Am Donnerstag, 3. Februar 2005 16:11 schrieb Rainer Frey:
A select should not lock a table even when it is not committed.
The SELECT obtains a read (shared) lock on the table, but the ALTER TABLE requires a write (exclusive) lock. This is certainly necessary because you don't want the structure of the table to be changed while you are reading it. Additionally, the locking protocol requires that all locks once obtained need to be held until the end of the transaction. Both of these issues together explain the problem you are seeing. There is nothing that can be done about it.
Thanks for the explanation, though I don't really get the necessity of a commit for a read-only statement. Can't a SELECT release its lock after it received the response?
Is there any possibility to set a timeout for the lock, after which the ALTER TABLE statement fails, instead of remaining in wait status (when calling with JDBC?
Rainer Frey
---------------------------(end of broadcast)--------------------------- TIP 6: Have you searched our list archives?
http://archives.postgresql.org
