Different results of a full_join in arrow and dplyr

Question

I get different results when using full_join on tibble and on arrow_table. Maybe somebody can give a hand on what is going on?

library(arrow)
library(dplyr)

xa1 <- arrow_table(x = 1L)
xa2 <- arrow_table(x = 2L)

x1 <- tibble(x = 1L)
x2 <- tibble(x = 2L)

full_join(xa1,xa2,on = c("x")) %>%  collect() %>% compute()
full_join(x1,x2)

# A tibble: 2 × 1
x
<int>
1     1
2    NA
full_join(x1,x2)
Joining, by = "x"
# A tibble: 2 × 1
x
<int>
1     1
2     2

Answer 1

There is no on argument in dplyr::.*_join. Usage according to ?dplyr::full_join is

full_join( x, y, by = NULL, copy = FALSE, suffix = c(".x", ".y"), ..., keep = NULL )

on is a data.table join argument. We need by here

library(arrow)
library(dplyr)
full_join(xa1, xa2, by = "x") %>%
     collect() %>% 
     compute()

-output

# A tibble: 2 × 1
      x
  <int>
1     1
2     2

By looking at the methods and source code

> methods("full_join")
[1] full_join.arrow_dplyr_query* full_join.ArrowTabular*      full_join.data.frame*        full_join.Dataset*           full_join.RecordBatchReader*
> getAnywhere(full_join.ArrowTabular)
function (x, y, by = NULL, copy = FALSE, suffix = c(".x", ".y"), 
    ..., keep = FALSE) 
{
    query <- do_join(x, y, by, copy, suffix, ..., keep = keep, 
        join_type = "FULL_OUTER")
    if (!keep) {
        query$selected_columns <- post_join_projection(names(x), 
            names(y), handle_join_by(by, x, y), suffix)
    }
    query
}

by is used in the functions that are called inside