A code smell is defined on Wikipedia as being a “symptom in the source code of a program that possibly indicates a deeper problem”. It’s a term commonly used by our code-writing brethren to describe sub-optimal code but I think the term can be applied equally well to SSIS packages too as I shall now explain
One of my pet hates about SSIS development is packages that throw warnings of the form:
The output column "ColumnName" (1358) on output "OLE DB Source Output" (1289) and component "OLE_SRC Name" (1279) is not subsequently used in the Data Flow task. Removing this unused output column can increase Data Flow task performance.
The warning is fairly self-explanatory – any column that appears in the data flow but doesn’t get used will throw this warning when the data flow is executed. Its not the negligible performance degradation that they cause that bothers me though, it’s the clutter that they cause in your log file/table. Take a look at the following screenshot if you don’t believe me:
There are 231409 such warnings in the system that I took this screenshot from, that is 231409 log records that should not be there. The most infuriating thing about this warning is that it is so easily avoidable; eliminating such columns is a very quick and easy thing to do in the SSIS Designer. The only problem I see is that the warnings don’t occur until you execute the package – it would be preferable for the designer to have an unobtrusive way of informing you of them as well. Anyway, I digress…
I consider such warnings to be a code smell because, to me, they’re symptomatic of a lack of due care and attention; a lack of developer discipline if you will. What other code smells can you think of when building SSIS packages? If I get a good list in the comments maybe I’ll compile them into a later blog post.