[dokuwiki] Re: Search Index

  • From: Harry Fuecks <hfuecks@xxxxxxxxx>
  • To: dokuwiki@xxxxxxxxxxxxx
  • Date: Thu, 11 Aug 2005 00:37:34 +0200

> I will try to build an demo index when I find the time just to have some 
> numbers. Maybe it could be possible to simply use serialize/unserialize for 
> it.

Could be a good starting place although serialize/unseralize might
lead to needing alot of memory, depending on how you use them. If you
serialize the entire index in one go, the entire index then first has
to be read into a string and from there, converted into a PHP data
structure. Using fopen / fgets, for example, you'd only be dealing
with a line at a time (perhaps a single line would be a seperate
serialized data structure?) - anything uninteresting could be
immediately discarded, free the associated memory.

> I think we have two restrictions to keep in mind: memory and execution time. 
> On a standard PHP install that's 8MB and 30 seconds per script run. I guess 
> the limit we hit first is memory. Executing the index update process in it's 
> own script could be a wise idea. Eg. using a method like pseudocron does with 
> an image tag that triggers it....

Might also be worth only processing one "index_update" message at a
time, while running the job frequently.
--
DokuWiki mailing list - more info at
http://wiki.splitbrain.org/wiki:mailinglist

Other related posts: