R2RML

A default, convention-based approach to generating RDF from relational data without a custom mapping document. The W3C Direct Mapping standard defines an automatic transformation where:

• Each table becomes a class
• Each column becomes a predicate
• Each row becomes an entity with a unique IRI

R2RML extends this by allowing customized, fine-grained control over the mapping process, overriding the default behavior when needed.

The core building block of an R2RML mapping document. Each Triples Map defines how to generate RDF triples from a logical source table or SQL query. It consists of three components:

• Logical Source: The input table or SQL view
• Subject Map: Generates the subject IRI for each row
• Predicate-Object Maps: Define the properties and their values

Multiple Triples Maps can target the same logical source to generate different RDF shapes.

The mechanism for generating RDF terms (IRIs, blank nodes, or literals) from database columns. R2RML defines three term types:

• Constant-valued: Always produces the same RDF term
• Column-valued: Uses the value from a specific SQL column
• Template-valued: Constructs IRIs using string templates with column placeholders like http://example.com/entity/{ID}

Term maps also specify the datatype and language tag for literals.

The R2RML mechanism for expressing foreign key relationships as RDF predicates. Instead of generating a literal value, a Referencing Object Map uses a join condition to identify the subject IRI of a related row. This enables:

• Parent-child relationships between entities
• Many-to-one navigation through foreign keys
• Generation of object properties rather than datatype properties

The join condition specifies the parent and child columns that define the relationship.

R2RML allows arbitrary SQL queries to serve as logical sources, not just base tables. This enables:

• Denormalization: Flattening joins into a single mapping source
• Computed columns: Generating derived values via SQL functions
• Filtering: Excluding rows that shouldn't appear in the RDF
• Aggregation: Pre-computing summary data before mapping

This flexibility means complex business logic can be pushed to the database layer before RDF generation.

While R2RML focuses on relational-to-RDF mapping, SPARQL-Generate extends SPARQL for mapping non-RDF data sources (XML, JSON, CSV) to RDF. Key distinctions:

• R2RML: Tightly coupled to SQL databases and JDBC connections
• SPARQL-Generate: Uses iterator patterns and binding functions for heterogeneous formats
• Both produce standard RDF graphs but target fundamentally different source paradigms

Choose R2RML for relational databases; SPARQL-Generate for APIs and document stores.