Re: Digest Number 492

Eric wrote:
> Thomas Beale wrote:
> > in YY_PERSER_SKELETON, the definition of syntax_error is as follows:
> >
> > syntax_error: BOOLEAN is
> > -- Has last parsing been unsuccesful?
> > do
> > Result := yy_parsing_status /= yyaccepted
> > end
> >
> > However, this means that syntax_error returns True even if no parsing
> > has yet been done, since just after creation, yy_parsing_status = 0, 
> not
> > yyAccepted (I just tested this, to be sure).
>
> I'm not sure I understand your point. The comment talks
> about the "last parsing". Therefore the value is undefined
> if no parsing occurred. To be more precise, the value is
> only relevant after a parsing occurred. Perhaps a precondition
> is missing, but there are many similar cases in EiffelBase,
> such as:
Eric,

I agree, if we believe the comment in the strictest possible way, but 
think of it in formal terms: if I call this routine before having done a 
parse (as is could be done in a generic parser handler routine), it 
returns True, meaning "syntax error occurred" even though this clearly 
is not the case. I think that the formal definition has to change, 
either with a precondition, or (my preference) relaxing the list of 
parsing_status values that is matched. I only believe in formal 
specifications! (and I wouldn't see ES's file handling routines as any 
kind of model to follow;-)

- thomas beale