Oh dear :unsure:. This makes me think maybe Paul was right when he said my post was unclear.
webrunner (4/12/2010)
IAfter all, I agree with Tom that it is counterintuitive to provide a function that ostensibly allows you to provide replacement text of your choice, when in reality that text may be truncated to the exact type AND length of the type specificed.
Actually I don't think it's counterintuitive - the type of the result is the same as the type of the check expression, and in the case of varchar, varbinary, and nvarchar the type includes specification of the maximum permitted length. It's the behaviour of LEFT that surprised me, because I expected it to deliver a result with the same type as its first argument, just as IsNull does.
2. Going back to the ostensible logic of the ISNULL behavior, I can certainly understand why Tom would be, if not flabbergasted, then baffled (as Paul put it) by this behavior, prior to learning more about it.
The IsNull behaviour is nice and clean and tidy, Paul and I have no disagreement on that. In fact what we mostly disagreed about was words (words badly chosen by each of us, I think) not anything in T-SQL.
On a different topic: my first QoTD is due to appear in 8 to 10 days time; I've tried to choose something that I can't get wrong, so that I don't suffer the bashing that many QoTD setters get, but I'm not at all sure I've succeded.
Tom