Forgot your password?
typodupeerror
Image

SQL in a Nutshell 86

Posted by samzenpus
from the read-all-about-it dept.
stoolpigeon writes "The cover of SQL in a Nutshell sports a chameleon, the little lizard well known for its ability to blend in just about anywhere. This is a great choice for the Structured Query Language. SQL has been around since the seventies, helping developers interact with the ubiquitous relational database management system. Thirty some years later, SQL grinds away in the background of just about any interactive web site and nameless other technologies. New alternatives are popping up constantly but I'm going to go out on a limb and say that SQL is going to be around for a long time. Anyone interacting with an RDBMS is in all likelihood going to need to use SQL at some point. For those who do, who also want a handy desktop reference, SQL in a Nutshell has been there for the last 9 years. The SQL language itself has not stood still over those years, and neither have the products that use SQL, and so now the book is available in a third edition." Read on for the rest of JR's review.
SQL in a Nutshell, 3rd Edition
author Kevin Kline, Daniel Kline, Brand Hunt
pages 590
publisher O'Reilly Media Inc.
rating 10/10
reviewer JR Peck
ISBN 978-0-596-51884-4
summary Covers the entire ANSI SQL2003 standard as well as how that standard is implemented
It's pretty easy to sum up what SQL in a Nutshell contains. It covers the entire ANSI SQL2003 standard as well as how that standard is implemented in MySQL 5.1, Oracle Database 11g, PostgreSQL 8.2.1 and Microsoft SQL Server 2008. There is a new ANSI standard more recent than the 2003 standard, ANSI SQL2006. This new standard does not change anything covered in the book, but introduced XML and XQuery which are not covered here. The format for conveying all this information mirrors that of the other "...in a Nutshell" books. There are four sections. The first is a very short (15 pages) history of SQL and the second is a summary of foundational concepts. The vast majority of the book is the third section, "SQL Statement Commands." These commands are given in alphabetical order. There is also a table at the very beginning of the chapter listing every command and showing how it is supported by the four platforms.

Each command is presented by starting with a short summary of what it does. This is followed by a table showing which RDBMS products support the command, the proper syntax for the command, key words, command rules, possible issues that may come up and implementation details and examples for each of the four RDBMS products represented. A couple of the differences between the second and third edition are that two RDBMS products were dropped and there are more examples. The products dropped allowed for there to be more examples while also making the book smaller than earlier editions. Anyone working with Sybase Adaptive Server or DB2 UDB will want to hold onto their second edition copy of this book if they want to have that platform specific content available, because it is not in this third edition.

The book states that the dropped platforms were the least popular of those in earlier editions. For those wondering why their favorite RDBMS is not in the list, that gives the answer. To keep length down the number of specific platforms covered was kept to four. Fortunately the books is still of high value for most readers as most decent RDBMS products will support ANSI SQL standards. On those occasions they do not, the reader would have to look to another resource for help. The length issue is easy to understand when looking at the GRANT statement and seeing that it covers over twenty pages. Most of this space is used to explain the various options available on each platform.

The last section SQL Functions documents all of the standard functions with examples and then contains a list of platform specific extensions, grouped by product. There is not a table showing platform support like there was for SQL statements. This section is much smaller, so it really isn't an issue. The single appendix that follows list standard and platform specific key words.

So who would benefit from SQL in a Nutshell? The most obvious to me is the DBA or developer working across more than one of the four platforms presented, especially if they don't move from one to the other too often. Like an Oracle DBA that needs to go do something in MS SQL Server every so often, or the same type of thing between any of the others. This makes for a quick resource that will sort out forgetting how one or the other does things rather quickly. But even if one isn't moving across multiple platforms, unless the whole standard has been memorized, this is a great help.

The second group I see gaining some real good from this book are those new to working with SQL. I've worked with all four platforms and others not covered in this book and on every single one of them I've hit error messages that were anything but helpful. Being able to go directly to a correct statement of syntax and usage is a real help when the system doesn't want to tell what is really going on. It is important to remember that this is a pure reference book. It is not written with the intent of teaching how to use SQL. That said, it covers the entire standard. Much like a dictionary can be used to increase one's knowledge of a language, reading through this reference can be a good way to learn more about SQL. Many introductory texts aren't going to cover the whole standard or as many platform specific details. The student of SQL would get a real jump by working through this book. It is compact enough that while it wouldn't be a thrilling read, it is completely doable.

