Re: legendre.m

[Marco Caliari]

Hi Ben.

For me, there is no need for checking under-overflows (even because it 
fails to detect legendre(151,eps). It is similar to

> exp(800)
ans = Inf

No warning.

Best regards,

Marco

On Wed, 15 Oct 2008, Ben Abbott wrote:

> On Oct 15, 2008, at 12:12 PM, Ben Abbott wrote:
>
> > On Wednesday, October 15, 2008, at 08:02AM, "Ben Abbott" <address@hidden> 
> > wrote:
> > > On Oct 15, 2008, at 3:24 AM, Marco Caliari wrote:
> > >
> > > > > | Moreover, what about the new legendre.m function described here
> > > > > | 
> > > > > http://www.nabble.com/normalized-ALF-(Assotiated-Legendre-Function)-td15278073.html#a15278073
> > > > >
> > > > > Please use separate threads for separate topics.
> > > > >
> > > > > I don't have an objection to including it, but I don't think it
> > > > > should
> > > > > unconditionally print a warning every time it is called.
> > > >
> > > > In the final message of that thread, I noticed that the warning was
> > > > misleading and suggested to remove it. I think Ben (who committed
> > > > the patch) did it. Ben?
> > > >
> > > > Marco
> > >
> > > I'll take a look.
> > >
> > > Ben
> >
> > I can confirm that committed version of legendre.m is the one I patched 
> > (with some changes for consistency with Octave's coding practices).
> >
> > I haven't been following along with this thread, but assume there is a 
> > desire to bring back the underflow error. Is that correct?
> >
> > Ben
>
> The attached changeset adds checks for both underflow and overflow and warns 
> on the first occurrence.
>
> Marco, I'm uncertain if you are using mercurial or if you are able to apply 
> patches, so I've also attached the complete legendre.m function.
>
> I also noticed a minor bug and fixed ... it is accompanied by an additional 
> test as well.
>
> Ben