[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: in defense of types



I don't quite understand Matthias here. Here is my view on types:


- advantages of static typing

  - performance: the original goal for static typing is to allow
  greater performance and memory consumption

  - static analysis: another goal is to ensure things at compile-time

  those 2 are usually mixed because of History. But these should now
  be split apart:

  - performance usually comes from a critical path, so some kind of
  type annotation or analysis can be done to have the performance
  advantage of static typing, but keeping most of the program
  dynamically typed (bigloo ?)

  - complex typing schemes (parametric polymorphism, runtime dispatch)
  do not allow the full speed of simple static typing, nor the perfect
  memory consumption (think boxing)

  - even more complex typing schemes (row subtyping, polymorphic
  variants) allow complex expressions but the runtime cost is
  important.


- some checks done in statically typed languages are not type related
(existence of a function/method, of a variable), but comes together
with the way compilation is done. Some tools can regain those checks
in dynamically typed languages (use strict, pychecker...)


- OO dynamically typed languages (Python/Ruby/Smalltalk) have types

  - objects are types: 
    - declaring an object is a type declaration
    - if you have a look at a corresponding program
  in Common Lisp, you'll see the many type annotations that are
  implicit (via sugar) in Python/Ruby/Smalltalk. 

  -> those languages have runtime type checks