R-alpha: christmas wish (partial support for single)

Friedrich Leisch Friedrich.Leisch@ci.tuwien.ac.at
Tue, 11 Nov 1997 10:33:23 +0100

The following is a re-post (after suggestion from martin) of a mail I
sent to r-core this morning, maybe some of you are interested:

>>>>> On Tue, 11 Nov 1997 08:34:16 +0100,
>>>>> Kurt Hornik (KH) wrote:

KH> As Andreas & Fritz & I have recently, in attempts to port various pkgs
KH> to R, come across the f2c problem again, I thought I'd ask ...

KH> Some time ago, someone (Ross?) suggested that we'd have an f2c add-on
KH> package.  Is this still in the making somewhere?

If we talk about porting packages to R ... well, christmas is
approaching rapidly, so it's that time of the year again where people
think a lot about wishes :-)

The only effort that's involved when porting most pkg's is that R
doesn't support single precision ... well I don't need single
precision through the whole thing, but perhaps .Fortran could be
enhanced to support it? 

I think of something like single() (and friends) attaching an
attribute single to a numeric variable, and if .Fortran detects it, it
converts the stuff internally to single before passing it to the
subroutine. We could also introduce mode "single" which is exactly the
same as "numeric", but .Fortran respects it.

... just an idea, there are probably 1000 other ways of doing it
without introducing single for all things ... but it would ease
porting packages considerably for people who are not used to
programming Fortran or C. I do speak C, but still I'm always afraid of
breaking some code when changing all those real definitions to double

That's not urgent, but maybe we could start thinking about it.


r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-devel-request@stat.math.ethz.ch