Two approaches
Choosing a small number of dimensions
- building a R-tree on these dimensions
- Problem: Which dimensions to choose according to incomplete subscriptions?
Choosing all dimensions
- Problem: generate 2d points queries
- S: (title = Hamlet ) & (price < $20)
- S: (title = Hamlet ) & (price < $20) & (authors = ?) & (category = ?)
- (Hamlet,7,Sheak.,book) ? (?,7,Sheak,book), (?,?,?,book), (?,?,??), (Hamlet,?,Sheak,book), (Hamlet,?,?,book), … (Hamlet,7,?,?), …
Only possible if no schema changes