2010/5/4 Peter Eisentraut <peter_e@gmx.net>:
> On tis, 2010-05-04 at 09:19 +0100, Grzegorz Jaśkiewicz wrote:
>> 2010/5/3 Peter Eisentraut <peter_e@gmx.net>:
>> > It was a convenient choice. You could propose a different method for
>> > generating the specific routine name, but given that it has to fit into
>> > an identifier and has to allow for function overloading, some kind of
>> > number makes the most sense, in absence of any other requirements.
>>
>> how about just a name, with no OIDs ?
>
> The "specific name" must be unique among functions with the same name.
>
>> I am trying to compare two databases, and this really does get in a
>> way. I think it defeats the purpose here, since I have to chop the
>> numbers off.
>
> If you want the plain name, join information_schema.parameters with
> information_schema.routines and use the column routine_name.
I basically need to compare the whole schema, as much detail as possible.
So for now I am just taking any views from information_schema. I see
fit, and remove any columns that are hardcoded to NULL (not
supported), store that in a bunch of temporary tables, and will try to
compare it. Which will lead me to primary key battle later on, so
thanks for the hint :)
--
GJ