Hi guys, I'm working on the final updates to hector-jpa for 0.8.2, and I've run into a bit of an issue. Now that forward and reverse comparators are natively supported, order by clauses on the indexes are natively sorted in Cassandra. I'm having some issues with my tests, here is my scenario.
An index for a "Sale" object is defined with the following format name, order by sale date desc. This translates into the following DynamicComposite. UTF8Type + BytesType(reversed=true) + BytesType (the RK that contains the sale item) When I'm creating the composite, everything appears to be constructed correctly, however when performing range scans where I only scan the first component in the composite, the rows are not returned in the reverse order as I expect, but rather in the order that would match a UTF8Type+BytesType+BytesType encoding. I've created a unit test that manually creates these columns the same as my plugin, and it passes. https://gist.github.com/c8517fb4799676a246d8 This leads me to believe that the serializiation of the column bytes on the client is incorrectly encoded. Where would be the best location in the DynamicCompoisteType comparator code to debug the columns as they are sorted? This will help me pin down the serialization problem, and ultimately get a 0.8.2 compatible version of the plugin deployed. Thanks, Todd
