Re: defining functions on-the-fly

[dirk(at)uwast.astro.wisc.edu (Dirk Fabian)]

In article <9eh94p$3ce$1@agate.berkeley.edu>,
Marshall Perrin <mperrin+news@arkham.berkeley.edu> wrote:
>Marc Schellens <m_schellens@hotmail.com> wrote:
>> try .comp:
>> 
>> IDL> .comp [ENTER]
>> - function f
>> - return,42
>> - end
>> % Compiled module: F.
>> IDL> print,f()
>>       42
>> IDL> 
>
>No, this doesn't work for what I have in mind - .comp is an executive command
>and so can only be used interactively. You can't use .comp in a procedure.  I
>want *my software* to be able to define functions on the fly, not myself.  So
>it looks like the best solution really is writing out a new .PRO file to the
>disk and compiling that. 

I faced the same problem a while back, and this writing out solution was
unacceptable as well.  The problem for me was that subsequent iterations
of the program would write a different equation in the .PRO file, but IDL
has already compiled a function by that name and doesn't look at or
recompile the new .PRO file.  Since .comp and .run are executive level,
you can't force it to either.

This is what I did:  (I had to interactively have N gaussians in a 
fitting program)

;prepare expressions for fitting and result plotting 
plotresult = 'model=convol(exp((-1.)*(0.0'
FOR i=0, ncomp-1 DO $
  plotresult = plotresult + ' + Gauss1(v,result['+strtrim(string(3*i), 2)+ $
   ':'+strtrim(string(2+3*i), 2)+'])'
plotresult = plotresult+')), normspreadfunc, /center, /edge_truncate)'

; then 
done = execute(plotresult)

; and you can do
oplot, v, model, color=200, thick=3 


I wasn't clear on why execute wasn't working for you... Perhaps this 
helps.

Cheers, Dirk