I have installed Lucene Search on a very small test bed (just a desktop Ubuntu server with an AMD3000+ processor, 1GB of RAM). I have successfully installed Lucene Search and got it to work. The site is completely indexed (there are a lot of documents with "lorem ipsum" content), with "Number of documents in index: 33", "Number of terms: 1537", "Number of fields: 28".
I have installed the MoreLikeThis module and pretty much left the settings by default, except that I added fields to take into account when extracting text (notably the Body field), and excluded content types. I was getting no results at all, despite the fact that there were lots of "lorem ipsum" articles on the site.
Now I have removed all the exclusions and it works (or at least the bloc pulls up some suggestions). Unfortunately some of the suggestions point to a CCK content type which has nothing to do with the other articles. When I exclude that content type from the search, then once again I get zero results. So my impression is that excluding content is the cause of the problem.
Moreover, I am getting a very noticeable performance hit now when I load a node. Since my target use case is a site whose main content is articles, where the Body contains between 2000 and 15000 words, this does not seem very realistic to do the search at the moment of the node load.
At present I am using the Related Links module which does pretty much the same thing, however it uses cron, I believe, to prepopulate a sort of index (or cache?) from which to serve the bloc, so load times are not really noticeable. Would it not make more sense to do something like this?
| Comment | File | Size | Author |
|---|---|---|---|
| #11 | luceneapi_morelikethis-750714-11.patch | 1006 bytes | cpliakas |
| #3 | luceneapi_morelikethis.patch | 950 bytes | jasperknops |
Comments
Comment #1
dgale commentedI also am noticing the problem of getting no results when excluding content types. In the morelikethis block, I am seeing a content type I don't want included, along with some others that I do. When I exclude that content type on the More Like This config page in Lucene Content and refresh my page in question, I get no results.
Comment #2
drupalsteve commentedI've observed the same problem. If any types are excluded from recommendations, no recommendations are provided.
Comment #3
jasperknops commentedI think I found the solution and created a patch. The problem is the way the arguments are passed to the query in the hook luceneapi_query_alter. Each term can be passed with property 'neither' but the multiterm must be passed as 'prohibited'.
Another thing I changed for myself, and is not included in the patch, is I added the nid to the $items that are passed to the theme function luceneapi_morelikethis_block so I can override this theme function and pass all the nids to a view I created.
Let me know if the patch solves your problem.
Comment #4
cpliakas commentedThanks for the patch jasperknops. At first glance this seems to be the correct approach, and I will apply it in my test installation to verify that it works as expected.
Thanks for the bug hunting,
Chris
Comment #5
daniel wentsch commentedThat patch did the trick for me, too. Thanks :)
Comment #6
daniel wentsch commented//sorry, I hit the save button twice.
Comment #7
joel_guesclin commentedAny ideas when this will be available? Also, I notice that the present version is in alpha and that the solution here is tagged beta - is there a timeline? And would it be of any help if I were to give some more time to making suggestions for improvement? I'm afraid I don't have the expertise to help with development on this.
Comment #8
cpliakas commentedHonestly, I've been remiss with development of this module in favor of porting Search Lucene API to D7 and completing the Facet API module. I will try to roll out a release so that some of the outstanding issues will be resolved, but my free time has completely diminished since I started working for Acquia.
Thanks, and sorry for the delays.
~Chris
P.S. #836760: Co-maintainer wanted
Comment #9
joel_guesclin commentedI would be grateful of an educated opinion on the status of D6 development (in particular and in general). I am still porting sites from D5 to D6, and don't plan to start porting to D7 probably for another year, for four reasons:
1) Lack of time
2) If it ain't broke, don't fix it
3) Waiting for the modules I use to be upgraded
4) Waiting for D7 itself to stabilise so I don't have to do endless patches and updates to Drupal Core (I'm going to have to start doing the rounds with D6.23 again already....
Comment #10
cpliakas commentedPatch needs to be re-rolled using the module's document root, not Drupal's.
Comment #11
cpliakas commentedRe-rolled patch using standards at http://drupal.org/node/323.
Comment #12
cpliakas commentedPatch committed in #391818.