M3 Types

[dukester]

Reading: 2.2 Types in Nelson's book

Smells of OOP to me.
It seems to me that in other languages a "type" is an "attribute" associated with a variable.
e.x. "age" can be declared as a variable to contain only integers.

In M3, it seems to me that a type e.g. INTEGER belongs to an object called TYPE.
Am I understanding M3 TYPE correctly?

[mikanystrom-intel]

Well, types are not only associated with variables. They are also associated with expressions.

I don't think the Modula-3 approach to types is that unusual. It basically rests on the idea that types are (restricted) sets. Restricted in the sense that you can't make a type that contains completely arbitrary sets of values (for example, both references and integers, or even both integers and single-precision floating-point numbers).

On the other hand, yes, the type system is designed to support OOP.

The best reference to Modula-3 types is

http://lucacardelli.name/papers/modula3typesystem.a4.pdf

[dukester]

I didn't realize that CM3 was an OOP-centric language. I have severe OOP-a-phobia! I quit Perl when Perl5 came out. Maybe I should stick with Ulm's Oberon-2 or Free Pascal. Thanks for the link and your input!

[mikanystrom-intel]

I wouldn't call it OOP-centric. The type system supports OOP. But it also retains the "classic" Modula types.

The included libraries do tend to use OOP internally, but they generally don't require users to interface with them that way.

For example, in the following code:

VAR 
   wr := FileWr.Open("myfile.txt");
BEGIN
   Wr.PutText(wr, "Hello myfile!\n");
   Wr.Close(wr)
END

the variable "wr" holds a writer object, which internally is implemented with all sorts of OOP trickery (there's a chapter about this in the green book). But as you can see, you can use this object entirely without descending into any OOP at all.

[dukester]

I see! I've had too many "oops" with OOP in the past, so your example is encouraging for me.

[norayr]

I didn't realize that CM3 was an OOP-centric language. I have severe OOP-a-phobia! I quit Perl when Perl5 came out. Maybe I should stick with Ulm's Oberon-2 or Free Pascal. Thanks for the link and your input!

Perl is a big language, and you can do the same thing by many ways in it. Plus it is not strongly typed. That is why I would stay away of Perl, not because of OOP.

FreePascal and Oberon-2 are languages with OOP support, both have so called virtual functions. In Oberon-2 they are called type-bound procedures.

Ulm's Oberon also supports OOP - Oberon-1 style. The RECORD will have a hidden field with type descriptor. But without VMT - you need to assign procedure fields yourself. Most of Ulm's Oberon library uses OOP, except a couple of lower level modules. Compiler itself won't be compiled without OOP.

FreePascal has even FCL and LCL - the libraries are OOP based.

As someone who mostly writes in Oberon, I can tell that OOP is not a reason to quit Modula-3. Modula-3 gives you choice to use OOP only when it is necessary. I would maybe say, Modula-3 is what would commercial, real world Oberon compiler look like. We have automatic memory management, today parallelism is necessary in multicore world, also exceptions and generics are usually required in industry- even Go introduced generics recently.

And with all that Modula-3 It is still very small language, no comparison with today's Object Pascal, but still very powerful. Part of its power is that it is small. Contrary to today's Pascal, which was, I would say, disfigured by corporate development.

If you'd like to avoid OOP, you may try GNU Modula-2, but even Modula-2 has some OOP features. Maybe P4 Pascal would be the choice for you?

[dukester]

As someone who mostly writes in Oberon
Vishap Oberon?? :)

Maybe P4 Pascal would be the choice for you?
Just tried to install it. Install chokes at first line of `configure'.

I'll continue with CM3 and maybe Oberon-2! Thanks for all your input!

[thielema]

On Thu, 2 Mar 2023, Duke Normandin wrote: I didn't realize that CM3 was an OOP-centric language. I have severe OOP-a-phobia! I quit Perl when Perl5 came out. Maybe I should stick with Ulm's Oberon-2 or Free Pascal. Thanks for the link and your input!

I would neither CM3 nor Modula-3 call OOP-centric. Instead, OOP is a pretty orthogonal feature of the type system in Modula-3. Modula-3 even has no special syntax for object methods. Instead you reuse procedures. That is, you can use a procedure for multiple method implementations (if types fit) or you can call a procedure behind a method without using object method call syntax at all. This is pretty useful if you want to pass a procedure as an argument to another procedure.

[dukester]

OOP is a pretty orthogonal feature of the type system in Modula-3
Right on! LOL

[RodneyBates]

On 3/2/23 14:01, Duke Normandin wrote: I didn't realize that CM3 was an OOP-centric language. I have severe OOP-a-phobia! I quit Perl when Perl5 came out. Maybe I should stick with Ulm's Oberon-2 or Free Pascal. Thanks for the link and your input!

