Path: news.cs.au.dk!news.net.uni-c.dk!howland.erols.net!newsfeed.cwix.com!206.172.150.11!news1.bellglobal.com!news.uunet.ca!not-for-mail
From: Karl Waclawek <waclawek@idirect.com>
Newsgroups: comp.lang.beta
Subject: Re: How to do low level stuff
Date: Thu, 28 Jan 1999 21:56:35 -0500
Organization: UUNET Canada News Transport
Lines: 81
Message-ID: <36B12363.12951978@idirect.com>
References: <368B8A70.CF9934FF@thestar.ca> <yz9g19nqf7e.fsf@heimdal.mjolner.dk> <36950D20.2FAD4D18@thestar.ca> <yz9r9t6hsww.fsf@fenris.mjolner.dk> <kolbjorn.aambo-2801991028020001@ubmac86.uio.no>
NNTP-Posting-Host: 207.136.101.92
Mime-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit
X-Mailer: Mozilla 4.5 [en] (Win95; I)
X-Accept-Language: en
Xref: news.cs.au.dk comp.lang.beta:11803



"Kolbjørn Aambø" wrote:
> 
> In article <yz9r9t6hsww.fsf@fenris.mjolner.dk>, Morten Grouleff
> <mg@mjolner.dk> wrote:
> 
> > Karl Waclawek <kwaclaw@thestar.ca> writes:
> >
> > [...]
> :
> > In the next release of the Mjolner System, some new basic types are
> > introduced. These include signed and unsigned integers of sizes
> > 8,16,32,64 bits.  There will also be a 16 bit wchar which is intended to
> > hold unicode chars.
> :
> This is just great.... asked for that some years a go by the way....
> 
> But there is one additional thing you just have to adress to get BETA to
> become a more useful language: Lets give variables an ENDIAN based type!
> 
> bsUInt64 (bigsmallUnsigned64bit int) Big Endian Unsigned 64 bit.
> sbUInt64 (smallbigUnsigned64bit int) Little Endian Unsigned 64 bit.
> same for all other types....
> 
> Cant you just rename wchar into Unicode:
> bsUnicode
> sbUnicode
> 
> If you followed that advice you could make an Internet Communication
> package where everything is done big Endian, for instance. Let the
> compiler do bs->sb and sb->bs implicit and issue warnings for converting
> overhead if neccesary....

I am not sure that that would be in the spirit of BETA, although I agree
that the memory layout of data should be something one should be able
to control. Once a BETA program has to face the real world, i.e. interfaces
to data with arbitrary storage formats and semantics, it becomes necessary
to convert between the nice and orderly academic world of BETA and the nasty
outside world. ;-)

Anyway, instead of special data types, would it not be more generic to provide
storage layout conversion functions for the basic data types?
Since basically all the world nowadays stores data in terms of 8-bit bytes,
once could get away with functions that simply convert any basic data type
into a byte array with specified layout.

Example for 64bit integers (let's call them Integer64):

Int64ToBytes:
  (# Int64: @Integer64;
     Layout: ^LayoutSpecifier;
     Bytes: [8] @Byte;
  enter (Int64, Layout[])                
  do <platform specific code>; INNER            
  exit Bytes
  #)

The "do" part would depend on the platform. You could also use inheritance
and extend the INNER part instead of using a layout specifier, like 

Int64ToBytesBE: Int64ToBytes(# do <big endian code> #);

So, if you want to place an Integer64 into a byte array at position 17
in big endian format, you would do something like:

  (MyInt64, BigEndianLayout[]) -> Int64ToBytes -> MyBuffer[17:24];
or
  MyInt64 -> Int64ToBytesBE -> MyBuffer[17:24];

Currently, however, it seems that you cannot assign to repetition slices,
another language feature that would be be useful.

Karl


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
   Karl Waclawek
   KD Soft Inc.
 * Phone:  (905) 579-3443
 * E-Mail: waclawek@idirect.com