Who wont like it? Probably anyone who doesn't like any of the other nutshell books from O'Reilly. This book is pretty much exactly like my Unix in a Nutshell, Linux in a Nutshell and MySQL in a Nutshell books. If the format and approach bothers you, don't look for any radical departure that will make it more palatable here. If you are like me and already know you like the format, then this is pretty much a sure thing. For the vast majority of us that work in the database world, this is the reference. I say this keeping in mind the scope of the book. Is this everything one needs to know about SQL? Obviously not. There is much more to be said about SQL as evidenced by all the words that have been said and are out there in print. But when one wants to know quickly about SQL statements and functions, I can't think of a better resource.

Slashdot welcomes readers' book reviews -- to see your own review here, read the book review guidelines, then visit the submission page.

*

This discussion has been archived. No new comments can be posted.

SQL in a Nutshell

Comments Filter:
  • by Anonymous Coward on Monday June 01, 2009 @03:06PM (#28171643)

    SQL> select COUNT(*) from 'posts';
    1

  • by CRCulver (715279) <crculver@christopherculver.com> on Monday June 01, 2009 @03:06PM (#28171655) Homepage
    While I got quite some milage out of my copy of Python in a Nutshell [amazon.com] back in the day, online documentation has much improved and I feel just as comfortable hitting a few keys to get the reference material I want as flipping through pages. O'Reilly Nutshell guides seem to me consigned to that most infamous category of tech reading printed material: the bathroom book.
    • Re: (Score:3, Informative)

      by greg1104 (461138)

      Typically the on-line reference manuals you'll find show how SQL works on one particular platform. The main value of the "SQL In a Nutshell" books has always been the way you can easily compare what's available on multiple database platforms. You will need to be aware of that sort of thing if you want to support more than one database in your code. And even if you're using some middleware to abstract that away, knowing which features work well and badly on various platforms can guide how you should imple

  • by Anonymous Coward on Monday June 01, 2009 @03:07PM (#28171669)

    the SQL to this book.

  • by Anonymous Coward on Monday June 01, 2009 @03:07PM (#28171675)

    That's not SQL in a nutshell... This is SQL in a nutshell:

    SQL: HELP! I'M TRAPPED IN A NUTSHELL
    SQL: WHAT KIND OF SHELL HAS A NUT LIKE THIS?!?

    (my 100 billion apologies, Mr. powers.)

  • If only... (Score:3, Informative)

    by Monkeedude1212 (1560403) on Monday June 01, 2009 @03:12PM (#28171765) Journal
    Oracle Documention wasn't as dry as a desert. I like O'Reilly because he's not afraid to converse like a regular human being in his books. I feel like I'm being taught something rather then being shown how to do it. Would I go out and by this book? If I used SQL - definately.
    • by ianezz (31449)

      If only Oracle documentation wasn't dry as desert

      Probably it's part of their strategy to sell you courses, in case you didn't notice.

  • by tcopeland (32225) <<tom> <at> <thomasleecopeland.com>> on Monday June 01, 2009 @03:12PM (#28171767) Homepage

    ...try Stephane Faroult's The Art of SQL [amazon.com]. I've read both that and his "Refactoring SQL Applications"; I think I got a little more out of the former.

    But anyhow, in both books he has a distinct and lively writing style and includes lots of anecdotes. His style kind of reminds me of Betrand Meyer... for those who have read Meyer's 1000 page tome "Object Oriented Software Construction".

  • Nutshell books (Score:5, Interesting)

    by gambit3 (463693) on Monday June 01, 2009 @03:17PM (#28171879) Homepage Journal

    while the "Nutshell" books make great gifts for someone somewhat experienced in a topic, it needs to be pointed out that they're not necessarily the best option for a beginner.

