Re: Ada95 BNF

["Martin Dowie" <martin.dowie@nospam.baesystems.com>]

Fair enough, but I still thnk it is strange that it
wasn't defined along the lines of:

defining_program_unit_name ::=
  [ parent_unit_name "." ] defining_identifier

parent_unit_name ::=
  identifier

given that "identifier" already exists and can be
navigated to through the existing definitions in
about 2 clicks and by-passing all the 'invalid'
options anyway...

:-)


Tucker Taft <stt@averstar.com> wrote in message
news:3AF80887.F8942BE3@averstar.com...
> Martin Dowie wrote:
> >
> > Ok... this seems a little strange to me - but if those are the rules!
:-)
> >
> > Is it really that hard to come up with a BNF that does actually match
> > what the language does allow?
>
> In many cases it is preferable to make the BNF somewhat
> "looser" than the "official" grammar, and then do the
> checking as part of semantic checking, because the errors
> tend to be handled a bit more gracefully.  In addition,
> in Ada, there are constructs which look identical
> syntactically but which are distinct semantically, such
> as a type conversion and an array indexing.  The grammar
> in the LRM distinguishes these, whereas the grammar used
> for an LR parser can't, since an LR parser requires that
> the grammar be context-free (i.e. independent of the
> "meaning" of particular identifiers).
>
> --
> -Tucker Taft   stt@avercom.net   http://www.averstar.com/~stt/
> Chief Technology Officer, AverCom Corporation (A Titan Company)
> Burlington, MA  USA (AverCom was formerly the Commercial Division of
AverStar:
> http://www.averstar.com/services/ebusiness_applications.html)