>> The problem appears to be in the sorting of nulls, which is used by
>> UNION ALL:
>>         test=> select null order by 1;
>>         ERROR:  type id lookup of 0 failed
>
>Hmm. And I've got trouble with the following when I assigned the type
>"UNKNOWNOID" to the null fields:

I think this is ok. Since the first select has to define the datatype, I think a forced
typecasting (like NULL::varchar) is perfectly OK in the above case.

Andreas

pgsql-hackers by date

Next:	From: Thomas G. Lockhart	Date: 1998-07-14 15:09:29
	Subject: Re: [HACKERS] atttypmod now 32 bits, interface change
Previous:	From: Tom Lane	Date: 1998-07-14 14:28:46
	Subject: Re: AW: [HACKERS] Sequence objects have no global currval operator?