Is `<-` also for data.frames using call by reference sometimes?

I have a simple problem that I wanted to solve using data.table. I was surprised about the following behaviour, as I thought, asignements in base R are always copying:

library(data.table)
df <- data.frame(
  t = 1:10,
  x = "x",
  y = "y"
)
df$z <- df$y # I was assuming this is a full blown copy (since done in base R)
setDT(df)
df[t>=5, `:=`(z=x)]
df

So in the end I want to have for the column z copy of x in case t>=5 and a copy of y otherwise. However, y is also changed, which I find surprising. What is the reason for that?

     t x y z
 1:  1 x y y
 2:  2 x y y
 3:  3 x y y
 4:  4 x y y
 5:  5 x x x
 6:  6 x x x
 7:  7 x x x
 8:  8 x x x
 9:  9 x x x
10: 10 x x x

>Solution :

data.table has its own way of doing things; which you discovered one example of, trying to save on memory because at the point where the setdf happens the columns are pointing to a single source.
To get around this issue, you might choose to create z in the data.table part ;

heres an example with chaining

library(data.table)
df <- data.frame(
  t = 1:10,
  x = "x",
  y = "y"
)

setDT(df)[, z := copy(y)][t>=5, z := x]
df