On Sat, 1 Jun 2019, Fred Engst wrote:
Hi Allin, Jack, and Peter,
I just sorted the mroz dataset by GUI, and the screenshot follows.
Thanks. This should be fixed in current git and snapshots. The problem
was fallout from our redefinition (last year) of "NA" (missing) from
the largest double-precision number to NaN (Not-a-Number). I hope and
think that it's the last instance of such fallout.
Our old NA unambiguously compared larger than any valid number on
sorting, while NaN does not compare equal to, greater than, or less
than any valid number. The glibc C library on Linux nonetheless
manages to sort correctly in the presence of NaNs (so Jack and I
didn't see any problem), but the Microsoft and Apple C libraries do
not. The fix was to handle NaNs explicitly in our sorting code.
BTY, as you might be able to see, email message from your responses
gets garbled when I gets them, particularly the scripts that you’ve
provided. Any reason why?
I don't know. Our responses look OK in the archive at
https://gretlml.univpm.it/hyperkitty/ . They also look OK to me in
Alpine and gmail via Firefox.
Allin