As a lurker and neophyte user of Numeric/numarray, I inject my perspective...
the existing documentation for NumPy is not fantastic - some aspects of it are positively obscure - but overall it is better than good-enough
I disagree here - I have found the docs to be a real stumbling block to learning and putting new (to me) methods to use. There is often a very high level of C and/or specific math familiarity assumed; I spend a large amount of time in those instances Googling for example code (usually not much Numpy code is just sitting, posted online) and Usenet examples and discussion. I then spent time at the prompt testing up my own examples.
The printed Python books are usually very good resources, unfortunately, there is too much competition for general, intro Python and not enough depth into specific areas. This is the classic Publishers Dilemma. I am buying the new wx book, myself. A similar, in-depth book on Numpy does interest me at $29-$39, and Ch2 seems to be nicely done. Hopefully the new scipy_core will have the low overhead of Numeric for small arrays. Does it have easy access to the pointer to the array memory, as in numarray? (for use with ctypes.memmove())
That said, I still like more examples. People learn in different ways. I still maintain some PHP web sites, and the best (some say only ;-) ) thing is their style of online docs with user comments, ex.: array-rand: http://www.php.net/manual/en/function.array-rand.php I you need to do a particular Thing, it is very likely to already be _in_ the docs. I think this is a better approach than Wikis, and is easier and more likely for users to contribute, like this: http://www.php.net/manual/add-note.php?sect=index&redirect=http://www.ph... (Notice also: "Please note that periodically, the developers may go through the notes and incorporate the information in them into the documentation. This means that any note submitted here becomes the property of the PHP Documentation Group.")
Ray Schumacher hobby: http://rjs.org/astro/1004x/Python/