[Bioc-devel] generics from Biobase not found during package installation

Hi everyone,
with the new bioC release I encounter a strange problem in my prada package:
I'm using some of the generics from Biobase (e.g. exprs, pData, 
phenoData, description) to build methods for my own objects of class 
cytoFrame (which describe data from FACS measurements).
e.g.
setMethod("description", signature="cytoFrame", 
definition=function(object) object at description, valueClass="character")

This used to work fine with all recent bioC releases, but now I get the 
following error when trying to install prada (and the same thing happens 
on the build test machines in Seattle):

* Installing *source* package 'prada' ...
** R
** inst
** save image
Loading required package: Biobase
Loading required package: tools

Welcome to Bioconductor

        Vignettes contain introductory material.
        To view, simply type 'openVignette()' or start with 'help(Biobase)'.
        For details on reading vignettes, see the openVignette help page.

Loading required package: RColorBrewer
Loading required package: grid
KernSmooth 2.22 installed
Copyright M. P. Wand 1997

** preparing package for lazy loading
KernSmooth 2.22 installed
Copyright M. P. Wand 1997
Error in methods:::mlistMetaName(mi, ns) :
        the methods object name for 'description' must include the name 
of the package that contains the generic function, but there is no 
generic function of this name
Execution halted
ERROR: lazy loading failed for package 'prada'

Obviously R can't find the generics from Biobase although it is 
attached. As a quick fix I tried to redefine those "missing" generics in 
my own package. By doing so I get rid of the error message during 
installation but I also break all functionality from Biobase. When 
adding require(Biobase) in my R code before setting the new methods this 
evaluates to TRUE but isGeneric(description) evaluates to FALSE during 
installation.
When starting R interactively an after attaching Biobase I can set my 
new methods without any problem simply by pasting in the code.

Does anybody have any idea what might be going wrong here?

My system setup is i686-pc-linux-gnu running with R 2.2.0 release and a 
fresh installation of bioC 1.7 release (Biobase 1.8.0, prada 1.6.1)

Thank's in advance,
Florian

Hi Florian,
  I suspect that the problem is that you will need to make use of 
.onLoad. One of the tricky bits with the "save image" technology is 
that any methods that you want to associate with generics from another 
package must have the setMethod code run each time the package is 
loaded (or attached). For any generics you define we save time by 
doing the association at package build time (because it cannot change).

  There is one example in Biobase (and I think some more substantial 
ones in iSPlot),

  Robert


Florian Hahne wrote:

Hi everyone,
with the new bioC release I encounter a strange problem in my prada 
package:
I'm using some of the generics from Biobase (e.g. exprs, pData, 
phenoData, description) to build methods for my own objects of class 
cytoFrame (which describe data from FACS measurements).
e.g.
setMethod("description", signature="cytoFrame", 
definition=function(object) object at description, valueClass="character")

This used to work fine with all recent bioC releases, but now I get 
the following error when trying to install prada (and the same thing 
happens on the build test machines in Seattle):

* Installing *source* package 'prada' ...
** R
** inst
** save image
Loading required package: Biobase
Loading required package: tools

Welcome to Bioconductor

        Vignettes contain introductory material.
        To view, simply type 'openVignette()' or start with 
'help(Biobase)'.
        For details on reading vignettes, see the openVignette help 
page.

Loading required package: RColorBrewer
Loading required package: grid
KernSmooth 2.22 installed
Copyright M. P. Wand 1997

** preparing package for lazy loading
KernSmooth 2.22 installed
Copyright M. P. Wand 1997
Error in methods:::mlistMetaName(mi, ns) :
        the methods object name for 'description' must include the 
name of the package that contains the generic function, but there is 
no generic function of this name
Execution halted
ERROR: lazy loading failed for package 'prada'

Obviously R can't find the generics from Biobase although it is 
attached. As a quick fix I tried to redefine those "missing" generics 
in my own package. By doing so I get rid of the error message during 
installation but I also break all functionality from Biobase. When 
adding require(Biobase) in my R code before setting the new methods 
this evaluates to TRUE but isGeneric(description) evaluates to FALSE 
during installation.
When starting R interactively an after attaching Biobase I can set my 
new methods without any problem simply by pasting in the code.

Does anybody have any idea what might be going wrong here?

My system setup is i686-pc-linux-gnu running with R 2.2.0 release and 
a fresh installation of bioC 1.7 release (Biobase 1.8.0, prada 1.6.1)

Thank's in advance,
Florian

Hi Florian,

I still don't understand why I don't see the error everywhere.  I can
only reproduce with our automted build account.  Everywhere else I
try, prada builds and checks just fine (prior to your changes).

I just did the following experiment using the build account where I do
see the failure on accessing the "description" generic:

Added to NAMESPACE the following line:

import("Biobase")

After that change, I was able to build without error.  I understand
what Robert is getting at with respect to SaveImage and what can and
cannot be done at build and/or install time vs load time.  I don't
know exactly what the NAMESPACE stuff is doing and perhpas it is
changing the behavior of SaveImage?

Can you see if simply modifying NAMESPACE to import Biobase solves the
problem on your systems?

Best,

+ seth

Hi Florian,

I still don't understand why I don't see the error everywhere.  I can
only reproduce with our automted build account.  Everywhere else I
try, prada builds and checks just fine (prior to your changes).

I just did the following experiment using the build account where I do
see the failure on accessing the "description" generic:

Added to NAMESPACE the following line:

import("Biobase")

After that change, I was able to build without error.  I understand
what Robert is getting at with respect to SaveImage and what can and
cannot be done at build and/or install time vs load time.  I don't
know exactly what the NAMESPACE stuff is doing and perhpas it is
changing the behavior of SaveImage?

Can you see if simply modifying NAMESPACE to import Biobase solves the
problem on your systems?

Best,

+ seth

Once this issue is settled for good would it be possible to push the
patched package to the release branch??