|P is for Practical|
The number of 0s in this sentence is _, of 1s is _, of 2s is _, of 3s is _, of 4s is _, of 5s is _, of 6s is _, of 7s is _, of 8s is _, and of 9s is _.
That's not how I recall reading the puzzle in G.E.B, (but then again it's over a decade since I last opened the book). As I recall the question, it was:
There are _ a's, _ b's and _ c's and _ d's [repeat for e..y] and _ z's in this sentence.
And of course the numbers are written in English 'one x, one y and one z'.
This is a much harder problem to solve. Changing one number value has a rippling effect on more than one other number value. IIRC Hofstadter tried to solve it in Lisp, gave up, and a friend built a hardware implementation to solve the question.
Bonus points for solving the general question: i.e. the phrase may contain any amount of fluff ("In this sentence, if you look carefully and count all the letters you will see that it contains _ a's...)
A short while later: come to think of it, the general problem is to solve this in any language. All you need is a module that spits out a string representation in your desired language (1 => 'ein', 2 => 'zwei', 3 => 'drei'). There are modules on CPAN that do that part for you Lingua::EN::Number, Lingua::FR::Number and Lingua::JA::Number (hmm, that last one should really be JP, but never mind). Then plug that into your algorithm, and solve the problem in any language.
Cette phrase contient _ a, _ b, ... et _ z.