On Wed, 24 Nov 2004, alan_at_erols.com wrote:
>
> "Galen Boyer" <galenboyer_at_hotpop.com> wrote in message
> news:uy8gsi3h1.fsf_at_standardandpoors.com...
>> On Tue, 23 Nov 2004, alan_at_erols.com wrote:
>> >
>> >> >> > This will not answer your question, but you have a
>> >> >> > bigger issue to solve.
>> >> >> >
>> >> >> > The first thing you need to know is that there is no
>> >> >> > such thing as the "first" or "newest" record unless
>> >> >> > you have time-dependent data of some kind (E.g., a
>> >> >> > timestamp). Tables are _sets_ of data, and as such,
>> >> >> > are technically unordered. There is no guarantee that
>> >> >> > records will be returned in the same order every time
>> >> >> > ( know- it sure looks like they are- don't be fooled).
>> >> >> > Rownum = 1 will not do what you need at all. Rownum
>> >> >> > can be used to reduce the number of records returned,
>> >> >> > but not to impose or infer an order.
>> >> >>
>> >> >> Well, almost. rownum will get him what he wants, as
>> >> >> long as it operates on the ordered set.
>> >> >>
>> >> >
>> >> > Only if the ordered set is ordered by something that is
>> >> > time dependent, or by whatever else it is that may be
>> >> > what he considers "first" (last name, E.g.). Anyway, I
>> >> > was trying to make the "set" point, as it did not appear
>> >> > that he understood the underlying problem.
>> >>
>> >> He is ordering on create date, so he at least seemed to
>> >> understand that side of it.
>> >>
>> >
>> > I don't see where he is using a date.
>>
>> ... order by t2.created ...
>
>
> You are assuming that "created" is a date.
Yes.
> I saw nothing to indicate that.
Seemed reasonable to me, but, yes I didn't ask for confirmation.
--
Galen Boyer
Received on Wed Nov 24 2004 - 09:49:15 CST
