From: James Britt <ruby@...>
Date: 2005-09-18T23:48:34+09:00
Subject: Re: proposed attr writer patch

Daniel Berger wrote:
> --- Mathieu Bouchard <matju@artengine.ca> wrote:
> 
> 
>>On Sun, 18 Sep 2005, James Britt wrote:
>>
>>
>>>attr_accessor :foo?, :bar!
>>>And would it not be better to just have people
>>
>>learn proper Ruby, rather than
>>
>>>add yet another exception to how Ruby behaves?
> 
> 
> The current behavior is broken.  

Broken? Does that not depend on a certain point of view?  Is it broken 
because Ruby does what it is told to do, rather than what the user may 
have meant it to do?

I believe "attr_accessor <some_symbol>" follows a simple definition.

It creates

def <some_symbol>=( new_val )
  @<some_symbol> = new_val
end

and

def <some_symbol>
  @<some_symbol>
end


But one has to use pick reasonable values for <some_symbol>, and 
understand that a method named foo?= is probably not what is wanted.

Basically, you get what you ask for, so ask wisely.


 > So, it should either
> do something useful or raise an error.  I vote the
> former, though I would accept the latter.

I'd prefer that latter, if the results are untenable Ruby, rather than 
masking the mistakes of a user by having the interpreter, um, interpret 
what the user *meant* to do.

>  
> 
>>It can be said that Ruby has so many exceptions that
>>just another 
>>one won't noticeably increase the number of them.
>>
>>It can also be said that Ruby has so many
>>exceptions, far too 
>>many of them, that it's important to stop the
>>invasion of more exceptions.
> 
> 
> Again, I consider this a bug fix, not an exception,
> since the current behavior creates syntactically
> invalid methods.  So, we can make it work as things
> stand now, raise an error, or we can alter Ruby itself
> to understand "?=" and "!=" as valid syntax (as you
> mention in ruby-core:4406).
> 
> So, unless either of you have got a better idea (and a
> patch), I'm not sure what you're complaining about.
> 

I'm generally against languages attempting to guess what I meant to do 
and then quietly doing something other than what I told it to do.  If 
this construct can create invalid Ruby, won't there be an error when 
someone tries to call the invalid Ruby?

I can see the value in raising an exception right off the bat, though 
I'm unclear what problem this is supposed to correct.

If someone is writing

   attr_accessor :foo?, :bar!

then the problem may not be that Ruby doesn't try to make sense of this, 
but rather that someone is writing code like that.


James