Source

extradoc / planning / micronumpy.txt

NEW TASKS
---------

- astype

- a good sort function

- indexing by arrays and lists

- endianness

- frompyfunc to create ufuncs from python functions

- more ufuncs

- linspace/other ranges

- more attributes/methods on numpy.flatiter

- expose ndarray.ctypes

- Make .T not force a lazy array.

- subclassing ndarray (instantiating subcalsses curently returns the wrong type)

  * keep subclass type when slicing, __array_finalize__

  * ndarray.view

OPTIMIZATIONS
-------------

- SSE

- count number of True's for bool arrays, so we don't have to recompute

- bridges are a tad too complicated - we need to store optimizer info
  in order to improve though

- iterators should possibly not require reinstantiating them.

- things like take/item/fancy indexing can use some knowledge about the density
  of data and either evaluate interesting points (without forcing) or
  do what they do now.

- counting by element_size instead of by 1 and then multiply sounds
  like a much faster option sometimes