[TYPO3-dev] Sphinx for indexed_search

[Ingo Renner]

Martin Kutschker wrote:

Hi Masi (and everybody else interested in cool numbers),

> I think we should rather implement nested lists for "pages" (as a TCA 
> feature!) than add this crutch. Other than that: why didn't you cache 
> in-memory?

no, because it would break too much stuff, you should know that yourself 
already.

> What is worth caching is the search result. Why redo the search every 
> time the user invokes paging in the browser?

if you think caching the search result is worth it, go ahead...

However, as I said: by caching the result of getTreeList all other 
extensions using pi_getPidList from pibase will benefit, too!

Here're some rough numbers from my test system:

indexed_search as in 4.2, pagetree with about 3000 pages, about 1000 
pages in index, search for a single word with about 250 results:
about 17 seconds

indexed_search with mysql fulltext index:
about 8 seconds

indexed_search with mysql fulltext index and getTreeList cached:
about 4 seconds

other extensions (general data storage folder with data structured in 
about 170 sub sysfolders, selecting the root data folder and setting 
recursion depth to infinite):
-------------------
tt_news treeList uncached: 0,4s
tt_news treeList cached: 0,05s

tt_address treeList uncached: 0,33s
tt_address treeList cached: 0,02s


BTW: with this scenario tt_addess would use the same tree list as 
tt_news (or the result of any other script using the function with the 
same root and depth)

Hope that makes clear how helpful this cache is in speeding up the whole 
system.
Now everybody smile please =)



Ingo

-- 
Ingo Renner
TYPO3 Core Developer, Release Manager TYPO3 4.2