Writing RDF in Apache Jena

This page describes the RIOT (RDF I/O technology) output capabilities.

See Reading RDF for details of the RIOT Reader system.

• API
• RDFFormat
• RDFFormats and Jena syntax names
• Formats
  • Normal Printing
  • Pretty Printed Languages
  • Streamed Block Formats
  • Line printed formats
  • Turtle and Trig format options
  • N-Triples and N-Quads
  • JSON-LD
  • RDF Binary
  • RDF/XML
• Examples
• Notes

API

There are two ways to write RDF data using Apache Jena RIOT, either via the RDFDataMgr

RDFDataMgr.write(OutputStream, Model, Lang) ;
RDFDataMgr.write(OutputStream, Dataset, Lang) ;
RDFDataMgr.write(OutputStream, Model, RDFFormat) ;
RDFDataMgr.write(OutputStream, Dataset, RDFFormat) ;

or the legacy way using the model API, where there is a limited set of "format" names

model.write(output, "format") ;

The format names are described below; they include the names Jena has supported before RIOT.

Many variations of these methods exist. See the full javadoc for details.

RDFFormat

Output using RIOT depends on the format, which involves both the language (syntax) being written and the variant of that syntax.

The RIOT writer architecture is extensible. The following languages are available as part of the standard setup.

• Turtle
• N-Triples
• NQuads
• TriG
• JSON-LD
• RDF/XML
• RDF/JSON
• TriX
• RDF Binary

In addition, there are variants of Turtle, TriG for pretty printing, streamed output and flat output. RDF/XML has variants for pretty printing and plain output. Jena RIOT uses org.apache.jena.riot.RDFFormat as a way to identify the language and variant to be written. The class contains constants for the standard supported formats.

Note:

• RDF/JSON is not JSON-LD. See the description of RDF/JSON.
• N3 is treated as Turtle for output.

RDFFormats and Jena syntax names

The string name traditionally used in model.write is mapped to RIOT RDFFormat as follows:

Formats

Normal Printing

A Lang can be used for the writer format, in which case it is mapped to an RDFFormat internally. The normal writers are:

Pretty printed RDF/XML is also known as RDF/XML-ABBREV.

Pretty Printed Languages

All Turtle and TriG formats use prefix names, and short forms for literals.

The pretty printed versions of Turtle and TriG prints data with the same subject in the same graph together. All the properties for a given subject are sorted into a predefined order. RDF lists are printed as (...) and [...] is used for blank nodes where possible.

The analysis for determining what can be pretty printed requires temporary datastructures and also a scan of the whole graph before writing begins. Therefore, pretty printed formats are not suitable for writing persistent graphs and datasets.

When writing at scale use either a “blocked” version of Turtle or TriG, or write N-triples/N-Quads.

Example:

PREFIX :      <http://example/>
PREFIX dc:    <http://purl.org/dc/elements/1.1/>
PREFIX foaf:  <http://xmlns.com/foaf/0.1/>
PREFIX rdf:   <http://www.w3.org/1999/02/22-rdf-syntax-ns#>

:book   dc:author  ( :a :b ) .

:a      a           foaf:Person ;
        foaf:knows  [ foaf:name  "Bob" ] ;
        foaf:name   "Alice" .

:b      foaf:knows  :a .

The default pretty printed output (shown above) aligns predicates and objects, which can result in wide lines. For a narrower indentation style, set ttl:indentStyle to long. See Turtle and Trig format options.

Pretty printed formats:

Streamed Block Formats

Fully pretty printed formats can not be streamed. They require analysis of all of the data to be written in order to choose the short forms. This limits their use in fully scalable applications.

Some formats can be written streaming style, where the triples or quads are partially grouped together by adjacent subject or graph/subject in the output stream.

The written data is like the pretty printed forms of Turtle or TriG, but without RDF lists being written in the ‘(…)’ form, without using [...] for blank nodes.

This gives some degree of readability while not requiring excessive temporary datastructure. Arbitrary amounts of data can be written but blank node labels need to be tracked in order to use the short label form.

Example:

PREFIX dc:  <http://purl.org/dc/elements/1.1/>
PREFIX :  <http://example/>
PREFIX foaf:  <http://xmlns.com/foaf/0.1/>
PREFIX rdf:  <http://www.w3.org/1999/02/22-rdf-syntax-ns#>

:book   dc:author  _:b0 .

