> > Otherwise we have to encode and decode a buffer of length zero. If no > > buffer is passed, we just drop that field all together. Also, I wasn't > > thinking that the second submitQuery() exists. The goal of the buffer > > field is to be able to keep data off heap. We probably shouldn't use > it > > for for metadata (such as the query plan). > - Not sure what you're saying… Are you suggesting that the query > itself should be kept in RunQuery >
Basically yes. I'd like to be strongly typed unless we're specifically dealing with data we want to manage off heap. As such, I was thinking that only record batches should be passed via the open ended bytebuf interface. I see that you've made the other updates. I'll merge those into the execwork wip branch. Thanks for all your effort. Jacques
