Mutated data.tables are not printed when enclosed in parentheses.

[eliocamp]

The walrus operator suppresses printing, but calling print explicitly does work.

library(data.table)
dt <- data.table(x = 1)

print(dt[, a := "a"])
#>        x      a
#>    <num> <char>
#> 1:     1      a

However, I woudl've assumed that enclosing the expression in parentheses would also work. It doesn't!

(dt[, a := "a"])

^{Created on 2025-10-05 with reprex v2.1.1}

[jangorecki]

Agree this is how base R normally does.
As a workaround for now you can use

dt[, a := "a"][]

Instead

[ben-schwen]

Isn't (dt[, a := "a"]) just syntactic sugar for dt[, a := "a"]?

[jangorecki]

But it prints (a <- 1)

[aitap]

As far as R is concerned, unlike base R assignments like a <- 1, which are properly invisible, dt[, a := "a"] is not an assignment and thus is visible and must be printed:

withVisible(dt[, a := "a"])
# $value
# 
# $visible
# [1] TRUE

R then calls print() and dispatches to print.data.table(), which contains special code to refuse being auto-printed. To print.data.table, dt[, a := "a"] and (dt[, a := "a"]) look the same when auto-printing.

Unfortunately, we also lack the ability to force the result of [.data.table to be invisible (see #6566 and links from there for more context).

[tdhock]

"This is a feature not a bug!"
we don't want to print for dt[, a := "a"], which silences printing, and is the same as (dt[, a := "a"]). It is normal/expected to have to add [] to get printing.

[jangorecki]

@tdhock but it is inconsistent to <- assignment which normally also does not print, unless wrapped in (

[MichaelChirico]

Isn't this related to #6631? After which adding [] to force print should be more rare?

[eliocamp]

From a user intent perspective, I would assume that (dt[, a := "a"]) indicates that the user wants the result to be printed, like if they did print(dt[, a := "a"]).

From @aitap post before, it seems that it's simply not possible to distinguish between (dt[, a := "a"]) (which should print) and dt[, a := "a"] (which shouldn't print). If that's the case, then feel free to close this issue :).

[ecoRoland2]

From @aitap post before, it seems that it's simply not possible to distinguish between (dt[, a := "a"]) (which should print) and dt[, a := "a"] (which shouldn't print).

Is this documented somewhere?

IMHO, print(dt[, a := "a"]) is much better coding style.

[MichaelChirico]

In fact, implicit_assignment_linter() lints this in the similar common use case of (x <- ...) "assign with print side-effect", though this has some pushback:

r-lib/lintr#2916

It is very common to find this in R's own manuals:

grep -Er "^\(.*" ~/svn/r-devel --include=*.Rd | wc -l
# 337

And yes, it's documented in ?"(":

(... has visibility set, so will auto-print if used at top-level.

[ecoRoland2]

No, I mean if it is documented by data.table that ( doesn't auto-print when using := assignment.

[aitap]

The := assignment making the result "invisible" and requiring [] to make it printable is mentioned in the "Reference semantics" vignette and in a comment in the ?data.table examples.