Persistent Mutate Error Caused by Previous group_by/summarise

Question

I'm seeing a mutate error caused by unrelated code that can either cause errors to be thrown or not based on unrelated code being run. For example,

I initialize data and run a block of mutate code successfully
I run a group_by, summarise code block that runs successfully but has warnings
The same original code with original data now throws a mutate error whereas it ran successfully before!
I run an unrelated mutate that doesn't affect the original data which runs successfully
I now run the mutate code a 3rd time, and the code runs successfully again now!

library(tidyverse)
library(scales)
#> 
#> Attaching package: 'scales'
#> The following object is masked from 'package:purrr':
#> 
#>     discard
#> The following object is masked from 'package:readr':
#> 
#>     col_factor

df_test <- tibble(group = c('a', 'a', 'b', 'b', 'b'), hour=parse_factor(as.character(c(1, 2, 1, 2, 1))), x = c(1,2,3, 4, 5), y=c(5, 6, 7, 8, 9))

return_data <- df_test %>%
  dplyr::mutate(
    hour = paste(hour, ':00'),
    across(.cols = c(x, y), scales::label_dollar())
  )

summarise_df_input <- function(.data, func, group_vars) {
  df_agg <- .data %>%
    group_by(across(all_of(group_vars))) %>%
    summarise(across(everything(), func))
  
  return(df_agg)
}

df_grouped <- df_test %>% summarise_df_input(mean, 'group')
#> Warning in mean.default(hour): argument is not numeric or logical: returning NA

#> Warning in mean.default(hour): argument is not numeric or logical: returning NA

return_data <- df_test %>%
  dplyr::mutate(
    hour = paste(hour, ':00'),
    across(.cols = c(x, y), scales::label_dollar())
  )
#> Error: Problem with `mutate()` input `..2`.
#> x subscript out of bounds
#> ℹ Input `..2` is `(function (.cols = everything(), .fns = NULL, ..., .names = NULL) ...`.

return_data <- df_test %>%
  dplyr::mutate(
    across(.cols = c(x, y), scales::label_dollar())
  )

return_data <- df_test %>%
  dplyr::mutate(
    hour = paste(hour, ':00'),
    across(.cols = c(x, y), scales::label_dollar())
  )

^{Created on 2021-03-05 by the reprex package (v1.0.0)}

Here is a reprex of what's going on. Does anyone have any idea what's happening? This is using dplyr 1.0.5 btw.

akrun akrun · Accepted Answer · 2021-03-05T19:56:24

it is using everything, instead it should be where(is.numeric) because other than the group column, there is an 'hour' column which is factor and mean works on numeric variables

summarise_df_input <- function(.data, func, group_vars) {
  .data %>%
      group_by(across(all_of(group_vars))) %>%
       summarise(across(where(is.numeric), func), .groups = 'drop')  
  
}

-testing

df_test %>% 
    summarise_df_input(mean, 'group')
# A tibble: 2 x 3
#  group     x     y
#* <chr> <dbl> <dbl>
#1 a       1.5   5.5
#2 b       4     8

Regarding the error in execution, it may be a bug. Changing the order of execution of across can bypass the error

return_data <- df_test %>%
    dplyr::mutate(
       across(.cols = c(x, y), scales::label_dollar()), hour = paste(hour, ':00')
  )

Persistent Mutate Error Caused by Previous group_by/summarise

1 Answers