[Numpy-discussion] help writing an array subtype

[Trevor Clarke]

I'm embedded python (and numpy) is a C++ app and I'm trying to access my
array data in numpy but I'm not sure where to start. Due to data sizes, I
don't have access to the entire array contiguously, my app implements a
virtual memory like paging system where variable sized pages of data are
brought into memory on demand. Each page has a contiguous block of data
containing at least the requested piece of the array (if it's more, it's
arranged so the appropriate data can be accessed with strides). I'd like to
load pages when they are accessed in an ndarray and when a new view is
requested. I was thinking of loading and holding the most recently accessed
page of the array on a per view basis. (i.e. if a view is created it can
access a different part of the array from the original array) I've ready
over the documentation on subclassing ndarray but there's a lot of
information there and I'm not quite sure the best place to start. Could
someone point me in the right direction or perhaps offer a better solution?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/numpy-discussion/attachments/20090205/3d26a68c/attachment.html>