Re: Named transaction

Tom Lane wrote:
> Andrew Dunstan <andrew(at)dunslane(dot)net> writes:
>
>> Alvaro Herrera wrote:
>>
>>> AFAIK that's an "autonomous transaction", at least to some other RDBMSs.
>>>
>
>
>> Right, but since I had to ask what that was recently I though I'd use a
>> bit more description :-)
>>
>
> Yes, but some other followups suggest that maybe a "named transaction"
> does something else entirely. Thus my request for a definition of what
> the OP is actually asking for.
>

According to the (hard to find) Firebird docs (or rather, the old
Interbase docs, which is all they have):

A single application can start simultaneous transactions. InterBase
extends transaction
management and data manipulation statements to support transaction
names, unique
identifiers that specify which transaction controls a given
statement among those
transactions that are active.

Transaction names must be used to distinguish one transaction from
another in programs
that use two or more transactions at a time. Each transaction
started while other
transactions are active requires a unique name and its own SET
TRANSACTION statement.
SET TRANSACTION can include optional parameters that modify a
transaction’s behavior.

There are four steps for using transaction names in a program:
1. Declare a unique host-language variable for each transaction
name. In C and
C++, transaction names should be declared as long pointers.
2. Initialize each transaction name to zero.
3. Use SET TRANSACTION to start each transaction using an available
transaction
name.
4. Include the transaction name in subsequent transaction management and
data manipulation statements that should be controlled by a specified
transaction.

cheers

andrew