[ruby-talk:18919] README for Dynamic-ORBit

If anyone wants to comment on my IDL mapping for Ruby, here's the README:


Author/maintainer: Tobin Baker
Email: [e-mail]

This library is licensed under the GNU Library General Public License.
See COPYING for details.

This is an alpha release of ORBit bindings for Ruby.  Most of the code
here has been stolen directly from Owen Taylor's excellent CORBA::ORBit
module for Perl.  Some other features (such as automatic loading of
IDL files) have been borrowed from Jason Tackaberry's ORBit-Python
module.  A feature that this library shares with these two modules is
dynamic loading of IDL files.  No intermediate IDL compilation step is
necessary!  You either specify the location of an IDL file directly with
CORBA._load_idl or simply require the top-level module from that file, and
the library will search all IDL files in the current directory, several
hard-coded paths, or the IDLPATH environment variable for an IDL file
containing that module.  The library parses the IDL file and dynamically
creates all Ruby objects corresponding to the IDL definitions in the file.

In this release, only client-side bindings are available.  Most of
the server-side code has been written and is currently being debugged.
I hope to release it in just a few weeks.

The IDL-Ruby mapping in this release is highly experimental.  Some
features are rather "baroque", and may well be withdrawn in the future,
depending on public feedback.

I should also say that I have gotten a few of the ideas for the
IDL-Ruby mapping from Dai, who began work on ORBit bindings for Ruby
long before I did.  Our approaches are rather different, but I hope that
by discussions on the newsgroups and mailing lists, we can eventually
arrive at a standardized CORBA mapping for Ruby that will be as portable
as Ruby itself.

Please email me with any bug reports or suggestions.  I hope CORBA
bindings bring us one step closer to world domination!  (Sorry, must
have been hanging around Linux zealots too long :-)


The IDL-Ruby Mapping


1. Modules

IDL modules map directly to Ruby modules.  The fact that Ruby has true
nested modules makes for an especially natural mapping from IDL to Ruby.
E.g.,

//IDL
module Foo {
	module Bar {
	};
};

#Ruby
module Foo
  module Bar
  end
end

It should be noted that since Ruby module names must begin with capital
letters, using IDL files with lowercase module names could be a problem.
I especially have in mind IDL written for use with Java, where package
names (which map to IDL modules) are conventionally in lowercase.  I have
done nothing to address this problem in this release, but some kind of
name mangling may be necessary for compatibility with existing IDL.
<soapbox>To me, this illustrates the kind of unintended consequences
that can arise from a well-meaning attempt to turn conventions into
rules.</soapbox>


2. Interfaces

IDL interfaces are also mapped to Ruby modules.  Each Ruby module
corresponding to an IDL interface includes the base module CORBA::Object,
which declares all the methods found in the IDL CORBA::Object interface,
prefixed with an underscore, along with an additional method, repo_id,
which returns the interface's Repository ID.  When an IDL interface
definition is encountered by the parser, a Ruby module corresponding to
the interface is created and each attribute and operation defined in the
interface is defined as an instance method in that module.  The Ruby
modules corresponding to base interfaces in the IDL definition are
included in the new module.  At runtime, an object reference entering
the ORB is demarshaled into an instance of CORBA::ORBit::Stub, which
wraps ORBit's native object reference type, and extended with the module
corresponding to the object reference's interface.  While interfaces are
never directly instantiated by the client, what happens looks something
like this:

//IDL
module Foo {
	module Bar {
		interface Baz {
			void do_it(in long dummy);
		};
	};
};

#Ruby
module Foo
  module Bar
    module Baz
      def do_it(dummy)
      end
    end
  end
end

myBaz = CORBA::ORBit::Stub.new()
myBaz.extend(Foo::Bar::Baz)

In the server-side mapping, a Ruby module with the suffix "_POA" is
created for each interface definition.	Any implementation class for
that interface must include this module.  E.g., assuming the same IDL
as the last example:

#Ruby
module Foo
  module Bar
    module Baz_POA
      def do_it(dummy)
      end
    end
  end
end

class Baz
  include Foo::Bar::Baz_POA
  def do_it(dummy)
    #method implementation
  end
end


3. Operations

IDL operations are mapped to Ruby method definitions.  As a rule,
in arguments are passed as ordinary parameters, return values and
out arguments are returned as a single array, and inout arguments are
passed in like in parameters and returned like out parameters, with no
"pass-by-reference" semantics involved.  E.g.,

