[Kde-pim] Review Request: Optimize Contact Search Job queries

[Commit Hook]

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://git.reviewboard.kde.org/r/107065/#review21332
-----------------------------------------------------------


This review has been submitted with commit 69f53dbf56194cd0633e7f5fede165c2cb7f6076 by Vishesh Handa to branch master.

- Commit Hook


On Oct. 26, 2012, 8:32 p.m., Vishesh Handa wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> http://git.reviewboard.kde.org/r/107065/
> -----------------------------------------------------------
> 
> (Updated Oct. 26, 2012, 8:32 p.m.)
> 
> 
> Review request for KDEPIM-Libraries, Tobias Koenig, David Faure, and Sebastian Trueg.
> 
> 
> Description
> -------
> 
> Some of the contact search job queries are pathological, and would result in virtuoso consuming a lot of cpu for large amounts of time. I've optimized the queries by doing the following -
> 
> * Not adding a "?r a nco:Contact" term. This is unnecessary as it results in an extra property being matched. Considering that Nepomuk data almost always follows the ontologies (the exception is legacy data), just using properties which have a domain of nco:Contact should guarantee the correct results.
> 
> * Avoid unions - Virtuoso cannot optimize the unions that well. Instead we use a FILTER(?p in (..)) instead.
> 
> * Avoding regex based search - Regex based search will always be terribly slow. It literally applies the regular expression on each candidate in order to filter them out. It's a lot better to use the full text index. This is done using 'bif:contains'. We do loose a little bit of accuracy, and we cannot match word boundaries. But I think have a good user experience trumps a little bit of accuracy. ( If you agree - then I'll use bif:contains everywhere )
>  
> 
> 
> Diffs
> -----
> 
>   akonadi/contact/contactsearchjob.cpp 5df3bfd 
> 
> Diff: http://git.reviewboard.kde.org/r/107065/diff/
> 
> 
> Testing
> -------
> 
> Not tested at all. I've just run some of the queries in virtuoso, and looked at the corresponding SQL. I've used that as a basis of optimization. I would like someone to test this out. Preferably people will a large quantities of indexed data. 
> 
> 
> Thanks,
> 
> Vishesh Handa
> 
>

_______________________________________________
KDE PIM mailing list kde-pim at kde.org
https://mail.kde.org/mailman/listinfo/kde-pim
KDE PIM home page at http://pim.kde.org/