Just noted that for the use-case of auto-completion for a certain value that is stored in a JCR property, it of course make sense to leverage the Lucene term space for that property. Using the index as I proposed is still good for performance then, and to handle different lists at "a" and "a/p" (the latter might have "new" terms compared to "a", since it could include terms with lower popularity starting with "ap" since it no longer includes all the other "a*" terms with *!=p.
Regards, Alex -- Alexander Klimetschek Developer // Adobe (Day) // Berlin - Basel
