qsort and qsortvec inconsistent with BAD values

[d-lamb]

The docs for qsort state that 'Bad values are moved to the end of the array:'

The docs for qsortvec state nearly the same thing: 'Vectors with bad components should be moved to the end of the array:'

However this does not seem to be borne out in practice in qsortvec.

pdl> p $p = pdl("[0 0] [-100 0] [BAD 0] [100 0]")

[
 [   0    0]
 [-100    0]
 [ BAD    0]
 [ 100    0]
]

pdl> p $p->qsortvec

[
 [ BAD    0]
 [-100    0]
 [   0    0]
 [ 100    0]
]

pdl> p $p->((0))->qsort
[-100 0 100 BAD]

My expectation is that the output of qsortvec (given the documentation, and the output of the qsort expression) should be:

[
 [-100    0]
 [   0    0]
 [ 100    0]
 [ BAD    0]
]

[mohawk2]

That is also my expectation! qsortvec confidently has HandleBad => 1 and the docs that you note, but doesn't have BadCode (or, as you rightly note, correct handling of bad values).

[mohawk2]

...and now it (and qsortveci) do. Thanks for the report!

[mohawk2]

qsort inplace with bad values was pretty underfunctional too, now fixed by the above commit. That was masked by the qsort testing code largely checking its work with all($pdl1 == $pdl2), and that does not do what you think if either ndarray has badvalues, including in different places from each other, but at least one matching non-bad value. Bad values are a bit like NaN, but not entirely.