Re: Compute attr_needed for child relations (was Re: inherit support for foreign tables) - Mailing list pgsql-hackers

(2014/08/21 13:21), Ashutosh Bapat wrote:

On Wed, Aug 20, 2014 at 3:25 PM, Etsuro Fujita

<fujita.etsuro@lab.ntt.co.jp <mailto:fujita.etsuro@lab.ntt.co.jp>> wrote:

    Hi Ashutish,

I am sorry that I mistook your name's spelling.

    (2014/08/14 22:30), Ashutosh Bapat wrote:

        On Thu, Aug 14, 2014 at 10:05 AM, Etsuro Fujita
        <fujita.etsuro@lab.ntt.co.jp
        <mailto:fujita.etsuro@lab.ntt.co.jp>

        <mailto:fujita.etsuro@lab.ntt.__co.jp

        <mailto:fujita.etsuro@lab.ntt.co.jp>>> wrote:

             (2014/08/08 18:51), Etsuro Fujita wrote:
              >>> (2014/06/30 22:48), Tom Lane wrote:
              >>>> I wonder whether it isn't time to change that.  It
        was coded
             like that
              >>>> originally only because calculating the values
        would've been a
             waste of
              >>>> cycles at the time.  But this is at least the third place
             where it'd be
              >>>> useful to have attr_needed for child rels.

             There was a problem with the previous patch, which will be
        described
             below.  Attached is the updated version of the patch
        addressing that.

        Here are some more comments

        attr_needed also has the attributes used in the restriction
        clauses as
        seen in distribute_qual_to_rels(), so, it looks unnecessary to call
        pull_varattnos() on the clauses in baserestrictinfo in functions

        check_selective_binary___conversion(),
        remove_unused_subquery___outputs(),
        check_index_only().

    IIUC, I think it's *necessary* to do that in those functions since
    the attributes used in the restriction clauses in baserestrictinfo
    are not added to attr_needed due the following code in
    distribute_qual_to_rels.

That's right. Thanks for pointing that out.

        Although in case of RTE_RELATION, the 0th entry in attr_needed

        corresponds to FirstLowInvalidHeapAttributeNu__mber + 1, it's

        always safer
        to use it is RelOptInfo::min_attr, in case get_relation_info()
        wants to
        change assumption or somewhere down the line some other part of code
        wants to change attr_needed information. It may be unlikely, that it
        would change, but it will be better to stick to comments in
        RelOptInfo
           443     AttrNumber  min_attr;       /* smallest attrno of rel
        (often
        <0) */
           444     AttrNumber  max_attr;       /* largest attrno of rel */
           445     Relids     *attr_needed;    /* array indexed [min_attr ..
        max_attr] */

    Good point!  Attached is the revised version of the patch.

If the patch is not in the commit-fest, can you please add it there?

I've already done that:

https://commitfest.postgresql.org/action/patch_view?id=1529

 From my side, the review is done, it should be marked "ready for
committer", unless somebody else wants to review.

Many thanks!

Best regards,
Etsuro Fujita