add_count overrides a column named n when counting a different column

[gsimchoni]

Hi,

I think this relates to #3838 #4076 #4284 , but none of them describes this unexpected behavior which changed and broke my code when I moved to version 0.8.3 (latest on CRAN):

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
tibble(n = c(1,1,1,2,3)) %>% add_count(n)
#> # A tibble: 5 x 2
#>       n    nn
#>   <dbl> <int>
#> 1     1     3
#> 2     1     3
#> 3     1     3
#> 4     2     1
#> 5     3     1
tibble(a = c(1,1,1,2,3), n = c(1,2,2,3,4)) %>% add_count(a)
#> # A tibble: 5 x 2
#>       a     n
#>   <dbl> <int>
#> 1     1     3
#> 2     1     3
#> 3     1     3
#> 4     2     1
#> 5     3     1

^{Created on 2019-08-05 by the reprex package (v0.3.0)}

I understand why in #3838 Winston suggested for count():

to name the column to nn only if there's another column n that would be present in the output

The problem is that with add_count() the user relies on an added column of counts. The first example shows how it adds nicely a nn column when it counts a n column. The second example shows how it overrides an existing non-related n column, when the user specifically asked to add a count column, not to override. This was different in previous versions.

Obviously I can not call that column n or pass an argument through name=:

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
tibble(a = c(1,1,1,2,3), n = c(1,2,2,3,4)) %>% add_count(a, name = "foo")
#> # A tibble: 5 x 3
#>       a     n   foo
#>   <dbl> <dbl> <int>
#> 1     1     1     3
#> 2     1     2     3
#> 3     1     2     3
#> 4     2     3     1
#> 5     3     4     1

^{Created on 2019-08-05 by the reprex package (v0.3.0)}

But it still breaks code, and it gets cumbersome for example when you want to chain counts, like so:

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
tibble(a = c(1,1,1,2,3), b = c(1,2,2,3,4)) %>% count(a, b) %>% add_count(a)
#> # A tibble: 4 x 3
#>       a     b     n
#>   <dbl> <dbl> <int>
#> 1     1     1     2
#> 2     1     2     2
#> 3     2     3     1
#> 4     3     4     1

^{Created on 2019-08-05 by the reprex package (v0.3.0)}

Why would it override the first n?

[romainfrancois]

Because name = "n" is the default, the magic never-ending appending "n" was not really predictable.