[Xapian-tickets] [Xapian] #394: Speed up phrase queries with a "settling pond"

Xapian nobody at xapian.org
Wed Apr 10 04:53:00 BST 2013


#394: Speed up phrase queries with a "settling pond"
-------------------------+--------------------------------------------------
 Reporter:  olly         |       Owner:  olly    
     Type:  enhancement  |      Status:  assigned
 Priority:  high         |   Milestone:  1.3.x   
Component:  Matcher      |     Version:  1.1.2   
 Severity:  normal       |    Keywords:          
Blockedby:               |    Platform:  All     
 Blocking:               |  
-------------------------+--------------------------------------------------

Comment(by olly):

 Gustavo Yoshizaki from Avature kindly provided the following benchmark
 results, and agreed to let me make them public:

 {{{
 These are the four configurations tested:
 A = No optimization
 B = Pond size optimization
 C = Phrase weight check optimization
 D = Both optimizations

 Average time in seconds to make more than 4000 searches (the same
 searches with the same dataset)
 A:    690.95
 B:    318.13
 C:    137.99
 D:    124.88

 Number of times in which the configuration was better than the other
 three configuration
 A: 464
 B: 728
 C: 1489
 D: 2183

 For the searches where configurations B, C and D is faster than
 configuration A, the % of speed up
 B: 62,94 %
 C: 59,58 %
 D: 57,87 %
 }}}

 So C is the patch which is in 1.2.14 and will be in 1.3.1.  B is the
 original pond size patch, I believe with the default pond size.

 The tl;dr version is that the patch now applied makes significantly more
 difference alone than the "pond" patch alone does (at least for these
 searches), but using both together would improve things further still.
 With the default pond size, by 10% on average, but I bet that's not the
 optimal size as it was just picked arbitrarily.

-- 
Ticket URL: <http://trac.xapian.org/ticket/394#comment:12>
Xapian <http://xapian.org/>
Xapian



More information about the Xapian-tickets mailing list