[JENA-1862] Query.cloneQuery is slow

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: Jena 3.14.0
• Fix Version/s: Jena 3.15.0
• Component/s: ARQ
• Assignee: Andy Seaborne
• Reporter: Claus Stadler
• Created: 13/Mar/20 15:04
• Updated: 07/Jul/20 16:23
• Resolved: 16/Apr/20 18:10

Description

This is a follow up to JENA-1861 which is about the thread safety of Query objects. As sharing a Query object across threads may introduce race conditions, the obvious workaround would be to just clone it. However, the current implementation serializes the query to string and then re-parses it which causes a very high overhead (for my use case it became the top entry in the VisualVM profiler).

Proposal:

• Extend ElementTransformCopyBase with a 'alwaysCopy' flag analogous to ExprTransformCopy
• Add a new constructor to ExprTransformApplyElementTransform(ElementTransform transform, boolean alwaysCopy) so that the alwayCopy flag of the underlying ExprTransformCopy can be set to true
• Implement clone using syntatic transforms as below

public static Query fastClone(Query query) {
    ElementTransform eltXform = new ElementTransformCopyBase2(true);
    ExprTransform exprXform = new ExprTransformApplyElementTransform2(eltXform, true);
    Query result = QueryTransformOps.transform(query, eltXform, exprXform);
    return result;
}

This approach 'works-for-me' and I can create a pull request for this, but maybe there are more subtleties to the outlined approach that need to be considered?

Issue Links

• relates to: JENA-1935 Query::cloneQuery produces invalid syntax
• links to: GitHub Pull Request #712