Soeren Sonnenburg wrote:

-- why do vectors have no 'orientation', i.e. there are only row but no column vectors (or why do you treat matrices/vectors differently, i.e. have vectors at all as a separate type)

I think the key to understanding this is that NumPy uses a fundamentally different data type that MATLAB ans it's derivatives. MATLAB was originally just what it is called: a "Matrix" laboratory. The basic data type of Matlab is a 2-d matrix. Even a scalar is really a 1X! matrix. Matlab has a few tricks that can make these look like row and column vectors, etc, but they are really always matrices. On the other hand, NumPy arrays are N-dimensional, where N is theoretically unlimited. In practice, I think the max N is defined and compiled in, but you could change it and re-compile if you wanted. In any case, many of us frequently use 3-d and higher arrays, and they can be very useful. When thought of this way, you can see why there is no such thing as a column vs. a row vector. A vector is a one-dimensional array: it has no orientation. However, NumPy does support NX1 and 1XN 2-d arrays which can be very handy:

import numarray as N a = N.arange(5) a.shape = (1,-1) a array([[0, 1, 2, 3, 4]]) b = N.arange(5) b.shape = (-1,1) a array([0, 1, 2, 3, 4]) b array([[0], [1], [2], [3], [4]])

So a is a row vector and b is a column vector. If you multiply them, you get "array broadcasting":

a * b array([[ 0, 0, 0, 0, 0], [ 0, 1, 2, 3, 4], [ 0, 2, 4, 6, 8], [ 0, 3, 6, 9, 12], [ 0, 4, 8, 12, 16]])

This eliminates a LOT of extra duplicate arrays that you have to make in Matlab with meshgrid. When you index into an array, you reduce its rank (number of dimensions) by 1:

a = N.arange(27) a.shape = (3,3,3) a array([[[ 0, 1, 2], [ 3, 4, 5], [ 6, 7, 8]],

[[ 9, 10, 11], [12, 13, 14], [15, 16, 17]], [[18, 19, 20], [21, 22, 23], [24, 25, 26]]])

a.shape (3, 3, 3) a[1].shape (3, 3) a[1][1].shape (3,)

When you slice, you keep the rank the same:

a[1:2].shape (1, 3, 3)

This creates a way to make row and column "vectors" from your 2-d array (matrix)

a = N.arange(25) a.shape = (5,5) a array([[ 0, 1, 2, 3, 4], [ 5, 6, 7, 8, 9], [10, 11, 12, 13, 14], [15, 16, 17, 18, 19], [20, 21, 22, 23, 24]])

To make a "row vector" (really a 1XN matrix)

a[0:1,:] array([[0, 1, 2, 3, 4]])

To make a "column vector" (really a NX1 matrix)

a[:,0:1] array([[ 0], [ 5], [10], [15], [20]])

I hope that helps: -Chris -- Christopher Barker, Ph.D. Oceanographer NOAA/OR&R/HAZMAT (206) 526-6959 voice 7600 Sand Point Way NE (206) 526-6329 fax Seattle, WA 98115 (206) 526-6317 main reception Chris.Barker@noaa.gov