The {marginaleffects} case in #6575 is pretty hard to tease apart, but I found this example that could be related where 1.16.4 and current master differ in behavior:
baz = function(x) setDT(x)
foo = function(x) {
bar = function() baz(x)
x = data.frame(a=1)
bar()
class(x)
}
foo()
## 1.16.4
# [1] "data.table" "data.frame"
## current master
# [1] "data.frame"From debugging of {marginaleffects} thus far, I only found that, in contrast to #6701 and #6725, {marginaleffects} (in example(comparisons) where R CMD check is failing, at least) only calls simply setDT(<name>).
The {marginaleffects} case in #6575 is pretty hard to tease apart, but I found this example that could be related where 1.16.4 and current
masterdiffer in behavior:From debugging of {marginaleffects} thus far, I only found that, in contrast to #6701 and #6725, {marginaleffects} (in
example(comparisons)whereR CMD checkis failing, at least) only calls simplysetDT(<name>).