Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!ucsd!ucsbcsl!eiffel!bertrand
From: bertrand@eiffel.UUCP (Bertrand Meyer)
Newsgroups: comp.lang.eiffel
Subject: Note on subsequent long message <>
Keywords: Eiffel type system, double precision, C interface, composite objects, copy semantics
Message-ID: <132@eiffel.UUCP>
Date: 2 Apr 89 01:32:37 GMT
Organization: Interactive Software Engineering, Santa Barbara CA
Lines: 58


	The message posted after this one, entitled ``Eiffel types''
and numbered <>, is a draft of an article on the
Eiffel type system, which will be adapted as a part of the 2.2 documentation.
The present message explains the purpose of ``Eiffel types'', which is long.

	I hesitated before posting such a long note as ``Eiffel types'' but it
seems that the potential benefits outweighed the standard objections.

	``Eiffel types'' is NOT the article on ``The complete rules for
type checking in Eiffel'' (``Type checking'' for short) which
many people (something like 300) requested after I mentioned it on
the net some time ago. ``Type checking'' complements ``Object-Oriented
Software Construction'' and the Eiffel User's Manual by explaining
the complete rules for Eiffel systems to be typewise correct. ``Type
Checking'' is mostly of theoretical interest; the problems it addresses,
although visible to people who look at Eiffel from an academic point
of view, arise very seldom in practical Eiffel usage.

	In contrast, ``Eiffel types'' deals with very important improvements,
both theoretical and practical, of the type system. These improvements
are part of version 2.2.

	The main theoretical advance is to make the notion
of type in Eiffel completely regular: all types are now classes; the basic
types (INTEGER, REAL, CHARACTER, BOOLEAN) become fully normal classes of
the Basic Eiffel Library. This approach contains the germ of a theory of
types for which (as is appropriate for an area of applied mathematics having
to do with automatic computing devices) the only predefined type is ``bit''.

	The practical advances are numerous: improvements in efficiency through
avoidance of unneeded dynamic allocations and indirections; double precision
reals (or, for that matter, arbitrary precision reals or integers); more
flexible interface with C and other languages; a simple form of operator
overloading; easier bit string manipulation; etc.

	Because ``Eiffel types'' has to do with the type system of Eiffel, I have
decided to merge it with ``Type checking'' but this will take some time.
This is the reason while I have not mailed ``Type checking'' yet to those
who requested it. That will come. Sorry for the delay. Since ``Eiffel types''
opens possibilities of immediate interest for Eiffel users, it appeared
appropriate to post it without delay.

	``Eiffel types'' is posted in <> in nroff-ed form. This is less
than perfect because the original uses italics, bold face, tables
and other typographical niceties. I considered posting the original troff form
but it depends to heavily on our local macro package.

	If there are any follow-ups they will be somehow forwarded to me
but I won't be able to respond soon, since I will be roaming the world
for the next few weeks with little likely access to news.

	Enough apologizing for one message.

-- 

-- Bertrand Meyer
bertrand@eiffel.com