CU Amiga Super CD-ROM 2/CU Amiga Magazine's Super CD-ROM 02 (1996)(EMAP Images)(GB)[!][issue 1996-04].iso/magazine/amiga_e/amigae.jan.archive/000052_crash!fwi.uva.nl!oortmers_Sun, 23 Jan 94 01:05:00 PST.msg

Text File | 1994-02-17 | 7KB | 157 lines

Received: by bkhouse.cts.com (V1.17-beta/Amiga)
      id <[e-mail]>; Sun, 23 Jan 94 01:05:00 PST
Received: from mail.fwi.uva.nl by crash.cts.com with smtp
    (Smail3.1.28.1 #18) id m0pNtwH-0001OKC; Sat, 22 Jan 94 17:48 PST
Received: from gene.fwi.uva.nl by mail.fwi.uva.nl with SMTP (5.65c/5.1)
          id AA17581; Sun, 23 Jan 1994 02:48:30 +0100
Received: by gene.fwi.uva.nl
          id AA25036; Sun Jan 23 02:48:28 1994
Message-Id: <[e-mail]>
Date: Sun, 23 Jan 1994 02:48:28 +0100
Return-Path: <[e-mail]>
X-Organisation: Faculty of Mathematics & Computer Science
                University of Amsterdam
                Plantage Muidergracht 24
                NL-1018 TV Amsterdam
                The Netherlands
X-Phone:        [phone]
X-Telex:        16460 facwn nl
X-Fax:          [phone]
From: [e-mail] (Wouter van Oortmerssen (Alfa_Inf92))
To: [e-mail]
Subject: E Statement, v5


                           STATEMENT, v5
           -> with additions about newest developments <-

                      The future of Amiga E

                    by Wouter van Oortmerssen

The question I get most from programmers is: when will the new version
of E be out, what it will cost and what are it's features? since the
answer is of interest to just about anybody working with E, I'll try to
summarise it here.

First, my aim was to release a v2.5, however, I implemented a lot
of stuff that was originally meant for v3.0, so now I'm skipping v2.5,
and v3.0 will be the first release after v2.1b.
The drawback ofcourse is that it will yet take a little longer before
a new version comes out, and the good news is that we'll have v3.0
features like full OO-ness relatively early. but more to that later.

The most current _beta_ version of E is v2.7a, and already has the
following features:

- compilation to modules, one of the most important enhancements.
  this allows you to write sources with OBJECTs, PROCs and CONSTs etc
  in them, and have them compiled to a ".m" file for use in other programs.
  any thinkable datahiding scheme supported.
  The module system is very flexible and easy to use, and allows for
  extremely fast linking.
- error reconstruction: EC now exactly pinpoints the spot in the
  source that caused the error
- optimisations in the compiler: EC is significantly faster than
  EC2.1b, even with all the new features and code-optimisations.
- various optimisations in code-generation
- Parts of the compiler have been rewritten and datastructures
  reorganised for efficiency, this results in a reduction in EC's
  memory consumption with 40% to 50%
- default arguments a la C++: allows you to give parameters default
  values incase the caller leaves out an argument.
  PROC bla(a,b=1,c=NIL:PTR TO extrainfo)
- kick 3.0 modules
- object inheritance (see below for other OO features)
  lets you inherit object members and methods from other objects,
  and create hierarchies of datatypes with shared behaviour
  OBJECT circle OF location
- multiple returnvalues and multiple-value assignments.
  functions can now return more than one value, and assignment
  statements can store in more than one var.
  a,b,c:=func(); RETURN 1,2,3
- symbolhunks for use with profiler, debuggers etc.
- .o to .m file converter (for assembly)
- numerous small bugs fixed and small enhancements made
- register allocation
  v2.1b did hardly put variables in registers, thus code that depends
  largely on the time needed to access a variabele would be slower.
  EC can either compute the best possible register usage itself,
  or the programmer can assign registers.
- various new code optimisations, for different sorts of expressions,
  library calls etc. also some special opti's for 020+ machines.
- module caching: this feature allows EC to keep modules
  in memory between compilations. effectively this means that
  the first time you compile a source which needs 10 modules
  EC will load them all from disk, the second upto the n-th
  time it doesn't have to load any. Because the new E can have
  huge apps seperated over many modules, compile/link times for
  large programs can be incredibly fast.
- the NEW operator allows for quick and easy dynamic
  memory allocation, with automatic exception raising.
  NEW p; NEW [1,a,bla()]:obj
- increasing number of great modules for all sorts of tasks. the plan
  is that E will provide an enormous amount of these modules to allow
  quick programming of large, quality apps. v3.0 will _probably_ already
  have modules for:
  * easy font sensitive resizable auto-organising GUI's
  * both simple 3d vectors and fully fledged realtime
    double-buffered 3d filled vectors
  * protracker player
  * iff loader/saver etc.
  * hashing
  * easy file access
  * lexical analysing
  * garbage collected LISP-lists
  * all sorts of Amiga.lib and amiga support routines
  and much more. (this is just an example selection)
- a new SELECT OF that understands complex CASE's like
  CASE "a" TO "z", "A" TO "Z", "0" TO "9", "_"
  and is also much faster than the other SELECT
- Unification: this is a feature normally only found in
  high-level functional and logic programming languages. It
  allows for the pattern-matching of nested structures in
  expressions, and bind variables along the way. When used
  properly, it can greatly reduce programming time to implement
  complex algorithms involving dynamic datastructures.
  IF exp <=> [BLA,[BLERK,y,0],x] THEN ...
- lots of great external utilities:
  * Build, a full featured Make utility for E
  * E-Yacc (and probaly also E-Lex), parser and lexical
    analyser generators for E
  * o2m, converts your macro-assembler .o files to modules
- Datahiding with public/private.
  allows members and methods to be private, i.e. invisable for
  the user of a class.
- virtual methods and method inheritance.
  in the E Object Model, object-polymorhism plays a bigger role than
  in languages like C++ which also have fake (non-virtual) methods.
  A method in E is a PROC defined on an object, however it functions
  quite different from a PROC since which methods are invocated on
  an objects is determined at run time.
- much more free format source
- ability to dereference pointers beyond just one ".", new types
  allowed in OBJECTs.
- (very likely) an absolutely GREAT editor written in E, called EE
  (courtesy of Barry Wills), from which you can edit and compile
  your E sources. It does all the great things any modern editor
  can do an besides that it has many E specific features, such as
  folding on PROCs etc. a must-have.

(still missing from this are REALs, but they'll definitly be in the
next release.)

Last but not least there is the question what 3.0 will cost. Well,
I don't know myself: I'd like to sell it for a very low-profit price
so that everybody can easily affort it. probably this won't be much
more than the price for a shareware compiler. Also I'll plan on
distributing a PD demo version of v3.0. When I feel like it, I might
also let a commercial publisher take care of distribution etc., when
a suitable one comes with a good deal.

so, I hope all this information gave some of you an insight of what's
coming...

Greetings to all E programmers in the world,

Wouter.