如何使用tidyr通过使用列名的字符串删除列来收集data.table？

Question

Seems to have found a bug in tidyr.

I have a piece of code like this

 rm(hello)
 a <- function() {  
  dt <- data.table(`*stat*` = c("mean","max"), time = c(11, 12), val1 = c(1,2), val2 =c(3,4))
  hello <- "^(time)$"
  dt2 <- dt %>% gather(dimension, value, -matches("\\*stat\\*"), -matches(hello))  
}

a()

But

    dt <- data.table(`*stat*` = c("mean","max"), time = c(11, 12), val1 = c(1,2), val2 =c(3,4))
hello <- "^(time)$"
dt2 <- dt %>% gather(dimension, value, -matches("\\*stat\\*"), -matches(hello))  

works.

The only difference is that one set of code is in a function but the other is not!

Answer 1

I'm just adding an answer here to show the function and standalone code have identical results once dt2 is added to the last line of the a() function (as suggested by @A5C1D2H2I1M1N2O1R2T1 in the comments).

library(tidyverse)

a <- function() {  
  dt <- data.table::data.table(`*stat*` = c("mean","max"), time = c(11, 12), val1 = c(1,2), val2 =c(3,4))
  hello <- "^(time)$"
  dt2 <- dt %>% gather(dimension, value, -matches("\\*stat\\*"), -matches(hello)) 
  dt2
}

a()
#>   *stat* time dimension value
#> 1   mean   11      val1     1
#> 2    max   12      val1     2
#> 3   mean   11      val2     3
#> 4    max   12      val2     4

dt <- data.table::data.table(`*stat*` = c("mean","max"), time = c(11, 12), val1 = c(1,2), val2 =c(3,4))
hello <- "^(time)$"
dt2 <- dt %>% gather(dimension, value, -matches("\\*stat\\*"), -matches(hello))  
dt2
#>   *stat* time dimension value
#> 1   mean   11      val1     1
#> 2    max   12      val1     2
#> 3   mean   11      val2     3
#> 4    max   12      val2     4

identical(a(), dt2)
#> [1] TRUE