select_if doesn't work if a column in the tibble is named "i"

[alejandroschuler]

This one took me a few minutes to track down... It appears that if a column in the tibble is named i, select_if will fail. This also applies to mutate_if and possibly others.

packageVersion("dplyr")
[1] ‘0.8.1’

iris %>% 
rename(i=Sepal.Length) %>% 
select_if(is.numeric)

Error in tibble_vars[[i]] : 
  attempt to select more than one element in vectorIndex

[barryrowlingson]

works for me in 0.8.0.1, so I upgraded and now its broken in 0.8.1.

[cderv]

This come from the internal dplyr:::tbl_if_vars, here is a reprex.

library(dplyr, warn.conflicts = FALSE)


.tbl <- iris %>% 
  rename(i = Sepal.Length)

.predicate <- dplyr:::as_fun_list(is.numeric, caller_env())
vars <- dplyr:::tbl_if_vars(.tbl, .predicate, caller_env(), .include_group_vars = TRUE)
#> Error in tibble_vars[[i]]: attempt to select more than one element in vectorIndex

^{Created on 2019-05-23 by the reprex package (v0.3.0.9000)}

There is a for if loop with i as iterator variable. I guess tidy eval is mixing things up. I'll dig a bit more ans see what I can find.

[romatik]

@cderv, you are right, the error is there. I suggest that this can be fixed like this:

  for (i in seq_len(n)) {
    selected[[i]] <- eval_tidy(quo(.p(.tbl[[tibble_vars[[!!i]]]], ...)), mask)
  }

This corresponds to this line:

dplyr/R/colwise.R

Line 209 in d13c66a

for (i in seq_len(n)) {

I can create a PR if this fix is OK.

[cderv]

Thanks @romatik for chiming in!
I planned initially to look for a fix tonight - glad to know it has been handled! 😊👍

[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/