Performance numbers for WebChanges on small _default web:

Performance numbers for WebChanges on large Main web on TWiki.org with 21K topics:

This is with 7493.

-- PTh

WebChanges is done by a search, so this is a duplicate of Item960 where search performance is discussed. Discarding as a duplicate.

Is the performance of the changes script an issue? It's rather better than using a search, as it shows what changed in each topic, though I expect the performance will be well down on Cairo (it does an awful lot more).

CC

No, this is a separate actionable item. In Cairo I put in a lot of efforts to optimize search. The optimization that affects this item here is gone. Cairo reads the file timestamp first (which is fast), sorts them, takes the limit into account with some slack, then sorts the short list based on the topic rev timestamp (slow operation on short list)

-- PTh

If you look at the code, Search.pm, lines 617-647, you will see that that particular optimisation was recoded slightly, but was preserved in Dakar. The stat was moved into the store because it assumes a file system implementation. I tried hard not to compromise any of your optimisations; if I destroyed any, it was accidental and needs to be identified.

CC