[R] `case_when()` sometimes doesn't find existing objects in the environment

[etiennebacher]

Describe the bug, including details regarding any error messages, version, and platform.

When running the example below foo.R with source("foo.R", local = new.env()), the object pattern is not found by arrow while it is defined in this new environment. It seems arrow doesn't search the objects in the correct environment when using case_when().

The same code using dplyr instead of arrow works fine. The arrow code also works fine interactively or with source() only.

foo.R:

library(arrow)
library(dplyr)

dat <- arrow_table(x = c("a", "b", "c"))
pattern <- "a"
dat |>
  mutate(x2 = case_when(pattern == "a" ~ "foo")) |>
  collect()

---

R 4.5.0
arrow 19.0.1.1

Component(s)

R

[thisisnic]

Thanks for reporting this @etiennebacher. Adding a reprex output here; executing this code in my R session gets the correct results, but when generating the reprex (with rexprex::reprex()) the results show the bug you mention.

library(arrow)
library(dplyr)

code <- '
arrow_df <- arrow_table(x = c("a", "b", "c"))

arrow_df |>
  mutate(x2 = case_when(pattern == "a" ~ "foo")) |>
  collect()
'

tf <- tempfile(fileext = ".R")
writeLines(code, tf)
source(tf, local = new.env())
#> Error in `case_when()`:
#> ! object 'pattern' not found

[thisisnic]

My best guess is that this line is the source of the error: https://github.com/apache/arrow/blob/main/r/R/dplyr-funcs-conditional.R#L116

Would you be interested in submitting a PR?

[etiennebacher]

Yes I can take a look later and let you know if I need help for this.

[etiennebacher]

Note that in your reprex, you never defined pattern in code (it should still error even after defining it though).

[thisisnic]

Oops, my bad, it was a trimmed down version of a bigger reprex, thanks for highlighting that!

Confirmed the same behaviour:

library(arrow)
library(dplyr)

code <- '
pattern = "a"

arrow_df <- arrow_table(x = c("a", "b", "c"))

arrow_df |>
  mutate(x2 = case_when(pattern == "a" ~ "foo")) |>
  collect()
'

tf <- tempfile(fileext = ".R")
writeLines(code, tf)
source(tf, local = new.env())
#> Error in `case_when()`:
#> ! object 'pattern' not found

[thisisnic]

Issue resolved by pull request 46667
#46667