Andrew Dunstan <andrew@dunslane.net> writes:
> Tom Lane wrote:
>> * Should CSV-mode logging include the virtual transaction ID (VXID) in
>> addition to, or instead of, XID? There will be many situations where
>> there is no XID.
> But will there be any where we care? Isn't the point of this to restrict
> allocation of a real XID to situations where having one might actually
> matter? All the rest seems to me just bookkeeping.
Well, the point is that a transaction might well emit log messages when
it hasn't changed anything and hence hasn't got an XID. If you would
like to be able to match up the messages generated by a single
transaction, you'd need to have VXID available to use as a join key.
In any case it seems a bit odd that we have a value that's available in
log_line_prefix and not available to users of CSV log output.
regards, tom lane
