mutate_if() w/ single column breaks in r-devel

[wibeasley]

In the recent r-devel, mutate_if() isn't working under some conditions when the dataset has only one column.

ds_1a <- data.frame(
  id               = 1L:5L,
  stringsAsFactors = FALSE
)

ds_1b <- dplyr::mutate_if(
  ds_1a,
  is.character,
  function(x) dplyr::coalesce(x, "-") #Replace NAs with blanks
)

The error is

Error in evalq(sys.calls(), <environment>) : 
  Vector 1 has type 'character' not 'integer'

But everything is fine with a second column:

ds_2a <- data.frame(
  id               = 1L:5L,
  name             = c("a", "b", NA_character_, "c", "d"),
  stringsAsFactors = FALSE
)
ds_2b <- dplyr::mutate_if(
  ds_2a,
  is.character,
  function(x) dplyr::coalesce(x, "-") #Replace NAs with blanks
)

output:

> ds_2b
  id name
1  1    a
2  2    b
3  3    -
4  4    c
5  5    d

I think the problem is with the predicate. That error is identical to if coalesce is called directly (when the two arguments have different data types):

> dplyr::coalesce(1L:5L, "")
Error: Vector 1 has type 'character' not 'integer'

The problem is avoided on Linux & Windows with 3.3.1. But it's occurring Travis and a local Windows machine:

> sessionInfo()
R Under development (unstable) (2016-09-06 r71220)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows >= 8 x64 (build 9200)

locale:
[1] LC_COLLATE=English_United States.1252 
[2] LC_CTYPE=English_United States.1252   
[3] LC_MONETARY=English_United States.1252
[4] LC_NUMERIC=C                          
[5] LC_TIME=English_United States.1252    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods  
[7] base     

other attached packages:
[1] REDCapR_0.9.5.9004

loaded via a namespace (and not attached):
 [1] httr_1.2.1     lazyeval_0.2.0 magrittr_1.5   assertthat_0.1
 [5] R6_2.1.3       DBI_0.5        tools_3.4.0    dplyr_0.5.0   
 [9] tibble_1.2     curl_1.2       Rcpp_0.12.7  

[lionel-]

I cannot reproduce on R-devel, osx. That looks like a bug in R. What do you get when you run

tbl_vars(df)

and

vapply(df, is.character, logical(1))

[wibeasley]

> tbl_vars(ds_1a)
[1] "id"

> vapply(ds_1a, is.character, logical(1))
   id 
FALSE 

> tbl_vars(ds_2a)
[1] "id"   "name"

> vapply(ds_2a, is.character, logical(1))
   id  name 
FALSE  TRUE 

I cannot reproduce on R-devel, osx. That looks like a bug in R.

This may be way off, but I'll swing away anyway. If it's not showing up on R-devel osx, could it be related to the new toolchain? I used the top link above. Maybe the Windows and Travis builds are using the new gcc toolchain, while th osx is not? I see that Travis build is using gcc 4.6.3.

Tell me if there's a different build of R you'd like me to try.

EDIT: I get the same error with yesterday's version of R-devel:

> sessionInfo()
R Under development (unstable) (2016-09-20 r71321)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows >= 8 x64 (build 9200)

locale:
[1] LC_COLLATE=English_United States.1252 
[2] LC_CTYPE=English_United States.1252   
[3] LC_MONETARY=English_United States.1252
[4] LC_NUMERIC=C                          
[5] LC_TIME=English_United States.1252    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods  
[7] base     

other attached packages:
[1] dplyr_0.5.0

loaded via a namespace (and not attached):
[1] lazyeval_0.2.0 magrittr_1.5   R6_2.1.3       assertthat_0.1
[5] DBI_0.5-1      tools_3.4.0    tibble_1.2     Rcpp_0.12.7

[krlmlr]

Are you still seeing these problems?

[wibeasley]

Sorry, it took a while to get back to a machine I could put r-dev on.

Yes, the error is still the same error.

Error in evalq(sys.calls(), <environment>) : 
  Vector 1 has type 'character' not 'integer'

> sessionInfo()
R Under development (unstable) (2016-11-29 r71698)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows >= 8 x64 (build 9200)

locale:
[1] LC_COLLATE=English_United States.1252  LC_CTYPE=English_United States.1252    LC_MONETARY=English_United States.1252
[4] LC_NUMERIC=C                           LC_TIME=English_United States.1252    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

loaded via a namespace (and not attached):
 [1] compiler_3.4.0 lazyeval_0.2.0 magrittr_1.5   R6_2.2.0       assertthat_0.1 DBI_0.5-1      tools_3.4.0    dplyr_0.5.0    tibble_1.2     Rcpp_0.12.8   

[hadley]

Given that we can't reproduce, I'm going to close. For future reference, I've included a simpler reprex below.

library(dplyr, warn.conflicts = FALSE)

df <- tibble(id = 1:5)
dplyr::mutate_if(df, is.character, coalesce, "-")
#> # A tibble: 5 × 1
#>      id
#>   <int>
#> 1     1
#> 2     2
#> 3     3
#> 4     4
#> 5     5