On Sat, Nov 22, 2008 at 13:19, Loïc BERTHE berthe.loic@gmail.com wrote:

I've encoutered an error during an ipython's session that I fail to understand :

In [12]: n = 4 In [13]: K = mat(diag(arange(2*n))) In [14]: print K [[0 0 0 0 0 0 0 0] [0 1 0 0 0 0 0 0] [0 0 2 0 0 0 0 0] [0 0 0 3 0 0 0 0] [0 0 0 0 4 0 0 0] [0 0 0 0 0 5 0 0] [0 0 0 0 0 0 6 0] [0 0 0 0 0 0 0 7]]

In [15]: o = 2*arange(n) In [16]: kc = 10 + arange(n) In [17]: K[ r_[o-1, o], r_[o, o-1] ] = r_[kc, kc] In [18]: print K [[ 0 0 0 0 0 0 0 10] [ 0 1 11 0 0 0 0 0] [ 0 11 2 0 0 0 0 0] [ 0 0 0 3 12 0 0 0] [ 0 0 0 12 4 0 0 0] [ 0 0 0 0 0 5 13 0] [ 0 0 0 0 0 13 6 0] [10 0 0 0 0 0 0 7]]

In [19]: K[ r_[o-1, o], r_[o, o-1]] += 10*r_[kc, kc]

<type 'exceptions.ValueError'> Traceback (most recent call last) /home/loic/Python/numpy/<ipython console> in <module>() <type 'exceptions.ValueError'>: array is not broadcastable to correct shape

In [20]: print K[ r_[o-1, o], r_[o, o-1]].shape, r_[kc, kc].shape (1, 8) (8,)

In [21]: print K[ r_[o-1, o], r_[o, o-1]].shape, r_[kc, kc][newaxis, :].shape (1, 8) (1, 8)

In [22]: K[ r_[o-1, o], r_[o, o-1]] += 10*r_[kc, kc][newaxis, :]

<type 'exceptions.ValueError'> Traceback (most recent call last) /home/loic/Python/numpy/<ipython console> in <module>() <type 'exceptions.ValueError'>: array is not broadcastable to correct shape

Could you explain me :

• Why do an assignment at line 17 works where an increment raises an

error (line 19) ?

matrix objects are a bit weird. Most operations on them always return a 2D matrix, even if the same operation on a regular ndarray would return a 1D array. In the assignment, no object is actually created by indexing K, so this coercion never happens, and the effect is the same as if you were using ndarrays. With +=, on the other hand, indexing K does return something that gets coerced to a 2D matrix. I'm not entirely sure where the ValueError is getting raised, but I suspect it happens when the result is getting assigned back into K.

Your code works fine if you just use regular ndarray objects. I highly recommend just using ndarrays, especially if you are going to be doing advanced indexing like this. The "always return a 2D matrix" semantics really get in the way for such things.