On 2016/02/12 6:38 PM, Michele Pradella wrote: > Already solved with UNION of SELECT >
It isn't solved, it is circumvented by trial and error without understanding why. I'm sure that works ok for you in this case, but the point is if you do study those documents a bit more, you may grasp the reason too, and perhaps make much better/faster queries in future without spending hours on trial and error. Save yourself a lot of time, we think. > Il 12/02/2016 17.03, Richard Hipp ha scritto: >> ...//so that it is easier to understand. Perhaps a table that shows the >> various combinations of COLLATE and PRAGMA case_sensitive_like work >> together. I dunno. I'll take the action item to revisit the >> documentation and try to improve it. @Dr.Hipp: Would it be possible / difficult to perhaps improve the like algorithm and pragma case-insensitive_like to take 3 values, such that 0 = OFF, 1 = ON and 2 = AS PER COLUMN DEF? I realize the like function maybe doesn't know which column, nor is the like operation guaranteed to happen on a column. Maybe there is another way? Perhaps only from inside the query planner when making plan decisions? (Currently it compares the ci-ness during planning to decide on best index to use, if I'm not mistaken). What I have done is simply added a custom LIKE function that takes a second parameter for case sensitivity to avoid pragma-juggling - but by this, the query planner is none the wiser I find myself sometimes needing the above, though not often enough to warrant requesting the feature, but it seems now maybe others have the need too. Thanks, Ryan