Re: Ada-95 Success Stories

[chris.morgan@baesema.co.uk (Chris Morgan)]

In article <31a4eb29.1625999@news.cais.com>, mdoernho@cais.com wrote:

> OK, OK, so maybe I was too quick in asking for Ada-95 success stories.

I don't think so, it's just that Usenet does not reach most of the "troops" :-)

My own small contribution is this :

I have experience in Ada and C++. Faced with a project to build a test
suite for our command and control system, I recommended Ada95 and GNAT in
the face of widespread expectation that we would have to use C since there
is a lot of low-level mucking about, and also scepticism that Ada95 is
ready yet.

So far my suggestion has been entirely vindicated. The work is going well.
Staff who are new to Ada find Ada95 and GNAT productive very quickly. Yes
they have some criticisms, but when they went off on a C++ course to help
them understand the C++ GUI the Ada will connect to, they came back
shaking their heads saying they were glad they didn't have to touch it!
(My reaction as well except I do have to touch it - yuck!)

Initially I was asked to stay within Ada83 programming paradigms as we
might have had to ditch GNAT and go back to a Ada83 compiler. Now however
we have confidence in GNAT (especially with the support we get from ACT)
and are gradually slipping in some new language features as appropriate.
So far we've used stream_io, the 'valid attribute, some text_io routines
that are new to Ada95, and I am experimenting with some tagged type
definitions to allow us to replace some horrible variant programming with
nice OOP.

Since the C++ compiler for the GUI is the Sun one (not G++) I have
connected it to the Ada with Ada stream_io. In the C++ I have overridden
each iostream operator for our base types with raw reads and writes rather
than ASCII and it works a treat. Simply using a named pipe in /tmp (Unix)
gives a shared memory connection for good performance.

About the only problem that has really held me up is that our Ada, being
strongly typed, uses a lot of fixed point types. These do not travel well
into the C or C++ domain since those languages do not support fixed point
types. I briefly thought of making a C++ template to support fixed point
types, but in the end my Ada thinks down to C level by converting into an
integer with appropriate explicit scaling in the Ada/C++ interface.

We're just getting into some real-time issues and here we are using some
protected objects, but I am also finding that my experience of Ada83
tasking is helping a lot. Certainly it's easier than using OS based
tasking from C.

This work is giving us as a company confidence to put our large projects
through an Ada95 compiler sooner rather than later as it looks like we
could switch some of our software quite easily, and it looks like there
could be major benefits.

My conclusions are as follows :

o Ada95 is great.

o GNAT is more than ready for serious work right now, and the price is right.

o If you know Ada, get familiar with Ada95 and the available
implementations   (GNAT, PowerAda, AdaMagic, ObjectAda) and consider
transitioning as soon as  possible - the next Ada95 success story could be
yours!

Naturally your mileage may vary (I think that's what the Americans would say).

Chris Morgan