From: Kristof Bastiaensen <kristof@...>
Date: 2004-04-26T18:34:17+09:00
Subject: Re: RCR: Comparing multiple values

>>For me that notation has one drawback: what should happend if
>>        y = 5
>>and you'll write
>>        [x, y == 1, 2]
>>?
>>
>>Is it an array with three elements
>>        [x, false, 2]
>>?
>>
>>or an array with one element which has value of "multiple comaprision"
>>        [false]
>>?

That would be a problem indeed.  Note that the same aplies
to method calling: foo(a, b == 1, 2)

The expansion of (x, y == 1, 2) should then occur only outside
parameter lists.

> 
> It should be  just 'false', as expressed in the first message (i.e.
> x==1 and 5==2)
> 

It surely shouldn't be false, because that would disable 
comparisons inside parameter lists.

> I mean, if we consider ',' as an expression that creates a list
> object, than 1,2 == could be just some kind of sugar.
> But chosing some kind of precedence would probably be impossible, or
> wqould break many of the actual behaviours.

I wouldn't consider ',' as an expression that creates any object,
but rather as a way to pass multiple arguments to methods and
continuations. This is close to the way it actually is in Ruby.

This would be a syntax extension.  However (x, y == exp1, exp2) 
would be different from (x == exp1 && y == exp2) because it would 
evaluate the parameters first, and then compare the results.

Another construct that might be useful:
x, y += 1, 2
#(meaning x += 1; y += 2)