Back in April I wrote up a notion I had on the usefulness of histograms. The main point was that histogram usefulness is not confined to skew on existing values — if there are values between the lowest and highest for a column that might exist but do not, then that is a form of skew that we can usefully document for the optimizer in the form of a histogram.
Did that makes sense? Oh, just read the article and the comments. And here is the one from before that was a little faulty.
While I hold my head high with the claim of “original thought” (on account of it being my own idea — the kids are really proud of me btw), maybe it’s one of those ideas that lots of people get at the same time. Like lightbulbs. Is it called Morphic Resonance?
Well anyway, here we go. I guess it might be spreading. If Arup, Jeff or Mike have anything to add that I missed I’d be happy to hear from them. I love a consensus.