Re: Language Translators

["Brian G. Holmes" <brian.holmes@gsc.gte.com>]

Robert Dewar wrote:
> 
> Brian says
> 
> <<If you can compile that C code as C++ then we have a translator that can
> convert it to Ada95.  Most C code will compile using a C++ compiler.
> When one says Ada I assume that the subject is Ada95 these days.  The
> Ada95 produced by our translator might vary from one that produced Ada
> from C code; the translation of a struct to an extendible type, for
> example.  Those variances would represent the C++ improvements made to
> C, so they might be helpful to those who enhance or maintain the Ada
> code.
> >>
> 
> I hope you don't translate *all* structs to tagged types, this would be
> an inappropriate choice.

All structs with non-private constructors may be created and extended. 
A strict translation would have to support extension for all of those
structs.  The translator cannot make subjective decisions regarding
whether or not a specific struct should or should not be extended.  As
such, a translation based on a tagged type is the appropriate
translation.

We have chosen to base our class and struct translations on controlled
types.  The advantages of such a translation are many, and include
strong support for the translation of constructors and destructors.  The
disadvantages of such a translation include the fact that we will be
unable to limit the creation and extension of all structs and classes in
the same way that a private constructor does.

BGH