Stephane,

> no, starting 'R' or like 'R%' is the same number of indexed read.

That is correct.

> but why the like 'R%' gave so much of indexed read (500 000 indexed

read

> when their is only 250 000 record that start with R)?

I can explain 250,000 reads -- the engine needs to check that each row
actually starts with R -- Firebird indexes are not "exact" (perhaps
someone can better describe this -- I have hit a mental block) so the
engine needs to read each row to double check the search conditions.


As for the other 250,000 reads, I thought I remembered something about
the engine double reading rows when evaluating LIKE/STARTING WITH
clauses, but I can't find a case in the tracker for this.

> is it a "lake" in firebird index engine ?

"lake"?? Do you mean "leak"?


Sean