//IDL
interface Foo {
	long do_it(in long in_arg, out long out_arg, inout long
		inout_arg);
};

#Ruby
#somehow get reference myFoo to Foo object
ret_val, out_arg, inout_arg_ret = myFoo.do_it(in_arg, inout_arg_par)

This treatment of parameters coincides with the Python Mapping
Specification.	This version, however, contains certain experimental
features which attempt to emulate the "call-by-reference" semantics
implicit in the IDL concepts of out and inout arguments.  Specifically, if
a block is supplied with a CORBA operation invocation, the block will be
executed with all block parameters set to the values of corresponding out
arguments returned by the invocation.  Since Ruby binds block parameters
to any local variables with the same name defined in the enclosing scope
of the block, this can be used to simulate "call-by-reference".  E.g.,

//IDL
interface Foo {
	void do_it(out long arg1, out long arg2, out long arg3);
};
//implementation sets arg1 = 1, arg2 = 2, arg3 = 3

#Ruby
#somehow get reference myFoo to Foo object
arg1, arg2, arg3 = nil
do_it() {|arg1, arg2, arg3|}
p arg1, arg2, arg3 #prints 1, 2, 3

Another feature emulates "call-by-reference" for inout parameters.  If a
parameter to an operation invocation is a Ruby symbol, it is interpreted
as an inout argument, provided the word "inout" is supplied as the last
parameter to the invocation:

//IDL
interface Foo {
	void do_it(inout inout_arg);
};
//implementation adds 1 to inout_arg

#Ruby
#somehow get reference myFoo to Foo object
inout_arg = 10
do_it(:inout_arg, inout)
p inout_arg #prints 11

It could certainly be argued that these features are more confusing
than useful.  I wanted to include them in this first release, however,
to see if anyone preferred operation semantics that more closely emulate
those implicit in IDL.


4. Attributes

IDL attributes are mapped to Ruby methods of the same name.  For ordinary
attributes, a "getter/setter" pair of methods is created; for readonly
methods, only a "getter" method is created.  These follow the ordinary
Ruby naming conventions for attributes, e.g.,

//IDL
interface Foo {
	attribute bar;
	readonly attribute baz;
};

#Ruby
module Foo
  def bar; end
  def bar=(val); end
  def baz; end
end

In the server-side mapping, attr_accessor or attr_reader methods will
be defined in the skeleton module corresponding to an interface:

#Ruby
module Foo_POA
  attr_accessor :bar
  attr_reader :baz
end


5. Constants

For at least two reasons, IDL constants are *not* mapped to Ruby
constants.  In the first place, as we all know, Ruby "constants" are not
really constant.  This is at variance with the "immutable" semantics
implicit in IDL.  In the second place, Ruby constants must begin with
a capital letter.  This is clearly an unreasonable requirement to place
on CORBA implementors who write their IDL with other languages in mind.
Therefore I implement constants as methods returning the value of the
constant.  E.g.,

//IDL
module Foo {
	const float pi	= 3.14159265;
};

#Ruby
module Foo
  def pi
    3.14159265
  end
end


6. Exceptions

Exceptions derive from CORBA::UserException (which derives from
CORBA::Exception, which derives from StandardError).  For each attribute
in the IDL definition, a matching parameter is added to the initialize
method for the corresponding Ruby class and an attr_reader is defined
with the same name.  E.g.,

//IDL
interface Foo {
	exception SystemFailure {
		short err_code;
		string desc;
	};
};

#Ruby
module Foo
  class SystemFailure < CORBA::UserException
    attr_reader :err_code, :desc
    def initialize(err_code, desc)
      @err_code = err_code
      @desc = desc
    end
  end
end


7. Structs

Structs derive, naturally enough, from the standard Ruby Struct class.
For each IDL struct definition, an anonymous Ruby struct class is created
and then installed as a constant in the appropriate module.  E.g.,

//IDL
module Foo {
	struct Person {
		short id;
		string first_name;
		string last_name;
		string address;
	};
};

#Ruby
module Foo
end

struct_class = Struct.new(nil, :id, :first_name, :last_name, :address)
Foo.const_set(:Person, struct_class)


