Implement nest_join

[hadley]

A nest_join() would create a list-column:

library(tibble)

df1 <- tibble(x = c(1, 2), y = c(2, 3))
df2 <- tibble(x = c(1, 1), z = c(2, 3))

# nest_join(df1, df2)
# Would yield:
out <- df1
out$df2 <- list(df2, NULL)
out
#> # A tibble: 2 x 3
#>       x     y df2             
#>   <dbl> <dbl> <list>          
#> 1     1     2 <tibble [2 × 2]>
#> 2     2     3 <NULL>

The output would always have exactly the same rows as the input, but with one new column, named (by default) with the name of y. unnest()ing this data frame would yield an inner_join() (or a left_join() with tidyverse/tidyr#358)

@romainfrancois how hard would this be to implement?

[romainfrancois]

I think we have most of the internal tools that we need. Would the nested tibbles have the x column in the example or just z ?

[hadley]

Hmmmm, good question; for now, I'd say let's drop the join columns from the nested column — that way unnest() will create the same thing as an inner join.

[romainfrancois]

That makes it easier to test it. I’ll 🤹‍♂️ with this, should be easy enough.

[romainfrancois]

Is NULL ok or should it be a 0 rows data frame of the right format ?

[hadley]

0-row data frame I think.

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/