🚀 go-pugleaf

I'm part of the IBM Prolog development group, so I guess I
should try to put the record straight on IBM Prolog.  
It's a lot better than what some people allege it to be.
In fact, I'd go so far as to say that it's the best Prolog
on the market in terms of features, performance, and robustness.
(But I'm biased ...)

--- Disclaimer: the following are my own opinions; they have
		not been approved by IBM management or IBM
		lawyers; and I take full responsibility
		for any mistakes. 

First of all, there are *two* IBM Prologs.  One is now obsolete.
It implemented Grant Robert's Waterloo Prolog syntax (although
it didn't use Grant's implementation techniques).  Remember
that this was done back in the Dark Ages when many people
had only punched cards for input.  So, the standard syntax was
UPPER CASE.  "Ugggghhhh", I hear you say.

So, there was a second syntax called MIXED.  It may come as
no great surprise that almost everyone used MIXED.  In fact,
the new IBM Prolog doesn't even support upper-case only
input (it doesn't require statement numbers in columns 73-80
either. :-))

So, let's move on to the new IBM Prolog.  Its second release
has just been released on CMS/SP, CMS/XA, MVS/XA and MVS/ESA
(translation: it can exploit up to 2**31 bytes of memory);
and has been announced for PS/2.  

	[I really want to tell you some performance numbers,
	 but I'll need legal clearance for that.  Suffice it
	 to say that I ran Peter Van Roy's tak benchmark and
	 got speed figures comparable to tak with an optimizing
	 C compiler; and that some benchmarks run in the 
	 2+ mega-LIPS range.]

Richard O'Keefe says:

> Admittedly, the manual I have is for Version 1 Release 1 on VM/SP,
> but if IBM's MVS release of IBM PROLOG differs significantly from
> IBM's VM/SP release, other Prolog vendors will have difficulty keeping
> the smiles off their faces.

First, a minor correction to Richard: your manuals are also for
VM/XA.  Anyway, you can stop smiling; the MVS release is as compatible
as can be after taking into account differences in operating systems
(e.g., file naming conventions are different).  I can readily pick
my CMS programs and run them on MVS.  What's more, IBM Prolog is
tightly integrated with REXX (the system command language; it's better
than C-shell + awk), GDDM (graphics display) and ISPF (dialogue
manager); it also has very good facilities for calling and being
called by C, Fortran, PL/I, Cobol and assembler and, last but no
least, it also integrates well with the SQL/DS and DB2 relational
databases.

-----

As Richard points out, there is no need to write an "Edinburgh"
translator; IBM provides one.   IBM provides the user with
the ability to tailor the syntax.  Suppose that you think that
lists should be written {a,b,c} instead of [a,b,c]?  No problem.
Would you like to use "if" instead of ":-"?  Again, no problem.

You get two pre-defined syntaxes: IBM standard ("synt1") and
Edinburgh ("synt2").  But there's more ...

> Better still, by using a Prolog program you can translate findall/3 to
> compute/4  ...

Why bother?  IBM provides a bagof and a setof (it might come as no
surprise that they translate almost directly into compute/4 or
compute/5).  What's more, they pass every evil test that Richard
has mentioned about correct behaviour (thank-you Richard: I appreciate
people who torture test programs.  eally).

IBM Prolog also provides user:expand_rule/2 which is a generalization
of the usual expand_term/2.  So, if you're really hung up on providing
a translator, you can have it done for you during "consult".

> The thing that puzzles me most is that Chapter 13 of the IBM PROLOG
> manual states clearly and explicitly in the very first paragraph that
> 	IBM PROLOG can process programs written with the Edinburgh
> 	syntax and using predicates that are normally part of an [sic]
> 	Prolog implementation for the Edinburgh syntax.
>	You can switch freely between syntaxes in your program,
>	and programs written in Edinburgh syntax can use all the
>	built-in predicates and data types of IBM PROLOG.
> IBM's "Edinburgh" syntax (called 'synt2') isn't _quite Edinburgh syntax,
> but it's close enough for government work.

Coming from Richard, I'll take the "close enough" comment as a great
complement.  Besides, it's a little difficult know what exactly "Edinburgh"
syntax is, what with Quintus, Arity, ALS, C-Prolog et al, not to mention ISO.

Anyway, you really can write some predicates in Edinburgh and call
them from IBM "standard" syntax (called "synt1").  What's more, you
don't need to put built2: in front of all the predicates.  If you're
in synt2 (Edinburgh), write(X) will do what you expect (that is, you
will get built2:write(X) which doesn't put a new-line after the term);
if you're in synt1 (IBM), write(X) will get built:write(X) which does
put a new-line after the term.

It's all done with "priority prefixes" which are automatically switched
when syntaxes are changed.  The description is in the manual, under the 
"declare" directive and the "prio_pref" global term.

Moreover, you can control in which syntax a term is written.  For example,
	write(f(','(a,b))) from synt1	==>	f(","(a,b)) 
	write(f(','(a,b))) from synt2	==>	f((a,b))
because "," is an operator only in synt2 and because '...' is the
delimiter for atoms in synt2 but "..." is the delimiter in synt1.
(This is a special case of `meta predicates'.)

There are lots more nice features in the implementation, but
I've rambled on enough.  Feel free to take cheap shots at JCL
(I don't use it any more; I use REXX); I think that you'll find
cheap shots at IBM Prolog a little more difficult.

Flames to me:

- Peter Ludemann	ludemann@mlpvm1.iinus1.ibm.com

	(disclaimer: the above are my opinions, and not IBM's)