I saw Alonso Robles of Headspring speak on RavenDB tonight at the Austin .NET User's group.
His talk (a good talk) followed a video he provided at http://info.headspring.com/ravendb-quickstart-video?utm_campaign=RavenDB-Video-Email&utm_source=Email which I was supposed to watch before attending. In writing this, I've still only watched about six minutes of it. In over my head a little bit, I gleamed that Raven uses Lucene for caching and that one when one writes to Raven and then references a Lucene-cached index, that the index may in fact be stale and not contain the update just made if the update just made would fall within the bounds of the select to populate the index (think Views in MSSQL).
The staleness should only linger for maybe two seconds, but it is something you will need to deal with. Naturally, you can refuse to deal with it, but Alonso argued that things work best if you embrace the Lucene caching. Write to Raven. Read from Lucene.
- Everything in Raven is JSON and thus everything is a string. If all of your datapoints for a search are floats, then you have to specify that you are searching as if the strings were floats.
- Searching by keywords? Use the StandardAnalyzer to crawl copy and break the copy into digestible words for matching.
- There is a GUI editor/record-browser for Raven that is written in Silverlight. (think MSSQL Management Studio Express)
- Cool stuff. I will take the time to finish the video clip at http://info.headspring.com/ravendb-quickstart-video?utm_campaign=RavenDB-Video-Email&utm_source=Email
No comments:
Post a Comment