Segfault on invalid group_by_ and mutate_impl

[karldw]

Debugging on some group_by_ code, I found I could generate a segfault with these three lines:

n_groups <- 3
sortvar <- 'Sepal.Length'
iris2 <- group_by_(iris, .dots = lazyeval::interp(~ntile(sortvar, n_groups), 
                   .values = list(sortvar=sortvar, n_groups=n_groups)))

(Sometimes the last line has to be run multiple times to segfault, but I can consistently generate a crash with fewer than six attempts.)

It's also interesting that the error from the group_by_ call is not consistent (in the cases where R doesn't segfault). The message seems to alternate between:

• CHAR() can only be applied to a 'CHARSXP', not a 'logical'
• CHAR() can only be applied to a 'CHARSXP', not a 'closure'
• CHAR() can only be applied to a 'CHARSXP', not a 'integer'
• CHAR() can only be applied to a 'CHARSXP', not a 'character'
• CHAR() can only be applied to a 'CHARSXP', not a 'NULL'

Here's the segfault traceback:

 *** caught segfault ***
address (nil), cause 'unknown'

Traceback:
 1: .Call("dplyr_mutate_impl", PACKAGE = "dplyr", df, dots)
 2: mutate_impl(.data, dots)
 3: mutate_.tbl_df(tbl_df(.data), .dots = dots)
 4: mutate_(tbl_df(.data), .dots = dots)
 5: as.data.frame(mutate_(tbl_df(.data), .dots = dots))
 6: mutate_.data.frame(.data, .dots = new_groups[needs_mutate])
 7: mutate_(.data, .dots = new_groups[needs_mutate])
 8: group_by_prepare(.data, ..., .dots = .dots, add = add)
 9: group_by_.data.frame(iris, .dots = lazyeval::interp(~ntile(sortvar,     n_groups), .values = list(sortvar = sortvar, n_groups = n_groups)))
10: group_by_(iris, .dots = lazyeval::interp(~ntile(sortvar, n_groups),     .values = list(sortvar = sortvar, n_groups = n_groups)))

The issue is the same for the CRAN and development versions of dplyr (versions 0.4.3 and 0.4.3.9001). Here's my sessionInfo (with the development versions of dplyr and lazyeval):

R version 3.2.4 Revised (2016-03-16 r70336)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 15.10

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

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

other attached packages:
[1] dplyr_0.4.3.9001

loaded via a namespace (and not attached):
[1] magrittr_1.5   R6_2.1.2       assertthat_0.1 DBI_0.3.1      tools_3.2.4   
[6] tibble_1.0-1   Rcpp_0.12.4    ulimit_0.0-3  

Possibly related: #1668, #1641.

This is not urgent! Obviously, fixing edge cases where users are writing buggy code shouldn't be a priority, but segfaults are always a little disconcerting.

[romainfrancois]

I would not expect this to work as the expression you are interpreting is:

> lazyeval::interp(~ntile(sortvar, n_groups), .values = list(sortvar=sortvar, n_groups=n_groups))
~ntile("Sepal.Length", 3)

See the quotes there. This however works:

> iris2 <- group_by_(iris, .dots = lazyeval::interp(~ntile(sortvar, n_groups), .values = list(sortvar=as.name(sortvar), n_groups=n_groups)))
> iris2
Source: local data frame [150 x 6]
Groups: ntile(Sepal.Length, 3) [3]

   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
          <dbl>       <dbl>        <dbl>       <dbl>  <fctr>
1           5.1         3.5          1.4         0.2  setosa
2           4.9         3.0          1.4         0.2  setosa
3           4.7         3.2          1.3         0.2  setosa
4           4.6         3.1          1.5         0.2  setosa
5           5.0         3.6          1.4         0.2  setosa
6           5.4         3.9          1.7         0.4  setosa
7           4.6         3.4          1.4         0.3  setosa
8           5.0         3.4          1.5         0.2  setosa
9           4.4         2.9          1.4         0.2  setosa
10          4.9         3.1          1.5         0.1  setosa
..          ...         ...          ...         ...     ...
Variables not shown: ntile(Sepal.Length, 3) <int>.

See the as.name.

[romainfrancois]

So the problem becomes:

> iris %>% mutate( a = ntile("Sepal.Length", 3 ) )
Error in mutate_impl(.data, dots) :
  CHAR() can only be applied to a 'CHARSXP', not a 'list'