Dear Patrick, many thanks for your prompt answer and your thesis document.
I'm not totally agree with your conclusions on ES. I did some tests and I found it very well. I did not had any difficulties with pyelasticsearch and found a lot (too much?) of documentations on elasticsearch.org and several mailing list. I agree that as the project is recent the community is smaller than Solr, but it grows more and more. My main difficulties was to configure correctly the stemmer, stoping words, etc, especially for multi-lingual documents. But I will have the same problem with any search engines. Moreover I tested, facets, snippets, filters without any problem on ES. Now I will try to interface it with Invenio. It seems not too difficult to use ES for fulltext, but what about facets. I did not see anything about facets in Invenio 1.1. When the facets will be available? In which version? Moreover replacing the Invenio indexer by Solr or ES seems more tricky no? Computing facets using Sorl or ES needs either: - replacing the Invenio indexer and do all the job with Solr or ES, but in that case what about all Invenio search facilies such as regexp and so on. I imagine that this imply a lot of deep modifications in Invenio, right? - using Solr or ES for facets computing but I need to give to it all recids on which we want to compute facets, right? Is it the role of java classes in modules/miscutil/lib/solrutils/org/invenio_software/solr ? Many thanks for your advises. Regards, ---------------------------------------------------------------------- Johnny Mariéthoz RERO, Av. de la Gare 45, CH - 1920 MARTIGNY Téléphone: +41(0)27 721 8579 Fax : +41(0)27 721 8586 Web : http://www.rero.ch ReroDoc : http://doc.rero.ch, [email protected] ---------------------------------------------------------------------- Le 10 oct. 2012 à 17:43, Patrick Oliver Glauner a écrit : > Hi Johnny > > In my undergrad thesis, I worked on the integration of information retrieval > systems with a strong focus on word similarity ranking: > <https://cdsweb.cern.ch/record/1456329>. It included a *quick* ElasticSearch > evaluation which was (unfortunately) not that positive. My main problem was > the lack of documentation. Following, the configuration was difficult which > seemed to result in arbitrary results. Hopefully, this has changed meanwhile. > > Please find the current Solr and an initial Xapian integration here: > <http://invenio-software.org/repo/personal/invenio-pglauner/log/?h=master-thesis> > It should be soon merged into master. With this bridge, an ES adapter can be > easily integrated. > > Cheers, Patrick > ________________________________________ > From: Johnny Mariéthoz [[email protected]] > Sent: Wednesday, October 10, 2012 5:34 PM > To: project-invenio-devel (Invenio developers mailing-list) > Subject: Invenio 1.1 and ElasticSearch > > Hello, > > does anyone has already implemented ElasticSearch in Invenio? > > After playing with it, it seems a good alternative to Solr as all > configuration is done using the REST API and thus can be done at client side > (Invenio). Moreover it support facets, filters, load balancing and so on > (probably as Solr). > > If I want to do that, do I have to create an elasticsearchutils.py in the > same manner as solrutils.py ? > > Are facets already supported to Invenio 1.1? > > Many thanks in advance. > > -- Johnny >
