"Gaurav Mathur" <[EMAIL PROTECTED]> wrote in message news:[EMAIL PROTECTED] > Suppose I have the following schema: > > CREATE TABLE t1 > ( > f1 integer, > f2 text > ); > > CREATE TABLE t2 > ( > val integer > ); > > Suppose the following inserts statement are executed by two processes > over the above schema. > > insert into t1 values ((SELECT val FROM t2 WHERE ROWID=1), "aa"); > UPDATE t2 SET val = val + 1 WHERE ROWID=1; > > The idea is to use the value 'val' from the table t2 and set the value > of f1 to that value on an insert. The question I have is whether I can > rely on the above sequence of SQL to reliably generate *unique* f2 > values for all INSERTs.
Why not just use ROWID as a unique identifier? > Basically, I am not sure whether there is a > possibility that a thread/process can be preempted after executing the > SELECT but before executing the INSERT in that thread/process. No. An implicit transaction is created for each statement. What is possible is for another thread to squeeze a statement between your insert and update. To prevent that, surround the two statements with an explicit transaction. Igor Tandetnik _______________________________________________ sqlite-users mailing list [email protected] http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
