Re: [ANN] an EBNF parser and coding pattern tool (LGPL)

[codeworker@free.fr (Cedric LEMAIRE)]

Anthony Williams <anthony.williamsNOSPAM@anthonyw.cjb.net> wrote in message news:<clcm-20030207-0005@plethora.net>...
> codeworker@free.fr (Cedric LEMAIRE) writes:
> 
> > And C++ templates are too restricted to do generic programming and
> > code generation in a large scope. I'll give examples later if some
> > people react to this assertion.
> 
> OK, I'll bite ... why do you think C++ templates are too restricted?

The C++ compiler doesn't give you introspection to implement C++
templates that
work on the composition of a class.

How to write with C++ templates an "Object Unmarshaller", for
instance? I mean, something that reads the serialized content of an
object and that builds the corresponding object in memory. Be careful!
No intervention must be done on these objects in C++ to facilitate
their integration into the "unmarshaller".

How to write an "Object Walker", something that enables to navigate
amongst the object attributes of the instances, for applying a task on
them (starting from a given object: to initialize all of them or to
serialize all of them or describing them in HTML to query the
application from a browser ...)?

I want to be able to write the generic code of these features as
easily as if I was implementing them by hand, for each object to
handle. But the 'coding pattern' must be written once only and must be
easy to support. And not to be an intellectual exercise for the
pleasure of the brain.

And it is why I said that C++ templates are too restricted (I could
have said: "... and not very readable on complex features" but I
haven't because it is a question of taste I imagine).

-- Cedric
-- 
comp.lang.c.moderated - moderation address: clcm@plethora.net