Data.table: Assignment operator := changes the body of a wrapper function

Created on 18 Sep 2019  ·  4Comments  ·  Source: Rdatatable/data.table

Hello!

Minimal reproducible example:

library(data.table)

f = function(flag = FALSE) {
  dt1 = data.table(a = 1)
  dt2 = data.table(a = 1)
  dt3 = dt1[, .(a, b = 0)]

  print(body(f))
  if (flag) dt3[dt2, b := 999, on = "a"]
  print(body(f))

  return(invisible(dt3))
}

f(flag = FALSE)
f(flag = TRUE)

> f(flag = FALSE)
{
    dt1 = data.table(a = 1)
    dt2 = data.table(a = 1)
    dt3 = dt1[, .(a, b = 0)]
    print(body(f))
    if (flag) 
        dt3[dt2, `:=`(b, 999), on = "a"]
    print(body(f))
    return(invisible(dt3))
}
{
    dt1 = data.table(a = 1)
    dt2 = data.table(a = 1)
    dt3 = dt1[, .(a, b = 0)]
    print(body(f))
    if (flag) 
        dt3[dt2, `:=`(b, 999), on = "a"]
    print(body(f))
    return(invisible(dt3))
}

> f(flag = TRUE)
{
    dt1 = data.table(a = 1)
    dt2 = data.table(a = 1)
    dt3 = dt1[, .(a, b = 0)]
    print(body(f))
    if (flag) 
        dt3[dt2, `:=`(b, 999), on = "a"]
    print(body(f))
    return(invisible(dt3))
}
{
    dt1 = data.table(a = 1)
    dt2 = data.table(a = 1)
    dt3 = dt1[, .(a, b = 999)]
    print(body(f))
    if (flag) 
        dt3[dt2, `:=`(b, 999), on = "a"]
    print(body(f))
    return(invisible(dt3))
}

As you can see, after assigning (:=) a new value to column "b" (when flag is TRUE), the body of "f" function is changed on the line "dt3 = ..." ("b = 0" is replaced by "b = 999"). The result of function run will then be different when executing it with FALSE flag value.

Output of sessionInfo():

R version 3.5.1 (2018-07-02)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 7 x64 (build 7600)

Matrix products: default

locale:
[1] LC_COLLATE=Russian_Russia.1251  LC_CTYPE=Russian_Russia.1251    LC_MONETARY=Russian_Russia.1251
[4] LC_NUMERIC=C                    LC_TIME=Russian_Russia.1251    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] data.table_1.12.3

loaded via a namespace (and not attached):
[1] compiler_3.5.1 tools_3.5.1    yaml_2.2.0
High bug

Most helpful comment

@mattdowle , Thanks!

After further investigations it comes out that there is no need for a wrapper-function.
Here is a concise example for this issue:

library(data.table)
value = 0
dt1 = data.table(a = 1)
dt2 = dt1[, .(a, b = ..value)]
dt2[1, b := 999]
print(value)

All 4 comments

Thanks for the great report! I can also reproduce in v1.12.2 too.

@mattdowle , Thanks!

After further investigations it comes out that there is no need for a wrapper-function.
Here is a concise example for this issue:

library(data.table)
value = 0
dt1 = data.table(a = 1)
dt2 = dt1[, .(a, b = ..value)]
dt2[1, b := 999]
print(value)

Thanks. I found the line at fault yesterday and did wonder that it might not be wrapper only. So it's really good to have an example already. WIP ...

Related #3766. Thanks @kirillmayantsev for not giving up on us!

Was this page helpful?
0 / 5 - 0 ratings