Re: [ferret_users] boundary condition for smoothing (@SPZ and friends)

[Ansley Manke <ansley.b.manke@xxxxxxxx>]

Hi Ryo,
We're thinking of new @ transformations.  As to phasing in or replacing 
the existing transformations, we'd probably start by making a new set, 
say @SBZe, for "edges" and recommend their use - keeping the old ones 
available so the results of people's scripts don't change unexpectedly.

Ansley

Ryo Furue wrote:

> Thanks, Billy and Ansley.
>
> | Yes, good suggestion. We should be able to write a set of new
> | functions that would parallel the smoothing transforms
> | 
> |  @SBX        box smoothed      
> |  @SBN        binomial smoothed
> |  @SWL        Welch smoothed
> |  @SHN        Hanning smoothed
> |  @SPZ        Parzen smoothed
> | 
> | but with the edges of the "windows" filled in so that these
> | smoothers use filled in so that data isn't lost at the edges.
>
> I'm not familiar with ferret functions, but. . . .
>
> What would the function version look like which correspond
> to this?
>
>    let g = f[x=1:10@sbx, y=1:20@sbx]
>
> Would it look like this?
>
>    let g = box_smoother(box_smoother(f, x=1:10), y=1:20)
>
> I guess the @ forms would be a bit more convenient.
> Would it be possible to introduce more @ transformations?
>
> Also, in a long run, that might be more benefitial to the users;
> that is, once we have new @ transformations, we could eventually
> declare the old ones "obsolete" and recommend the users to
> migrate to the new ones. (I'm assuming that having edge values is
> always better, and that assumption might be wrong.)
>
> Regards,
> Ryo
>