[sqlite] Suggestions for Fast Set Logic?

2015-07-12 Thread James K. Lowden
On Sat, 11 Jul 2015 19:02:59 -0600 Scott Robison wrote: > > I don't follow you. A complex query is an assemblage of clauses. > > Whether or not broken down "to individual sub queries", the search > > arguments are the same. They are few in kind, even though they may > > be large in number.

[sqlite] Suggestions for Fast Set Logic?

2015-07-12 Thread Simon Slavin
On 12 Jul 2015, at 2:13am, Scott Robison wrote: > If you can know that, great. I'm just taking him at his word that arbitrary > complexity could result in table scans which are unacceptable. I don't think he has tried it. I think he has concluded that it will be too slow without having tried

[sqlite] Suggestions for Fast Set Logic?

2015-07-12 Thread Simon Slavin
On 12 Jul 2015, at 2:02am, Scott Robison wrote: > I'm just saying that knowing which index to create in advance is impossible > if the user can specify arbitrarily complex where clauses. You don't do that. You create a few indexes in advance based on the search combinations users are likely

[sqlite] Suggestions for Fast Set Logic?

2015-07-12 Thread Simon Slavin
On 12 Jul 2015, at 1:16am, James K. Lowden wrote: > Scott Robison wrote: > >> As described, the user will be able to create arbitrarily >> complex queries. Since it is impossible to know in advance what >> indexes might be required, breaking it down to individual sub queries >> with simple