MidCOM search engine in production

We’ve today moved the MidCOM search tool into production together with the development release 2.3. Some statistics from Torben:

M-P.org needs 242 MB (Apache+PHP+data) to do a complete reindex, which means that PHP used around 220 MB of memory for this. It didn’t bother the indexer much, it stayed at around 20 MB memory load.

During the complete reindex, which ran around 10-15 minutes we indexed 1,583 articles and perhaps one or two hundred attachments.

Not bright, but could have been worse, especially if a reindex should be neccessary very rarely.

Once the initial reindexing has been run all new changes in the Midgard database will be automatically indexed. This means that the search tool will always have fresh results, even if new documents have been added, or old ones have been changed or deleted.

Other advantages from this new search system include the fact that the search results are component-aware, enabling us to display thumbnails for image results etc. The searches are also aware of the Midgard permission system, ensuring that users will get results from all documents they’re allowed to see. The possibilities of the Lucene search syntax are also promising.

Searching Midgard site for Seoul

Try out the new tool by searching the Midgard site. It can also be deployed by installing the indexer and setting MidCOM template to use the xmltcp indexer plugin. I’ve already done this for my site and it seems to work really well.

Great work, Torben!

In the other news, I was interviewed today about flight training for the Finnish youth TV program Buusteri. The show should air on week 14.

Read more Midgard posts.