[cx-oracle-users] Re: Need fetchall invalidate LOB locators?

[Hamish L. <ham...@gm...>]

Anthony Tuininga wrote:

> without looking at the code I would guess that [DCOracle2] define new
> LOB locator variables for each of the rows that are fetched. That
> seems to me to be a lot of overhead for very little gain -- unless you
> can give me some valid reason for why fetchall() is a good idea when
> dealing with LOBs.

The main reason I'd offer is completeness and to avoid special-casing.
I've been working on the contribution of an Oracle driver to the PyDO
2 ORM project, and I've been obliged to reimplement one of the
inherited core methods just because it uses fetchall; I can envisage
other situations where code that is not our own might expect to be
able to call fetchall. You are right of course that there is a memory
overhead in allocating LOB locator variables for each row. But any
time the programmer chooses between fetchall and fetchone/iteration,
they are deciding on a trade-off between memory and time costs. Why
for the special case of the resultset's having LOBs should cx_Oracle
remove that choice from the programmer?


Hamish Lawson