n_distinct() on sqlite with more columns returns an error

@edoardomichielon commented on Jun 20, 2018, 11:10 AM UTC:

---

When I use a connection to a sqlite on disk, the verb n_distinct() returns an error if there are two or more columns. Same code, if I select just one column or collect data before counting record, it works properly.

# remove objects
rm(list =  ls())

# require packages
require(dplyr)
require(dbplyr)
require(RSQLite)

# create a sqlite db and connect to it
con <- DBI::dbConnect(RSQLite::SQLite(), ":memory:")

# copy data into sqlite
copy_to(con, mtcars)

# point to table 
my_tbl <- tbl(con, "mtcars")

# n_distinct with one column (THIS WORKS)
my_tbl %>% group_by(gear) %>% summarise(n_distinct(mpg))

# if I use the local (or collected) data it is ok (THIS WORKS)
my_tbl %>% collect() %>% group_by(gear) %>% summarise(n_distinct(mpg, cyl))

# n_distinct with two columns (THIS DOES NOT WORKS)
my_tbl %>% group_by(gear) %>% summarise(n_distinct(mpg, cyl))

## Error in result_create(conn@ptr, statement) : 
## wrong number of arguments to function COUNT()

The code is correctly translated into Sql

# Show query
my_tbl %>% group_by(gear) %>% summarise(n_distinct(mpg, cyl)) %>% show_query()

## SELECT `gear`, COUNT(DISTINCT `mpg`, `cyl`) AS `n_distinct(mpg, cyl)`
## FROM `mtcars`
## GROUP BY `gear`

This issue was moved by batpigandme from tidyverse/dplyr/issues/3687.

[hadley]

Minimal reprex:

library(dplyr, warn.conflicts = FALSE)

mf <- dbplyr::memdb_frame(x = c(1, 1), y = c(2, 2), z = 1:2)
mf %>% group_by(x) %>% summarise(n_distinct(y, z))
#> Error in result_create(conn@ptr, statement): wrong number of arguments to function COUNT()

^{Created on 2019-01-02 by the reprex package (v0.2.1)}

[Jollywatt]

Why was this closed as completed? 1b58b68 simply makes dbplyr return a different error message. Multi-column n_distinct() is still not supported...

[hadley]

Because COUNT() only supports a single argument...