[Framework-Team] plip9327 (plone.app.contentlisting) review and performance

Geir Bækholt plone at baekholt.com
Fri Sep 4 16:05:17 UTC 2009


David has reviewed plip9327 at
https://dev.plone.org/plone/browser/buildouts/plone-coredev/branches/4.0/plips/plip9327-review-davisagli.txt?rev=29499
and points at some worry for performance loss.

After some profiling, i found the 15% performance loss in the fact that 
we now look up fullnames instead of usernames in the searchresults 
listing. This was already listed as a risk in the original search 
results PLIP. https://dev.plone.org/plone/ticket/9352

Thus, the performance problem is due to added functionality to PLIP 
9352, and really shouldn't affect the acceptance of PLIP 9327.

I have now also added a local cache (i'll improve it tomorrow) for the 
userdata-lookups so the current performance overhead is at around 2-5% 
now, not 15%.

so 1) the overhead is not really relevant to the reviewed PLIP
and 2) it is now quite smaller than originally thought.

Now that we have usable (though crude) performance tests at 
https://svn.plone.org/svn/plone/plone.app.search/trunk/plone/app/search/tests/test_search_performance.py
it is easy for me to keep profiling, running comparisons and adding 
improvements.


-- 
Geir Bækholt
http://www.jarn.com





More information about the Framework-Team mailing list