[Octave-bug-tracker] [bug #47515] median(integers) returns a floating point answer

[Rik]

Update of bug #47515 (project octave):

                  Status:         Patch Submitted => Fixed                  
             Open/Closed:                    Open => Closed                 

    _______________________________________________________

Follow-up Comment #1:

This is good.  I made a few changes that are shown below so you can get the
flavor of what I was doing.  The extra documentation is correct, but not
really necessary.  Pretty much all functions operate along the first
non-singleton dimension unless specified differently.  Rather than using
elseif, it felt better to contain all the code that belonged to the else block
in a single else block.  I then re-used the existing sum code rather than
using a different function for logicals.  I got away with a single isfloat()
test rather than two tests.  Checked in here
(http://hg.savannah.gnu.org/hgweb/octave/rev/ffb4770ba079).


 ## return them in a row vector.
 ##
 ## If the optional @var{dim} argument is given, operate along this
dimension.
-##
-## If @var{x} is a multi-dimensional array and @var{dim} is not specified,
-## the median is taken along the first non-singleton dimension.
 ## @seealso{mean, mode}
 ## @end deftypefn
 
@@ -85,14 +82,14 @@ function retval = median (x, dim)
   k = floor ((n+1) / 2);
   if (mod (n, 2) == 1)
     retval = nth_element (x, k, dim);
-  elseif (islogical (x))
-    retval = max (nth_element (x, k:k+1, dim), [], dim);
   else
     retval = sum (nth_element (x, k:k+1, dim), dim, "native") / 2;
+    if (islogical (x))
+      retval = logical (retval);
+    endif
   endif
   ## Inject NaNs where needed, to be consistent with Matlab.
-  if (!(isinteger (x) || islogical (x)))
-    ## This converts from discrete types, even if isnan(x) is empty
+  if (isfloat (x))
     retval(any (isnan (x), dim)) = NaN;
   endif

 



    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?47515>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/