Re: [Gnumed-devel] pooling connections

[Julio Jiménez]

Horst Herb wrote:

> On Sun, 1 Sep 2002 09:33 pm, Julio Jiménez wrote:
>
>  
>
> > It's true.. "commits" are at connections levels. But I think you musn't
> > have 20 connections for it. you
> > must issolate and serialize the transactions (don execute cursors for
> > insert, update... until really need to be upgraded, then apply
> > connection.commit()) this is the way i usually use.
> >    
>
> Let us assume you have the processes A and B.
> Both request a connection C, and get the same connection.
> A begins a long transaction, B a short one.
> B commit fails, rollback is triggered.
> If A and B share the same connection, *both* transactions would be rolled 
> back, wouldn't they? 
>
>  
well, of course all pendings transactions will rolled back..., but what 
I say is that serialize and isolate the transactions...
to know... don't send any transaction (cursor.execute()) for any cursor 
untill you really need to finish.

I suppose A and B are cursors for C conecction... and wat two transactions

you can

A.execute('insert....')
A.execute('delete....')
A.execute('update....')
[...]
fo finish...
C.commit()


B.execute('insert....')
B.execute('delete....')
B.execute('update....')
[...]
fo finish...
C.commit() or C.rollback() (only B cursor is afected)

You can't
A.execute('insert....')
A.execute('delete....')
B.execute('insert....')
B.execute('delete....')
B.rollback()  #wrong
then
c.rollback() finish all transactions for A and B


Can you isolate transactions in this way?
Do you need keep insert... delete... pending for commit while the progam 
is running?

When you say...

"the processes A and B."

Do you mean cursors or multy-threaded procceses...


Julio Jiménez