David Mertz, Ph.D.
Color Commentator, Gnosis Software, Inc.
November, 2001
Since my last book roundup in Charming Python #12, a number of interesting books on Python have been published. This installments provides new comparative reviews of recent Python titles (or titles missed in the last roundup).
Python is a freely available, very-high-level, interpreted language developed by Guido van Rossum. It combines a clear syntax with powerful (but optional) object-oriented semantics. Python is available for almost every computer platform you might find yourself working on, and has strong portability between platforms.
This installment is intended to round out the book roundup I wrote for this column at the beginning of this year. Since then, some new books have been published, and some other books have made it to my desk that I had not yet had a chance to look at 11 months ago. There are still a few ommissions of books I have not quite got my hands on, but between the two installments, almost every book you might consider buying on Python today is discussed.
By the time next year rolls around, even more books will join the Python bookshelf. Surely there will be new titles I have not heard of; but some I know about look promising.
Python stalwart and brilliant Usenet essayist Alex Martelli has apparently contracted to write two separate books for O'Reilly--one titled, Python in a Nutshell (which one waits with anticipation if only to see Martelli's prolific writing inclinations squeezed into a compact size). Bruce Eckel, whose Thinking in Java has been very well received, has had not-yet-realized plans to write a Thinking in Python for a while; perhaps he will for 2002. Christopher Jones and Fred Drake have another O'Reilly title Python & XML underway, which looks promising. And, of course, I cannot let this column go by without mentioning that your humble author is in the process of writing the title Text Processing in Python for Addison-Wesley. 2002 looks to be another good year for Python.
The few currently in print Python books that I have not had the chance to look at should be mentioned briefly. No slight is intended towards these titles, but neither can I make any particular recommendation. These are:
* Core Python Programming, Wesley J. Chun, Prentice Hall (December 2000).
* Sams Teach Yourself Python in 24 Hours, Ivan Van Laningham, Sams (May 2000).
* Python: Developer's Handbook, Andre dos Santos Lessa, Sams (December 2000).
If I have missed anything altogether, I apologize to the appropriate publishers and authors. Let me know; I'll try to work something in at some future point.
While books are a wonderful way to get started with most topics (including Python), there is also very good documentation that accompanies Python itself. Take a look at: http://python.org/doc/ to get yourself started. The Python team have written a quite excellent introductory tutorial, and the quality of the the module index, library reference, language reference are much better than most programming languages, or free software in general, has (but the distinction between the latter two items is always fuzzy to me when I go to look something up... try both if in doubt).
Six months ago, the one Python book I made sure to keep within arm's reach while programming was David Beazley's Python Essential Reference. Mind you, the rest are all within a few steps of my chair, but Beazley's is the one I didn't have to stand up for. That's changed a bit in the last few months--now the book in arm's reach is the second edition of Python Essential Reference. The main change between editions is simply an update to coverage from Python 1.5.2 to Python 2.1. The organization and most of the content remain the same. Apparently in response to some user complaints, the font size was increased slightly (and the length went from 318 to 396 pages); unfortunately, at the same time the 2nd edition is printed on paper with a much lower brightness--readability (from a purely physical perspective) has actually gotten slightly worse.
Another title that was mentioned but not reviewed in the previous installment is Fredrik Lundh's Python Standard Library. At that time, only an encumbered electronic version was available, but since then O'Reilly has published a printed version. Python Standard Library is a thin volume--281 pages in physical size, but more lines of the books are taken up with source code and vertical whitespace than with prose description. The total words do not add up to a large number.
Python Standard Library has a fairly narrow target, but succeed well in fulfilling its goal. The target is experienced Python programmers who also have either the standard module documentation and/or one of the thicker books at hand. Almost all of the modules in Python 2.0's standard library are covered, but the coverage is not aimed at documenting every constant, function and class inside each module. Instead, each module is introduced with extremely clear and articulate prose descriptions, followed by well chosen typical usage examples.
Many of the other books I have looked at also use extensive source code examples; but in many of those, one has to re-read and ponder for a while to figure out exactly what an example is meant to illustrate, which part of it illustrates the topic at hand, and why it is an illustration. In contrast, the source code examples in Python Standard Library are only as long as necessary, are clear, have adequate inline documentation, and get right to the point being illustrated. Likewise, the prose introductions provide exactly the sort of clear guidance one might ask of the more experienced programmer in the cubicle next to yours (if you are lucky enough to have such a co-worker). The descriptions, however, by intention do not try to teach you basic concepts--they assume you basically know what you are doing, and just need to refine or direct how you do it.
