Re: Should interp1.m function allow jumps in X-values?

[Ben Abbott]

On Aug 22, 2012, at 9:29 AM, Michael D Godfrey wrote:

> On 08/22/2012 08:33 AM, Ben Abbott wrote:
>> The reason for the difference is subtle (and sorry for not pointing it out 
>> earlier).  The sort() function does not change the order of the duplicate X 
>> values.  Thus, the continuity condition is not switched from right to left.  
>> The flipup() function is used in interp1 to do that.  Below I used the 
>> fliplr() to ensure the proper continuity.
>> 
>>      X = [ 2 1 3 2];
>>      Y = [ 9 1 3 10]; 
>>      % These are right-continuos (returns 10)
>>      y1 = interp1 (X, Y, 2)
>>      [~, n] = sort (X);
>>      y2 = interp1 (X(n), Y(n), 2)
>>      [~, n] = sort (X, "descend");
>>      y3 = interp1 (X(n), Y(n), 2)
>>      y4 = interp1 (X, Y, 2, "-right")
>>      % These are left-continuous (returns 9)
>>      [~, n] = sort (fliplr (X), "descend");
>>      y5 = interp1 (X(n), fliplr(Y)(n), 2)
>>      y6 = interp1 (X, Y, 2, "-left")
>> 
>> Ben
>> 
> Very good.  I could have thought of that!  So, that gets rid of an important
> problem.  And this adds a bit in favor of your addition of -right and -left,
> or at least another point to be made clear in the documentation.  This
> is definitely a subtle point about sorting.  Time to look up what Knuth
> has to say about this.
> 
> By the way, the current devel interp1.m does not do the flipr().  Did you
> add that in your local patch?
> 
> Michael

I haven't pushed the patch that includes the flip part yet.

It has beed posted to the bug-tracker.

        https://savannah.gnu.org/bugs/?36732

See attachement #26413

Ben