by Steve Bolton
…………As expected, I’ve run into quite a few obstacles while trying to deploy and debug my first plug-in algorithm. I’ve finally emerged from “DLL Hell” and have successfully deployed the first algorithm I’ll be using in this tutorial, but still have fatal errors lurking somewhere in the AlgorithmMetadataBase class which are preventing it from loading at server startup. These types of bugs are apparently quite difficult to pin down, since they occur before a debugger can be attached. As soon as I find a way through this conundrum, I’ll finish this series on plug-in algorithms with blog posts introducing the object model, discussing common “gotchas” like this in the deployment process and covering advanced functionality like feature selection and customer parameters. I want to make sure the plug-in I wrote actually works in practice before resuming this mistutorial series, so I may not be able to finish the series until sometime in August. Stay tuned though – barring unforeseen circumstances (I don’t want to jinx myself) it will get finished. I’ve climbed so far up this rickety stairway that I’m not going to let the locked attic I’ve found at the top deter me indefinitely from finishing the series. The last installments will be on writing custom data mining viewers for SQL Server Data Mining (SSDM), which should be a simpler matter.