Learn to Program Using Python: Learn to Program Using Python: A Tutorial for Hobbyists, Self-Starters, and Those Who Want to Learn the Art of Programming is probably not a book for most readers of this column. As the title promises, Alan Gauld's title is for non-programmers who want to become programmers, or at least to understand what it is programmers do. For the people the book is for, Gauld has written a delightful book.
I tried to recruit one non-programmer "informant" (as one says in anthropology) to give an opinion on Learn to Program Using Python. My recruitment was only a mixed success, but not because of any failing of the book, merely my choice of informant. Nonetheless, the testimony I got regarding the first few chapters was very positive--the basic concepts of flow control, branching, variables, and the like seemed pitched right for my semi-voluntary beginner.
Gauld walks through a reasonable range of topics, and keeps them in a good pedagogical sequence. The reader of this book should read the book from one cover to the other. There is a good index, but jumping around is not what this type of book is for. The source code examples do an excellent job of progressively presenting new concepts while remaning clear and to the point. Even though many books' use of sidebars and boxes is just busy and distracting, Gauld strikes a good balance with his "Things to Ponder" and "Points to Remember" that appear at the end of each chapter. Overall, this looks like a just plain fun book to read.
There is a certain style of programming book that I have trouble seeing the point of. Sample source code is important to learning a programming language or technique; but a certain type of book consists primarily of source, with the prose existing mostly to explain the source code in a line-by-line way. It is not merely the quantity of source code that I find ineffective. Python Standard Library has a high proportion of source code--but the source code exists as a continuation and elaboration of the prose; the source code explains rather than merely being explained. In another style of book, the source code just lays there as a dead mass, with the prose vainly trying to vivify it.
Two of the books in the current roundup are in the described
source-code-with-commentary. One is Python Annotated
Archives. This is one of those heavy tomes that its
publisher, Osborne, is fond of (720 pages). I cannot really
recommend this book to much anyone. Experienced programmers
will not be able to find reference information easily, and
beginners will lack the background to jump into any particular
script. Moreover, it almost seems to be cheating to publish
a book that just reproduces (albeit also discusses) all the
scripts in the Demo/ directory of every Python distribution
(copyrights are in order, I mean it in a "scruples" way). What
are we really getting by buying this book?
A lot of concepts inevitably occur in 700+ pages. But there is
no particular order apparent to where they occur. Here is one
arbitrary example: the use of _doc_ strings is a fairly
important convention in Python programming. A beginner would
not know about it from this book unless she happened to turn to
page 162 to find an explanation of them--not because that page
is where documentation conventions are discussed, but just
because the script being discussed happens to have some
_doc_ strings. A handful of other pages scattered through
the book also mention, "by the way, this is a _doc_ string."
An experienced programmer is unlikely to need the explanation,
a beginner is unlikely to find it.
Moreover, the explanatory prose is haphazard (and not
infrequently wrong or incoherent--e.g. I cannot make heads or
tails of the lead-in to the very first example prime.py).
Martin Brown just sort of throws whatever explanation comes to
mind at each block of three or four lines of each script (which
are redundantly printed, then printed again with annotation).
But if someone doesn't understand what a loop or an numeric
operation does, the place to explain it isn't next to each
loop, but in an "about loops" or "about operations" section of
a book.
Programming with Python suffers a similar weakness to Python Annotated Archives. This title advertises on its cover "filled with over 40 full-length, annotated scripts." It lives up to its promise; but the Vaults of Parnassus have many more than 40 full-length scripts, covering a far wider range of purposes. If you need a Python script to accomplish some particular moderately complex purpose, you are a lot better off finding one to download from a Python internet archive, like Parnassus (and a few dollars the richer for not buying a book).
There is a little bit of other stuff in Programming with Python than the scripts themselves. The scripts have a little bit of explanation--but only a little bit. The beginning bit of the book has brief summary of the statements (with built-in functions mistakenly called statements), modules, and some Python syntax. But the Python tutorial is better. The last few pages of the book have a pretty good set of FAQ-type questions about Python. I haven't seen anything exactly like that on the web, and the questions presented are good ones, with accurate answers. But that is only 57 pages worth of material, and your particular questions will only be a few of those listed.
