Matrixify Q matrix before comparison with a matrix

[dkarrasch]

This came up in a nanosoldier run in JuliaLang/julia#46196. Independently from that PR, it is advisable to matrixify the Q matrix before comparing it via getindex, which is slow for Q's. With the proposed change, the tests in this package become independent from whether JuliaLang/julia#46196 gets accepted or not.

[fkastner]

Thanks for looking after the ecosystem, Daniel!
It is good to know that also such small packages are notified/updated when Base changes.

Just one question: Could we also use Matrix(Q), which would better convey the intention here?
Actually, I just tested it and it seems that Matrix(Q) != Q*I. Is that a bug?

[dkarrasch]

No worries, Felix. This package is registered, so this is as big as it can get. 😄 The difference you mention is known, intentional and documented, see the docstring of qr. In your case, as you noted, Matrix(Q) returns a thin matrix, but on the rhs, you have the full, square matrix. Alternatively, you could use Q*Matrix(I, size(Q)) as suggested in the docstring, but Q*I gives the same result and doesn't allocate the intermediate Bool matrix.

[fkastner]

Ok, that makes sense then. Thank you again!