Open Centril opened 2 weeks ago
Oh, I see. I missed the fact that the row in question is not within the range.
If we had BTreeIndexIter::<K, V>::skip_to_next_greater_than(&mut self, min_bound: &K)
(the leapfrog join operator) we could use multi-column indexes to answer range queries...
Description of Changes
ColId
inIndexJoin
andJoinInner
as opposed toFieldName
. This avoids queryingHeader
inbuild_query
.Header
s for inner joins and projections.QueryExpr::head
now provides the end-result header for the entire query, which is the same header as you'd get frombuild_query
. This means that we can eventually stop making aHeader
inbuild_query
and thereby move more runtime to query compilation and spend less in query building / evaluation.build_query
functions.API and ABI breaking changes
None
Expected complexity level and risk
2