Re: Skip ExecCheckRTPerms in CTAS with no data

On Tue, Sep 29, 2020 at 5:09 PM Bharath Rupireddy
<bharath(dot)rupireddyforpostgres(at)gmail(dot)com> wrote:
>
> On Mon, Sep 28, 2020 at 7:48 PM Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> >
> > Bharath Rupireddy <bharath(dot)rupireddyforpostgres(at)gmail(dot)com> writes:
> > > In case of CTAS with no data, we actually do not insert the tuples
> > > into the created table, so we can skip checking for the insert
> > > permissions. Anyways, the insert permissions will be checked when the
> > > tuples are inserted into the table.
> >
> > I'd argue this is wrong. You don't get to skip permissions checks
> > in ordinary queries just because, say, there's a LIMIT 0 on the
> > query.
> >
>
> Right, when there's a select with limit 0 clause, we do check for the
> select permissions. But my point is, we don't check insert
> permissions(or select or update etc.) when we create a plain table
> using CREATE TABLE test_tbl(a1 INT). Of course, we do check create
> permissions. Going by the similar point, shouldn't we also check only
> create permission(which is already being done as part of
> DefineRelation) and skip the insert permission(the change this patch
> does) for the new table being created as part of CREATE TABLE test_tbl
> AS SELECT * FROM test_tbl2? However select permission will be checked
> for test_tbl2. The insert permissions will be checked anyways before
> inserting rows into the table created in CTAS.
>

Added this to the commitfest for further review.

https://commitfest.postgresql.org/30/2755/

With Regards,
Bharath Rupireddy.
EnterpriseDB: http://www.enterprisedb.com