    • Re:Nutshell books (Score:4, Insightful)

      by gubers33 (1302099) on Monday June 01, 2009 @03:28PM (#28172077)
      You hit that one on the head. I purchased PHP in a nutshell a few months ago, if I didn't have previous experience in PHP from an HTML and a database class I took in college, I would not have been able to understand the book very well at all.
  • As has been noted, it's been around for quite some time now. There was enough uncovered material remaining to justify another book?
  • I've always thought the books I see in people's offices are just trophies they're showing off stating "I know this language" or "I know that topic". I've actually never seen a co-worker use one of the books either. I agree with some of the other posters, that it's just easier to search Google for something.
    • by Bigbutt (65939) on Monday June 01, 2009 @04:15PM (#28172823) Homepage Journal

      One of the problems with google is that if it's sufficiently generic, you'll get 3,600,000 pages. I can just grab the appropriate book and have a better chance of finding my answer. Another issue is the number of spammers trying to catch your attention by snatching search results just to be able to point you to their site.

      O'Reilly and Addison-Wesley have good reputations for putting out quality books. Searching and wading through blogs and difficult to navigate web sites (I'm looking at you Sun) to find the right answer can be lengthy and a crap shoot too. Picking up the subject matter book gives a good chance of having the right answer immediately.

      So yea, I have a bunch of books here and at home. I also subscribe to O'Reilly's Safari Bookshelf. I haven't read each and every one, but they've all been helpful at one point or another in my career.

      [John]

      • Re: (Score:3, Interesting)

        Not only that, but books are good for things that you wouldn't think to Google. I noticed recently that a coworker did this (slightly simplified to illustrate the point):

        public void doSomethingOrOther( int[] a )
        {
        int b[] = new int[ a.length ];
        for ( int i = 0; i < a.length; i++ )
        {
        b[i] = a[i];
        }
        ...
        }

        I pointed him to "System.arraycopy" which is quite a bit faster and does the same thing on a single line of code. This just isn't something you'd find out about unless (a) you read a book o

        • by kv9 (697238)

          I pointed him to "System.arraycopy" which is quite a bit faster and does the same thing on a single line of code. This just isn't something you'd find out about unless (a) you read a book or (b) had a coworker who read a book. (Well, I guess if you read all the online javadoc line-by-line, you'd discover this, too...)

          here's a hint for your fellow janitor: before doing something stupid, google it [google.com]. what's he gonna do next? copy a string byte by byte? shit, even Java programmers should know these basic things.

      • I second what you said about google; you get too many hits, which is as bad if not worse than getting none. As an aside, I've also noticed that recently the hits it returns often don't contain what I searched for.

        I still have Unix in a nutshell, it saved my life when I made the move from a mainframe environment.

    • Re: (Score:3, Interesting)

      by Rary (566291)

      I have a large library of technical books, but I also use Google as my primary reference source. The reasons I continue to buy books are:

      1. I read them on the bus to and from work.
      2. Occasionally I can find something quicker in a book that I'm familiar with which is focused on a single topic, whereas Google is generic and often full of useless crap.
      3. I get my books paid for by the company I work for, and they make nice trophies. ;)
    • I've always thought the books I see in people's offices are just trophies they're showing off stating "I know this language" or "I know that topic". I've actually never seen a co-worker use one of the books either. I agree with some of the other posters, that it's just easier to search Google for something.

      I keep books around because sometimes you don't have access to an internet connection and you want to keep working without worrying about being able to look stuff up when you want.

    • by turing_m (1030530)

      I've actually never seen a co-worker use one of the books either.

      I find a book can be helpful when I'm getting a general overview of a complex topic. You can lay in bed (or on the floor) and read, which is harder to do with google. I found it difficult to grok how to design database tables for example, without sitting down, reading and thinking. And falling asleep, and getting up and doing it all over. It's a lot harder to learn databases than say, learn enough Perl to do something useful. Hence why the on

  • by Brandybuck (704397) on Monday June 01, 2009 @03:48PM (#28172375) Homepage Journal

    There are a gazillion programming languages, with new ones added every day. C, C++, Java, C#, Objective C, Pascal, Modula 3, Ada, Ocaml, Haskell, Lisp, Scheme, Python, Ruby, Perl, Lua, Javascript, etc. There's even a choice of shell scripts: sh, csh, bash, ksh, zsh, etc.

