Re: CREATE INDEX and HOT - revised design

Simon Riggs wrote:
> On Wed, 2007-03-28 at 22:24 +0530, Pavan Deolasee wrote:
>
>> Just when I thought we have nailed down CREATE INDEX, I realized
>> that there something more to worry. The problem is with the HOT-chains
>> created by our own transaction which is creating the index. We thought
>> it will be enough to index the tuple at the head-of-the-chain since
>> that
>> would be the visible copy once the transaction commits. We thought
>> of keeping the index unavailable for queries in pre-existing
>> transactions
>> by setting a new "xid" attribute in pg_index. The question is what
>> value
>> to assign to "xid". I though we would assign ReadNewTransactionId().
>
> If you are indexing a table that hasn't just been created by you, set
> the xcreate field on pg_index at the *end* of the build using
> ReadNewTransactionId(). Any xid less than that sees the index as
> invalid. If you created the table in this transaction (i.e.
> createSubId != 0) then set xcreate to creating xid.

Couldn't you store the creating transaction's xid in pg_index, and
let other transaction check that against their snapshot like they
would for any tuple's xmin or xmax? (With one exception - the creating
transaction would consider indices it built itself invalid, which
is not how things usually work for xmin/xmax).

This would mean that any transaction that believes that the creating
transaction has committed also consideres the index to be valid.

greetings, Florian Pflug