Re: Bug? CREATE TABLE AS (... UNION ...)

Lists: pgsql-hackers
From: Gregory Stark <stark(at)enterprisedb(dot)com>
To: PostgreSQL Development <pgsql-hackers(at)postgresql(dot)org>
Subject: Bug? CREATE TABLE AS (... UNION ...)
Date: 2007-01-30 11:00:16
Message-ID: 87fy9saif3.fsf@stark.xeocode.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-hackers

I think I found a bug, or at least a discrepancy. Afaict the
transformSetOperationsStmt function should have identical code to
transformSelectStmt outside of the operations affected by set operations. If
that's the case then the SELECT INTO/CREATE TABLE AS code was not updated when
last it was touched for regular queries.

I think this means WITH[OUT] OIDS and WITH <definition> won't currently work
correctly if the select query involves a UNION or other set operation. Also
temporary tables created with an ON COMMIT option will ignore it and any
tablespace directive will be ignored.

Should I just copy the same code over or is anyone interested in refactoring
this? Or do I have it wrong somehow?

TransformSelectStmt:

/* handle any SELECT INTO/CREATE TABLE AS spec */
if (stmt->into)
{
qry->into = stmt->into;
if (stmt->intoColNames)
applyColumnNames(qry->targetList, stmt->intoColNames);
qry->intoOptions = copyObject(stmt->intoOptions);
qry->intoOnCommit = stmt->intoOnCommit;
qry->intoTableSpaceName = stmt->intoTableSpaceName;
}

transformSetOperationStmt:

/*
* Handle SELECT INTO/CREATE TABLE AS.
*
* Any column names from CREATE TABLE AS need to be attached to both the
* top level and the leftmost subquery. We do not do this earlier because
* we do *not* want sortClause processing to be affected.
*/
if (intoColNames)
{
applyColumnNames(qry->targetList, intoColNames);
applyColumnNames(leftmostQuery->targetList, intoColNames);
}

--
Gregory Stark
EnterpriseDB http://www.enterprisedb.com


From: Gregory Stark <stark(at)enterprisedb(dot)com>
To: "PostgreSQL Development" <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Bug? CREATE TABLE AS (... UNION ...)
Date: 2007-01-30 11:19:13
Message-ID: 87bqkgahji.fsf@stark.xeocode.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-hackers

"Gregory Stark" <stark(at)enterprisedb(dot)com> writes:

> Should I just copy the same code over or is anyone interested in refactoring
> this? Or do I have it wrong somehow?

Hm, it appears I have this wrong somehow since I can create tables using
CREATE TABLE AS specifying tablespaces just fine. But I do't see how it can
work.

--
Gregory Stark
EnterpriseDB http://www.enterprisedb.com


From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Gregory Stark <stark(at)enterprisedb(dot)com>
Cc: "PostgreSQL Development" <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Bug? CREATE TABLE AS (... UNION ...)
Date: 2007-01-30 17:02:14
Message-ID: 3312.1170176534@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-hackers

Gregory Stark <stark(at)enterprisedb(dot)com> writes:
> Hm, it appears I have this wrong somehow since I can create tables using
> CREATE TABLE AS specifying tablespaces just fine. But I do't see how it can
> work.

Look at the first few lines of transformSetOperationStmt.

regards, tom lane