#909 Method 'infinity_norm' does not actually compute the infinity norm
|Reported by||Oliver Sander (firstname.lastname@example.org)|
|Reported at||Apr 26, 2011 21:01|
|Version||Git (pre2.4) [autotools]|
|Operating System||Unspecified / All|
I would expect a method called 'infinity_norm' of a matrix to compute the infinity norm: the maximum of the absolute values of all entries. However at least the implementations in DenseMatrix, BCRSMatrix, and Matrix compute something else, namely the maximum over the 1-norms of the matrix rows. This is a flagrant violation of the principle of least surprise.
The fix is easy, however apparently the behavior is intentional. At least it is mentioned in the documentation. Why would I need such a strange norm? And isn't there a better name for it?