[CZT-Devel] Re: [Fwd: Re: Bug in parser?]

[Petra M. <pe...@cs...>]

Hi,

isn't is sufficient to use the Mixfix attribute for this as well?
e1 = e2 ->  <MemPred Mixfix="true">...</MemPred>
e1 \in {e2} -> <MemPred Mixfix="false">...</MemPred>

A MemPred is an equality iff Mixfix=="true" and the second (right) 
expression is a SetExpr
(usually the second argument is a relational operator).

Do I miss something? (perhaps it is too early to think about this ;-)

Petra

Mark wrote:

> Tim wrote:
>
>>>I think the way that the parser does it is correct. If you look at rule 
>>>12.2.10.3, it suggests that e1 = e2 is transformed to e1 \in {e2}, 
>>>because they are equivalent predicates. Also, I've just realised that 
>>>this could cause some problems when "unparsing" a tree... how do we know 
>>>whether the original expression is e1 = e2 or e1 \in {e2}?? Maybe I 
>>>should add an annotation to the any MemPreds that record it if it is an 
>>>equals sign?
>>>      
>>>
>
>I think Tim is right.   (looking at 12.2.10.3)
>
>But he is also right that it is impossible to know how to
>'unparse' this back to an equality.
>
>Also, it seems a bit messy for visitors to figure out that
>they are looking at an equality when they see  E1 \in {E2}
>
>
>This makes me think that, instead of Mixfix (boolean), MemPred should 
>have an attribute something like:
>
>    Kind = Infix | Member | Equals
>
>We need to think about this a bit more before doing anything...
>  
>