk load, while no server runs
tripletrail serve --location ./data --dataset kg
The server binds to 127.0.0.1 and has no authentication until you configure it. Before
anyone else can reach it, read Running it in production.
| Command | What it does |
|---|---|
tripletrail serve |
Serves datasets: SPARQL, Graph Store Protocol, admin API, web UI, OGC API, Linked Data. See server. |
tripletrail load |
Bulk-loads RDF or GeoParquet files into an on-disk dataset. |
tripletrail lift |
Turns CSV, JSON, XML, spreadsheets and other files into RDF with Facade-X queries and RML mappings. See Facade-X and RML. |
tripletrail validate |
Validates data against SHACL shapes. See validation and reasoning. |
tripletrail infer |
Runs SHACL-AF rules until they infer nothing new. |
tripletrail entail |
Writes the RDFS or OWL 2 RL entailment of data to a graph or a file. |
tripletrail export |
Writes the results of a SELECT as GeoParquet. See GeoParquet. |
tripletrail canon, tripletrail diff |
RDFC-1.0 canonical N-Quads and a digest of a dataset, and what changed between two releases. See canonicalization. |
tripletrail pipeline run |
Runs an ETL pipeline.yaml (lift, load, update, infer, reason, validate, export, publish and python steps) headless. See pipelines. |
tripletrail test |
Runs the regression tests of a workspace, for CI. See tests. |
tripletrail health, tripletrail hash-key |
A health check for containers, and the hash of an API key for the auth configuration. |
tripletrail <command> --help lists every option. The options of serve can also be set with
an environment variable (--timeout is TRIPLETRAIL_TIMEOUT, and so on), except --load and
the --fx-* options.
Routing. The routing functions take any edge set as a sub-select, as pgRouting takes an SQL query, so they work on any data model:
PREFIX pgr: <http://w3id.org/tripletrail/pgr#>
PREFIX ex: <http://example.com/>
SELECT ?seq ?node ?edge ?agg_cost WHERE {
SERVICE <http://w3id.org/tripletrail/pgr> {
[] pgr:function pgr:pgr_dijkstra ;
pgr:start_vid ex:a ; pgr:end_vid ex:d ;
pgr:seq ?seq ; pgr:node ?node ; pgr:edge ?edge ; pgr:agg_cost ?agg_cost .
{ SELECT ?id ?source ?target ?cost WHERE {
?id a ex:Track ; ex:from ?source ; ex:to ?target ; ex:length ?cost .
} }
}
}
PostGIS functions. st: has the PostGIS names and arguments:
PREFIX st: <http://w3id.org/tripletrail/st#>
PREFIX geo: <http://www.opengis.net/ont/geosparql#>
SELECT ?line (st:ST_LineInterpolatePoint(?wkt, 0.5) AS ?midpoint) WHERE {
?line geo:hasGeometry/geo:asWKT ?wkt .
}
Full-text search. tt:search ranks resources by their labels, titles and descriptions
(more):
PREFIX tt: <http://w3id.org/tripletrail/fn#>
SELECT ?s ?score WHERE { ?s tt:search ("central" ?score) } ORDER BY DESC(?score)
Files as RDF. A SPARQL Anything query reads a file and builds triples from it:
PREFIX xyz: <http://sparql.xyz/facade-x/data/>
PREFIX fx: <http://sparql.xyz/facade-x/ns/>
CONSTRUCT { ?station <http://ex.org/name> ?name }
WHERE {
SERVICE <x-sparql-anything:location=stations.json> { ?s xyz:code ?code ; xyz:name ?name }
BIND(fx:entity("http://ex.org/station/", ?code) AS ?station)
}
tripletrail lift queries/ --output stations.nt
New to TripleTrail? The user guide takes you through it step by step, on a small sample railway network. The pages below are the reference.
| Topic | Page |
|---|---|
| Step-by-step user guide: install, data, spatial, routing, validation, lifting, pipelines, Workbench, maps and APIs, Fuseki, access control, production | docs/guide/ |
| Running the server: endpoints, datasets, admin API, access control, backups, web UI, Workbench, pipelines, describe pages, OGC API, Linked Data, coming from Fuseki | docs/server.md |
| Spatial functions, set-returning functions, vector tiles, linear referencing | docs/spatial.md |
| Routing: pgRouting functions and what goes beyond them | docs/routing.md |
| SHACL validation, SHACL-AF rules, RDFS and OWL 2 RL entailment | docs/validation-and-reasoning.md |
Full-text search (tt:search, Jena's text:query) |
docs/full-text-search.md |
| Facade-X: querying files with SPARQL Anything queries | docs/facade-x.md |
| R2RML, RML and YARRRML mappings | docs/rml.md |
| GeoParquet import and export | docs/geoparquet.md |
| Canonical N-Quads, dataset digests and diffs between releases | docs/canonicalization.md |
| Rust library | docs/library.md |
| TripleTrail Enterprise: the enterprise build, licence keys and updates | docs/enterprise.md |
| Python package | python/README.md |
| Every function and its status | docs/function-matrix.md |
| W3C and OGC test suite results | docs/conformance.md |
| Benchmarks against PostGIS, QLever, Jena, GraphDB and others | tripletrail-bench |
| Design: architecture, storage, spatial index | docs/architecture.md, docs/store.md, docs/spatial-index.md |
| Changes per release | CHANGELOG.md |
| Crate | What it does |
|---|---|
tripletrail |
Engine: a store with everything below registered. |
tripletrail-server |
The tripletrail command and server. |
tripletrail-store |
Quad store with 64-bit term ids, eight compressed sorted quad indexes, geometries parsed on first use and kept by id, and a checkpoint plus write-ahead log on disk. See docs/store.md. |
tripletrail-geo |
geof: (GeoSPARQL 1.1) and st: (PostGIS-named) functions on geo:wktLiterals. |
tripletrail-spatial |
R-tree over all geometry literals, and the query rewrite that answers spatial filters and joins from it. |
tripletrail-routing |
Data-model agnostic routing core on a pgRouting edge set (id, source, target, cost, reverse_cost). Knows nothing about RDF. |
tripletrail-pgr |
pgRouting functions as a SPARQL SERVICE, with runtime edge queries and named networks. |
tripletrail-shacl |
SHACL Core and SHACL-SPARQL validator, SHACL-AF rules, and the compat: vehicle rules used by routing. |
tripletrail-fx |
Files as RDF through SERVICE <x-sparql-anything:…>, following the Facade-X model. |
tripletrail-rml |
R2RML, RML and YARRRML mappings. |
tripletrail-testsuite |
The W3C and OGC test suites. The benchmarks are in tripletrail-bench. |
The queries are parsed, optimized and evaluated by patched copies of Oxigraph's spargebra,
sparopt and spareval crates under vendor/, each with a TRIPLETRAIL.md that lists
the changes.
TripleTrail is licensed under the Apache License 2.0. Third-party components and
their licenses are listed in NOTICE; cargo deny check (see deny.toml)
keeps dependencies to permissive and weak-copyleft licenses.
TripleTrail is developed by Matdata as open core: this repository is the open-source engine, and Matdata offers commercial editions and support on top of it. Contributions need a signed Contributor License Agreement; see CONTRIBUTING.md. Vulnerabilities go through SECURITY.md.
TripleTrail is built on Oxigraph and is not affiliated with or endorsed by the Oxigraph project.