    But only one SQL. I'm sure there are some other database query languages, but they are so obscure that no one but the longbeards have ever heard of them. Why is that? Why are there no alternatives to SQL? Not just minor variants, but actual alternatives.

    • Re: (Score:3, Funny)

      by Jarlsberg (643324)
      It works really well, it's an open standard, and it has been with us for about as long as we've had Personal Computers. Let's all be glad Microsoft Access didn't dethrone SQL.
      • by Brandybuck (704397) on Monday June 01, 2009 @03:58PM (#28172549) Homepage Journal

        You can say that for several programming languages as well. There's got to be something else to it.

        • by Jarlsberg (643324)
          Point taken. Perhaps SQL, being backend, is less interesting to replace than the programming layer itself? That said, I've seen and tried many different database solutions, going back to Commodore 128 and Geofile, to stuff like Filemaker on Windows, but nothing quite beats SQL.
        • by Abcd1234 (188840) on Monday June 01, 2009 @04:08PM (#28172721) Homepage

          Creating a new query language is *hard*.

          I mean, I can sit down and create a new programming language fairly easily. Hell, most computing science students write a compiler at some point during education. But a new query language? That requires a DB engine, a query optimizer, and who knows what else. All to replace a language that, thus far, has worked exceedingly well.

          That said, as another poster points out, there are other languages out there, XPath being the most notable (CSS selectors also come to mind). But none of them are as clear, simple, and straightforward as good ol' SQL, which, I think, says something about its design.

          • Re: (Score:3, Insightful)

            I think you very much overstate the "clarity," "simplicity" and "straighforwardness" of SQL; there are many well-studied ways in which a relational language could be better than SQL. But I do think you've hit the nail on the head otherwise: designing and implementing a credible RDBMS is extremely hard compared to designing and implementing a programming language, and no relational query language is going to go anywhere without being paired to a good RDBMS.

          • by DragonWriter (970822) on Monday June 01, 2009 @05:34PM (#28174149)

            Creating a new query language is *hard*.

            No more so than creating any other new programming language; most functional programming languages (and plenty of not-particularly-functional programming languages that have functionally-inspired features) have querying constructs roughly comparable to those in SQL or other dedicated query languages (and often clearer, more straightforward, and more expressive than SQL.)

            What's hard to build is efficient storage engines and query optimizers, not query languages, but once those are built, the language you express the queries in shouldn't matter much as long as the what is expressible is the same.

            The really hard part though is finding a market for a new dedicated query language; a big selling point of a dedicated query language is because it will be generally familiar for most users regardless of what prior products from a large set the user has used, and anything that isn't SQL, whatever it might have going for it, is going to lack that advantage as a query language.

            • by Abcd1234 (188840)

              What's hard to build is efficient storage engines and query optimizers, not query languages, but once those are built, the language you express the queries in shouldn't matter much as long as the what is expressible is the same.

              I disagree. Existing query optimization and execution engines are built with the capabilities and constructs of SQL in mind. A new query language, on the other hand, presumably exists to enable things which existing languages (like SQL) don't... otherwise, why would you bother? So

              • Re: (Score:2, Insightful)

                I disagree. Existing query optimization and execution engines are built with the capabilities and constructs of SQL in mind. A new query language, on the other hand, presumably exists to enable things which existing languages (like SQL) don't... otherwise, why would you bother? So, given that, it's very likely that the language would require new, novel technologies for optimization and implementation... and those things ain't easy to get right.

                I think DragonWriter is going from the core assumption (which I

                • by Tablizer (95088)

                  Closer adherence to the relational model (no NULLs or three-valued logic, no row duplication allowed, etc.).

                  In practice, row duplication and well-placed nulls can simplify a lot of things. I believe the purists go overboard sometimes.
                         

                  • by The_Noid (28819)

                    I can imagine the null bit, but duplicates? Duplicates make a mess of things and complicate the language more than they solve.

