> This seems not be the case in ECPG. In each COMMIT, the cursors are
> closed (they dont even need to close cursors in Oracle!). And at each
> BEGIN TRANSACTION PostgreSQL seems to compute again parsing and query
> plan..
I am still convinced that there is room for interpretation of the
standard here, Michael.
Since we have "begin work" all cursors that were opened outside
a tx block (before "begin work") should imho in no way be affected by a commit.
(e.g. Informix does it like that)
Someone who wants more conformant behavior would need to use the mode
of operation where you are always in a tx anyway, thus loosing the
above feature :-)
Unfortunately I think the backend currently would lack the necessary
support for this, since commit does the cleanup work for the cursor ?
Such a cursor would need an explicit close or open on the prepared
statement to be cleaned up.
Andreas