I am doing this. Let's assume that you can make a query so that each row of the result is one document that you want to index. Create an XML document in Solr update format in a StringBuilder, then POST that to the Solr instance. After about 1000 documents, post a "<commit/>". After all the rows, post a "<commit/>", then an "<optimize/>".
My documents are rather small, so I batch twenty of them in a single post. That runs faster than one document per post. The whole thing is about two pages of Java. It will get bigger when I start checking the return codes from the commits. I would share it, but it uses a custom DB layer, so it wouldn't work for anyone else. wunder == Walter Underwood Seach Guru, Netflix On 11/24/06 8:14 AM, "Nicolas St-Laurent" <[EMAIL PROTECTED]> wrote: > Thank you Bertrand. > > The documentation on Solr is still sparse. I've already looked in > SolrResources, find some idea, but not exactly what I need. When my > solution will work, I will document it into the wiki. > > Nicolas > > Le 06-11-24 à 02:48, Bertrand Delacretaz a écrit : > >> On 11/23/06, Nicolas St-Laurent <[EMAIL PROTECTED]> wrote: >> >>> ...I index huge Oracle tables with Lucene with a custom made >>> indexer/search engine. But I would prefer to use Solr instead... >> >> Instead of using Lucene's API directly, with Solr you'll have to add >> your documents to the index using HTTP POST messages. >> >> There are a few Java clients for Solr floating around on the wiki and >> in Jira IIRC, but you just need a POST, any way of doing it is fine >> (using jakarta httpclient for example). >> >> See http://wiki.apache.org/solr/SolrResources for more info. >> >> -Bertrand
