I'm very excited about the JIT query compilation stuff finally making it into postgres! A big win for OLAP queries, although it looks like they compile specific expressions within the query rather than the whole query operator, but that could change. I'm very curious to see how this interacts with the query planner (e.g. when it's worth the overhead or not).
Right, it's "just" expressions and tuple deforming right now. Especially the former required significant refactoring (landing in v10), but that's nothing against the refactoring required to do proper whole query compilation. I'm working on the necessary changes (have posted prototype), which have independent advantages as well.
There's planning logic to decide whether JIT is worthwhile, but that's probably the weakest part right now.
