Bug in s3_dispatch

[Tazinho]

(Probably due to a change in R 4.0; I discovered this when solving the regarding exercise in Adv R)

# Output in R 4.0.3
x <- structure(1:10, class = "test")

sloop::s3_dispatch(t(x))
#> => t.test
#>  * t.default
t(x)
#>      [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10]
#> [1,]    1    2    3    4    5    6    7    8    9    10
#> attr(,"class")
#> [1] "test"
t.test(x)
#> 
#>  One Sample t-test
#> 
#> data:  x
#> t = 5.7446, df = 9, p-value = 0.0002782
#> alternative hypothesis: true mean is not equal to 0
#> 95 percent confidence interval:
#>  3.334149 7.665851
#> sample estimates:
#> mean of x 
#>       5.5

^{Created on 2020-12-31 by the reprex package (v0.3.0)}

[tslumley]

I came here to report this

[moodymudskipper]

s3_dispatch() names and looks up the methods naively without looking at the global tables, I think this would solve most issues, right now as soon as standard conventions are not followed it seams to break.

here's another use case, rare but for one of the most used functions:

s3 <- getNamespaceInfo("utils", "S3methods")
s3[paste0(s3[,1], ".", s3[,2]) != s3[,3],]
#>      [,1]   [,2]         [,3]         [,4]
#> [1,] "head" "data.frame" "head.array" NA  
#> [2,] "tail" "data.frame" "tail.array" NA  
#> [3,] "tail" "table"      "tail.array" NA

sloop::s3_dispatch(head(iris))
#> => head.data.frame
#>  * head.default

And another example, not related to global tables here, but from the fact foo's name is used directly rather than finding the method in the body.

foo <- function() {
  UseMethod("head")
}
sloop::s3_dispatch(foo(letters))
#>    foo.character
#> => foo.default

This can be used, rather than converting the function symbol to character

names(utils::isS3stdGeneric(foo))
#> [1] "head"