Refactoring SSIS codebase

Oct 31, 2008 at 8:50 PM
My preliminary source download, build, installation and testing of the component have demonstrated its value.

Any thoughts on approaches to converting existing SSIS packages from the "out of the box" SCD wizard to this component.? Promoting the adoption of the component will involve convincing IT decision-makers that conversion projects can be accomplished as painlessly as possible.

(It's quite interesting to note your performance improvement benchmark of a factor of 100. This will certainly carry weight in any conversion/adoption discussions.)
Coordinator
Nov 1, 2008 at 10:55 PM
Thanks for the review!  (Now go rate it 5-stars and review the release :) )

My thoughts on "conversion"... It is quite likely that having the component "upgrade itself" can not happen - but it would be quite (theoretically) possible to write a conversion program to do so.  All of the configuration information for the components is stored in XML - so theoretically an XSLT transform could convert a package from the stock component(s) to this one.  I'm not an XML guy, but I think I know one... :)  At the very least, I could probably write a C# program to wok some magic.

The difficulties:
1. Such a program would have to know (reliably) the metadata semantics for each component.  I know that I'd like to reformat my metadata in order to make it more transparent and obvious, and I haven't looked at the stock one sufficiently to see if it's metadata is transparently accessible.
2. The stock SCD component isn't really one component, is it?  (One of the reasons you're here, isn't it?)  I'm not sure if there's really any valuable "information" in the metadata of the components attached to the SCD - but since you can rename/delete/relink those components, it would likely be next to impossible to retrieve any of that metadata.

But I will surely take a little look at that idea - How about you post up an Issue?  (If you don't, I will)

BTW - I would LOVE a little backup or contrary evidence to my claims of performance increases.  Since it sounds like you have a sizeable set of packages, and you are evaluating this for use in replacing the stock component, please let me know what (if any) performance improvements you see.  I believe the primary source of the performance is the "full caching" nature of the component, compared to no cache for the stock SCD.