Re: Smooth()

[jeyadev(at)wrc.xerox.com (Surendar Jeyadev)]

In article <3A385A8B.B4CCE3C5@wellesley.edu>,
Richard G. French <rfrench@wellesley.edu> wrote:
>Out of curiosity, can anyone think of a good reason that the
>smooth() function bombs when you ask it to smooth something by 1?
>
>Ex:
>
>print,smooth([1.,3.,4.,2.],3)
>   1.00000      2.66667      3.00000      2.00000
>print,smooth([1.,3.,4.,2.],1)
>% SMOOTH: Width must be > 2 and smaller than array dimensions: <INT     
>(       1)>
>% Execution halted at:  $MAIN$     
>
>I have lots of instances where the amount of smoothing I want to 
>do is a variable, and it seems silly to have to do my own checking
>to see if I actually need any smoothing or not. 
>It certainly makes sense to me that if the
>smoothing width is only one bin, the smooth() function should just
>be a no-op and return the input array. This is what the REBIN()
>function does, after all - it does not complain if you tell it you
>want the array to be rebinned to its actual size!
>
>I've had to construct a 'mysmooth()' function that checks to see
>if the number of points by which to smooth things is less than 2,
>in which case I just return the calling array. 
>
>Are there other functions out there that you can think of that
>don't have a sensible default evalution for limiting cases like this?
>I would love to see this one changed.

See my posting yesterday about the AVG function. When it gets a
scalar, it spews out an error message as the argument expected is
an array. Yes, one would assume that some elementary checking to 
catch degenerate cases is done. I am using PV-Wave CL Version 6.01.
-- 

Surendar Jeyadev         jeyadev@wrc.xerox.com