Let me state up-front that this is an indispensable book for me, and, as a reference, or as a source for expanding a basic C++ capability, it's easily worth 5 stars, not 4. In fact, as an intermediate-level C++ user, this is the most indispensable book that I have. I have 15 C++ books on my shelf, and this one (the fourth edition) is the most comprehensive, even surpassing Stroustrup's tome in most areas. (Stroustrup's influence on Primer is pretty obvious in spots.) The book's depth and its style of presentation, however, will mean that it's not the best primer for newcomers, and that's why I only gave it a 4-star rating. Look at the painting metaphor expressed on the front cover of the book. This book is more than a primer: it's a full pallet of colors, and as such is more than a beginner "painter" will be able to put to use quickly. I need to clarify what I'm saying here.
You might ask how a book can be the most comprehensive available without being longer in page count. The answer comes down to style. This book rarely expends space on full-length, executable code examples. The treatment of nine out of ten subjects offers only snippets of code - not complete executable examples. Many books offer examples that take up multiple pages for the code and multiple pages for the line-by-line explanations of the code. This difference in presentation is the single biggest reason why C++ Primer can offer so much more thorough coverage of C++ language features in fewer than 900 pages. The downside is, if you intend to learn C++, you absolutely must experiment with working code - there is no other way. The fact that this book does not offer much complete working code means that you will have to spend more time getting yourself set up to experiment. This will be especially true if you are a beginner, just because it will be harder to get things working.
The conciseness of code examples is matched by conciseness of explanations. In this regard, however, I have found no downside: the vast majority of explanations are extremely clear to the point of elegance. I've found only a few flaws in clarity, which are probably left over from earlier editions.
The tremendous depth of this book will be another weakness for newcomers to C++ who need to get started quickly. In my opinion, you just can't get started quickly when every treatment of every subject has to cover all of the ins and outs of that subject. There are too many subjects and too many ins and outs. If you try to go straight through the book, you'd better be prepared to work lots and lots of problems along the way: it will take so long to get through it that if you don't practice as you go, you may find yourself forgetting almost as fast as you're learning.
The other side of the coin is this: if you really do take this book, work through it from start to finish, and really master the material, you'll have a fantastic knowledge of C++. I seriously doubt that three good college courses in C++ - assuming no prior programming experience - would teach as much. The fact that you'll have to develop your own working code to demonstrate each feature will, like the depth of treatment, make you a better programmer, even though it seriously slows you down in the process.
You can conclude, then that if you are a newcomer to C++, and especially if you are inexperienced with programming in general, this book will be a long, hard climb. I'd be willing to bet that fewer than 5 percent of purchasers of this book start at the beginning and work straight through. Even so, if you are a newcomer taking a C++ programming class or just using another book to get started, it will be valuable for parallel reading and will come in very handy when you want to do a deep dive. Finally, don't forget that if you have a basic grounding in C++ fundamentals and you want to grow from there, this is a dynamite, five-star opportunity.
I feel like I have to elaborate here a bit as well. Time after time after time, when I've wanted a better understanding of a really fine point on the language, I've turned to this book and been rewarded for doing so. Every single subject is explained fully and, in most cases, eloquently. You can learn to do a lot quickly in C++ using other resources, but C++ offers the ability to fine tune and get close to the hardware. This book will make you better understand what the compiler is doing. The authors took the time to make extensive use of cross-references, which are invaluable in a subject this complex. The explanations, bolstered by cross-references, tend to leave me more than satisfied that I've gotten a grip. When I finish a section and have some example code executing, I feel like I have a complete understanding: I rarely find myself asking, "... but what about ...." I could go on and on, but you get the message. (One reason I'm taking the time to write this is that I feel indebted to the authors!)
The book builds up a consistent set of semantics - more so than any other book I've read. One example is their thorough explanation and consistent use of the terms "declare" and "define." Another is example is "argument" and "parameter list." Unambiguous semantics are essential to unambiguous communication, and, as you know, programming is one of the most unforgiving disciplines there is; ambiguity is intolerable. This attention to detail pervades the entire book, beyond just semantics.
Also, like with Stroustrup's book, I don't have to worry about whether the authors got it right or wrong or only half-right. This is an authoritative work: these guys know their subject as well as they know how to communicate it.
I especially like the dark page summaries and glossaries that act as chapter boundaries. I like the highlighted notes, tips, "bewares," and best practices. I like the blocks of text that address "cautions." These features are value-multipliers for me.
I also appreciate that there is absolutely no subconscious assumption that you already know C. That assumption seems to creep into so many other books, even though the authors deny it.
I agree with other reviewers that this edition is a huge improvement over previous editions. (I have the second edition beside me as I write this.) Frankly, I don't like the earlier editions. My guess is that the new authors, who came on board for this edition, should get a lot of credit. (I know an absolutely brilliant guy who has done some real innovation with C++. For example, he wrote a C++ library that allows you to write C++ in Python style, while getting C++ performance. This same guy was so intimidated by an earlier edition of Primer that he put off trying to learn C++ for several years! I'm still trying to convince him to get the fourth edition!)
So the fourth edition is a vast improvement, but, assuming that we don't try to make it all things to all readers and stick with the objective of compact comprehensiveness, how could we improve it further? The code examples could be simpler. I think Schildt is the master of communicating through code examples, just because his examples do the best job of isolating on the subject at hand. C++ is an extremely broad language, and when I read about a concept, I don't want to be expected to know everything, especially code examples, that preceded it in the book. Primer could do a better job - at least for me - in that respect. Examples should not be codependent: each example should lean only on the preceding material as much as is necessary to get the point across. Stated another way, OO programmers are big on encapsulation; OO tutorial writers should use the same rigor in encapsulating modules of tutorial text. All experts agree that the best code is code that is easily readable. In a tutorial, the meaning of the code examples should jump off the page. Primer is not bad in this regard, but it could be better. Some will argue that leaning on previous material reinforces the learning of the material, and they will be right. In my opinion, however, that argument applies mostly to course textbooks where the learning of diverse features can be forced into a tightly channeled sequence. I often use Schildt and Holtzner for reference when I need a quick answer. (As an aside, Stroustrup's book is the worst, for me, in terms of code examples being unnecessarily complex and codependent, even though I consider his book second only to this one as an indispensable reference.)
Bottom line: C++ Primer is an amazing accomplishment. The authors have done the best job to date of sorting out and presenting in lucid fashion a vast and complex network of features. I speak from experience as a technical writer as well as a student of C++. I feel guilty for assigning only four stars, but I'm concerned about the newbie who sees a 5-star rating on a "primer," and rushes to spend his limited budget on this book as a sole reference. C++ Primer is a very wise investment if you are really serious about C++, whether newbie or expert, but most newbies will also need a primary resource that is more beginner-oriented. Other books I recommend, in addition to the ones mentioned previously, include Prata (a good tutorial and a good resource for complete code examples, although there is significant codependence of examples); Lafore; Josuttis' classic STL book (definitely not for beginners); Schildt's STL book, also for its code examples; Keough and Gray (for quick reference only); and Safir and Brown (which does assume that you know C).