This is one of those books that are perhaps nice to have IN ADDITION to something better. Btw, do NOT overestimate the "for smarties" part in the title: the book is not all that advanced: it's more like an extensive cookbook with a lot of personal opinion thrown in (not always consistent; for example, in one place he inveighs against the evils of using sequential-number sequences as primary keys -- 'cause a table is not a sequence, you see, we're talking about sets here, who, by definition, are unordered -- OK, fine. Ten pages later he blasts the GUID type -- why? Because it's not inherently sequential and it's hard to spot the gaps in the sequences. But hey, why do we care about gaps? All we care about is that the field values be unique, which they are, gaps or no gaps. Seems like GUIDs should be perfect from the set-theoretical point of view, but no, he doesn't like them -- precisely because of the presence of those aspects, the lack of which he bemoaned one chapter back in the IDENTITY type. It's like he wrote these two chapters one ten years after the other, and forgot what he was talking about in one when writing the other.
The content (or rather the intent behind it) is very good. There's a logical progression from the overall-schema things, to tables, and so on, including such esoterica as hierarchies and graphs (which is good not only, or even not so much because of the topics themselves, but because nice recent SQL features like CTEs are used a lot in the sample code thus demonstrating their non-trivial use).
The downsides: the main flaw in Celko's writing is that whatever he writes reads like a two-page journal article, by which I mean it's all a perfunctorily dashed-off collection of tidbits; the overall structure is very tenuous (for example, he starts the hierarchies' chapter by saying they're a sort of graphs -- but the graphs proper chapter comes afterwards: wouldn't it make more sense to switch their order in the book then?) Some chapters are borrowed from other writers: for example, the chapter on temporal databases is taken, or rather squeezed out of Snodgrass's book (which I happened to be reading in parallel and thus was able to notice that). I'm not hinting at plagiarism here: I'm sure Snodgrass was aware of this borrowing and had OK'ed it, but first it would be nice to mention the fact of borrowing (I think), and second, and most important, when you compress a book into a chapter, you gotta do it very carefully so as to keep the material connected, coherent, and clear. This is not the case here (go for the original: it's good, and can be downloaded for free; google on the name).
In general, Celko's writing, while not abhorrent, is mostly (though not everywhere) very sloppy; everything reads like a first draft never touched again by either the author or an editor. A fair amount of typos, and a lot of unclear, careless pages that make you struggle for meaning (not always successfully). Mangled French again: Joseph, if there's an accent over the last 'e' in a masculine form of a participle, it's gotta be an accent aigu, not accent grave (feminine simply adds an extra 'e' w/o changing anything else); thus it's 'née' not 'nèe'. Maybe it's a typo, but it's consistently repeated thoughout the book (and actually present in his other books!). While we're here: "Borland (née Inprise)" is actually the other way around: Inprise (née Borland).
There's no question that the author is a smart guy, but writing isn't his forte and he should be less casual about it -- and then, he should also insist that his publisher provide good editorial oversight. I think MK is a good publisher, and I think they ought to be able to do a better job helping their writers achieve readability.
Bottom line: I don't regret having this book; it's friendly and chatty (in a good sense); the inherently dry material is livened up a bit by a sprinkling of curious trivia; it's been somewhat enlightening on the first read, and repeatedly useful as a reference afterwards. Otoh, it's written sloppily and I feel that, inasmuch as I benefitted from it, I did only because I happen to have enough foundation to compensate for its flaws on my own. But I'm not looking for this kind of effort when reading technical books though, so three stars. Could be more, but for that the book needs to be aggressively edited and restructured. It's the third edition, btw: it would seem that there's been plenty of time to do just that.
All in all, I recommend it, but only half-heartedly: as long as the prospective reader understands that this is not a terribly advanced book, as well that the reading won't be easy. The book's OK, but not on par with what you get from writers like Gray or Date.
