Re: Preserving signbit for range operator

[Daniel J Sebald]

On 02/25/2013 02:17 PM, Michael D. Godfrey wrote:
> On 02/25/2013 03:03 PM, Daniel J Sebald wrote:
> > [-3:-0] -> [-3 -2 -1 0]
> > [-0:3] -> [-0 1 2 3]
> This is at least "controversial." Sorts treat -0 and 0 as equal, but
> they have to be recognized as different for various purposes such
> as getting the directions right for functions of complex variables.

The complex number field is not an ordered field.  What does direction mean?


> So, it seems clear to me that the "right" result for the first line
> above is: [-3:-0] -> [-3 -2 -1 -0].
>
> I cannot think of an argument for not doing this.

Well, the definition of range needs to be tweaked then.  To me the 
current definition reads something analogous to [a:z:b] translates to

a a+1z a+2z ... a+Lz

where L is the largest integer for which a+Lz <= b.  However, 
technically that is not correct.  It is not the maximum, necessarily, 
but "maximum or minimum (whichever is pertinent based upon the order of 
the limits a and b)".  Consider:

octave:6> [0:-3:-5]
ans =

   0  -3

No matter the choice, the description for the range could use some 
mathematical accuracy.

Dan

PS: A bit arbitrary, but this result