[JENA-1290] Slow Model.createList()

• Type: Improvement
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: Jena 3.0.1, Jena 3.2.0
• Fix Version/s: Jena 3.3.0
• Component/s: Core

Environment

OpenJDK 1.8.0.121-b13

• Assignee: Andy Seaborne
• Reporter: Alexis Armin Huf
• Created: 19/Feb/17 21:19
• Updated: 10/May/17 15:35
• Resolved: 20/Feb/17 22:38

Description

Model.createList(Iterator) is slow for large list sizes. Starting at 128 nodes, creating a list from an ArrayList iterator takes ~3.6ms and the time goes up to 182ms for 1024 nodes. Creating the RDF list by hand and then doing .as(RDFList.class) for the 1024-nodes tests takes only 1.2ms. I will attach a MWE.

Going through the code, the cause seems to be that RDFListImpl.with() , which is called for every member added, will always walk the entire list to get the last node. It mentions a possible cache for that, but it is not implemented. I wrote a patch for this issue replacing ModelCom.createList() implementation. I could not find tests for createList, so I added three simple ones to TestModel.

PS: I'm having a classpath issue with jena-arq tests on mvn clean install, so I could only run tests for jena-core. Are there more specific instructions somewhere for building and testing jena?