_:b0    rdf:rest   _:b1 ;
        rdf:first  :a .

:a      foaf:knows  _:b2 ;
        foaf:name   "Alice" ;
        rdf:type    foaf:Person .

_:b2    foaf:name  "Bob" .

:b      foaf:knows  :a .

_:b1    rdf:rest   rdf:nil ;
        rdf:first  :b .

Formats:

Line printed formats

There are writers for Turtle and Trig that use the abbreviated formats for prefix names and short forms for literals. They write each triple or quad on a single line.

The regularity of the output can be useful for text processing of data.
These formats do not offer more scalability than the stream forms.

Example:

The FLAT writers abbreviates IRIs, literals and blank node labels but always writes one complete triple on one line (no use of ;).

PREFIX :  <http://example/>
PREFIX dc:  <http://purl.org/dc/elements/1.1/>
PREFIX foaf:  <http://xmlns.com/foaf/0.1/>
PREFIX rdf:  <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
_:b0 foaf:name "Bob" .
:book dc:author _:b1 .
_:b2 rdf:rest rdf:nil .
_:b2 rdf:first :b .
:a foaf:knows _:b0 .
:a foaf:name "Alice" .
:a rdf:type foaf:Person .
_:b1 rdf:rest _:b2 .
_:b1 rdf:first :a .
:b foaf:knows :a .

Turtle and Trig format options

Some context settings affect the output of Turtle and TriG writers. Unless otherwise noted, the setting applies to both Turtle and TriG.

Format Option Usage

Setting directive style

    riot --set ttl:directiveStyle=rdf11 --pretty Turtle file1.rdf file2.nt ...

and in code:

  RDFWriter.source(model)
     .set(RIOT.symTurtleDirectiveStyle, "sparql")
     .lang(Lang.TTL)
     .output(System.out);

Setting indent style

    riot --set ttl:indentStyle=long --formatted=ttl file1.rdf file2.nt ...

and in code:

  RDFWriter.source(model)
     .format(RDFFormat.TURTLE_LONG)
     .output(System.out);

or:

  RDFWriter.source(model)
     .set(RIOT.symTurtleIndentStyle, "long")
     .lang(Lang.TTL)
     .output(System.out);

Base URI

Output can be written with relative URIs and no base. Note: such output is not portable; its meaning depends on the base URI at the time of reading.

Turtle and Trig can be written with relative URIs by setting the base URI for writing and switching off output of the base URI.

  RDFWriter.create()
     .base("http://host/someBase")
     .set(RIOT.symTurtleOmitBase, true)
     .lang(Lang.TTL)
     .source(model)
     .output(System.out);

N-Triples and N-Quads

These provide the formats that are fastest to write, and data of any size can be output. They do not use any internal state and formats always stream without limitation.

They maximise the interoperability with other systems and are useful for database dumps. They are not human readable, even at moderate scale.

The files can be large but they compress well with gzip. Compression ratios of x8-x10 can often be obtained.

Example:

The N-Triples writer makes no attempt to make it’s output readable. It uses internal blank nodes to ensure correct labeling without needing any writer state.

_:BX2Dc2b3371X3A13cf8faaf53X3AX2D7fff <http://xmlns.com/foaf/0.1/name> "Bob" .
<http://example/book> <http://purl.org/dc/elements/1.1/author> _:BX2Dc2b3371X3A13cf8faaf53X3AX2D7ffe .
_:BX2Dc2b3371X3A13cf8faaf53X3AX2D7ffd <http://www.w3.org/1999/02/22-rdf-syntax-ns#rest> <http://www.w3.org/1999/02/22-rdf-syntax-ns#nil> .
_:BX2Dc2b3371X3A13cf8faaf53X3AX2D7ffd <http://www.w3.org/1999/02/22-rdf-syntax-ns#first> <http://example/b> .
<http://example/a> <http://xmlns.com/foaf/0.1/knows> _:BX2Dc2b3371X3A13cf8faaf53X3AX2D7fff .
<http://example/a> <http://xmlns.com/foaf/0.1/name> "Alice" .
<http://example/a> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> <http://xmlns.com/foaf/0.1/Person> .
_:BX2Dc2b3371X3A13cf8faaf53X3AX2D7ffe <http://www.w3.org/1999/02/22-rdf-syntax-ns#rest> _:BX2Dc2b3371X3A13cf8faaf53X3AX2D7ffd .
_:BX2Dc2b3371X3A13cf8faaf53X3AX2D7ffe <http://www.w3.org/1999/02/22-rdf-syntax-ns#first> <http://example/a> .
<http://example/b> <http://xmlns.com/foaf/0.1/knows> <http://example/a> .

