Purely Functional Data Structures

This is a great book. An improved version of Okasaki's PhD thesis, the book introduces the reader to the notion of efficient data structures which still get to be purely functional (i.e. not imperative, no side-effects).
A common problem with purely functional data structures is that, since there is no state, we can't change the value of a variable once it has been assigned. We can always create a new variable with the same exact name, but it will be a completely new variable and any pointer to the previous variable would be lost. This is called Persistency. Persistency makes building efficient purely functional data structures and algorithms, even for basic tree searching and sorting, an intricate craft.
In this book, we are presented with an arsenal of techniques to build purely functional data structures which are efficient in the sense that they compete well with imperative counterparts. The book uses lazy evaluation and memoization as the keys to unlock efficiency. The algorithms are presented as working code in Standard ML (and in Haskell in the appendix) which is great. Everything is very well explained and rigorous. Techniques for computing the complexity of algorithms are already presented (banker's method and physicist method).
Make no mistake. This is not an introduction to functional languages. Knowledge of those languages is assumed and, although an experienced programmer in other kind of languages may profit from it without knowing any functional languages, if you are looking for an introduction, buy something else.
Having said that, if you are familiar with Haskell or ML (any dialect), this is book is a great way to both (1) learn important techniques that you can apply when building your own data structures and algorithms and (2) as a reference for basic functional data structures and algorithms (ordered sets, red-black trees, etc).
I did buy this book and I am very pleased with it.