Hi, I'm finding that if I compile with either SQLITE_ENABLE_STAT3 or SQLITE_ENABLE_STAT4, repeatedly executing a prepared SELECT statement is substantially slower.
I have a silly benchmark app that populates a database with test data, creates some pertinent indexes, runs ANALYZE, then executes a particular SELECT statement against the database several thousand times. Executing all these SELECTs normally takes a few seconds, but it's 10x slower when SQLite is compiled with SQLITE_ENABLE_STAT4. Curious, I ran my test app under Visual Studio's profiler, and saw that a substantial amount of time is spent in calls to sqlite3Reprepare(). Reading the docs at http://www.sqlite.org/c3ref/prepare.html, I assume SQLlite is recompiling the statement because it thinks it can find a better query plan based on the parameters I'm binding to it. Unfortunately, the time needed for this recompilation is dramatically outweighing any potential savings that a better query plan could offer! Has anyone else had a similarly negative experience with SQLITE_ENABLE_STAT3/4? Cheers, Nick _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
