Although we do not have the "Advanced Search" capabilities yet, I thought that having a list of the functions of the Lucene-based search engine would be useful. So, I put them in a Wiki... Hope that helps someone...
Current search options
- 7,440 Posts. Joined 9/2004
- Location: An innocent prisoner will become more angry by the hour due to...
- Select All Posts By This User
Did you actually test any of that before you posted it?
None of it appears to work. The search code is taking all queries and enclosing them in quotes, which makes it behave in very odd ways if you try to use different features of the search. Sometimes it just gives up:
Fatal error: Maximum execution time of 30 seconds exceeded in /usr/share/php/versionedIncludes/zend/1.7.2/Zend/Search/Lucene/Storage/File.php on line 268
I spent several minutes looking for a thread from perhaps November or December earlier, and I was completely unable to come anywhere close to finding it...because the search is pretty much useless. So I used Google, and came to the startling realization that no one has fed google any content older than this transition. And/or Google isn't eating it. Isn't this (one reason) why vBB has a low fi version with long thread lists paginated to eternity?
So not only did this transition break the already not very good search capabilities we had through vBB, it also destroyed all evidence on Google of threads older than a week. Awesome.
Edited by Garrett - Fri, 06 Feb 09 15:08:54 GMT
We're actively working on replacing our existing search library with a much more performant (and feature-rich) version of the same library. In addition, we will provide a more advanced interface, similar to what vBulletin offered.
In regard to your other point, there's a delay between when google spiders new URLs, and when they become available in search results. Google is very actively crawling the site, as evidenced by our logs. The drop in results you're seeing now happens to every site that undergoes a major redesign. Google's index will catch up over the coming weeks.
Edited by frankfarmer - Fri, 06 Feb 09 20:00:59 GMT
It is??? When I do a search, rather than giving me NEW topics listed first, it gives me ancient topics listed first. What changed? How do we get to the features which will give us posts instead of threads? What am I missing?
The new engine is active. It is able to search the entire database and return what it thinks to be the most likely 1000 results. You can use the directives outlined in that wiki to narrow your search more easily. The dev team is going to be working on adjusting the pre-defined relevance.
The next step (hopefully this week, but perhaps next) will be to allow for the "Advance Search" selection capabilities, including limiting results to certain forums and/or certain posters. We will also get to the point where you can choose threads or posts. Those features are still to come.
Is this search engine active in the Forums or just the Wiki? For instance, if I search in the Forums for "Whitefish", it brings back TONS of information. I just want to see recent discussions, though, so I try
whitefish mod_date:[20080101 TO 20090301]
And it returns zip. I've tried other dates, still zip. I've tried quotes around "whitefish". Still zip. The only example I saw for using the date modified doesn't have it limited to only posts with a certain word in it.
We haven't put much up regarding use of the advanced lucene query syntax, because it's fairly complex, the index schema is going to change rapidly while we add features, and we certainly don't expect most Bears to make use of it in the long run -- simply because we'll provide a much easier-to-use interface in the next week or so. So, if you want to experiment, by all means, go for it -- but for most Bears, it'll make more sense to just wait until we've got an interface that hides all the arcane lucene syntax.
With that out of the way, to answer your question...
You can't really take the mod_date example literally (much like you wouldn't want to use any of the other examples literally either) -- it assumes a field name (mod_date) and date format (YYYYMMDD) that we don't use.
The two fields you will probably be interested in are:
creationDate, which stores the date a wiki was created, a user joined, or a thread was started, etc.
lastModification, which stores the date of a thread's last post, or a wiki's last edit
Regarding the date format we use, see: http://wiki.apache.org/solr/SolrQuerySyntax
It's pretty ugly, but fortunately, once we've got the advanced search interface up, we'll handle all the date manipulation behind the scenes.
So, the query you tried:
Also, keep in mind that by default, terms are combined with OR -- so the query as written will find documents that contain "whitefish" OR were modified in the time range in question -- although those that match both terms should be ranked most relevant.
Still, you probably want only documents that both terms, so you'll need to specify that explicitly, in either of the following forms:
After our next release, the following, slightly more intuitive syntax will also work:
Note that this currently returns zero results, because the index that's up right now doesn't have compatibly formatted dates. See the comment about a rapidly changing index schema earlier in this post ;)
At any rate, for those of you interested in tinkering with this stuff, that should be enough to get you going; for everyone else, our #1 priority is putting together a nice interface for all this stuff so you never have to deal with this arcane gibberish.