[Rd] trouble with package loading: Function found when exporting methods from the namespace ‘raster’ which is not S4 generic: ‘all.equal’

Patrick Giraudoux p@tr|ck@g|r@udoux @end|ng |rom un|v-|comte@|r
Sun Jan 30 18:52:51 CET 2022

Dear listers,

After having exchanged on the R-package-devel using r-project.org list, I feel 
that the question should now be adressed on the R developper list.

You'll find below the trouble I met, and one of the answers that Zhian 
Kamvar has been kind enough to provide.

Il looks like the origin might be found in the way R deploys methods and 
namespace is managed, which is by far out my skills...

Suspect some bug somewhere (?)





One user (in cc) has signaled a problem installing the package pgirmess 
(version 1.7.1 on CRAN), with this message on loading:

> library(pgirmess)

Error: package or namespace load failed for ‘pgirmess’:

Function found when exporting methods from the namespace ‘raster’ which 
is not S4 generic: ‘all.equal’

In addition: Warning message:

no function found corresponding to methods exports from ‘raster’ for: 
‘direction’, ‘gridDistance’

I cannot identify where the trouble comes from since:

- pgirmess goes through OK in all the checks on CRAN

- I can install/remove it with correct loading on my own platform 
(Windows 10)

- pgirmess 1.7.1 does not import 'raster' and the functions mentioned above

The user has installed the last R version on Windows.  However, she gets 
a correct loading only if she loads 'Matrix' before (or 'lme4' - but to 
my knowlege, lme4 actually loads Matrix).

She meets also the same issue with pgirmess 2.0 (not on CRAN yet, see 
https://github.com/pgiraudoux/pgirmess however this version (still on 
test) has deprecated/dropped all functions related to dealing with rasters.

This occurs in R directly (not necessarily within RStudio).

On the web, Iook likes this issue has is already been spotted for some 
other packages (e.g. mapview, tmap, etc.). Checking help-lists I saw 
that installing rtools might solve the problem (e.g. 
however, it seems to me a strange workaround since rtools should not be 
made necessary for lay R user (not developers).




Le 28/01/2022 à 17:35, Zhian Kamvar wrote :
I've run into deep dependencies registering methods that mess things up 
before: https://github.com/thibautjombart/adegenet/issues/308.
AFAICT, it has something to do with the way R deploys methods where they 
can be available if the namespace is available. In this case, it's via 
spdep -> spData -> raster. That being said, I am NOT an expert on this 
and would like to know more if anyone has better insight.

	[[alternative HTML version deleted]]

More information about the R-devel mailing list