Обсуждение: Re: [PATCHES] Patch for - Change FETCH/MOVE to use int8

Alvaro Herrera <alvherre@commandprompt.com> writes:
> I don't think this is the right approach.  Maybe it would be reasonable
> to add another arm to the %union instead, not sure.  The problem is the
> amount of ugly casts you have to use below.  The scanner code seems to
> think that a constant larger than the biggest int4 should be treated as
> float, so I'm not sure why this would work anyway.

I'm not sure that I see the point of this at all.  ISTM the entire
reason for using a cursor is that you're going to fetch the results
in bite-size pieces.  I don't see the current Postgres source code
surviving into the era where >2G rows is considered bite-size ;-)

I thought the int8-LIMIT patch was equally pointless, btw, but at
least it was not very invasive.  This one is not passing the minimum
usefulness-to-ugliness ratio for me.

            regards, tom lane

Tom Lane wrote:
> Alvaro Herrera <alvherre@commandprompt.com> writes:
> > I don't think this is the right approach.  Maybe it would be reasonable
> > to add another arm to the %union instead, not sure.  The problem is the
> > amount of ugly casts you have to use below.  The scanner code seems to
> > think that a constant larger than the biggest int4 should be treated as
> > float, so I'm not sure why this would work anyway.
>
> I'm not sure that I see the point of this at all.  ISTM the entire
> reason for using a cursor is that you're going to fetch the results
> in bite-size pieces.  I don't see the current Postgres source code
> surviving into the era where >2G rows is considered bite-size ;-)

Think MOVE to a specific section of the cursor > 2gig.  I can see that
happening.

--
  Bruce Momjian   bruce@momjian.us
  EnterpriseDB    http://www.enterprisedb.com

  + If your life is a hard drive, Christ can be your backup. +

Bruce Momjian <bruce@momjian.us> writes:
> Tom Lane wrote:
>> I'm not sure that I see the point of this at all.  ISTM the entire
>> reason for using a cursor is that you're going to fetch the results
>> in bite-size pieces.  I don't see the current Postgres source code
>> surviving into the era where >2G rows is considered bite-size ;-)

> Think MOVE to a specific section of the cursor > 2gig.  I can see that
> happening.

Yeah, and by the time it happens you'll have gotten bored and found
something else to do.  With no support in the system for random access
to a cursor result, this is just about as useless as the FETCH case.

In any case I agree with Alvaro's comment: the way to support int8 in
a FETCH/MOVE command is not to try to convert the entire rest of the
grammar to int8 instead of int4 as its native datatype.

            regards, tom lane