Closed zoemarschner closed 2 weeks ago
Looks great, thank you so much for this! I haven't tested this code, but the problem makes sense and the fix looks reasonable so I'm going to merge it.
Also, I love that this removes a // TODO what is this E doing here?
comment 😂
Currently, solving underdetermined systems using SuiteSparse by calling
solve
does not work. For example,will print out
1 1 1
, but the correct answer is0.5 2 0.5
. This is because SuiteSparse's QR decomposition includes a permutation matrix (i.e.,E
such thatQR=AE
), but geometry central does not properly account for this in the call toSuiteSparseQR_solve
. (Thank you @MarkGillespie for helping to find the bug!)