                    • by Tablizer (95088)

                      An example would be sending large volumes of data to another company. Suppose the data set has a compound key(s), but the customer doesn't want the keys to keep it slim. We want to do a SELECT that does not include the key(s). This may result in duplicate rows.

                    • by The_Noid (28819)

                      In that case those duplicate rows are useless anyway and should be removed directly. That will also save on the amount of data that needs to be transferred.

                  • It depends on what one means by "eliminating nulls"; there's a reason I mentioned NULL and 3-valued logic in the same breath. I certainly would welcome a 2-valued system where NULL was an optionally allowed value for every column (treated as a true value, and not the funny "no value" semantics).

                    Or even better, we could just have a simplified version of something like Haskell's labeled union types. This would take care of NULL and enumerated types in one feature.

              • I disagree. Existing query optimization and execution engines are built with the capabilities and constructs of SQL in mind. A new query language, on the other hand, presumably exists to enable things which existing languages (like SQL) don't... otherwise, why would you bother?

                The same reason people bother with new Turing-complete programming languages, which can't do anything that isn't possible in existing Turing-complete programming languages: expressiveness, clarity, etc. Usually, new languages don't e

          • Re: (Score:3, Insightful)

            by lakeland (218447)

            Nah, it's easier than that.

            Firstly, writing a DB engine is pretty easy. B-Tree implementation is a standard 2nd year assignment, hash is first year, and trying them into a database with persistent storage would probably still be 2nd year or 3rd year at worst. Transations (undo segments and the like) are harder, but you could leave them out.

            Query optimisation firstly isn't in SQL, though I'd be lying if I said it was in relational algebra. Certainly it is in an intermediate language that is convertible to

        • by Rary (566291)

          Programming languages do many, many, many different things in many, many, many different environments. Database query languages do, more or less, one thing in one environment. If you've built a good database query language, there's no need to build another one. However, if you've built a good programming language, somebody else will find a situation in which it's not so good, and will therefore want/need a different one.

    • by rubycodez (864176) on Monday June 01, 2009 @03:57PM (#28172529)

      both XQuery and XPath 2.0 are used in the "real world". Unidata Query Language is used in many enterprises, I worked at a manufacturing plant where the MRP system used Unidata as back end.

      so quit yer bellyachin'

      • Re: (Score:2, Insightful)

        by AkiraRoberts (1097025)
        Having spent a painful few months with Unidata Query Language, I have to say I much prefer SQL. That experience was somewhat less than fun. As for SQL, it's been quite a while since I picked it up, but I recall that it took me all of an hour to get comfortable enough to do damage. And now, some 10 years later, I haven't even come close to exhausting its possibilities. Perhaps that's the reason for its popularity - a nice balance of ease and extensibility.

        When it comes to books though, I have a fondness fo
    • Re: (Score:3, Informative)

      by abigor (540274)

      It's mainly because SQL was the first (only? someone correct me) language to implement Codd's relational model, via the tuple calculus. The relational model is of course the basis for relational databases, so the idea was SQL would be provably correct in its representation of the relational model. There is a document called The Third Manifesto that details why this is not the case, and makes some suggestions for the way forward, but I don't remember much else about it.

      • by hey hey hey (659173) on Monday June 01, 2009 @06:28PM (#28174913)
        It's mainly because SQL was the first (only? someone correct me) language to implement Codd's relational model, via the tuple calculus.

        Hardly. Quel [wikipedia.org] predates SQL, and was superior in almost every way. However, SQL had IBM behind it, and Quel just had UC Berkeley (guess who won that battle).

        • by abigor (540274)

          Yes, I knew someone would correct me, thanks ;)

          Anyway, my point was that any new language would have to have a similar relationship to the relational model that SQL and I assume Quel do.

    • Re: (Score:3, Interesting)

      by TCM (130219)

      There's even a choice of shell scripts: sh, csh, bash, ksh, zsh, etc.

      Objection!

      You use different shells for different interactive properties. Scripts should be written for #!/bin/sh. And yes, that's /bin/sh to you, no /bin/sh == /bin/bash perversion.

