Yes. I almost completely agree. But this brings up one of the main problem with the puzzles we are set here, and one of the main problems I have in my career (touched on by "Know Thy Data"). Without a solid amount of data to test problems on, it's really easy to make assumptions.   In a perfect world, with lots of data, I would check the cardinality of the text field to check whether it was free-form text or not. If I could confirm that it was the case (very high cardinality), I would tailor my code to fit. With a small set of data, I made an assumption based on my understanding of the spec, and I would argue that I was correct, even if I wasn't right <grin>.   But that doesn't matter. I have (mostly silently) disagreed with some of the answers that people have given in the past, but the question setter has been happy so it is of no import. I think it is less likely that a single slanted interpretation of a specification is going to be accepted if there are extensive source data from which to formulate a solution. I blanch when I'm given test data which has been put together by a business analyst which may only have a hundred or a thousand observations in it. If it's a banking system, give me all the customers, all the accounts and all their transactions. It'll take me a wee bit longer to analyse it, but I will eventually come up with a solution which is more likely to last. ... View more