Re: firls.m, part 2

[Nicholas Jankowski]

On Sun, Jun 18, 2017 at 10:55 AM, je suis <address@hidden> wrote:
> Given the answers in the previous threads I assumed there is some
> interest in this, but now that days seem to pass, I have to ask: is
> there anymore left?
>
> Vlad

Well, a number of people are working on a number of their own
projects, it appears that you were point person on this one.  Looking
back through your last request for input was:

"Please check the two scripts for whatever input you can think of."

Well, this isn't a routine that I currently have an immediate
technical use for, so the only inputs I'd have to test are the
examples we've already gone through. With the modified expint, do they
reproduce well? That was an interesting find and might warrant its own
discussion. If there actually is a deficiency in expint, and your
version improves it I think fixing the main one would be preferable to
just having your fix as a subfunction to firls.

I notice there are still no built-in tests at the end of the current
version.  I'd recommend adding a bunch of those for basic input/output
form checking as well as a few (simple if possible) expected numerical
outputs. I think one of my past emails may have included some
examples.

I'm thinking the best thing might now be to create two separate
submissions to the Octave bug tracker, one for firls (if you haven't
done this already) and one for expint at bugs.octave.org. You can link
the github archive and mention the dependency between the two bug
reports.

 I've cc'd Mike, who is maintaining the signals package.  He may have
suggestions on what else is needed for this function prior to
inclusion. There is no single maintainer for the specfun package,
which contains expint, but maybe he has suggestions there.

Nick J.