Skip to content

The RMLMapper executes RML rules to generate high quality Linked Data from multiple originally (semi-)structured data sources


Notifications You must be signed in to change notification settings


Repository files navigation


Maven Central

The RMLMapper executes RML rules to generate Linked Data. It is a Java library, which is available via the command line (API docs online). The RMLMapper loads all data in memory, so be aware when working with big datasets.

Want to get started quickly? Check out Releases on where to find the latest CLI build as a jar, and see Usage on how to use the commandline interface!

Table of contents



  • local data sources:
    • Excel (.xlsx)
    • LibreOffice (.ods)
    • CSV files (including CSVW)
    • JSON files (JSONPath)
    • XML files (XPath)
  • remote data sources:
    • relational databases (MySQL, PostgreSQL, Oracle, and SQLServer)
    • Web APIs with W3C Web of Things
    • SPARQL endpoints
    • files via HTTP urls (via GET)
      • CSV files
      • JSON files (JSONPath (@ can be used to select the current object.))
      • XML files (XPath)
  • functions (most cases)
    • For examples on how to use functions within RML mapping documents, you can have a look at the RML+FnO test cases
  • configuration file
  • metadata generation
  • output formats: nquads (default), turtle, trig, trix, jsonld, hdt
  • join conditions
  • targets:
    • local file
    • VoID dataset
    • SPARQL endpoint with SPARQL UPDATE


  • functions (all cases)
  • conditions (all cases)
  • data sources:
    • NoSQL databases
    • TPF servers


The standalone jar file (that has a commandline interface) for every release can be found on the release's page on GitHub. You can find the latest release here. This is the recommended way to get started with RMLMapper. Do you want to build from source yourself? Check Build.


The RMLMapper is built using Maven. As it is also tested against Oracle (check here for details), it needs a specific set-up to run all tests. That's why we recommend to build without testing: mvn install -DskipTests=true. If you want, you can install with tests, and just skip the Oracle tests: mvn test -Dtest=!Mapper_OracleDB_Test.

A standalone jar can be found in /target.

Two jars are found in /target: a slim jar without bundled dependencies, and a standalone jar (suffixed with -all.jar) with all dependencies bundled.

Building with profile no-buildnumber disables using and updating (and uses 0 as build number), e.g.:

mvn clean package -P no-buildnumber

outputs for example target/rmlmapper-<version>-r0.jar



The following options are most common.

  • -m, --mapping <arg>: one or more mapping file paths and/or strings (multiple values are concatenated).
  • -o, --output <arg>: path to output file
  • -s,--serialization <arg>: serialization format (nquads (default), trig, trix, jsonld, hdt)

All options can be found when executing java -jar rmlmapper.jar --help, that output is found below.

usage: java -jar mapper.jar <options>
 -b,--base-iri <arg>                 Base IRI used to expand relative IRIs
                                     in generated terms in the output.
 -c,--configfile <arg>               path to configuration file
 -d,--duplicates                     remove duplicates in the HDT,
                                     N-Triples, or N-Quads output
    --disable-automatic-eof-marker   Setting this option assumes input
                                     data has a kind of End-of-File
                                     marker. Don't use unless you're
                                     absolutely sure what you're doing!
 -dsn,--r2rml-jdbcDSN <arg>          DSN of the database when using R2RML
 -e,--metadatafile <arg>             path to output metadata file
 -f,--functionfile <arg>             one or more function file paths
                                     (dynamic functions with relative
                                     paths are found relative to the cwd)
 -h,--help                           show help info
 -l,--metadataDetailLevel <arg>      generate metadata on given detail
                                     level (dataset - triple - term)
 -m,--mappingfile <arg>              one or more mapping file paths and/or
                                     strings (multiple values are
                                     concatenated). r2rml is converted to
                                     rml if needed using the r2rml
                                     arguments.RDF Format is determined
                                     based on extension.
 -o,--outputfile <arg>               path to output file (default: stdout)
 -p,--r2rml-password <arg>           password of the database when using
                                     R2RML rules
 -psd,--privatesecuritydata <arg>    one or more private security files
                                     containing all private security
                                     information such as usernames,
                                     passwords, certificates, etc.
 -s,--serialization <arg>            serialization format (nquads
                                     (default), turtle, trig, trix,
                                     jsonld, hdt)
    --strict                         Enable strict mode. In strict mode,
                                     the mapper will fail on invalid IRIs
                                     instead of skipping them.
 -t,--triplesmaps <arg>              IRIs of the triplesmaps that should
                                     be executed in order, split by ','
                                     (default is all triplesmaps)
 -u,--r2rml-username <arg>           username of the database when using
                                     R2RML rules
 -v,--verbose                        show more details in debugging output                                                      

