Calcite SPARQL adapter

This adapter can be used to execute SQL queries over a SPARQL endpoint via Apache Calcite and Apache Jena. The project is a work in progress and it is not meant to be run in production.

Description

The adapter can be used in several modes to expose the RDF data in a SPARQL endpoint as data:

• the property mode, which exposes every RDF property (the ?p of a ?s ?p ?o pattern) as a table with two columns s (subject) and o (object);
• the class mode, which exposes every RDF class (the ?cl of a ?s rdf:type ?cl pattern) as a table whose columns are all the properties that link items of such class (the ?s of the same pattern) to objects;
• the mapping mode, where selected properties are directly mapped to columns of a specified table.

Example of the model file

The SPARQL endpoint in the following examples is a Jena in-memory endpoint created from a local file. Replace with e.g. "jdbc:jena:remote:query=https://dbpedia.org/sparql" to use a remote endpoint.

1. Using the property mode:

{
  "version": "1.0",
  "defaultSchema": "sparql",
  "schemas": [
    {
      "name": "sparql",
      "type": "custom",
      "factory": "com.datagrafting.sql2sparql.calcite.SparqlSchemaFactory",
      "operand": {
        "endpoint": "jdbc:jena:mem:dataset=data.nq",
        "tableMode": "property"
      }
    }
  ]
}

2. Using the class mode:

{
  "version": "1.0",
  "defaultSchema": "sparql",
  "schemas": [
    {
      "name": "sparql",
      "type": "custom",
      "factory": "com.datagrafting.sql2sparql.calcite.SparqlSchemaFactory",
      "operand": {
        "endpoint": "jdbc:jena:mem:dataset=data.nq",
        "tableMode": "class"
      }
    }
  ]
}

3. Using the mapping mode:

{
  "version": "1.0",
  "defaultSchema": "sparql",
  "schemas": [
    {
      "name": "sparql",
      "type": "custom",
      "factory": "com.datagrafting.sql2sparql.calcite.SparqlSchemaFactory",
      "operand": {
        "endpoint": "jdbc:jena:mem:dataset=data.nq",
        "tableMode": "mapping",
        "tableMappings": [
          {
            "name": "Person",
            "columns": [
              {
                "name": "name",
                "property": "http://xmlns.com/foaf/0.1/name"
              },
              {
                "name": "age",
                "property": "http://xmlns.com/foaf/0.1/age"
              }
            ]
          }
        ]
      }
    }
  ]
}

Basic Usage

• Java example
• Python (Jupyter notebook) examples using jaydebeapi and Apache Arrow

Running the examples

Compile the main library using Java 11:

cd core
mvn clean package -DskipTests
cd ..

or using Java 8:

cd core
mvn clean package -P java1.8 -DskipTests
cd ..

Please note that compiling with Java 8 forces Jena version to 3.17.0 (last version before switching to Java 11).

For the Java examples:

cd example
mv clean package
/path/to/java -cp target/classes:../../core/target/calcite-sparql-core-0.0.1-SNAPSHOT.jar com.datagrafting.sql2sparql.examples.SparqlClassTableRemote

For the Python examples, Python 3 and Jupyter Notebook need to be installed first. Then, the notebooks can be run from the examples/python directory.

Pushed-down SQL constructs

• SELECT * and SELECT with any number of columns
• WHERE (with =, <>, <, >, <=, >=)
• ORDER BY
• LIMIT

Releasing

When releasing, please make sure to update both the changelog and the citation file.

Citing the project

If you use this project in an article, please cite it as specified in the CITATION.cff file.

Acknowledgements

Many thanks to:

• Paul Jackson for interesting discussions on the mapping from RDF to SQL schema.