A popular type of computer programming book is the thick tome dedicated to being a complete reference for a language or tool. I've had a chance to look at three such books.
A first option for the "about everything" Python reference is the second edition of Programming Python. I have already written about the first edition, and most of what I wrote there is still true. Programming Python has grown to an even larger 1255 pages, and now covers Python 2.0. A lot of new material is in it, but the same strategy of building up examples is used. Mark Lutz' writing continues to be clear and informative, and his examples well written. Programming Python is more of a very thorough course in Python than it is a reference one returns to. There is an adequate index, and it is not unduly difficult to find things. But the main point is to walk the reader through aspects of programming in Python.
The second option for a complete reference to Python is called, straightforwardly enough, Python: The Complete Reference, by Martin Brown. As in Python Annotated Archives, Brown has a habit of interspersing short snippets of source code with prose describing the code (this time with a distracting dark gray bar to the left of the code). I still do not really like that style, but most of the sins of the other book are absent from this one. The Complete Reference covers pretty much the same ground as Programming Python, both are more about learning Python (as a programmer experienced in other languages) than they are about acting as a reference. But The Complete Reference has its material divided into 27 well chosen chapters, with both a detailed TOC and a good index. You may not be able to turn to the exact page that answers your question (as you can in Python Essential Reference, but at least you can easily find the right chapter to read. At a bit under 700 pages, reading 1/27th of the book to brush up on a topic is a reasonable task. The prose is also better organized and better written than Brown's other title.
In the "complete reference" category, I like the Python 2.1 Bible the best. If nothing else, it is more up-to-date in terms of Python version than the other options. But even beyond that, it is a nicer book for the particular slant. In the Hungry Minds tradition, the book is a bit too "loud" typographically--I can hardly focus on the cover without getting dizzy, and the text is littered with a few too many icons, tables, dingbats and the like. That said, I find the bright sans-serif header font and vertical white spacing to be very attractive; and the clean gothic fixed font makes for much more easily scanned source code samples than does the courier most books use. And even the judicious use of 10% screens draws the eye helpfully.
Aside from the (important) physical apprearance of Python 2.1 Bible, the organization is the most useful for programmers with a bit of experience (the first few chapters develop the necessary basic concepts pretty well). The book is broken into more chapters (38), and the chapters better focussed on the most important Python concepts than are the other "big" books. Moreover, the TOC and index are both of extremely good detail and quality. Brueck and Tanner's prose is furthermore the best in this category of books--factual, accurate and to the point (Lutz, and especially Brown, tend a bit towards "chatty").
Python & XML, Christopher A. Jones & Fred L. Drake, Jr., O'Reilly 2001 (est). ISBN: 0-596-00128-2
Bruce Eckel's "Thinking in Python" website:
http://www.mindview.net/Books/Python/ThinkingInPython.html
Python Essential Reference, Second Edition, David M. Beazley, New Riders 2001. ISBN: 0-7357-1091-0.
Python Standard Library, Fredrik Lundh, O'Reilly 2001. ISBN: 0-596-00096-0.
Learn to Program Using Python: A Tutorial for Hobbyists, Self-Starters, and Those Who Want to Learn the Art of Programming, Alan Gauld, Addison-Wesley 2001. ISBN: 0-2017-0938-4
Python : Annotated Archives, Martin C. Brown, Osborne/ McGraw-Hill 2000. ISBN: 0-07-212104-1.
Programming With Python, Tim Altom, Premier Press 1999. ISBN: 0-7615-2334-0
Programming Python, Second Edition, Mark Lutz, O'Reilly 2001. ISBN: 0-596-00085-5.
Python: The Complete Reference, Martin C. Brown, Osborne/ McGraw-Hill 2001. ISBN: 0-07-212718-X.
Python 2.1 Bible, David Brueck, Stephen Tanner, Hungry Minds 2001. ISBN: 0-7645-4807-7.
David Mertz' failures as a hunter, fisherman, and shepherd have
led him to his life of critical criticism. Tomorrow he may try
something else. David may be reached at [email protected]; his
life pored over athttp://gnosis.cx/publish/. Suggestions and
recommendations on this, past, or future, columns are welcomed.