Re: ode45 gives solution in the "Dark Side," but not in Octave

[Thomas Treichl]

John B. Thoo schrieb:
> Hi, Thomas.  The function "heaviside" is part of SpecFun package:
>
> <http://octave.sourceforge.net/doc/funref_specfun.html>
>
> I will try the other solvers you suggest.  Thanks.
>
> ---John.

Ok, I was able to start your script - thanks for the hint. You're using fixed 
time steps (tspan) with the ode45 solver which is absolutely okay, but the next 
argument you use (w0) already is complex even before ode45 can do its first 
step, ie. the initial values already are complex. I don't expect that these 
values should be complex?! Here is what I see if I remove the semicolon at the 
end of the line w0=...

octave-3.2.0:1> run_twoway
  w0 =
    -0.537882842739990 + 0.000000000000000i
    -0.537882789290712 + 0.000000000000000i
    -0.537882628942835 + 0.000000000000000i
    -0.537882361696230 + 0.000000000000000i
    -0.537881987550684 - 0.000000000000000i
    -0.537881506505897 - 0.000000000000000i
    -0.537880918561486 + 0.000000000000000i
  <SNIP>

I think that this is the problem you have - ode45 thinks your problem is complex 
and then tries to solve your complex problem, which it is able to if the complex 
problem makes sense... So the origin of this problem seems to be somewhere else 
in your initialization routines. Can you confirm?

Best regards
  Thomas

PS. Just because of interest why you don't have ode5r, did you install OdePkg 
with the 'pkg' command or did you just unpack the *.tar.gz into some directory? 
ode5r should be there if you installed OdePkg (it is globally installed on my 
system but it can also be somewhere in your local home directory)

octave-3.1.55:7> which ode5r
`ode5r' is a function from the file 
/Users/Thomas/bin/Octave.app.3.2.0/Octave.app/Contents/Resources/libexec/octave/packages/odepkg-0.6.7/i386-apple-darwin8.11.1-api-v37/dldsolver.oct