I'm using R's object system to create a modify objects with a number of functions. I am not sure the best way to do it for either S3 or S4 systems, or if there's any reason to prefer one over the other.
For example:
letters <- function(a) {
rtn <- list('a'=a)
class(rtn) <- 'letters'
rtn
}
b <- function(letobj, b) {
letobj$b <<- b
}
Will get the job done. But surely this isn't the preferred way to do this in S3. And what's the equivalent in S4? And which is preferred?
Questons soliciting opinion based answers are off topic on SO; however, this is how the code shown in the question would normally be written using S3. The key point is that we return the object in b and then the caller can set L to the modified version.
lets <- function(a) structure(list(a = a), class = "letters")
b <- function(x, b) { x$b <- b; x}
L <- lets("A")
L <- b(L, "B")
L
giving:
$a
[1] "A"
$b
[1] "B"
attr(,"class")
[1] "letters"
or we can use a replacement function:
"b<-" <- function(x, value) { x$b <- value; x }
L <- lets("A")
b(L) <- "B"
L
Alternately don't define b() at all and just use:
L$b <- "B"
If you want to support inheritance then we need to define a generic and then the method.
b <- function(x, b, ...) UseMethod("b") # b generic
b.letters <- function(x, b, ...) { x$b <- b; x) # b's letters method
Now it would be possible to define additional b methods for other classes.