On 11-09-19 4:48 PM, Stephanie M. Gogarten wrote:
I am trying to build a package (GWASTools, submitted to Bioconductor)
that uses the "sandwich" package.  I have references to "sandwich" in
DESCRIPTION:
Imports: methods, DBI, RSQLite, sandwich, survival, DNAcopy

and NAMESPACE:
import(sandwich)

In the code itself is a call to vcovHC:
Vhat<- vcovHC(mod, type="HC0")

I have sandwich version 2.2-7 installed.

When I run R CMD check on my package, I get the following error during
checking of examples:
* using R Under development (unstable) (2011-08-29 r56828)
* using platform: x86_64-apple-darwin9.8.0 (64-bit)
...
* checking examples ... ERROR
...
Error in estfun.glm(x) : could not find function "is.zoo"
Calls: assocTestRegression ... meatHC ->  rowMeans ->  is.data.frame ->
estfun ->  estfun.glm

I import sandwich, sandwich depends on zoo, but a function in zoo cannot
be found during execution.

I tried to get around this by explicitly including a "require" statement
in assocTestRegression:
require(sandwich)

The example now runs, but I have a warning in R CMD check:
* checking for unstated dependencies in R code ... WARNING
‘library’ or ‘require’ call not declared from: ‘sandwich’

I am not sure why this is a problem, as there is a "require" statement
(for a different package) in another function which does not cause any
warnings.

Can anyone advise on how to get my package to successfully pass R CMD check?


If you said that your package Depends on sandwich, you'd probably pass checks, but this is arguably a bug in the sandwich package.

The problem is that since you only import sandwich, it never gets attached. So it never attaches its dependency zoo. It should import zoo (or import the functions in zoo that it uses) to work properly the way you are calling it.

You can see the same error as follows:

1.  Install sandwich, but don't attach it or zoo.

2.  Run

example(glm)

to create the "glm" object glm.D93.

Try to call

sandwich::estfun(glm.D93)

You'll get the same error as before because is.zoo() will not be loaded.

It's also arguably a design flaw in R. For testing sandwich, zoo would generally end up being attached, because the testing of sandwich would attach it. However, examples like the one above are never tested.

Duncan Murdoch

______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

Reply via email to