Re: BUG: Struct.new(:a?).instance_methods

On Thu, 10 Feb 2005, Hugh Sasse Staff Elec Eng wrote:
> On Thu, 10 Feb 2005, Yukihiro Matsumoto wrote:
> > Right.  "a?" should not be allowed for a struct member name, since
> > there cannot be a "a?=" method.  Here's the patch.
> Matz, would it be possible to change the message to guide people to
> what is correct, please?
>   + rb_raise(rb_eNameError, "`%s' is not a proper name for a struct 
> member, use /\w+/",
>   or something of that sort?

Also, attr_accessor and its relatives should be subject to the same thing.
Currently (I use 1.8.1), attr_accessor :foo? creates both a :foo? method
and a :"foo?=" method, which shows the latter is possible, although it's
cumbersome to call, because not supported directly by ruby syntax.

But isn't it better to instead change the implementation so that a?= is
allowed?... It would be also good to allow @foo? instance variables. All
those things are possible, it's just that the parser puts unnecessary
complications in what's allowed and what's not.

I say that because currently the "?" suffix is much less useful than it
could be. In fact, if I really insisted on using a suffix for predicates
(boolean functions) I'd use the letter p instead (for example), because at
least it can be used uniformly in any context.

In short, according to POLS (Principle of Least Segregation), I would like
the visible minority of ?-names to be treated with equality.

_____________________________________________________________________
Mathieu Bouchard -=- Montr饌l QC Canada -=- http://artengine.ca/matju