On Fri, Apr 17, 2015 at 1:30 AM, Andres Freund <andres@anarazel.de> wrote:
>> However, like my second approach, there would be a "speculative
>> affirmation" WAL record.
>
> I think there should be one, but it's not required for the approach. The
> 'pending speculative insertion' can just be completed whenever there's a
> insert/update/delete that's not a super deletion.
>
> I.e. in the REORDER_BUFFER_CHANGE_INSERT/... case ReorderBufferCommit()
> just add something like:
>
> if (pending_insertion != NULL)
> {
> if (new_record_is_super_deletion)
> ReorderBufferReturnTupleBuf(pending_insertion);
> else
> rb->apply_change(rb, txn, relation, pending_insertion);
> }
> ...
>
> You'll have to be careful to store the pending_insertion *after*
> ReorderBufferToastReplace(), but that should not be a problem.
Okay. It seems like what you're saying is that I should be prepared to
have to deal with a REORDER_BUFFER_CHANGE_INTERNAL_SNAPSHOT change (or
multiple such changes) from within ReorderBufferCommit() between a
speculative insertion and a super deletion, but that I can safely
assume that once some other insert/update/delete is encountered (or
once all changes have been drained from the reorder buffer), I can
then apply the speculative insertion as a regular insertion.
Is that what you're saying, in a nutshell? IOW, when you said this:
"""
I earlier thought that'd not be ok because there could be a new catalog
snapshot inbetween, but I was mistaken: The locking on the source
transaction prevents problems.
"""
What you meant was that you'd decided that this pattern is not broken,
*provided* that the implementation simply account for the fact that a
REORDER_BUFFER_CHANGE_INTERNAL_SNAPSHOT change could come before some
other (non-REORDER_BUFFER_CHANGE_INTERNAL_DELETE, A.K.A.
non-superdelete) change came? And that we might need to be more
"patient" about deciding that a speculative insertion succeeded (more
"patient" than considering only one single non-superdelete change,
that can be anything else)?
--
Peter Geoghegan