Construct a new object with ggproto
, test with is.proto
,
and access parent methods/fields with ggproto_parent
.
ggproto(`_class` = NULL, `_inherit` = NULL, ...) ggproto_parent(parent, self) is.ggproto(x)
_class | Class name to assign to the object. This is stored as the class
attribute of the object. This is optional: if |
---|---|
_inherit | ggproto object to inherit from. If |
... | A list of members in the ggproto object. |
parent, self | Access parent class |
x | An object to test. |
ggproto implements a protype based OO system which blurs the lines between classes and instances. It is inspired by the proto package, but it has some important differences. Notably, it cleanly supports cross-package inheritance, and has faster performance.
In most cases, creating a new OO system to be used by a single package is not a good idea. However, it was the least-bad solution for ggplot2 because it required the fewest changes to an already complex code base.
ggproto methods can take an optional self
argument: if it is present,
it is a regular method; if it's absent, it's a "static" method (i.e. it
doesn't use any fields).
Imagine you have a ggproto object Adder
, which has a
method addx = function(self, n) n + self$x
. Then, to call this
function, you would use Adder$addx(10)
-- the self
is passed
in automatically by the wrapper function. self
be located anywhere
in the function signature, although customarily it comes first.
To explicitly call a methods in a parent, use
ggproto_parent(Parent, self)
.
Adder <- ggproto("Adder", x = 0, add = function(self, n) { self$x <- self$x + n self$x } ) is.ggproto(Adder)#> [1] TRUEAdder$add(10)#> [1] 10Adder$add(10)#> [1] 20Doubler <- ggproto("Doubler", Adder, add = function(self, n) { ggproto_parent(Adder, self)$add(n * 2) } ) Doubler$x#> [1] 20Doubler$add(10)#> [1] 40