From maintainers-request at octave dot org Thu Mar 30 10:03:47 2006
Subject: Re: Bi variate to mono variate functions
From: David Bateman <David dot Bateman at motorola dot com>
To: "John W. Eaton" <jwe at bevo dot che dot wisc dot edu>
Cc: Anglade Pierre-Matthieu <anglade at gmail dot com>, maintainers@octave.org
Date: Thu, 30 Mar 2006 17:59:17 +0200

John W. Eaton wrote:

>On 30-Mar-2006, David Bateman wrote:
>
>| I defined the function
>| 
>| function y = myfun2(x,c)
>|   global p;
>|   p = 10;
>|   y = 1./(x.^3-2*x-c);
>| end
>| 
>| and then did the following in matlab
>| 
>| >> global p
>| >> p = 5;
>| >>  Q = quad( at (x)myfun2(x,p),0,2)
>| 
>| Q =
>| 
>|    -0.4605
>| 
>| >> p = 10;
>| >>  Q = quad( at (x)myfun2(x,p),0,2)
>| 
>| Q =
>| 
>|    -0.2043
>| 
>| >>
>| 
>| I think this makes it pretty clear that the anonymous variable is
>| evaluated when the anonymous function is declared...
>
>I think your test is not quite right.  You are not using P in the
>expression inside of myfun2, and you are redefining the anonymous
>function in each case (so how can you tell whether the parameter is
>evaluated when the function is defined or evaluated?).
>
>Maybe a better test would be
>
>  >> type myfun2
>
>  function y = myfun2(x,c)
>	 y = 1./(x.^3-2*x-c);
>
>  >> p = 5;
>  >> f =  at (x) myfun2 (x, p);
>  >> quad (f, 0, 2)
>
>  ans =
>
>     -0.4605
>
>  >> p = 10;
>  >> quad (f, 0, 2)
>
>  ans =
>
>     -0.4605
>
>so the function is defined just once, and it keeps the value
>available at the time of the definition instead of picking up the
>modified value that is defined later, or
>
>  >> clear all
>  >> f =  at (x) myfun2 (x, p);
>  >> quad (f, 0, 2)
>  ??? Undefined function or variable 'p'.
>
>  Error in ==>  at (x) myfun2 (x, p)
>
>
>  Error in ==> quad at 63
>  y = f(x, varargin{:});
>
>  >> p = 10
>
>  p =
>
>      10
>
>  >> quad (f, 0, 2)
>  ??? Undefined function or variable 'p'.
>
>  Error in ==>  at (x) myfun2 (x, p)
>
>
>  Error in ==> quad at 63
>  y = f(x, varargin{:});
>
>
>I suppose the way to do this is to modify the function
>make_anon_fcn_handle in parse.y to perform some magic necessary to
>assign initial values to any symbols that are present in the anonymous
>function symbol table and are defined as variables in the current
>symbol table, skipping those symbols that appear in the argument list
>of the anonymous function.  Would that work?  Would it cause trobule
>for inline functions (which are implemented using an anonymous
>function).
>
>jwe
>
>  
>
I don't see why it would cause trouble for inline functions as long as
the octave_fcn_inline class doesn't call make_anon_fcn_handle, but
rather the
octave_fcn_handle class directly..

On another point I don't think you responded to the proposal of syntax like

function y = myfun(x), y = ...; endfunction
function y = myjac(x), y = ...; endfunction
y = fsolve({ at (x)myfun(x),@(x)myjac(x)}, ...);

to allow function handles and inline functions for fsolve and family,
without the pain of caching the values of the user function I previously
had. Jacobian and constraint functions would be strings in a cell array,
which would be fairly easy to implement.

D.

-- 
David Bateman                                David dot Bateman at motorola dot com
Motorola Labs - Paris                        +33 1 69 35 48 04 (Ph) 
Parc Les Algorithmes, Commune de St Aubin    +33 6 72 01 06 33 (Mob) 
91193 Gif-Sur-Yvette FRANCE                  +33 1 69 35 77 01 (Fax) 

The information contained in this communication has been classified as: 

[x] General Business Information 
[ ] Motorola Internal Use Only 
[ ] Motorola Confidential Proprietary