[#3109] Is divmod dangerous? — Dave Thomas <Dave@...>

14 messages 2000/06/06

[#3149] Retrieving the hostname and port in net/http — Roland Jesse <jesse@...>

Hi,

12 messages 2000/06/07

[#3222] Ruby coding standard? — Robert Feldt <feldt@...>

16 messages 2000/06/09

[#3277] Re: BUG or something? — Aleksi Niemel<aleksi.niemela@...>

> |I am new to Ruby and this brings up a question I have had

17 messages 2000/06/12
[#3281] Re: BUG or something? — Dave Thomas <Dave@...> 2000/06/12

Aleksi Niemel<aleksi.niemela@cinnober.com> writes:

[#3296] RE: about documentation — Aleksi Niemel<aleksi.niemela@...>

> I want to contribute to the ruby project in my spare time.

15 messages 2000/06/12

[#3407] Waffling between Python and Ruby — "Warren Postma" <embed@...>

I was looking at the Ruby editor/IDE for windows and was disappointed with

19 messages 2000/06/14

[#3410] Exercice: Translate into Ruby :-) — Jilani Khaldi <jilanik@...>

Hi All,

17 messages 2000/06/14

[#3415] Re: Waffling between Python and Ruby — Andrew Hunt <andy@...>

>Static typing..., hmm,...

11 messages 2000/06/14

[#3453] Re: Static Typing( Was: Waffling between Python and Ruby) — Andrew Hunt <andy@...>

32 messages 2000/06/16

[#3516] Deep copy? — Hugh Sasse Staff Elec Eng <hgs@...>

Given that I cannot overload =, how should I go about ensuring a deep

20 messages 2000/06/19

[#3694] Why it's quiet — hal9000@...

We are all busy learning the new language

26 messages 2000/06/29
[#3703] Re: Why it's quiet — "NAKAMURA, Hiroshi" <nahi@...> 2000/06/30

Hi,

[#3705] Re: Why it's quiet — matz@... (Yukihiro Matsumoto) 2000/06/30

Hi,

[ruby-talk:03601] Re: Interface polymorphism

From: Aleksi Niemel<aleksi.niemela@...>
Date: 2000-06-22 00:23:16 UTC
List: ruby-talk #3601
> I have not given up on the idea of implementing a DBC module in
> Ruby. But every time I get into it, it makes my brain itch. I
> want to (have to!) make the conditions heritable, but I want to
> make it easy to use as well...

I've given some thought to this too and my conclusion, so far, has been to
change the Ruby interpreter code.

class Gem
  use DBC
  @carats
  attr_accessor :carats
  def invariant
    @carats > 10
  end
  def pre_value(price)
    price > 0
  end
  def value( price )
    @carats * price
  end
  def post_value(price, ret)
    ret > 40  # a little bit made up condition
  end
end

class Ruby < Gem
  def value( price )
    @carats ** price
  end
end

class PieceOfWorthlessGlass < Gem # like in nethack
  def invariant
    @carats > 0
  end
end 

stone = Gem.new
stone.carats = 5  # should raise InvariantFailed
stone.carats = 15
stone.value( 0 )  # should raise PreConditionFailed
stone.value( 1 )  # should raise PostConditionFailed
stone.value( 2 )  # should raise PostConditionFailed
stone.value( 3 )

stone = Ruby.new
stone.carats = 5  # should raise InvariantFailed
stone.carats = 15
stone.value( 0 )  # should raise PreConditionFailed
stone.value( 1 )  # should raise PostConditionFailed
stone.value( 2 )  
stone.value( 3 )

stone = PieceOfWorthlessGlass.new
stone.carats = 5  
stone.value( 0 )  # should raise PreConditionFailed
stone.value( 1 )  # should raise PostConditionFailed
stone.value( 2 )  # should raise PostConditionFailed
stone.value( 3 )  # should raise PostConditionFailed
stone.value( 9 )  

To make this happen we can't use the trick to execute some
'changeMethodCallsToDBCStubs()' directly (as Ruby code) as we discussed in
earlier threads. It will break with inheritance and even when we add
something to the class.

So the options are (should I say include):
1) to change the method calling to check for DBC functions (pre_method,
post_method and invariant) and the normal first singleton, self, modules and
parents order works out the inheritance of DBC.
2) to change the inner workings of 'def' to replace automatically real
methods with {pre(args) method(args) post(args) invariant} type of code 

I think I prefer the second one if it's applicable at all. The performance
penalty would be minimal (or as small as possible :?).

For those who're going to note that the second way wouldn't work when we add
special DBC-checkers to singleton classes or just afterwards, I could say
that it does work. 

So imagine there has been no pre_value DBC precondition checker written
earlier (I'm too tired to write real example). When "executing" the 'def' in

class << stone
  def pre_value(price)
    price > 100
  end
end

will restub the value by executing

class << stone
  def value(price)
    raise PreConditionFailed unless pre_value     # ###
    orig_value(price)
    raise PostConditionFailed unless post_value
    raise InvariantFailed unless invariant
  end
end

behind the curtains. The earlier stub didn't have the marked line before,
but now we see that there exists a proper pre_condition checker we can use,
thus the new version incorporates new call.

Even this version isn't elegant for various reasons, but most notably
because Class.methods would list DBC-methods too. But this is probably minor
thing now (and probably can be hacked easily).

In any case, I'd like to make it clear that I would be really happy to see
Ruby with more or less built-in DBC support (widely used).

	- Aleksi

In This Thread

Prev Next