Re: high resolution floating point

[Carnë Draug]

On 4 February 2014 07:57, CdeMills <address@hidden> wrote:
> I looked at the "vpa" implementation in symbolic:
> 1) it is based upon CLN and Ginac, CLN using gmp for multiresolution integer
> 2) CLN does not compile under cygwin, there seems to be changes in the gmp
> API
>
> My plan is to take inspiration from vpa, with the substitution of MPFR to
> Ginac and CLN. Would it be OK for the multi-resolution package to shadow the
> vpa found in symbolic ? On the long run, we could drop the vpa support from
> symbolic, or decide that only one implementation could be active in an
> Octave session.

I really don't think that's a good idea. We should not be shadowing
functions of other packages (exception made for packages whose purpose
is to actually do it such as NaN). That's what dependencies are for.
And the vpa function belongs to the symbolic package (following the
logic that packages will mimic the functions in Matlab's toolboxes) so
it should not be moved from one place to other.

Now, I understand that the symbolic package is unmaintained and
severely broken. If nothing else besides vpa works, just remove them
(hey! we have a revision control system, let's use it) and make a
release with what works if you need it.

Carnë