df2 <- df <- data.frame(a=1) can have different columns after setDT

[franknarf1]

Cath brought this up in SO chat:

library(data.table)
df2 <- df <- data.frame(a=1)
setDT(df)[, b := 2]
df2
# the column was not added
#    a
# 1: 1

dt2 <- dt <- data.table(a=1)
dt[, b := 2]
dt2
# the column was added
#    a b
# 1: 1 2

It looks like the issue is that after setDT, address(df) != address(df2). df2 stayed at the original address while df moved to a new one...? I'm not sure if it's a problem, but it makes the copy-vs-reference thing even harder to explain, I guess.

[franknarf1]

@DavidArenburg pointed out that this is part of alloc.col's behavior. Assuming it's deliberate, maybe the setDT docs could use a note about the <- case along the lines of "weird things happen if you have other pointers to the table, so be wary." I think this has come up before in the context of passing DFs to a function that has setDT inside -- it will set the passed DF to DT, but not add columns, etc., based on the code inside the function.

f = function(DF) setDT(DF)[, b := 2]
f(myDF <- data.frame(a = 1))
myDF
# intuitively, one might expect a "b" column here:
#    a
# 1: 1

Related #1628