[R] nls and R scoping rules
luke at stat.uiowa.edu
Thu Jun 10 14:39:11 CEST 2004
On Thu, 10 Jun 2004, Prof Brian Ripley wrote:
> On Thu, 10 Jun 2004, Prof Brian Ripley wrote:
> > Around R 1.2.x the notion was introduced that variables should be looked
> > for in the environment of a formula. Functions using model.frame got
> > converted to do that, but nls did not. I guess that the best way forward
> > is to ensure that nls (and nlsModel) does search the environment of the
> > formula for functions.
> It transpires that is rather easy to achieve. At the top of nlsModel and
> nlsModel.plinear use
> env <- new.env(parent=environment(form))
> instead of
> env <- new.env()
> This then automatically searches for objects used in the formula
> 1) nls is in a namespace, so you need to fix the copy in the namespace or
> fix the source code and rebuild.
> 2) This will add to the baggage needed when you save() a nls fit in a
> workspace. I think that is inevitable as we cannot just identify the
> funtions used in the formula (as they might call other functions in the
> local environment), and it is necessary to capture the objects needed to
> evaluate the formula for the predict() method to be reliable.
> We could call all.names() to get the names of functions used directly in
> the formula, but that seems not good enough (previous para).
> Question to the cognescenti: is the price in 2) too great for this to be
> done for 1.9.1? I will put the change in R-devel for now.
I don't think so. The fit objects already contain the formula (via
the environment of the functions in the structure). Since environment
sharing is preserved within a serialization, this means the change
would only alter the parent of the env environment, not add anything
University of Iowa Phone: 319-335-3386
Department of Statistics and Fax: 319-335-3017
241 Schaeffer Hall email: luke at stat.uiowa.edu
Iowa City, IA 52242 WWW: http://www.stat.uiowa.edu
More information about the R-help