From: Michael Neumann <mneumann@...>
Date: 2007-09-22T17:27:20+09:00
Subject: Re: CplusRuby - Gluing C and Ruby

Konrad Meyer schrieb:
> Quoth Konrad Meyer:
>> Quoth Daniel Berger:
>>> On Sep 21, 1:01 pm, Michael Neumann <mneum...@ntecs.de> wrote:
>>>> Hi all,
>>>>
>>>> I am proud to announce my latest project called CplusRuby.
>>>> Below I pasted it's README for further explanation.
>>>> Alternatively read the following blog article:
>>>>
>>>> http://www.ntecs.de/blog/articles/2007/09/21/cplusruby-gluing-c-and-ruby
>>> <snip>
>>>
>>> First patch!
>>>
>>> C:\Documents and Settings\djberge\My Documents\My Downloads\Ruby>diff -
>>> u cplusruby.orig cplusruby.rb
>>> --- cplusruby.orig      Fri Sep 21 15:50:38 2007
>>> +++ cplusruby.rb        Fri Sep 21 15:49:04 2007
>>> @@ -11,6 +11,10 @@
>>>  #
>>>  # Implements a simple ordered Hash
>>>  #
>>> +require 'rbconfig'
>>> +require 'win32/process' if RUBY_PLATFORM.match('mswin')
>>> +include Config
>>> +
>>>  class OHash < Hash
>>>    def []=(k, v)
>>>      @order ||= []
>>> @@ -146,20 +150,28 @@
>>>      dir = File.dirname(file)
>>>      mod, ext = base.split(".")
>>>
>>> +    make = RUBY_PLATFORM.match('mswin') ? 'nmake' : 'make'
>>> +
>>>      File.open(file, 'w+') {|f| f << self.generate_code(mod) }
>>>      Dir.chdir(dir) do
>>> -      system("make clean")
>>> +      system("#{make} clean")
>>> +
>>>        pid = fork do
>>>          require 'mkmf'
>>>          $CFLAGS = cflags
>>>          $LIBS << (" " + libs)
>>>          create_makefile(mod)
>>> -        exec "make"
>>> +        exec "#{make}"
>>>        end
>>>        _, status = Process.waitpid2(pid)
>>> -      raise if status.exitstatus != 0
>>> +
>>> +      if RUBY_PLATFORM.match('mswin')
>>> +         raise if status != 0
>>> +      else
>>> +         raise if status.exitstatus != 0
>>> +      end
>>>      end
>>> -    require "#{dir}/#{mod}.so"
>>> +    require "#{dir}/#{mod}." + CONFIG['DLEXT']
>>>    end
>>>
>>>    def self.generate_code_for_class
>>>
>>> Mostly fixes MS Windows issues, but the DLEXT bit also fixes platforms
>>> that don't generate .so files (OS X, HP-UX).
>>>
>>> Regards,
>>>
>>> Dan
>> Does the C->ruby mapping map between char * and Strings?
>>
>> Thanks,
>> -- 
>> Konrad Meyer <konrad@tylerc.org> http://konrad.sobertillnoon.com/
> 
> Okay, so I have added the following to DEFAULT_OPTIONS:
> 
>     :int => {
>       :default => '0',
>       :ctype => 'int %s',
>       :ruby2c => '(int)NUM2INT(%s)',
>       :c2ruby => 'INT2NUM(%s)'},
> 
>     :char_p => {
>       :default => '""',
>       :ctype => 'char *%s',
>       :ruby2c => 'StringValuePtr(%s)',
>       :c2ruby => 'rb_str_new2(%s)'} 
> 
> My code is like this:
> 
> #---------------------start
> 
> require 'cplusruby'
> 
> class BankAccount < CplusRuby
>   property :balance,  :int
>   property :name,     :char_p
> 
>   method_c :withdraw, %{int dollars}, %{
>     if (dollars <= selfc->balance){
>       selfc->balance -= dollars;
>       return true;

I think you shouldn't return "true" here. At least it's
not indended, as the withdraw function is "void".


>     }
>   }
> 
>   def initialize
>   end
> end
> 
> CplusRuby.evaluate("foo.c", "-Os", "")
> 
> n = BankAccount.new
> puts n.balance
> puts n.name
> 
> #---------------------end
> 
> When I run it with ruby I get a whole bunch of errors stemming from
> the fact that it generates C code that uses a type called
> "BankAccount" without defining it anywhere (struct BankAccount is
> defined though).

I think C does not yet work. Try C++ :)

CplusRuby.evaluate("foo.cc" ...)

But it should be easy to fix, by adding a "typedef struct", and
"VALUE(*)()" instead of "VALUE (*)(...)".

Ah, and returns values are not yet implemented. Nevertheless you can
assign to a property and use it as a return value.


Thanks,

   Michael