On 2015-04-15 17:58:54 +0300, Heikki Linnakangas wrote:
> When the speculative insertion is finished, write a new kind of a WAL record
> for that. The record only needs to contain the ctid of the tuple. Replaying
> that record will clear the flag on the heap tuple that said that it was a
> speculative insertion.
>
> In logical decoding, decode speculative insertions like any other insertion.
> To decode a super-deletion record, scan the reorder buffer for the
> transaction to find the corresponding speculative insertion record for the
> tuple, and remove it.
>
> BTW, that'd work just as well without the new WAL record to finish a
> speculative insertion. Am I missing something?
I'm, completely independent of logical decoding, of the *VERY* strong
opinion that 'speculative insertions' should never be visible when
looking with normal snapshots. For one it allows to simplify
considerations around wraparound (which has proven to be a very good
idea, c.f. multixacts + vacuum causing data corruption years after it
was thought to be harmless). For another it allows to reclaim/redefine
the bit after a database restart/upgrade. Given how complex this is and
how scarce flags are that seems like a really good property.
And avoiding those flags to be visible to the outside requires a WAL
record, otherwise it won't be correct on the standby.
Andres