8. Unions

Unions are a bit more complicated.  They derive from a special
CORBA::Union class, whose definition is:

module CORBA
  class Union
    def initialize(discriminator=nil, value=nil)
      @discriminator = discriminator
      @value = value
    end
    def _d
      @discriminator
    end
    def _v
      @value
    end
  end
end

Each subclass of Union has methods corresponding to each attribute in the
IDL definition, which either fetch the value provided the discriminator is
set to the appropriate value, or set the value to the argument supplied
and set the discriminator to the appropriate value for the attribute.
The discriminator cannot be set at all except at initialization.  E.g.,

//IDL
union Foo (switch long) {
	case 0: short a;
	case 1: float b;
	case 2: string c;
	case 3: boolean d;
};

#Ruby
class Foo < CORBA::Union
  def a
    @value if @discriminator = 0
  end
  def a=(value)
    @value = value
    @discriminator = 0
  end
  def b
    @value if @discriminator = 1
  end
  def b=(value)
    @value = value
    @discriminator = 1
  end
  def c
    @value if @discriminator = 2
  end
  def c=(value)
    @value = value
    @discriminator = 2
  end
  def d
    @value if @discriminator = 3
  end
  def d=(value)
    @value = value
    @discriminator = 3
  end
end

No type checking is done when attributes are assigned to; type checking
is currently reserved for marshaling.  This may change in the future,
but seems in accord with Ruby's weakly typed nature.


9. Sequences

Sequences are mapped to Ruby arrays, except in the case of octet or
character sequences, which are mapped to strings for efficiency.  Bounded
sequences are currently not checked for overflow at assignment, but this
could change.  Bounds checking currently only occurs at marshaling.


10. Arrays

Arrays are mapped to Ruby arrays, except in the case of octet or
character arrays, which are mapped to strings for efficiency.  They are
currently not checked for overflow at assignment, but this could change.
Bounds checking currently only occurs at marshaling.


11. Enumerations

IDL enumerations are mapped to two entities in Ruby; first each constant
in the enumeration is mapped just as an ordinary constant would be if
it were defined in the same IDL module, with consecutively increasing
integral values starting from 0; then an array is created in the Ruby
module corresponding to the IDL module, with indices matching the constant
values and string values matching the name of the constant.  E.g.,

//IDL
module Foo {
	enum Color { red, green, blue };
};

#Ruby
module Foo
  def red; 0; end
  def green; 1; end
  def blue; 2; end
  Color = ["red", "green", "blue"]
end

It will be noticed again that the name of an enumeration must begin with a
capital letter, since it is defined as a constant in the enclosing module.
This remains to be resolved by some kind of name mangling mechanism.


12. Typedefs

Typedefs are "unwound" to a definition composed of basic types or
non-typedef user-defined types.  They never should appear in Ruby code.


13. Any

IDL Any values are mapped to objects of class CORBA::Any.  They have
two members, a TypeCode and a value, which are available as read-only
attributes.  Any objects delegate all method calls to the enclosed value,
which will raise a TypeError if it does not respond to the method invoked.
Any objects may be generated by the client by instantiating a TypeCode
appropriate for the value to be inserted, and then calling CORBA::Any.new
with the TypeCode and the value as arguments.  Type checking for Any
currently occurs only at marshaling.


14. TypeCodes

IDL TypeCodes are mapped to a simple Ruby "wrapper" class for the
ORBit native type.  They may be instantiated from Ruby by calling
CORBA::TypeCode.new with the Repository ID of the TypeCode as an argument.
Currently none of the operations in the OMG CORBA::TypeCode interface
are available in Ruby, but this may change if there seems to be reason
to do so (for instance querying the TypeCode member of an Any value
for various attributes).  TypeCode constants corresponding to all the
TypeCodes for basic types are defined in the class CORBA::TypeCode,
with the same names as in the C mapping.


15. The Client-Side Mapping for the ORB Pseudo-Object

The ORB pseudo-object is instantiated by calling CORBA::ORB.init with
any legal arguments defined in the CORBA specification or by ORBit
itself. All the operations defined in the OMG CORBA::ORB interface are
available from Ruby, except for the various TypeCode creation operations.


Examples

Examples are available in the files "test-suite.idl", "test-server.py",
and "test-client.rb".  Since server-side bindings are not yet complete,
you will have to install ORBit-Python to run this test suite (which is
stolen directly from that project, due to lack of time and imagination on
my part).  If there is popular demand, I may also include a server-side
script for Perl (CORBA::ORBit) as well.