It looks like the definition of the diagonal changed somewhere between Numeric 24.0 and numpy:<br>
<br>
In Numeric:<br>
<br>
&gt;&gt;&gt; x = Numeric.arange(2*4*4)<br>
&gt;&gt;&gt; x = Numeric.reshape(x, (2, 4, 4))<br>
&gt;&gt;&gt; x<br>
array([[[ 0,&nbsp; 1,&nbsp; 2,&nbsp; 3],<br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; [ 4,&nbsp; 5,&nbsp; 6,&nbsp; 7],<br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; [ 8,&nbsp; 9, 10, 11],<br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; [12, 13, 14, 15]],<br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; [[16, 17, 18, 19],<br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; [20, 21, 22, 23],<br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; [24, 25, 26, 27],<br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; [28, 29, 30, 31]]])<br>
&gt;&gt;&gt; Numeric.diagonal(x)<br>
array([[ 0,&nbsp; 5, 10, 15],<br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; [16, 21, 26, 31]])<br>
<br>
But in numpy:<br>
<br>
&gt;&gt;&gt; import numpy as Numeric<br>
&gt;&gt;&gt; x = Numeric.arange(2*4*4)<br>
&gt;&gt;&gt; x = Numeric.reshape(x, (2, 4, 4))<br>
&gt;&gt;&gt; Numeric.diagonal(x)<br>
array([[ 0, 20],<br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; [ 1, 21],<br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; [ 2, 22],<br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; [ 3, 23]])<br>
<br>
The old logic seems to be clear: x is a pair of matrices and diagonal
returns a pair of diagonals, but the new logic seems unclear: the
disagonal returns the first rows of the two matrices transposed.<br>
<br>
Does anyone know when this change was introduced and why?<br>
<br>
<br>