R7: A first look

Transcript

1. Hadley Wickham on behalf of
   

   The R Consortium Working Group on OOP
   

   https://rconsortium.github.io/OOP-WG/
   R7:
   
   
   A
   fi
   rst look at a new OOP system for R
   June 2022
   

   View Slide

2. bizarro
   < -
   function(x) {
   
   
   if (is.numeric(x)) {
   
   
   -
   x
   
   
   } else if (is.logical(x)) {
   
   
   !x
   
   
   } else if (is.character(x)) {
   
   
   str_reverse(x)
   
   
   } else if (is.factor(x)) {
   
   
   levels(x)
   < -
   rev(level(x))
   
   
   x
   
   
   } else {
   
   
   stop("Not supported")
   
   
   }
   
   
   }
   Why do we need OOP?
   This is even more
   important for functions
   in base R, like print()
   

   View Slide

3. S3
   S4
   English Fairy Tales (1918) by Flora Annie Steel, illustrated by Arthur Rackham
   R7 R7 = S3 + S4
   

   View Slide

4. library(R7)
   
   
   bizarro
   < -
   new_generic("bizarro", "x")
   
   
   With R7, we
   fi
   rst de
   fi
   ne a generic
   Name of the generic
   Name of the argument used for dispatch
   

   View Slide

5. method(bizarro, class_numeric)
   < -
   function(x) {
   
   
   -
   x
   
   
   }
   
   
   method(bizarro, class_logical)
   < -
   function(x) {
   
   
   !x
   
   
   }
   
   
   method(bizarro, class_character)
   < -
   function(x) {
   
   
   str_reverse(x)
   
   
   }
   Then we de
   fi
   ne methods
   Name of the generic Built-in class
   

   View Slide

6. bizarro
   
   
   #> function (x,
   . . .
   ) with 4 methods:
   
   
   #> 1
   :
   method(bizarro, "integer")
   
   
   #> 2
   :
   method(bizarro, "double")
   
   
   #> 3
   :
   method(bizarro, "character")
   
   
   #> 4
   :
   method(bizarro, "logical")
   
   
   

   View Slide

7. bizarro(1)
   
   
   #> [1] -1
   
   
   bizarro(TRUE)
   
   
   #> [1] FALSE
   
   
   bizarro(mean)
   
   
   #> Error: generic `bizarro()` can't f
   i
   nd method
   
   
   #> for function object
   

   View Slide

8. Generics de
   fi
   ne behaviour
   Classes de
   fi
   ne data
   

   View Slide

9. range
   < -
   new_class("range",
   
   
   properties = list(
   
   
   start = class_numeric,
   
   
   end = class_numeric
   
   
   )
   
   
   )
   
   
   Creating a new class
   Name of the class
   Property type
   Property name
   Constructor function
   

   View Slide

10. x
    < -
    range(start = 1, end = 10)
    
    
    x
    
    
    #>
    
    
    #> @ start: num 1
    
    
    #> @ end : num 10
    
    
    x@start
    
    
    x@end
    < -
    20
    
    
    x@end
    < -
    "x"
    
    
    #> Error: @end must be , not
    

    View Slide

11. r
    < -
    range(start = 10, end = 1)
    

    #>
    
    
    #> @ start: num 10
    
    
    #> @ end : num 1
    But…
    

    View Slide

12. range
    < -
    new_class("range",
    
    
    properties =
    . . .
    ,
    
    
    validator = function(self) {
    
    
    if (length(self@start)
    ! =
    1) {
    
    
    "@start must be length 1"
    
    
    } else if (length(self@end)
    ! =
    1) {
    
    
    "@end must be length 1"
    
    
    } else if (self@end < self@start) {
    
    
    "@end must be greater than or equal to @start"
    
    
    }
    
    
    }
    
    
    )
    Can add a validator to prevent that from happening
    

    View Slide

13. range
    < -
    new_class("range",
    
    
    properties = list(
    
    
    start = class_double,
    
    
    end = class_double,
    
    
    length = new_property(
    
    
    getter = function(self) self@end - self@start,
    
    
    )
    
    
    ),
    
    
    …
    
    
    )
    Can also create dynamic properties
    

    View Slide

14. S3
    S4
    English Fairy Tales (1918) by Flora Annie Steel, illustrated by Arthur Rackham
    R7
    

    View Slide

15. View Slide

16. 1. Team e
    ff
    ort.
    
    
    2. Formal de
    fi
    nition of class and
    properties (like S4).
    
    
    3. Embrace data access.
    
    
    4. Backward compatible with S3.
    
    
    5. Pit of success.
    Why do we hope this will succeed?
    

    View Slide

17. Please try it out!
    https://rconsortium.github.io/OOP-WG/
    remotes
    : :
    install_github("rconsortium/OOP-WG")
    What doesn’t work?
    

    What don’t you understand?
    These slides at https://speakerdeck.com/hadley/r7-a-
    fi
    rst-look
    

    View Slide

18. View Slide