NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/rexx/130

Internet Message Format | 1996-08-06 | 5.2 KB

Path: news.ox.ac.uk!news
From: [e-mail] (Ian Collier)
Newsgroups: comp.lang.rexx,comp.lang.perl.misc
Subject: Re: Perl to Rexx Conversion
Date: 8 Jan 1996 16:57:54 GMT
Organization: Oxford University Computing Laboratory
Distribution: inet
Message-ID: <[e-mail]>
References: <4aq2ad$[e-mail]> <4at9hp$[e-mail]> <[e-mail]> <4b5cmg$[e-mail]>
NNTP-Posting-Host: boothp2.ecs.ox.ac.uk
X-Local-Date: Monday, 8th January 1996 at 4:57pm GMT

Well it's been a while but I haven't seen a detailed answer to this so
here goes.  (Actually I did reply to the email copy of this but it got
returned a few days ago, so this is more or less the same as that).

In article <4b5cmg$[e-mail]>, Tom Christiansen
<[e-mail]> writes:
> [courtesy cc of this posting sent to cited author via email]
[obviously I haven't bothered resending this by email :-) ]

> In comp.lang.Perl.misc,
>     [e-mail] (Ian Collier) writes:

> Nonetheless, Rexx is not so clearly a structured programming language,
> you know.  What about this:
>
>      do i=1 to 10
>         if i<8 then
>            iterate i
>         else
>            signal somewhere
>
>      somewhere:

Ahem, you can write spaghetti code in almost any language!  That doesn't
prove much.  For what it's worth, "signal" is supposed to be for an
*abnormal* change in the flow of control and is deprecated as a "goto"
instruction.  You cannot, for example, jump to a point that is inside a
loop (even the same one that you are in at the moment) and expect it to
continue looping.

> Since you bring up the matter, I was wondering about a few things:

As you probably know, Rexx has been fixed by an ANSI standard at
approximately the level it was at in 1990 with a few small extensions.
Object Rexx has been developed by IBM and is in beta testing (a public
beta is on IBM's DevCon CD).  IBM plans to release a free source code
version of Object Rexx for Linux.  I have seen demonstrations, but I
have not used the product.

> Have they nailed down what lines() does yet, or is it still challenging
> to check for eof()?  Can you explicitly open() and close() files and
> pipes yet?

The standard specifies an action for lines(), although it will be some
time before the majority of implementations become conforming.  There
is no standard way to open and close files, though the stream() function
provides a hook for the implementor to provide this, and the OS/2 syntax
will probably become a defacto standard.  Checking for eof is still messy,
though it can be done for example by calling stream() to get the file's
status.

> Is it easy to load various and multiple functions from other files
> yet?

If these files are Rexx files and if you want more than one of them in a
file, then the answer is no, unless you are using Object Rexx, in which
case I believe you can.

> Does Rexx have lexical scoping yet, or is all dynamic still?

I think one would say it is all dynamic, although in most programs it is
hard to tell the difference.

> What about real arrays?  I don't just mean using word() or words() or
> wordposs() and splitting on nothing but white space.  I mean real first
> class lists.  It would also be nice to split on a pattern, like say
> /\s*[;,]\s*/.  Can you sort your arrays?

Object Rexx has array objects but I don't know much about them.  I assume
operations like split, join and sort are possible on these but I've no idea
whether they are built in.

> Are there associative arrays yet?  Do you have things like keys(),
> values(), each(), defined(), exists(), and delete()?  What about
> defining a key retrieval ordering?

Associative arrays have always existed, as has defined() (though it is
called symbol()).  It has always been possible to delete variables (or
elements of an associative array) using the drop instruction.  Exists()
I've never heard of.  Object Rexx has a "do over" construction which could
be used to emulate the others.  It works for all kinds of objects, not just
associative arrays.

> Can you pass things by reference or by value at your discretion?
> Is this implicit or explicit?

Object Rexx has argument passing by reference.  The caller has no special
syntax but the callee uses different syntaxes depending on whether it wants
the parameters by value or by reference.  Of course the only way to pass a
complex object like an associative array is by reference.

> Does it have regexps yet or is it just that clunky parse stuff?

No regexps, I'm afraid.  Some day, someone somewhere will write some.  Until
then, parse isn't _that_ bad.

> Can you dynamically load C functions?  Can you test for whether
> a function is defined or not?

There is no standard syntax for doing it while the Rexx program is
running, but several implementations provide some.  The OS/2 syntax
will again probably prevail.  It is also specified in the standard that
the interpreter should have an API which allows the programmer to add some
functions to the interpreter before starting it up.  The API has a routine
which tests whether a function has been dynamically loaded; this call is
also available to the OS/2 Rexx programmer.

Ian Collier - [e-mail] - WWW Home Page (including REXX section):
http://www.comlab.ox.ac.uk/oucl/users/ian.collier/index.html