Comment on Peter's suggestion: I am no expert on these things, but it is my understanding that for MySQL LAST_INSERT_ID() fetches the last autoincrement value made by the current insert, so the outcome is not affected by virtually simultaneous requests. And I don't think MySQL accepts a value for the autoincrement key. I am a bit surprised that the code you quote is reliable - surely there is a possibility of an insert after calling GetNextID and before calling SQLInsert in the python code? Comment on Tino's suggestion: I think that LAST_INSERT_ID() is MySQL specific and I guess CURRVAL() is Postgres specific. But the principle is the same: calling within the same Z SQL Method ought to be safe. Anyway, I hope John has had is question answered. Cliff Tino Wildenhain wrote:
Peter Bengtsson wrote:
What if you have 1,000,000 requests/sec? What if between the INSERT and the LAST_INSERT_ID() another INSERT is made?
I use PostgreSQL and with postgres you can always ask the sequence what the next id is going to be. It goes something like this::
next_id = context.GetNextId()[0].next_id context.SQLInsertUser(uid=next_id, name='Peter')
where 'GetNextId' is a ZSQL method that looks like this::
<params></params> SELECT NEXTVAL('users_uid_seq') AS next_id
Its even easier: one ZSQL Method:
INSERT INTO foo (foo_id,blah,bar) VALUES (nextval('foo_foo_id_seq'),<dtml-sqlvar blah> ... ); SELECT CURRVAL('foo_foo_id_seq') as foo_id;
But your above solution is valid too.
Regards Tino _______________________________________________ Zope maillist - Zope@zope.org http://mail.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://mail.zope.org/mailman/listinfo/zope-announce http://mail.zope.org/mailman/listinfo/zope-dev )