Eclipse Community Forums: EclipseLink » HashPartitioningPolicy fails for large IDs of type long

Home » Eclipse Projects » EclipseLink » HashPartitioningPolicy fails for large IDs of type long(reason: remainder function returns negative result)

HashPartitioningPolicy fails for large IDs of type long [message #916123]

Tue, 18 September 2012 11:19

Eclipse User

I'm toying with eclipse link partitioning for my current project. we have an id generator that is base on the twitter id schema (first x bits = timestamp (seconds), 10 bits machine id, rest counter per second)

this means our ids are pretty large by default. the hashpartitioningpolicy uses the following line to find the target server:

int index = value.hashCode() % this.connectionPools.size();

in java % is the remainder function that can be negative IF value.hashCode() returns a negative value (its not a real modulo). since our Long ids are large the resulting hash code can indeed become negative, which causes a negative index, which causes an ArrayIndexOutOfBoundsException.

I'm going to work around this by implementing my own custom policy but this really should be fixed.

Greetings,
Alexander

Re: HashPartitioningPolicy fails for large IDs of type long [message #923001 is a reply to message #916123]

Tue, 25 September 2012 10:27

Eclipse User

This is a bug.

https://bugs.eclipse.org/bugs/show_bug.cgi?id=371514

It will hopefully be fixed soon.

[Updated on: Tue, 25 September 2012 10:28] by Moderator

Previous Topic:	NPE in MethodAttributeAccessor.getAttributeValueFromObject()
Next Topic:	EclipseLink/Toplink Cache coordination using JMS for manual DB updates

Goto Forum:

-=] Back to Top [=-

Current Time: Sat Jul 26 23:29:52 EDT 2025

.:: Contact :: Home ::.

Breadcrumbs

Sign up to our Newsletter