Could not find function "lead" with latest version of dplyr

[md0u80c9]

Hi all,

The following line of code works perfectly in the stable version of dplyr. In the development version however (with the newer version of bindrcpp) this generates an error:

Error in mutate_impl(.data, dots) : could not find function "lead"

I'm not aware of lead being affected by the tidy select changes so is the function being parsed differently in some way?

  importedData <- dplyr::mutate(importedData,
    InpatientReadmission = cumsum(
      (dplyr::lead(.data$TeamCode,
                  n = 1,
                  default = -1) > 0) & (.data$TeamCode < 0) * 1))

[krlmlr]

Thanks for the report, this is really important to fix before release. Could you please create a reproducible example that uses a small toy dataset or a commonly available one?

[md0u80c9]

Hi,

It looks as though this is as simple as:

test <- tibble::tibble(foo = c(1,2,3))
test <- dplyr::mutate(test, bar = dplyr::lead(.data$test, n = 1, default = -1))

Fails.

test <- tibble::tibble(foo = c(1,2,3))
dplyr::lead(test, n = 1, default = -1)

Works.

[krlmlr]

Thanks. I can't run your first example, I'm seeing:

test <- tibble::tibble(foo = c(1,2,3))
test <- dplyr::mutate(test, bar = dplyr::lead(.data$test, n = 1, default = -1))
#> Error in mutate_impl(.data, dots): Evaluation error: Column `test` not found in `.data`.

Created on 2018-03-24 by the reprex package (v0.2.0).

Also, for the second example I'm seeing:

test <- tibble::tibble(foo = c(1,2,3))
dplyr::lead(test, n = 1, default = -1)
#> # A tibble: 3 x 1
#>     foo
#> * <dbl>
#> 1   -1.
#> 2   NA 
#> 3   NA

Created on 2018-03-24 by the reprex package (v0.2.0).

What do you expect lead() on a data frame to return?

[md0u80c9]

Apologies - that first example should be:

test <- tibble::tibble(foo = c(1,2,3))
test <- dplyr::mutate(test, bar = dplyr::lead(.data$foo, n = 1, default = -1))

On dplyr 0.7.4 you get:

> test
# A tibble: 3 x 2
    foo   bar
  <dbl> <dbl>
1    1.    2.
2    2.    3.
3    3.   -1.

Output from dplyr 0.7.4.9002:

> test <- tibble::tibble(foo = c(1,2,3))
> test <- dplyr::mutate(test, bar = dplyr::lead(.data$foo, n = 1, default = -1))
Error in mutate_impl(.data, dots) : could not find function "lead"

[krlmlr]

Thanks, confirmed:

@romainfrancois: Could you please take a look?

[md0u80c9]

Just followed the thread (#3456). Don't think the current fixes address the above? Certainly I can still produce the original error that led me to the report with the latest version of dplyr (pulled from GitHub just now). Just tested with the reprex and I still see the same too.

[romainfrancois]

@md0u80c9 did you try with the fix-issue-3456 branch?

install_github( "tidyverse/dplyr", ref = "fix-issue-3456" )

[md0u80c9]

Tested with that branch now - and that seems to work both on the reprex and on my original code. Great stuff :-).

[romainfrancois]

Thanks for taking the time to try it.

Just for the record, a commit message in the branch would have closed the issue automatically on merge.

Not a big deal because we usually merge fast, but potentially a problem on others repos.

Thanks again for reporting the problem in the first place.

[krlmlr]

Reopening, will close when PR is merged.

[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/