A database management system for ontologies

OntoDB is a database management system for ontologies. Data is RDF, the query language is SPARQL, and a database can live in a directory with a write-ahead log and snapshots.

$ brew install pkgconf raptor cmake ninja
$ git clone https://github.com/lamng3/OntoDB.git && cd OntoDB
$ cmake --preset dev && cmake --build --preset dev
Get started Playground SPARQL

Query a real database

output of the ontodb shell

Try it: Pick a query, press Run for its rows, then Explain analyze for the plan the optimizer chose.


    

    
Press Run.

    
Each query ran in the real shell on the Pizza ontology (loaded 27 triples). The page shows what the shell printed, with IRI prefixes shortened. \explain analyze prints each plan node with its estimated and actual rows.

  





  
What it does

  
RDF in, SPARQL out, with the machinery of a database underneath.

  
PlaygroundRun SPARQL over the Pizza ontology or your own Turtle, right in the browser.TutorialLoad an ontology, query it, update it, and restart the database.SPARQLWhat the query engine accepts, and how it is planned.Write-ahead logEvery transaction is logged before it is applied; recovery replays the log.Buffer poolPages, pins, and replacement, with a hit rate you can read.OptimizerJoin order from estimated rows, checked against actual rows.RecoveryCrash the shell on purpose, then verify the database rebuilt from its log.




This manual describes how to install OntoDB, load ontologies, query and update them, keep a database on disk, and recover it after a crash. The last parts describe how OntoDB works inside and how the project is developed.
For the theory behind each part (pages, buffer pools, B+ trees, query plans, locking, the write-ahead log, and ARIES), read the companion book, OntoDB: Database Internals by Example. Its design-decision boxes explain why OntoDB is built the way it is: each names the choice, the alternatives, and the cost.
What OntoDB does today
  • Loads Turtle, N-Triples, and RDF/XML. OWL files load as RDF triples; there is no reasoner.
  • Answers SPARQL SELECT queries over a basic graph pattern with FILTER, and runs INSERT DATA and DELETE DATA. See SPARQL reference.
  • Keeps a database on disk when started with a directory: every update is a transaction, a checksummed write-ahead log records it, \checkpoint writes a snapshot, and reopening the directory restores every committed transaction. See Persistent databases.
  • Checks itself: \verify rebuilds the database from its files and compares it with the live one.
Indexes, the cost-based optimizer, concurrent transactions, and the page-based disk store are designed and being built. See Roadmap.
Quick start
cmake --preset dev && cmake --build --preset dev
./build/dev/ontodb mydb
OntoDB> \load data/tiny.owl
loaded 14 triples in 5.2 ms
OntoDB> SELECT ?kind WHERE { ?kind <http://www.w3.org/2000/01/rdf-schema#subClassOf> <http://example.org/materials#Material> . }
?kind
<http://example.org/materials#Metal>
<http://example.org/materials#Polymer>
(2 rows)
Installation has the prerequisites, and the Tutorial walks through a full session.
How this manual is organized
Getting startedInstall, build, and a first session from load to recovery.Using OntoDBThe shell, SPARQL, loading data, persistent databases, and settings.ReferenceOn-disk file formats, messages, and troubleshooting.InternalsQuery engine, dictionary, indexes, B+ tree, buffer pool, concurrency, and recovery.ProjectBuilding and testing, the roadmap, and how results are reported.
Conventions
OntoDB> is the shell prompt. Commands that start with \ are shell commands; everything else is SPARQL. A Status box says what works today. A Planned box describes work that is designed but not built. Features that are not built yet answer not built yet (PLAN x.y), where x.y is the item in the project's PLAN.md.