if you put much of this into your
etl-process-files - stacking together can get a little bit messy if the
etl-process is not linear (which leads to many etl-files containing
similar processing code). Considering to build a system or file format
specific driver to seperate data flow instructions (with a domain
specific language) and processing by the driver could be much cleaner
because you stick closer to the predefined scriptella ways of querying
and scripting to system- or file format connections...