--- In firebird-support@yahoogroups.com, Helen Borrie <helebor@...> wrote:

>
> At 05:09 AM 8/03/2008, you wrote:
> >i am trying to mark (via a column that reserved for this purpose)
> >updated/inserted records in a trigger with name of the current

savepoint.

> >
> >How can i get current savepoint's name?
>
> A user savepoint's "name" is a client-side symbol (identifier) for

the argument to an SQL call from the client, in the context of the
current transaction. So, while the client can create a savepoint
between two other statements and refer back to it by using that
symbol, it is unknown to any subsequent code that will execute on the
server. Internal savepoints are not available at all.

>
> Assuming that your purpose is to have some string that you can write

to the affected records during an AFTER trigger, you could write a
stored procedure that you can call from your After Update or Insert
triggers to construct a varchar variable according to the information
you want to store, and call it at whatever point you need during your
write operation, i.e. outside or inside your FOR loop.

>
> For example:
>
> create procedure GetContextID
> returns (context_id varchar(92))
> as
> declare stub varchar(72);
> declare context_id varchar(92);
> declare time_now varchar(30);
> begin
> stub = cast (current_connection as varchar(20)) || ':'
> || cast (current_transaction as varchar(20)) || ':'
> || cast (current_user as varchar(20)) || ':' ;
>
> time_now = cast(cast('NOW' as timestamp) as varchar(30));
> context_id = stub || time_now;
> end
>
> Your triggers would just need to declare a variable for the context

id and execute the SP at the appropriate time, e.g.,

>
> create trigger ...
> active after insert
> as
> declare variable vcontext_id varchar(92);
> .....
> begin
> .....
> .....
> execute procedure GetContextID returning_values(:vcontext_id)
>
> and you would just plug the variable :vcontext_id into your INSERT

or UPDATE statement.

>
> A different approach might be to maintain a table for each savepoint

created by your application, that stores (without committing) a
varchar containing the name of the most recent SAVEPOINT argument
along with sufficiently context-unique keys that you can retrieve the
savepoint name via a query when the After trigger runs. From V.2.1
onward you could use a transaction-level GTT for this if you didn't
need persistence.

>
> ./heLen
>

Thank you for your answer.
I am trying to mark affected records with save point logic in order to
provide undo capability to my software.
and While doing that i must know which records
updated/deleted/inserted in which savepoint.
Because in client-side, i can change the dataset (fibplus dataset)
with its direct chache modify functions with querying only marked
records . In that way i don't need to close/open whole dataset which
is pretty big.
Now i think it is obvious that i should shift this logic to client-side.