    • by nurb432 (527695)

      Because it works, well.

    • by Tablizer (95088)

      There are a gazillion programming languages...Why are there no [common] alternatives to SQL?

      Indeed! Over at the c2.com wiki we've been wondering that also. We've listed several limitations or annoyances of SQL that are not inherent to relational theory itself.

      I've even created my own draft relational query language called SMEQL ("smeegol") roughly based on an old IBM experimental language called "Business System 12" that has a functional programming feel and is more re-composable than SQL.

      Another option is

    • Re: (Score:3, Informative)

      by Hognoxious (631665)

      But only one SQL.

      I know for a fact that isn't true. I work with SAP, which uses a slightly odd dialect/subset of it.

  • New alternatives are popping up constantly but I'm going to go out on a limb and say that SQL is going to be around for a long time.

    That's pretty much guaranteed. COBOL is still around.

    There absolutely are better alternatives, for almost every situation. No one in their right mind starts a new system in COBOL [coboloncogs.org], when they have a choice.

    Yet COBOL is still around, and will be still around for awhile. So will SQL.

    The only question is whether SQL will be like COBOL or like C. I could make a similar case for C being obsolete, and there are certainly many cases where a performance penalty is well worth it to get some other desired feature -- fo

  • SQL is just a TLA (Score:1, Interesting)

    by Anonymous Coward

    Actually according to both the ANSI and ISO stds, SQL has no translation from the acronym. Now before flames start, please read the standards. A friend of mine has been an active member of the ANSI committee since the early '90's, and I have presented proposals to the committee.

  • SELECT * FROM table WHERE field LIKE condition;

    INSERT INTO table (field1, field2, field3) VALUES(value1, value2, value3);

    That's SQL in a nutshell. I think anything more than that is a fairly detailed book.

    • by Samah (729132)
      You're not going to accomplish much with a select and insert. As long as you know all the CRUD syntaxes (Create, Read, Update, Delete) you can use an existing database, and although I'd hardly call it "in a nutshell", it's the first thing anyone dabbling with SQL should know before they start reading a few chapters into "a detailed book" and get to creating and modifying schemas.
  • by FranTaylor (164577) on Monday June 01, 2009 @06:04PM (#28174575)

    If it were, there would not need to be vendor-specific examples in every SQL book.

    Why can't people just implement standard ANSI SQL and be done with it?

    I am really tired of vendors (MySQL) and their non-standard SQL. I want my JDBC applications to just work and not have special-case code for each database.

    • by Tablizer (95088)

      Why can't people just implement standard ANSI SQL and be done with it?

      Partly because the standard is too complex, and partly because non-standard extensions reduce the chance of people switching vendors. DBMS vendors don't want people to migrate to other DBs. Thus, they implement proprietary features and sometimes nifty shortcuts that get you hooked because converting would be too difficult and/or you'd lose the nifty shortcuts.

  • No DB2? (Score:3, Interesting)

    by metamatic (202216) on Monday June 01, 2009 @06:31PM (#28174939) Homepage Journal

    The book states that the dropped platforms were the least popular of those in earlier editions.

    Last I checked, IBM DB2 had the biggest market share of any SQL database. (Link to 2003 Gartner Study [programminglearn.com], and I don't think the situation has changed much.)

    So do DB2 users just not buy books like SQL In A Nutshell? Or have O'Reilly made a serious mistake here?

    From my point of view it looks like a mistake, as I'm only interested in PostgreSQL and DB2... but then again, I work for IBM, so maybe I'm a special case?

    [Opinions mine, not IBM's.]

    • by greg1104 (461138)

      The last Gartner Study [reuters.com] I read analyzing 2007 put DB2 as basically tied for 2nd place with Microsoft at around 20% each, and both combined are behind Oracle. But market share is not the same as user base, and there I'd bet DB2 is far behind MySQL at least. You might have a case for putting DB2 ahead of PostgreSQL as far as user base goes, but O'Reilly does have heavy open-source roots in their publishing line that I suspect swayed their focus against IBM's product here.

Never test for an error condition you don't know how to handle. -- Steinbach

Working...