[Building Sakai] sakai-dev Digest, Vol 39, Issue 24

[Ian Boston]

Colin,
Nice work.
Have you had a look at ElasticSearch ?
It has most the the features that Solr has but it is much easier to
make reliable in a cluster where the application is making constant
updates.

It has a json based interface, is dead easy to install and run and is
fully elastic. Add more ElasticSearch instances on the same subnet
auto discover others and forms one cluster with automated routing.
Even understrands EC2 interconnect out the box.

Quite a lot of sites are starting to use it.

HTH
Ian

>
> Message: 11
> Date: Thu, 24 May 2012 16:25:31 +0100
> From: Colin Hebert <colin.hebert at oucs.ox.ac.uk>
> Subject: [Building Sakai] Sakai search - A solr implementation
> To: sakai-dev at collab.sakaiproject.org
> Message-ID:
>        <CAM09BhtUmdfse3H2GjinqKY1rxPvMM5ZoreTpFqAGWE5dxq3bA at mail.gmail.com>
> Content-Type: text/plain; charset="utf-8"
>
> Hello,
>
> I've worked the past few weeks on a different implementation of the Sakai
> Search API using solr : https://github.com/ColinHebert/Sakai-Solr.
> It is still a work in progress, but it's functional; I didn't have much
> time to do a proper documentation, but it's pretty easy to set up.
>
> The solr configuration used in unit tests (
> https://github.com/ColinHebert/Sakai-Solr/tree/master/impl/src/test/resources/solr)
> is the one I currently use when I run solr with sakai and works (so far)
> fine.
>
> To enable search with solr, the sakai.properties file has to be modified to
> contain this:
> solr.server=http://localhost:8983/solr/
> search.enable=true
>
> And of course, the project has to be deployed. With all that it should run.
>
>
> I intend to add some user documentation later, but this early version has
> some nice features:
>
> - faceted search, searching for "tool:content myKeyword" will search
> "myKeyword" only in documents indexed by the content module;
> "title:presentation.pdf" will search every indexed element named
> presentation.pdf, etc.
> - the possibility to index and search metadata on documents indexed by
> "content", (dublin core metadata or any property stored in
> ResourceProperties), or simply hide it (from the solr configuration).
> - all the goodness <http://wiki.apache.org/solr/ClusteringComponent>
> of solr<http://lucene.apache.org/solr/>(it works with 3.5, 3.6 and I'm
> working on making it compatible with 1.4)
> and tika <http://tika.apache.org/> with solrCell
> - an entity provider for entity broker (for some reasons, the current
> implementation doesn't seem to work right now)
>
>
> Some classes aren't meant to stay in this project but should be either in
> their own project or could be integrated in already existing projects
> (CitationContentProducer, SiteContentProducer,
> BinaryContentHostingContentProducer) as they're not specific to my
> implementation.
>
>
> So far, it's more a developer preview than anything, but as I said, it's
> working pretty well and only needs a few adjustments.
>
> If you have some time to review the code, test it, give some feedback,
> contribute or find some bugs, I'll be glad to hear of it.
> Everything is on github, so feel free to fork the project.
>
> ---
>
> Colin