I have a slew of Springer books on specific packages, but my library is lacking a good solid book on R the language. While a language spec book would be great, it would be nice if I could find a book that has a solid founding in the language itself but couched in the examples that typify the day to day work of R users - statistics, visualization, analysis & modelling.
What is the one R book that you would take to a desert island?
When I read John Chambers 'Software for Data Analysis: Programming with R' (Springer, 2008) last year, I really benefited from it. Even though I had used S and R for a decade, a couple of things became a lot clearer after reading this book by the (co-)creator of the language.
Other books that come to mind are the Braun and Murdoch's 'First Course on Statistical Programming with R' (which I haven't read), or Gentleman's 'R Programming for Bioinformatics' which is well regarded (and not at all about Bioinformatics!).
I enjoyed "Programming With Data" by John Chambers (1998), which was really about S but did a good job helping me understand the R language itself. I'm not sure if I would recommend it to a non-programmer or how it compares with his recent book, but I benefited from reading it.
Phil Spector's "Data Manipulation with R" is really good in the day to day sense, only not modeling and visualization, but, well, data manipulation and exploration. It's a slim thing, and gives you a good handle on the various data types and common data munging techniques.
On language foundations, I recommend "S Programming" by Venables and Ripley.
After several years, I still come back to Statistical Computing: An Introduction to Data Analysis using S-Plus by Michael J Crawley. It's very comprehensive and the examples are written in a style that is easy to understand, even for those with limited statistical knowledge.
There is an updated version that is more R-centric, called The R book, which covers much the same content. (I haven't got a copy myself, so I can't accurately review the contents.)
My Desert Island book is "Statistics With R", a free e-book by Vincent Zoonekynd. I usually refer to it online but he's prepared a pdf version, available on the first page of his Site, all 1266 pages.
This is the rare book that seems to be a great reference and an excellent source of advanced tips. It's the best R resource i've come across, free or not.
VZ relies almost exclusively on the source code (he mentions a number of interesting and useful but undocumented functions).
One of the chapters for instance, "From Data to Graphics", is an in-depth tutorial on R graphics presented by examples on an astonishingly wide range of subjects, e.g., L1-L2 regularization, tree maps, normalization, etc., which includes detailed discussion of critical topics that usually don't get proper coverage in other texts (QQ plots, for instance). Another theme of the same chapter is an exposition on data visualization--i.e., which plotting formats best illuminate which types of data. A good example here is his polar chart near the middle of the chapter, showing visitor numbers to a website by hour of the day; VZ uses this example recommend this chart type for periodic data.
Personally, I've gotten a lot of mileage out of R in a Nutshell. There are areas that it doesn't cover, like the esoterica of R's syntactic objects, but it provides a good overview of the practical aspects of R in a unified presentation. I found that most documentation online was less thorough and missing major concepts.