Accessing Web APIs with authentication

The W3C Web of Things Security Ontology is used to describe how Web APIs authentication should be performed but does not include the necessary credentials to access the Web API. These credentials can be supplied using the -psd <PATH> CLI argument. The PATH argument must point to one or more private security files which contain the necessary credentials to access the Web API.

An example can be found in the test cases src/test/resources/web-of-things.


An example of how you can use the RMLMapper as an external library can be found at ./src/test/java/be/ugent/rml/readme/



We publish our Docker images automatically on Dockerhub for every release. You can find our images here: rmlio/rmlmapper-java.

Build image

You can use Docker to run the RMLMapper by following these steps:

  • Build the Docker image: docker build -t rmlmapper ..
  • Run a Docker container: docker run --rm -v $(pwd):/data rmlmapper -m mapping.ttl.

The same parameters are available as via the CLI. The RMLMapper is executed in the /data folder in the Docker container.

Including functions

There are three ways to include (new) functions within the RML Mapper

  • dynamic loading: you add links to java files or jar files, and those files are loaded dynamically at runtime
  • preloading: you register functionality via code, and you need to rebuild the mapper to use that functionality
  • add as dependency

Registration of functions is done using a Turtle file, which you can find in src/main/resources/functions.ttl

Dynamic loading

Create a Turtle file that describe the functions that need to be included and add the jar which contains those functions.

Note: the java or jar-files are found relative to the cwd. You can change the functions.ttl path (or use multiple functions.ttl paths) using a commandline-option (-f).

For example the snippets below dynamically link an fno:Function to a library, provided by a jar-file (CustomFunctions.jar). The example links a function that parses the latitude (50.2) out of the following string "POINT (50.2 5.3)".

functions.ttl contains the description of the function in Turtle:

@prefix dcterms: <> .
@prefix doap:    <> .
@prefix fno:     <> .
@prefix fnoi:    <> .
@prefix fnom:    <> .
@prefix grel:    <> .
@prefix grelm:   <> .
@prefix rdfs:    <> .

grel:parsePointLat a fno:Function ;
  fno:name "parsePointLat" ;
  rdfs:label "parsePointLat" ;
  dcterms:description "Parse the latitude from a point." ;
  fno:expects ( grel:valueParam ) ;
  fno:returns ( grel:stringOut ) .

    a                  fnoi:JavaClass ;
    doap:download-page "CustomFunctions.jar" ;
    fnoi:class-name    "CustomFunctions" .

    a                    fno:Mapping ;
    fno:function         grel:parsePointLat ;
    fno:implementation   grelm:javaString ;
    fno:methodMapping    [ a                fnom:Function ;
                           fnom:method-name "parsePointLat" ] .

The accompanying java file

public class CustomFunctions {
  public static String parsePointLat(String s) {
    return s.replace("POINT ", "").replace('(', ' ').replace(')', ' ').trim().split("\\s+")[0];

To dynamically include the custom function, compile the java-file and include functions.ttl with the -f option:

javac && jar cvf CustomFunctions.jar CustomFunctions.class
java -jar mapper.jar -f functions.ttl <other options>


This overrides the dynamic loading. An example of how you can use Preload a custom function can be found at ./src/test/java/be/ugent/rml/readme/

Adding as dependency

This is most interesting if you use RMLMapper as a library in your own project. Just add the dependency to the function library you want to use in your project.

You can also add a function library as a Maven dependency in pom.xml of RMLMapper. You'll have to rebuild RMLMapper to use it.

Generating metadata

Conform to how it is described in the scientific paper [1], the RMLMapper allows to automatically generate PROV-O metadata. Specifically, you need the CLI arguments below. You can specify in which output file the metadata should be stored, and up to which level metadata should be stored (dataset, triple, or term level metadata).

