Comment 15 for bug 798255

I concur that Thomas's branch is much faster, and acceptably faster at that, on my development server.

Using a copy of our production database with 4,544,711 circ transactions, the upgrade script has gone from taking over a day to just three hours to run. This is on a virtual machine with 6GB of RAM allotted.

I'll sign off on Thomas's branch, since it is a performance improvement of something that already exists and not necessarily a new feature. (We have a rule between us about things like that to avoid any problems with the community.)

It would be good if others could test the script, too. Just to make sure that we haven't somehow magically optimized it for our data, but left others broken. I seriously doubt it, but you never know until you try.