Rationale for Ada 2005
1.4 Conclusions
This overview of Ada 2005 should have given the reader
an appreciation of the important new features in Ada 2005. Some quite
promising features failed to be included partly because the need for
them was not clear and also because a conclusive design proved elusive.
We might think of them as Forthcoming Attractions for any further revision!
Some esoteric topics have been omitted in this overview;
they concern features such as: streams, object factory functions, the
partition control system in distributed systems, partition elaboration
policy for high integrity systems, a subtlety regarding overload resolution,
the title of Annex H, quirks of access subtypes, rules for pragma Pure,
and the classification of various units as pure or preelaborable.
The remaining chapters will expand on the six major
topics of this overview in more detail.
It is worth briefly reviewing the guidelines (see
Section
1.2 above) to see whether Ada 2005
meets them. Certainly the Ravenscar profile has been added and the problem
of mutually dependent types across packages has been solved.
The group A items were about real-time and high-integrity,
static error checking and interfacing. Clearly there are major improvements
in the real-time area. And high-integrity and static error checking are
addressed by features such as the overriding prefix, various pragmas
such as Unsuppress and Assert
and additional Restrictions identifiers. Better
interfacing is provided by the pragma Unchecked_Union
and the Mod attribute.
The group B items were about improvements to the
OO model, the need for a Java-like interface feature and better interfacing
to other OO languages. Major improvements to the OO model are brought
by the prefixed (Obj.Op) notation and more
flexible access types. The Java-like interface feature has been added
and this provides better interfacing.
The final direct instruction was to incorporate the
vectors and matrices stuff and this has been done. There are also many
other improvements to the predefined library as we have seen.
It seems clear from this brief check that indeed
Ada 2005 does meet the objectives set for it.
© 2005, 2006 John Barnes Informatics.
Sponsored in part by: