Search Refresh

[RangeBob]

I suspect the indexing of posts is running more infrequently than normal.
My search won't turn up anything more recent than 24 hours ago.

[Doug_M]

Is there even an indexing? Doesn't it just directly search the database?

[Rory McCanuck]

I noticed that too.

[linung]

Is there even an indexing? Doesn't it just directly search the database?

indexing is a must.

[jwirecom109]

Change a few settings, Let me know if its still a problem

[RangeBob]

It's working again.
I just searched for "settings" and found post #5 in this thread, which was 3 minutes ago.
Good good.

[RangeBob]

Is there even an indexing? Doesn't it just directly search the database?

I assume the forum software includes something like Lucene/SOLR, which would have posts sent to it every ten minutes or so.
Searches would sent to Lucene, which would return the best matching posts, which in turn are presented from the database.

[Doug_M]

indexing is a must.

Yes but it is done at the database level (concurrent with the text of the posts themselves) with indexed fields as opposed to some "web crawler" that creates a separate database or table of indexed data.

[Doug_M]

I assume the forum software includes something like Lucene/SOLR, which would have posts sent to it every ten minutes or so.
Searches would sent to Lucene, which would return the best matching posts, which in turn are presented from the database.

Those are really fast. I remember when I posted the LGR that La Presse had it was on a PostgreSQL database. Search wasn't slow but it wasn't fast. Then someone (sorry, can't remember your name if you're still here) did the same but used Lucene (or SOLR) and it was near instant. See http://gank.ca/lgr

Still I assume that would be a custom solution for VB and that VB would have a built-in search that just used the database's existing indexed columns. I've never installed or managed VB though so I'm just guessing based off other systems like Wordpress.

[RangeBob]

Yes but it is done at the database level (concurrent with the text of the posts themselves) with indexed fields as opposed to some "web crawler" that creates a separate database or table of indexed data.

Not with large text fields with multiple words (sentences, paragraphs).
Typical database indexes, such as b-tree (or n-ary-tree), work well with single values;
but with a typical database index, finding a word within a sentence means a table scan rather than using the index.

Inverted indexes -- such as Lucine, or Oracle Text, or Word-In-Text, or similar -- work well with sentences and documents. But they're often performed asynchronously outside of the database and well after the commit, often several minutes later. And apparently the past couple of days until jwirecom109 fixed it, the inverted index was not being updated at all, even though the posts were still visible within threads.

[RangeBob]

Then someone (sorry, can't remember your name if you're still here) did the same but used Lucene (or SOLR) and it was near instant. See http://gank.ca/lgr

alter3d


SOLR is leverages Lucene.
Considering how it works, it's surprising that it offers acceptable performance, but it does.