Unexpected behavior with inheritance and using S3 classes as slots in S4 class

Ezra,

I think it's just the fact the you specified the wrong class inheritance in setOldClass() - it has to match you S3 definition, so it should be:

setOldClass(c("b","a"))

In which case it works:

n(s1 = b_1)

An object of class "n"

Slot "s1":

[1] "world"

attr(,"class")

[1] "b" "a"

Cheers,

Simon

On Jan 28, 2021, at 15:55, Ezra Tucker ezra at landtucker.com wrote:

Hi all,

I have a situation where I'm trying to use S3 classes as slots in an S4 class and am running into some problems with inheritance.

My example:

with S3 classes
---------------

a <- function(x) structure(x, class = "a")

b <- function(x) structure(x, class = c("b", "a"))

setOldClass(c("a", "b"))

a_1 <- a("hello")

b_1 <- b("world")

n <- setClass("n", slots = c(s1 = "a"))

n(s1 = a_1) # works as expected

is(b_1, "a") # yields TRUE

n(s1 = b_1) # doesn't work-- unexpected

Error message is:

Error in validObject(.Object) :

invalid class ?n? object: invalid object for slot "s1" in class "n": got class "b", should be or extend class "a"

Digging in, validObject looks like it's calling getClassDef, which gives me

getClassDef("b")

Virtual Class "b" [in ".GlobalEnv"]

Slots:

Name: .S3Class

Class: character

Extends: "oldClass"

-- it extends "oldClass" but it doesn't extend class "a" despite the fact the S3 version does.

Now, In this simple example, I could have defined a and b as s4 classes, but in the real world, I didn't define them and have to use setOldClass.

I'm currently getting around this issue by doing setClassUnion("ab", c("a", "b")) and then specify to use "ab" as the type for the slot, but since I already know b inherits from a, this ought not to be necessary (and there are other reasons this is undesirable).

1.  Is there anything I missed, in setting up this simple example, that would cause n(s1 = b_1) to work properly? Or is there anything I should do other than setClassUnion?
2.  Is this a shortcoming of validObject or getClassDef?

Thank you all!

-Ezra

[[alternative HTML version deleted]]

R-devel at r-project.org mailing list