josefondrej / advanced-r

My notes and exercises from the great book Advanced R by Hadley Wickham

Advanced R

Collection of my notes and exercises from the great book Advanced R by Hadley Wickham. Intended for personal use only. However if you want to learn about programming in R, then I highly recommend the sources mentioned in this README.md

Other useful R resources

• R packages -- book on how to create package in R. For quick reference see this sample github repo.

• tidyverse R Style Guide, Google has it's own which is based on the tidyverse. I think Google makes the following good points:

  • do not use attach
  • do not use -> (right hand assignment)
  • always use explicit returns (do not rely on R's implicit return feature)
  • qualify namespaces - e.g. purrr::map()

IDE for R

Recently the company JetBrains started to support R language in their products. As far as I know this is the case for at least IntelliJ IDEA an PyCharm. I was using both of these for a couple of years to develop Java / Python and I was super happy with them.

I tried the R support only for couple of weeks in PyCharm and while there are still some minor bugs but I would say that it works pretty well and together with the other features the IDE offers makes it a clear choice for me.

You might also want to check out RStudio, which has more R focused features, but the tools for code refactoring and overall user-friendliness are nowhere near the JetBrains products.

Notes

2. Names & Values

• copy on modify (objects are immutable - exceptions)
• pass by value
• lobstr::obj_addr(z)

3. Vectors

• logical, integer, double, string
• list = atomic(reference)
• attr(object, "key") = "value", attributes(object),
• data.frame = named list of vectors

4. Subsetting 5. Control Flow 6. Functions

• function = (base::formals(f), base::body(f), base::environment(f) = evaluation env)
• lazy argument evaluation
• base::stop("error"), base::on.exit()

7. Environments

• environment = rlang::env(a = 1, b = 2)
• rlang::global_env() -> env(package_1) -> ... -> env(package_n) -> rlang::base_env() -> rlang::empty_env()
• <<-
• function captures/binds environment (=looks for the variables in it) in which it is defined (closure)
• execution environment
• (namespace and import environment)
• rlang::caller_env

8. TryCatch

• tryCatch(error = function(cnd) {}, code)

9. Functionals

• purrr::map(1:10, function, arg_1 = "fixed_value")
• purrr style: base::split(mtcars, mtcars$cyl) %>% map(func_1) %>% ... etc. (split = pandas.groupby)
• purrr::reduce, purrr::accumulate

10. Function Factories

• use base::force(arguments) in fn factories (to avoid hell caused by lazy evaluation)
• the created fn keeps track of it's environment -- gc() won't collect any garbage you leave there

11. Function Operators

12. Base vs OO Objects

• objects = base objects + oo objects (base::is.object / base::attr(object, "class") == NULL)
• sloop::s3_class(object) -- used by S3,S4 to pick methods

13. S3(🤮)

• sloop::s3_dispatch(generic_function(object))

• S3 object definition: new_myclass(args) + validate_myclass(object:myclass)

  • myclass(args){helpers + return new_myclass(x)}

• S3 object method definition: function.myclass(), calling function(object:myclass) then calls function.myclass which is done by base::UseMethod

• S3 generic: my_new_generic <- function(x) {UseMethod("my_new_generic")}

  method must have same arguments as it's generic

• S3 inheritance:

  • class(object) can be c("first class", "second class" ...)
  • base::NextMethod

  base type of subclass = type of superclass

  attributes of subclass = superset of superclass

• S3 subclassable constructor:

  new_subclassable_myclass <- function(x, ..., class = character()) {
    return(structure(x, ..., class=c(class, "subclassable_myclass")))
  }
  

• S3 classmethod - use vctrs::vec_restore

14. R6

MyClass <- R6::R6Class(
  classname = "MyClass",
  inherit = SuperClass,
  public = list(
    field = 123,
    func = function(x) {
      super$func(x = (x + 1))
    }
  ),
  private = list(
    .field = 234
  )
)

myclass = MyClass$new()
myclass$field
myclass$func()

15. S4

TODO: Finish writing summary