Very fast requests beating ZODB commits...
I'm seeing the following behaviour in Zope2.7b2 (python 2.3.2) on a relatively fast computer (Athlon 1.8GHz): 1. request comes in which modifies ZODB 2. code handling request replies with REQUEST.RESPONSE.redirect() 3. redirected page uses data input at step 1 (specifically, it's auth info) but that info hasn't been committed yet, so we get errors So I've modified the method at step #1 to get_transaction().commit() before doing the redirect(), but I'm usually a little concerned when I have to invoke the transaction machinery directly like this... Any thoughts? Is there some way to hold the redirect off until after the transaction is committed normally? Richard
On Tue, 14 Oct 2003 02:25 pm, Richard Jones wrote:
1. request comes in which modifies ZODB 2. code handling request replies with REQUEST.RESPONSE.redirect() 3. redirected page uses data input at step 1 (specifically, it's auth info) but that info hasn't been committed yet, so we get errors
Ehem. Operator error (confusion) ... nothing to see, move along. Richard
On Tue, 2003-10-14 at 07:31, Richard Jones wrote:
On Tue, 14 Oct 2003 02:25 pm, Richard Jones wrote:
1. request comes in which modifies ZODB 2. code handling request replies with REQUEST.RESPONSE.redirect() 3. redirected page uses data input at step 1 (specifically, it's auth info) but that info hasn't been committed yet, so we get errors
Ehem. Operator error (confusion) ... nothing to see, move along.
Its funny nonetheless that you should report such a thing: we have a customer who reports similar issues in a bit of their CMS, where a popup submits its form (in Javascript) and then refreshes its parent window; if the commit takes long enough (e.g., it hits a conflict), the parent can end up showing the "old" data after the refresh. Tres. -- =============================================================== Tres Seaver tseaver@zope.com Zope Corporation "Zope Dealers" http://www.zope.com
Tres Seaver wrote:
Its funny nonetheless that you should report such a thing: we have a customer who reports similar issues in a bit of their CMS, where a popup submits its form (in Javascript) and then refreshes its parent window; if the commit takes long enough (e.g., it hits a conflict), the parent can end up showing the "old" data after the refresh.
Damn these concurrent transactional object databases ;-) Chris
participants (3)
-
Chris Withers -
Richard Jones -
Tres Seaver