/strongtyping

Strong typing for Ruby

Primary LanguageC

= Description
  The strongtyping library is a provides a convenient way for Ruby methods
  to check parameter types, and also dynamically query them. In addition to
  merely checking a single set of types, it allows easy overloading based on
  a number of different templates.

= Installation
  gem install strongtyping

= Synopsis
  Let's say you have the following function:

    def foo(a, b)
      ...
    end

  Now let's say this function wants 'a' to always be a String, and 'b'
  should be Numeric:

    require 'strongtyping'
    include StrongTyping

    def foo(a, b)
      expect(a, String, b, Numeric)
      ...
    end

  If 'a' or 'b' is of the wrong type, an ArgumentTypeError will be raised.

  Overloading is just as easy:

    require 'strongtyping'
    include StrongTyping

    def bar(*args)
      overload(args, String, String){ |s1,s2|
        ...
        return
      }

      overload(args, String, Integer){ |s,i|
        ...
        return
      }

      overload_default args
    end

  If someone calls 'bar' with two Strings, or a String and an Integer,
  the appropriate block will be called.  Otherwise, an OverloadError
  is raised.

  How about default parameters? Say we have the following function:

    def baz(a, b = nil)
      action  = "You baz #{a}";
      action += " with a #{b}"  if b

      print action, "\n"
    end

  Now, b can either be nil or a String.  We don't want to have two
  full overload cases... that would duplicate code.  So, expect()
  allows an array of types:

    expect(a, String, b, [String, NilClass]);

  This takes care of the above case nicely.

  What if your code is curious about which types are allowed?  The
  get_arg_types function is provided for just this purpose.  Given the
  above definitions for 'foo' and 'bar', consider the following code:

    p get_arg_types(method(:foo))  # => [[String, Numeric]]
    p get_arg_types(method(:bar))  # => [[String, String], [String, Integer]]
    p get_arg_types(method(:baz))  # => [[String, [String, NilClass]]]
      
  This is useful if you're converting user input into a form that the
  method expects.  (If you get "1234", should you convert it to an
  integer, or is it best left a string? Now you know.)

  What if you have an array of arguments, 'arr', and you're worried
  that the method 'bar' won't accept them?  You can check ahead of
  time:

    if not verify_args_for(method(:bar), arr)
      print "I can't let you do that, Dave\n"
    end

= Reference
  Module: StrongTyping

  Methods:
  
    expect(obj0, Module0[, obj1, Module1[,...objN, ModuleN]])

       Verify the parameters obj0..objN are of the given class (or
       module) Module0..ModuleN

    overload(args, [Module0[, Module1[,...ModuleN]]]) { | o0, o1,..oN | }

       Call the block with 'args' if they match the pattern
       Module0..ModuleN.  The block should _always_ call return at the
       end.

    overload_exception(args, [Module0[,...ModuleN]]]) { | o0, o1,..oN | }

       This acts identically to overload(), except the case specified
       is considered invalid, and thus not returned by get_arg_types().
       It is expected that the specified block will throw an exception.

    overload_default(args)       
    overload_error(args)
    
       Raise OverloadError.  This should _always_ be called after the
       last overload() block.  In addition to raising the exception,
       it aids in checking parameters.  As of 2.0, the overload_error
       name is deprecated; use overload_default.

    get_arg_types(Method)

       Return an array of parameter templates.  This is an array of
       arrays, and will have multiple indices for functions using
       multiple overload() blocks.

    verify_args_for(method, args)

       Verify the method 'method' will accept the arguments in array
       'args', returning a boolean result.

  Exceptions:

    ArgumentTypeError < ArgumentError

      This exception is raised by expect() if the arguments do not
      match the expected types.

    OverloadError < ArgumentTypeError

      This exception is raised by overload_default() if no overload()
      template matches the given arguments.
       
= Q & A
  This section written by Ryan Pavlik.

  Q: Why?
  A: Because it was originally needed for the Mephle library.

  Q: No really, why bother with static typing? Isn't ruby dynamic?
  A: This is not 'static typing'.  This is 'strong typing'.  Static
     typing is what you get when a variable can only be of a certain
     type, as in C or C++.  Strong typing is enforcing types. These may
     seem similar, but they are actually not directly related.

     Some other languages, such as Common Lisp, allow for dynamic,
     strong typing. Strong typing and dynamic typing are not mutually
     exclusive.

  Q: Yeah, but really, why bother?  Why not just let ruby sort out the
     errors as they occur?  
  A: This is incorrect thinking. Allowing errors to just occur when
     they happen is naive programming. Consider the following:

       # Wait N seconds, then open the bridge for M seconds
       def sendMsg(bridge, n, m)
         sleep(n)
         bridge.open
         sleep(m)
         bridge.close
       end

     Now say 'm' is pased in as a string.  Oops!  A TypeError is
     raised. Now the bridge is open, and somewhere (hopefully!) someone
     caught the exception so the program didn't crash, but the bridge
     opening wasn't reversed, so it's going to stay open and back up
     traffic until someone fixes the problem.

