I'd suggest putting this into the documentation of sqlite3_last_insert_rowid(), that the call is not reliable in scenarios such as this one.
-----Ursprüngliche Nachricht----- Von: D. Richard Hipp [mailto:[EMAIL PROTECTED] Gesendet: Sonntag, 28. Oktober 2007 17:48 An: sqlite-users@sqlite.org Betreff: Re: AW: AW: [sqlite] INSERT OR IGNORE and sqlite3_last_insert_rowid() On Oct 28, 2007, at 10:59 AM, Michael Ruck wrote: > Yes, I am well aware of this possibility as I've written in my > initial mail. > It just doesn't fit with the > description of sqlite3_last_insert_rowid() in my understanding. I > think this > is a bug - either in the documentation > or in the implementation. sqlite3_last_insert_rowid() should return > the > correct id, no matter what and it doesn't. > Consider this scenario: CREATE TABLE ex1(id INTEGER PRIMARY KEY, b UNIQUE, c UNIQUE); INSERT INTO ex1 VALUES(1,1,1); INSERT INTO ex1 VALUES(2,2,2); INSERT INTO ex1 VALUES(3,3,3); Now you do your INSERT OR IGNORE: INSERT OR IGNORE INTO ex1 VALUES(1,2,3); Three different constraints fail, one for each of three different rows. So if sqlite3_last_insert_rowid() were to operate as you suggest and return the rowid of the failed insert, when rowid would it return? 1, 2, or 3? D. Richard Hipp [EMAIL PROTECTED] ---------------------------------------------------------------------------- - To unsubscribe, send email to [EMAIL PROTECTED] ---------------------------------------------------------------------------- - ----------------------------------------------------------------------------- To unsubscribe, send email to [EMAIL PROTECTED] -----------------------------------------------------------------------------
