Re: [HACKERS] Re: ALTER TABLE DROP COLUMN

Hiroshi Inoue wrote:

> > > Wow, that almost seems to easy to be true. I never thought that having
> > > tuples of different structures in the table at the same time would be
> > > possible. If so then I don't see a reason why this would be too hard to
> > > do.
> >
> > If the transaction is not committed, I don't think anything actually
> > reads the tuple columns, so you are safe.
> >
>
> Hmm,tuples of multiple version in a table ?
> This is neither clean nor easy for me.
> There's no such stuff which takes the case into account,AFAIK.
>
> Seems no one but me object to it. I'm tired of this issue and it's
> painful for me to continue discussion further in my poor English.
> I may be able to provide another implementation on trial and it
> may be easier than only objecting to your proposal.
> Is it OK ?

    Consider me on your side.

    For some good reasons, I added a

        ReferentialIntegritySnapshotOverride

    mode,  that  causes  any  tuple to be visible when fetched by
    CTID.  Actually, there will be at least a read lock on  them,
    so  locking  will  prevent  damage.  But I can think of other
    situations where this kind of "read whatever I want  you  to"
    could be needed and would fail then.


Jan

--

#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#========================================= wieck@debis.com (Jan Wieck) #