knutwalker / ntparser

A simple and performant N-Triples parser for the JVM

Version Matrix


A simple and well-performing N-Triples parser for the JVM.


NtParser is published to Maven Central:

  • group = de.knutwalker
  • artifact = ntparser_2.10 or ntparser_2.11
  • version: 0.3.0
<!-- for Scala 2.10 -->
<!-- for Scala 2.11 -->
libraryDependencies += "de.knutwalker" %% "ntparser" % "0.3.0"

You need at least one modelfactory instance as well. Available are:

  • de.knutwalker:ntparser-model_2.1x:0.3.0
  • de.knutwalker:ntparser-jena_2.1x:0.3.0


import de.knutwalker.ntparser.StrictNtParser
import de.knutwalker.ntparser.model.ntModel

StrictNtParser("/path/to/my/graph.nt") foreach { stmt 
import de.knutwalker.ntparser.model.NtModelFactory;
import de.knutwalker.ntparser.model.Statement;
import de.knutwalker.ntparser.StrictNtParser;
import java.util.Iterator;

Iterator<Statement> statementIterator = StrictNtParser.parse("/path/to/my/graph.nt", NtModelFactory.INSTANCE());
while (statementIterator.hasNext()) {
  Statement stmt =;

in-depth usage

There are two parsers, de.knutwalker.ntparser.StrictNtParser and de.knutwalker.ntparser.NonStrictNtParser.

Both provide a number of possibilities to parse a graph and will always return an Iterator of a statement. The apply methods are Scala API, the parse methods are Java API (and are thus consuming and producing Java Iterators).

The StrictNtParser will halt with an Exception at the first parse error. The NonStrictNtParser will just log exceptions and continue parsing.

In addition to that, you need to provide an instance of a typeclass or factory, that will be used to create the resulting node graph. There are two implementations available in the artifacts ntparser-model and ntparser-jena. The first will use a very simple Scala ADT, the second produces a Jena model. You can also implement the interface de.knutwalker.ntparser.ModelFactory to include different backends.

If you provide a String, it is treated as the filename. NtParser transparently handles gzipped and bzipped files as well, if you use this overloading. Other options include providing a, for the Scala API (apply), File, Path for the Java API (parse), as well as InputStream and the respective Iterable and Iterator for both Scala and Java. The Iterable and Iterator should represent single lines, that is, one N-Triple per element.

As an alternative, you can instantiate an de.knutwalker.ntparser.NtParser directly. The NtParser has several parse* methods, that each take a String and assume, that this is one line of some bigger document and that is correctly delimited. One invocation of any parse* method will parse at-most one statement. There are two flavours, NtParser.strict and NtParser.lenient. The lenient one supports grammar features, that are not officially specified, such as long quotes from Turtle.

Specification conformity

NtParser passes all tests provided by the W3C