[SciPy-dev] Dangerous Matrix +/- Scalar behavior

[Norbert Nemec]

Hi there,

just got bitten by the same bug as often before:

In physics, it is not unusual to abbreviate the notation
    E*eye(n) - H
for a scalar value E and a n*n-matrix H by simply writing
    E - H

With the current definition of the matrix class, however, this will be
interpreted as
elementwise subtraction and lead to a hard-to-find error.

Since elementwise addition/subtraction of a scalar value to a matrix
appears extremely
rarely in linear algebra, wouldn't it be an idea to disallow this
operation on matrices
(which usually indicates an error) and raise an exception instead?

Greetings,
Norbert