     This is an academic example, but there are many cases when just
     letting an error happen will lead to an inconsistent system state.
     Ruby (and most systems) are not transactional, and inconsistent
     states are unacceptable.

     In addition, it is desireable to know _programmatically_ why
     something failed, as specific action can be taken if desired.

     "Wait," someone in the audience says, "you could just check to see
     if 'm' and 'n' are of the correct type!"

     Yes, yes you could.

     That's what this module is for. ;-)

  Q: Isn't it up to the caller to call my function correctly?
  A: The caller cannot know and deal with errors that may occur in your
     code.  That's your job.  Checking for errors ahead of time and
     informing the caller about problems is also your job.  This module
     just makes it easy.

     In addition, it's nice for the caller to be able to ask and check
     what your method expects ahead of time to guard against error.
     The strongtyping module also provides functionality for this.

  Q: OK, but strong typing is baaad.  What if I want to pass something
     that acts like something else, or responds to a given symbol?
     Doesn't ruby have "duck" typing?
  A: First, what you're suggesting is evil. If you want that, go
     write C++. :-)

     Second, you should never depend on a function's implementation.  If
     the documentation says "pass me a hash" and you pass it anything
     that responds to :[], your code may break when the next version
     comes out.

     Third, if you pass something that responds accurately to the
     _interface_ (methods provided by class or module) specified, then
     that should be _of_ that class or module.  This may not be the case
     with all ruby objects yet; for instance, anything responding to :[]
     being something like a Mappable.  You can make this the case in
     your code, or urge developers to create a standard set of interface
     mixins for just this purpose.

     "Duck" typing just a term for this sort of "maybe" behavior, much
     like what C++ STL templates use.  However, the problem is that even
     if an object responds to a method, there is no guarantee that the
     method acts in an expected manner---and the interface may still
     change without notice.  "Duck" typing sounds much like 'duct
     taping' depending on your accent, and I think duct-taping is a good
     description of this is in practice. :-)

     Another argument is that ruby allows one to change the behavior of
     methods at any time:

       a = String.new;
       def a.split
         print "hello world\n"
       end

     For this, I have two responses:  first, if a method is deprecated
     or changed dramatically, strongtyping can aid in letting the code
     know:

        a = String.new;
        def a.split(*args)
           overload(args) { print "hello world\n" }
           overload_default args
        end

     This case will drop any normal calls through to overload_default,
     raising an exception, which can be caught and analyzed.  You can
     even provide another case that calls the superclass.

     Second, either you're changing the method in a subtle manner (it
     does what it used to, with added effect), or an outrageous manner
     (it acts nothing like it did before).  In the former case, code
     should work fine anyway.  In the latter case, as in the above
     example, you should ask yourself why you're changing it.  The
     function no longer splits, why is it called split?  This is not
     good design; the strongtyping module is here to aid in good design,
     not prevent poor design.

     A more realistic example would be the academic "Shape" class
     example of inheritance, with "Ellipse" and "Circle".  Ruby properly
     allows one to make Circle a subclass of Ellipse, and redefine
     "setSize" to the constrained definition of a circle.  This change
     is visible to code---an ArgumentError will be raised (2 arguments
     for 1), or setSize can throw a ConstraintError.  strongtyping
     provides a useful function, overload_exception, for just this case:

        class Circle < Ellipse
           :
           def setSize(*args)
              overload(args, Integer) {
                 | r |
                 @radius = r
                 return
              }

              overload_exception(args, Integer, Integer) {
                 | a, b |
                 raise ConstraintError
              }
              
              overload_default args
           end
           :
        end

     Of course, there are a number of good choices for handling
     this... you may still allow #setSize(a, b) if a == b.  The
     important part is that the change in behavior can now be determined
     by code.

  Q: But I always write perfect code.  I know what my functions do, and
     what they take, and what I'm passing them.
  A: No one writes perfect code.  Additionally, not all environments are
     as controlled as yours may be.  Especially in a networked
     environment when someone may be invoking a method remotely, you
     can't depend on calling code not to be malicious.

  Q: OK, OK.  But, uh... what is Mephle?
  A: Mephle is a soon-to-be-released network-transparent persistant
     object system written in ruby.  It uses many of the Unity concepts
     (http://unity-project.sf.net/).  It will be on the RAA when
     released.

= License
  strongtyping - Method parameter checking for Ruby
  Copyright (C) 2003-2011 Ryan Pavlik

  This library is free software; you can redistribute it and/or
  modify it under the terms of the GNU Lesser General Public
  License as published by the Free Software Foundation; either
  version 2.1 of the License, or (at your option) any later version.

  This library is distributed in the hope that it will be useful,
  but WITHOUT ANY WARRANTY; without even the implied warranty of
  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  Lesser General Public License for more details.

  You should have received a copy of the GNU Lesser General Public
  License along with this library; if not, write to the Free Software
  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307  USA

= Future Plans
  There's some odd C code that is causing a couple of warnings, but it
  does not appear to be harmful. I want to clean this code up eventually.

= Authors
* Ryan Pavlik (original author)
* Daniel Berger (maintenance)