I tried to grok Kettle, but at some point it seemed easier to just program a solution and this has worked well for what we're doing. I think if you're doing so

I tried to grok Kettle, but at some point it seemed easier to just program a solution and this has worked well for what we're doing. I think if you're doing something simple that you're going to have to repeat a lot, and possibly train others to do, then rolling your own makes sense. If you're dealing with a lot of variance in terms of the format, volume, and schema of the data you're transforming then a generalized ETL tool might make more sense. We're mostly transforming CSV, to facilitate migrations to our product (Qubit Toolkit: an open source information management used predominantly by archivists and written in PHP), so I've been developing a CSV-specific utility class for our needs with the idea of possibly making it into a generalized CSV transformation tool (where non-programmers can use something like XML or YAML to define transformations): https://github.com/mcantelon/qubit-csv-transform https://github.com/mcantelon/qubit-csv-transform