> On Mar 31, 2017, at 1:03 PM, Hugi Thordarson <h...@godurkodi.is> wrote: > > Hi Andrus, sorry about the belated reply. > >>> Hi all. >>> I’m getting some unexpected behaviour when applying a fetch limit and doing >>> joint prefetching on a to-many relationship. In short, the number of >>> objects returned is always lower than the actual fetch limit I set. Problem >>> goes away if I do a disjoint prefetch or no prefetching at all. >>> >>> I created a tiny self-contained maven project that demonstrates the problem: >>> >>> https://bitbucket.org/hugi/prefetch-dilemma >>> >>> The source for the main class in that project (that can be run to show what >>> happens): >>> >>> https://bitbucket.org/hugi/prefetch-dilemma/src/master/src/main/java/prefetchdilemma/Main.java >>> >>> Anyone run into this before? >>> >>> Cheers, >>> - hugi >> >> >> Yeah, that's a limitation described here: [1]. Also mentioned in the docs, >> though it may be hard to spot [2]: >> >> "Disjoint-by-ID Prefetching Semantics [..] Moreover this is the only type of >> prefetch that can handle SelectQueries with fetch limit. Both joint and >> regular disjoint prefetches may produce invalid results or generate >> inefficient fetch-the-entire table SQL when fetch limit is in effect.” > > Maybe Cayenne should throw an exception (or log a warning) if prefetch > semantics other than Disjoint-by-ID is used with a fetch limit? This is > something I can see users burning themselves on (I know I did). > > Cheers, > - hugi
I'd go with a warning initially. Though we will need to identify all specific cases to print it. E.g. "fetchLimit|offset + joint + to-many". Andrus
