Re: Transaction id inrement - Mailing list pgsql-hackers

From Tom Lane
Subject Re: Transaction id inrement
Date
Msg-id 8801.998320965@sss.pgh.pa.us
Whole thread Raw
In response to Transaction id inrement  ("maruthi maruthi" <maruthi49@hotmail.com>)
List pgsql-hackers
"maruthi maruthi" <maruthi49@hotmail.com> writes:
>    Transaction id is incremented even in sql queries like
> "select"  which does not change the state of database, is it not
> unnecesary?.

No, it's not unnecessary.  Every DB operation has to have a transaction
ID; what's more, we have to assign one long before we have any idea
whether the transaction will prove to be read-only.

It's at least theoretically possible that we could recycle the
transaction ID of a completed transaction that's proven to be read-only,
but the bookkeeping involved would be far more trouble than it's worth.
Not least because it would break MVCC assumptions about transactions
starting in sequence number order.
        regards, tom lane


pgsql-hackers by date:

Previous
From: Peter Eisentraut
Date:
Subject: Re: Locale by default?
Next
From: Bruce Momjian
Date:
Subject: Re: Guide to PostgreSQL source tree