Possible setClassUnion Unloading Issue

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

Possible setClassUnion Unloading Issue

R devel mailing list
I'm encountering a problem caused by class unions from unloaded
packages that are referenced by the still-loaded subclasses.  In
essence, when the class union is loaded initially, it registers itself
as a super class of the component classes in the S4 cache.  When the
package the class union lives in is unloaded, it does not undo that
registration.  Here is an MRE of the problems it causes:

     ## the character and NULL basic types, and install it
     testpkg <- tempfile()
     dir.create(file.path(testpkg, 'R'), recursive=TRUE)
     writeLines('', file.path(testpkg, 'NAMESPACE'))
     description <- 'Package: testpkg\nTitle: Textpkg\nVersion: 0.1\n'
     writeLines(description, file.path(testpkg, 'DESCRIPTION'))
     code <- 'setClassUnion("chrOrNULL", c("character", "NULL"))'
     writeLines(code, file.path(testpkg, 'R', 'union.R'))
     install.packages(testpkg, repos=NULL, type='src')

     ## Load the package and check validity of a character object
     library('testpkg')
     validObject('hello')
     ## [1] TRUE

     ## Detach, unload, and confirm that superclass is still attached
     ## to character
     detach('package:testpkg', unload=TRUE)
     getClassDef('character')
     ## Class "character" [package "methods"]
     ##
     ## No Slots, prototype of class "character"
     ##
     ## Extends: "vector", "data.frameRowLabels", "SuperClassMethod",
"chrOrNULL"
     ##
     ## Known Subclasses:
     ## Class "signature", from data part
     ## Class "className", from data part
     ## Class "ObjectsWithPackage", from data part

     ## `validObject` now fails:
     validObject('hello')
     ## Loading required package: testpkg
     ## Error in validObject("hello") :
     ##   invalid class "character" object: superclass "chrOrNULL" not
defined
     ##   in the environment of the object's class

     ## But if we run it again it works
     validObject('hello')
     ## [1] TRUE

There are two issues:

1. Whether the dangling superclass association in the `character` (and
    `NULL`) classes after the detach/unload is desirable.
2. That the attempt to reload the package to retrieve the class
    definition fails after the first call to `validObject`, though not
    the second.

I can see how it may be desirable to keep true superclasses associated
with objects, even if somehow the package that contains them is
unloaded.  The case for class unions _seems_ much weaker. I should not
need to reload my package to check validity of something like
'character'.  In my actual use case I had a class that contained
'character' among other things that I was checking with
`validObject(..., complete=TRUE)`.

I attach a patch that adds a cleanup step to `cacheMetaData` which is
kicked off by the package unload process. I grant I'm not an expert in
these matters and there could very well be an important use case where
removing the references is undesirable.

This brings us to the second issue: even if the first patch is
undesirable, `validObject` fails while attempting to reload the class
definition.  I believe the error that occurs when we try to run
`validObject` the first time happens in the `validObject` ->
`getClassDef` -> `.requirePackage` set of calls.  When
`.requirePackage` does not find our package already loaded, it loads,
attaches it, and returns the _package_ environment.  This is different
to what happens when the package namespace is already loaded, in which
case the _namespace_ environment is returned.

The problem with the _package_ environment is that any unexported
class definitions will not be found there.  This is why `validObject`
fails the first time, but works the second.  The first time the
package is loaded and `getClassDef` is given the package env to look
in, but it can't find 'chrOrNULL' because it is not exported.  The
second time `.requirePackage` finds the namespace from the prior load
and returns that, and this does have the class definition.

Since `.requirePackage` is used much more broadly than
`cacheMetaData`, I am reluctant to submit a patch without feedback
from those more knowledgeable of the inner workings of S4.

I ran `make check-all` on a recent version of R-devel with the patch
applied, and everything came back OK.  However, I'm guessing there
isn't much natural testing of unloading packages, so this probably
requires extra scrutiny.

Best,

Brodie


______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

methods-patch.txt (1K) Download Attachment