incubator-jspwiki-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Janne Jalkanen <janne.jalka...@iki.fi>
Subject Re: [jira] Created: (JSPWIKI-246) Remove/replace IndexPlugin
Date Thu, 17 Apr 2008 14:40:39 GMT
> Yes, upon wiki initialization, the first time each of the respective
> managers is called upon for their respective indices. For example,
> here's an output from the IndexPlugin:
> 
>     2254 pages in PageManager; 2365 pages in ReferenceManager;
>     4524 pages in com.ecyrd.jspwiki.providers.CachingProvider
> 
> Now, I freely admit I may have something awry. I'm mucking about building
> pages via batch processes, etc., but it did strike me as a bit strange
> seeing the discrepancies.

Hmhmhm...  I think this needs a bit of investigation.  The PageManager
figure should be the correct one.  It might be that some of the
checkEnglishPlurals magic that ReferenceManager is doing might be
causing some problems here - because that number should always be less
than or equal to what PageManager is doing.

CachingProvider figures sound strange.  But that could be an artifact
of the Cache mechanism - maybe you're not calling it right?

> >Because by the time RefMgr startup is complete, findCreated() *should* 
> >return a list of all pages, though it does not necessarily reflect the 
> >latest state of the repository.
> 
> I don't follow. Isn't the latest state of the repository the same thing
> as the list of all pages? What I'm finding is that there are three
> different numbers sometimes.

The repository might be changed outside of the wiki process, in which
case ReferenceManager will not be notified of the changes until
someone actually tries to access the page, CachingProvider notices
that the page is changed, and requests ReferenceManager to do a new
reindex.

> Understood. I kinda want a simple boolean that gets set true following
> a run of ReferenceManager's buildKeyLists(Collection) method. Not sure
> how special-case that really is...

I think it is a bit of a special case.  The case where those two
should be differing significantly is during the bootup/reindex time,
but at that time you do not generally need accurate output from
plugins in any case.

> I think I'm using a very nearly current svn version of 2.6.2. I may
> have forgotten to kill the work directory -- but gotta run now, will
> try again tomorrow.

This is the change I am talking about.  It's not been overly well
tested, but it should make the process a bit more robust.

2008-04-12  Janne Jalkanen <jalkanen@apache.org>

        * 2.7.0-svn-11

        * Improved restart time at the expense of more disk being
        used and the first startup taking a bit longer.

/Janne

Mime
View raw message