Josh Berkus <[EMAIL PROTECTED]> writes:
> ILIKE is not somehow aware that it is equivalent to lower().
Is it? Given the wild and wonderful behaviors of locales here and
there, I wouldn't want to assume that such an equivalence holds.
In particular I note that iclike() seems to be multibyte-aware
--On tisdag, mars 30, 2004 16.56.09 -0800 Josh Berkus <[EMAIL PROTECTED]>
wrote:
Palle,
But what about ILIKE. It does not take advantage of indices built with
lower():
Nope. If you want to use a functional index, you'll need to use the
function when you call the query. ILIKE is not somehow a
Palle,
> Too bad... that was my idea, that it would somehow be aware that it is
> equivalent to lower() like. It really is, isn't it? I would have though
> they where synonymous. If not, makes ILIKE kind of unusable, at least
> unless you're pretty certain the field will never indexed.
Yup.
Palle,
> But what about ILIKE. It does not take advantage of indices built with
> lower():
Nope. If you want to use a functional index, you'll need to use the function
when you call the query. ILIKE is not somehow aware that it is equivalent
to lower().
--
-Josh Berkus
Aglio Database Sol
--On tisdag, mars 30, 2004 19.16.44 -0500 Tom Lane <[EMAIL PROTECTED]>
wrote:
Palle Girgensohn <[EMAIL PROTECTED]> writes:
Shouldn't the optimizer use indices if the like condition does not have
any wildcards?
I can't get excited about this; if you are depending on LIKE to be fast
then you sho
Palle Girgensohn <[EMAIL PROTECTED]> writes:
> Shouldn't the optimizer use indices if the like condition does not have any
> wildcards?
I can't get excited about this; if you are depending on LIKE to be fast
then you should have locale-insensitive indexes in place to support it.
Switching the tes
