Re: [HACKERS] "op must return bool to be used with subquery"? - Mailing list pgsql-hackers

From Tom Lane
Subject Re: [HACKERS] "op must return bool to be used with subquery"?
Date
Msg-id 7679.926354677@sss.pgh.pa.us
Whole thread Raw
In response to Re: [HACKERS] "op must return bool to be used with subquery"?  (Bruce Momjian <maillist@candle.pha.pa.us>)
List pgsql-hackers
Bruce Momjian <maillist@candle.pha.pa.us> writes:
>> Given
>> create table t1 (name text, value float8);
>> 
>> this fails:
>> SELECT name, value FROM t1 as touter WHERE
>> (value/(SELECT AVG(value) FROM t1 WHERE name = touter.name)) > 0.75;
>> ERROR:  parser: '/' must return 'bool' to be used with subquery

> Aren't you really saying WHERE col / (subselect).  That doesn't return
> bool, so the message seems correct.

No, look again: the result of the subselect is being used as an operand
within the WHERE clause:WHERE (value/(SUBSELECT)) > 0.75;

If the / were the toplevel operator in the WHERE then the message would
make sense, because the WHERE clause as a whole must yield boolean.
But that doesn't mean that the operator immediately above the subselect
must yield boolean.

Besides, I already fixed this ;-)

> We don't allow subselects in target lists, or inside expressions.

We don't allow 'em in target lists, I think (anyone understand why not?)
but they work fine inside expressions in WHERE or HAVING.
        regards, tom lane


pgsql-hackers by date:

Previous
From: jwieck@debis.com (Jan Wieck)
Date:
Subject: Re: [HACKERS] create view as select distinct (fwd)
Next
From: Bruce Momjian
Date:
Subject: Re: [GENERAL] cluster truncates table name?