If we have {$or:[
{a:/^abc/},
{a:/^ab/}]} we scan the overlap part twice because we would have to split the range [ "ab", "ac" ] in two to do so and we can't handle that yet. Note that range limiting that doesn't require splitting a range is handled nicely now. ( {$or:[{a:{$lt:"abc"}},
{a:/^ab/}]} does not do any double scanning. )
- depends on
-
SERVER-1212 tighter index bounds for negated regular expression
- Closed