The main N-Triples and N-Quads writers follow RDF 1.1 and output using UTF-8.
For compatibility with old software, writers are provided that output in ASCII (using \u escape sequences for non-ASCI characters where necessary).

JSON-LD

Caution. This section describes features that may be removed.

Jena uses different third party processors for JSON-LD 1.0 and JSON-LD 1.1.

This section describes support for passing configuration to the JSON-LD 1.0 processor only. It does not apply to the JSON-LD 1.1 processor.

It is planned that support for JSON-LD 1.0 will be removed in Jena5.

The project is looking for contributions for passing framing configuration to the JSON-LD 1.1 processor, which is titanium-json-ld.

JSON-LD output is supported, in its various flavors (“compacted”, “expanded”, “flattened”, “framed”), by using one of the following RDFFormats:

The default registration for JSONLD is JSONLD_PRETTY. JSONLD_PRETTY is identical to JSONLD_COMPACT_PRETTY.

Output can be customized, passing more info to the writer by using the “Context” mechanism provided by Jena. The same mechanism is used to pass the “frame” in the JSONLD_FRAME_PRETTY and JSONLD_FRAME_FLAT cases.

What can be done, and how it can be, is explained in the sample code.

RDF Binary

This is a binary encoding using Apache Thrift or Google Protocol Buffers for RDF Graphs and RDF Datasets, as well as SPARQL Result Sets, and it provides faster parsing compared to the text-based standardised syntax such as N-triples, Turtle or RDF/XML.

RDF_THRIFT_VALUES and RDF_PROTO_VALUES are variants where numeric values are written as values, not as lexical format and datatype. See the description of RDF Binary. for discussion.

RDF/XML

RIOT supports output in RDF/XML. RIOT RDFFormats defaults to pretty printed RDF/XML, while the jena writer name defaults to a streaming plain output.

More details RDF/XML Output.

Examples

Example code may be found in jena-examples:arq/examples.

Ways to write a model

The follow ways are different ways to write a model in Turtle:

  Model model =  ... ;

  // Write a model in Turtle syntax, default style (pretty printed)
  RDFDataMgr.write(System.out, model, Lang.TURTLE) ;
    
  // Write Turtle to the blocks variant
  RDFDataMgr.write(System.out, model, RDFFormat.TURTLE_BLOCKS) ;
    
  // Write as Turtle via model.write
  model.write(System.out, "TTL") ;

Ways to write a dataset

The preferred style is to use RDFDataMgr:

  Dataset ds = .... ;
  // Write as TriG
  RDFDataMgr.write(System.out, ds, Lang.TRIG) ;

  // Write as N-Quads
  RDFDataMgr.write(System.out, dataset, Lang.NQUADS) ;

Additionally, a single model can be written in a dataset format - it becomes the default graph of the dataset.

Model m = ... ;
RDFDataMgr.write(System.out, m, Lang.TRIG) ;

might give:

PREFIX :      <http://example/>
PREFIX dc:    <http://purl.org/dc/elements/1.1/>
PREFIX foaf:  <http://xmlns.com/foaf/0.1/>
PREFIX rdf:   <http://www.w3.org/1999/02/22-rdf-syntax-ns#>

{
    :book   dc:author  ( :a :b ) .

    :a      a           foaf:Person ;
            foaf:knows  [ foaf:name  "Bob" ] ;
            foaf:name   "Alice" .

    :b      foaf:knows  :a .
}

Adding a new output format

A complete example of adding a new output format is given in the example file: RIOT Output example 7.

Notes

Using OutputStreams is strongly encouraged. This allows the writers to manage the character encoding using UTF-8. Using java.io.Writer does not allow this; on platforms such as MS Windows, the default configuration of a Writer is not suitable for Turtle because the character set is the platform default, and not UTF-8. The only use of writers that is useful is using java.io.StringWriter.