Wrong full matrix for Q of QR-factorization in sparse

[KlausC]

There seems to be a bug in the constructor Matrix(::SuiteSparse.SPQR.QRSparseQ).

julia> versioninfo()
Julia Version 0.7.0-DEV.4473
Commit 2f826b55d6 (2018-03-06 05:55 UTC)
Platform Info:
  OS: Linux (x86_64-redhat-linux)
  CPU: Intel(R) Core(TM) i7-3610QM CPU @ 2.30GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-3.9.1 (ORCJIT, ivybridge)
Environment:

julia> A = sparse([0.0 1 0 0; 0 0 0 0])
2×4 SparseMatrixCSC{Float64,Int64} with 1 stored entry:
  [1, 2]  =  1.0

julia> F = qrfact(A);

julia> Matrix(F.Q)
2×0 Array{Float64,2}

julia> Matrix(F.Q')
2×2 Array{Float64,2}:
 1.0  0.0
 0.0  1.0

Size calculation of the result matrix in LinearAlgebra/src/qr.jl:530 is not ok

Matrix(A::AbstractQ{T}) where {T} =
 lmul!(A, Matrix{T}(I, size(A.factors, 1), min(size(A.factors    )...)))

The number of columns of Matrix(F.Q) should be the same as the number of rows of F.R, or it should be square.

[andreasnoack]

This is an unfortunate consequence of the storage format for the householders being different between dense and sparse QR here but the Matrix(AbstractQ) constructor assumes they are the same. The constructor should probably not access the fields directly.