bugs related to Data.Matrix.inverse

[LATBauerdick]

Hi -- hope this is the right way to report bugs for the Data.Matrix package!

There are two issues showing up when matrix inversion runs into problems (e.g. for nearly singular matrices):

in function ref:

goodRow = case listToMaybe (filter significantRow [1..ncols mtx]) of
            Nothing -> error "Attempt to invert a non-invertible matrix"
            Just x -> x

1. ncol needs to be nrows, current code crashes with access error instead of returning error.

2. as ref is called from inverse (and rref), which promises to return Either String (Matrix a), calling error in case of non-invertible matrices is not a very nice thing to do. ref should be modified to also return Either String (Matrix a), so inverse can return Left "can't invert". It's important that user code can handle singular or numerically unstable matrices themselves.

Thanks for your consideration (and thanks for providing Data.Matrix, which is otherwise a joy to use).

Best regards, LatB

[octopuscabbage]

Good suggestions & Report, I can prepare a PR handling these later, I have a pretty free day today.

[octopuscabbage]

ncol needs to be nrows, current code crashes with access error instead of returning error.

Could you provide a test case where this is true? Would make my life a bit easier and make sure i solve it.

[octopuscabbage]

Issue should be resolved in this PR #45 waiting on your test case to make sure I have it solved.