Skip to content

Submit a package which use doParallel

6 messages · Brian G. Peterson, Duncan Murdoch, Etienne Lord +1 more

#
Hi,

I'm trying to submit my first package which depends on doParallel:

Depends: R (>= 3.0), igraph, doParallel

Running hadley devtools: devtools::check() and devtools::release() result
in no problem (no ERROR nor NOTE on Linux, Mac and Windows).

However, when in use the devtools::build_win() command, it results in the
following note:

* checking R code for possible problems ... NOTE

complete_network: no visible global function definition for '%dopar%'
complete_network: no visible global function definition for '%:%'
complete_network: no visible global function definition for 'foreach'
sample_network: no visible global function definition for '%dopar%'
sample_network: no visible global function definition for 'foreach'

Googling for " no visible global function definition for '%dopar%' " result
in a few CRAN packages with
similar note e.g.
http://www.r-project.org/nosvn/R.check/r-devel-linux-x86_64-fedora-clang/penDvine-00check.html

1) Is there a special way to include the doParallel package so that this
error do not occur?

2) Installing the doParallel package "before" running any test on windows
seems to remove this note. Is there any way to specify it in the
DESCRIPTION?

Best regards,

Have a nice day.
#
On Thu, 2015-03-05 at 18:03 -0500, Etienne Lord wrote:
add foreach to your Depends.  That should resolve the error you're
seeing.
<...>

  
    
#
On 05/03/2015 6:03 PM, Etienne Lord wrote:
It's much better to import what you need.  If someone calls one of your
functions using :: notation, it will fail, because it won't know where
to find the doParallel objects.
List it in the DESCRIPTION as "Imports:  doParallel" and in your
NAMESPACE file import the functions you need.
That's a problem with testing within a session.  It's best to run checks
on a vanilla system, using R CMD check.

Duncan Murdoch
#
On 05/03/2015 6:16 PM, Brian G. Peterson wrote:
Please don't do that.  Use Imports.

Duncan Murdoch
#
Thanks for the quick reply.

Adding all the dependencies of doParallel (foreach, parallel, iterators) in
the DESCRIPTION and in import statements in NAMESPACE resolved the
build_win problems. Don't know why this is required for Windows build.

Thanks again.

2015-03-05 18:22 GMT-05:00 Duncan Murdoch <murdoch.duncan at gmail.com>:

  
    
#
On Mar 5, 2015 5:01 PM, "Etienne Lord" <m.etienne.lord at gmail.com> wrote:
in
Most likely not a Windows specific issue,. Instead it could be because the
latter uses http://win-builder.r-project.org/ and that uses newer versions
of R than what you use locally. I'd guess you get the same NOTEs if you
test with R devel (now 3.2.0rc) locally and maybe even R 3.1.2 patched.

Henrik