Book Review of LAN TIMES GUIDE TO SQL

Ocelot's now-defunct newsletter, THE CAT'S MEOW, was mostly about Ocelot, but also had lots of expert information about the broader SQL world. The following article is an excerpt from THE CAT'S MEOW's occasional book review section. (Copyright© 1998 by Ocelot Computer Services Inc.)

... Extract begins. Title: "Peter's Column: Ask the Swami" ...

James Groff and Paul Weinberg
LAN Times Guide To SQL
Osborne-McGraw-Hill, 1994.

Experts should never review books -- they'll just conclude "I learned nothing from this book so it's no good". Amateurs should review books -- they'll better represent the attitude and preconception of the typical reader. The exception that proves this rule is: if the book is about SQL, an expert should do the review too, because all SQL books are lies and only experts can tell. There, you see -- I have become cynical since reading SQL For Dummies. And yet I carry on, perhaps out of a need to convince myself that not all is lost, that the light of SQL truth is not confined in one tiny candle (C. J. Date and H. Darwen's A Guide to the SQL Standard). With this hope, I picked up LAN Times Guide To SQL -- which I will hereinafter refer to as "LANTIMES".

Let me first dispose of the blurb-writer drivel that comes with the preface and on the back cover ("covers SQL2 ... the complete guide ... describes the SQL database language specified by the ANSI/ISO SQL standard ... comprehensive in-depth ...") and so on.

The book's coverage is of about 10% of SQL2, which is more often called SQL-92 these days. The coverage of domains consists of three sentences ("you can create a domain" is about it). I'm sure that readers would need to know more than that NATIONAL CHAR means "fixed-length national character string", but that's all they get (there's no explanation of character set objects). There is no example of a date or time literal using the standard syntax (e.g. DATE '1994-01-01'). The dizzying complexities of table reference (about 15 pages in Date+Darwen's book) are slashed because "table <name>", "except", "union join", "(join table)" -- indeed everything except SELECT ... FROM TABLE -- are ignored.

I forgive omissions, though. The publisher doubtless insisted on words like "complete" or "comprehensive" while Groff and Weinberg winced and bleated in the background. Actual bloopers are rare, and are usually just a reference to "the standard" when they refer to an outdated SQL-86 item (for example we read that SQL/DS allows column names within double quotes to contain special characters, "in violation of the ANSI/ISO standard") (for another example we read that to use calculated columns in ORDER BY we "must specify a column number instead of a column name") (for another example we read that SELECTs within UNIONs cannot contain expressions) (for another example we read that a subquery can only be used in a search condition) (for another example we read that the creator of a table is automatically the owner).

One difference between LANTIMES and Date+Darwen is in the rigour of terminology. Date would add here: "the reader is severely warned (Achtung!) that the word "rigour" here denotes formalist obsession to be precise about meanings". LANTIMES is much, much looser than that. Groff and Weinberg have no qualms about saying that a SQL statement fails if an expression is "not TRUE", or about saying that a LIKE expression can return NULL. In both these cases the authors should have used the word "unknown". Imprecise language can lead to misconceptions, and the DBMS does not forgive tiny bits of sloppiness.

LANTIMES does beat Date+Darwen when it comes to diagrams (I have to allow that this must be a plus for many people, though I never read diagrams myself). Plus, Groff and Weinberg have done a fair amount of survey so they can tell you whether, say, Sybase has a quirk re: datetime data types (I have to allow that this must be a plus for many people, though I can't figure out why non-standard dialects still exist myself). The LANTIMES discussion of programmed (CLI and embedded) SQL is about as good as Date+Darwen's, and for coverage of the INFORMATION_SCHEMA they're both even (i.e. both bad).

So, I give this book a double rating. As a read-on-the-airplane executive's guide, this book rates an A. However, as a description of standard SQL it does not match its claims and no programmer could possibly use it for a reference guide, so it's a C in that respect.

... end of book-review extract from THE CAT'S MEOW ...