[Rd] FW: [Q] Package NS Hook Lint NOTE (unanswered)

Roebuck,Paul L proebuck at mdanderson.org
Mon Sep 19 19:30:57 CEST 2011

Still have a question about the following NOTEs appearing on CRAN checks:

> .onLoad calls:
> require(methods)
> Package startup functions should not change the search path.
> See section ŒGood practice¹ in ?.onAttach.

For years, I have had the following in my S4-requiring packages.

.onLoad <- function(libname, pkgname) {
    ## In case namespace is loaded (via import) by package that
    ## doesn't depend on S4 methods and used in a session with
    ## non-default set of packages

It was all based on an explanation Prof. Ripley gave to someone
about some package loading corner case (reference lost to history
though circa R-2.7 or before). Is it no longer an issue?

The 'methods' package is (in all cases) already listed in each package's
DESCRIPTION Depends entry. Under normal circumstances, is the require()
method in the .onLoad() actually altering the search path (since the
'methods' package was already a dependency)?

More information about the R-devel mailing list