 -e,--metadatafile <arg>          path to output metadata file
 -l,--metadataDetailLevel <arg>   generate metadata on given detail level
                                  (dataset - triple - term)


Command line

Run the tests via


Right-click src/test/java directory and select "Run 'All tests'".

Derived tests

Some tests (Excel, ODS) are derived from other tests (CSV) using a script (./


Make sure you have Docker running. On Unix, others read-write permission (006) is required on /var/run/docker.sock in order to run the tests. The tests will fail otherwise, as Testcontainers can't spin up the container.


  • A problem with Docker (can't start the container) causes the SQLServer tests to fail locally. These tests will always succeed locally.
  • A problem with Docker (can't start the container) causes the PostgreSQL tests to fail locally on Windows 7 machines.


Dependency License
ch.qos.logback logback-classic Eclipse Public License 1.0 & GNU Lesser General Public License 2.1
com.github.fnoio function-agent-java MIT
com.github.fnoio grel-functions-java MIT
com.github.fnoio idlab-functions-java MIT
com.github.rdfhdt hdt-java GNU Lesser General Public License v3.0
com.github.tomakehurst:wiremock-jre8 Apache License 2.0 mssql-jdbc MIT
com.mysql mysql-connector-java GNU General Public License v2.0 Oracle Free Use Terms and Conditions
net.minidev json-smart Apache License 2.0
org.apache.jena fuseki-main Apache License 2.0
org.eclipse.rdf4j rdf4j-client Eclipse Distribution License v1.0
org.junit.jupiter junit-jupiter-api Eclipse Public License v2.0
org.junit.jupiter junit-jupiter-engine Eclipse Public License v2.0
org.junit.jupiter junit-jupiter-params Eclipse Public License v2.0
org.junit.vintage junit-vintage-engine Eclipse Public License v2.0
org.postgresql postgresql BSD
org.testcontainers jdbc MIT
org.testcontainers junit-jupiter MIT
org.testcontainers mssqlserver MIT
org.testcontainers mysql MIT
org.testcontainers oracle-xe MIT
org.testcontainers postgresql MIT

Commercial Support

Do you need...

  • training?
  • specific features?
  • different integrations?
  • bugfixes, on your timeline?
  • custom code, built by experts?
  • commercial support and licensing?

You're welcome to contact us regarding on-premise, enterprise, and internal installations, integrations, and deployments.

We have commercial support available.

We also offer consulting for all-things-RML.


Typed spreadsheet files

All spreadsheet files are as of yet regarded as plain CSV files. No type information like Currency, Date... is used.

XML file parsing performance

The RMLMapper's XML parsing implementation (javax.xml.parsers) has been chosen to support full XPath. This implementation causes a large memory consumption (up to ten times larger than the original XML file size). However, the RMLMapper can be easily adapted to use a different XML parsing implementation that might be better suited for a specific use case.

Language tag support

The processor checks whether correct language tags are not, using a regular expression. The regex has no support for languages of length 5-8, but this currently only applies to 'qaa..qtz'.

Duplicate removal and serialization format

Performance depends on the serialization format (--serialization <format>) and if duplicate removal is enabled (--duplicates). Experimenting with various configurations may lead to better performance for your use case.

I have a question! Where can I get help?

Do you have any question related to writing RML mapping rules, the RML specification, etc., feel free to ask them here: ! If you have found a bug or need a feature for the RMLMapper itself, you can make an issue in this repository.


Generate static files at /docs/apidocs with:

mvn javadoc:javadoc

UML Diagrams

Architecture UML Diagram

How to generate with IntelliJ IDEA

(Requires Ultimate edition)

  • Right click on package: "be.ugent.rml"
  • Diagrams > Show Diagram > Java Class Diagrams
  • Choose what properties of the classes you want to show in the upper left corner
  • Export to file > .png | Save diagram > .uml

Sequence Diagram

Edit on
  • Go to
  • Click on 'Open Existing Diagram' and choose the .html file

[1]: A. Dimou, T. De Nies, R. Verborgh, E. Mannens, P. Mechant, and R. Van de Walle, “Automated metadata generation for linked data generation and publishing workflows,” in Proceedings of the 9th Workshop on Linked Data on the Web, Montreal, Canada, 2016, pp. 1–10. PDF


The RMLMapper executes RML rules to generate high quality Linked Data from multiple originally (semi-)structured data sources







No packages published