Woody Wu wrote: > On 2014年5月8日 GMT+08:00AM1:51:43, Clemens Ladisch <clem...@ladisch.de> wrote: >> Simon Slavin wrote: >>> You can let SQLite make a better estimate by using the 'ANALYZE' >>> command. >> >> Yes. However, even if a better estimate were to show that this index is >> not very useful, there would exist, at the moment, no better index (with >> id1 or id2 as the first indexed column). > > Do you mean that my index for this case is already the best choice?
No. In a database that has only this index, it is the only choice. But you should create a better index. > And, as a general question, for a query in the form of > select max(a) from table where a < InF and b=B and c=C. > what's the best index in the case of (1) B is an existed one, or (2) B is not > exists ? That depends. Create two indexes on (b,c,a) and (c,b,a), run ANALYZE, and check which one gets actually used. Regards, Clemens _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