I would call it OOP-optional.  If you don't declare anything with the type constructor OBJECT, you will not have to deal with inheritance, dispatching calls, or any of that stuff.  Just use RECORD and REF RECORD to do, for example linked, heap-allocated data structure. Do realize that the common word "object" is often used to mean any heap-allocated chunk of data, or even a simple declared variable.  When OOP got really faddish, there were many people trying to be really with-it, just by redefining what they had long been doing as "object-oriented", when it was not.  For example, many falsely redefined modular programming, with equivalents to interfaces-and-modules as OOP, even C header files, if used in certain ways. Not wanting to appear dated, the C++ folks conflated objects and structs into a single messy thing they called a class, with a mix of both, with the result that you can't tell a dispatching method call from a conventional procedure call by looking at it.  Each kind can optionally look like the other, and it's often a long quest through transitive #include files to ascertain it. See: https://queue.acm.org/detail.cfm?id=1028906. In Modula-3, these are syntactically distinct, so you know right away whether a call will go predictably to a fixed place or somewhere dynamic. Don't get me wrong.  I am a big advocate for using OOP techniques for certain situations, and they can be very helpful in such cases.  I don't remember a software project of moderately large size that didn't benefit from OOP somewhere, or would have, had the language supported it. I was an early adopter of object orientation, acquiring and porting SIMULA-67 compilers for the machines available to me at the time.  Anybody here old enough to remember SIMULA-67?  The 67 is a clue to its age.  It was the first object-oriented programming language, but it wasn't called that.  Only later was that term coined, and applied retroactively to it and newer. As I say, OOP has some very appropriate applications, but OOP seems to have engendered a very misguided cult.  "Objects first", as a pedagogical method (heh, heh) is, IMO, a disaster, as OOP objects have as prerequisites, all manner of simpler constructs, and then add additional complexity on top of them. So enjoy Modula-3 without OOP, if you like.

…

— Reply to this email directly, view it on GitHub <#1159 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABSVZNH3QA722UMKAU2E2LDW2D4CDANCNFSM6AAAAAAVNYKM2I>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[dukester]

So enjoy Modula-3 without OOP, if you like.

I'll stick with it.

BTW Rodney, everything you say about OOP I've also heard proclaim by the FP crowd. Just lurk in the Lisp irc channel for a while and you;ll get an ear full. LOL IMHO, they all their pros and cons; all are more or less suited for certain projects and each will appeal to an individual depending on how intuitive that language resonates with him.

I punched ny first PL/I code at McGill in 1968. 😱

[RodneyBates]

On 3/2/23 15:31, Duke Normandin wrote: |So enjoy Modula-3 without OOP, if you like.| I'll stick with it. BTW Rodney, everything you say about OOP I've also heard proclaim by the FP crowd. Just lurk in the Lisp irc channel for a while and you;ll get an ear full. LOL IMHO, they all their pros and cons; all are more or less suited for certain projects and each will appeal to an individual depending on how intuitive that language resonates with him.

I am a big fan of doing certain things in functional style, whether the language is a so-called functional language or not.  Modula-3 also supports this too, although with less syntactic sugar than a functional language.  Pointers to records, heap allocation, and GC are a lot of what is really required.  I have done a couple of instrumented experiments on data structure implementations, functional and imperative style, side-by-side and found worthwhile reductions in memory allocation, deallocation, total copying (sum of client and implementation side), and retention, as well as reduced compute time, in thousands of randomly generated operations. I do, however, like the ability to occasionally do something imperative down at the implementation level, while keeping the abstract view entirely functoinal, as seen by client code.  Partially lazy evaluation and exploitation of partially overlapping objects are examples. And, for client code, it can be dream, compared to imperative.

…

I punched ny first PL/I code at McGill in 1968. 😱 — Reply to this email directly, view it on GitHub <#1159 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABSVZNDHHZK6BJ5TN5CVHMTW2EGT3ANCNFSM6AAAAAAVNYKM2I>. You are receiving this because you commented.Message ID: ***@***.***>

[dukester]

I don't mind scheme; but I enjoyed messing with newlisp
Not a fan of OCAML or Haskell. A few people I've run across write functionally using C. In Modula-3 to do the same I suppose you'd need an interface(s) for the function "prototypes; an implementation module(s) to flesh them out; and a Main.m3 to daisy-chain the lot. I suppose.

[RodneyBates]

I like ML. For a big M3 example in functional style, see m3-libs/ordsets

…

On 3/2/23 21:24, Duke Normandin wrote: I don't mind scheme; but I enjoyed messing with newlisp <http://www.newlisp.org/> Not a fan of OCAML or Haskell. A few people I've run across write functionally using C. In Modula-3 to do the same I suppose you'd need an interface(s) for the function "prototypes; an implementation module(s) to flesh them out; and a Main.m3 to daisy-chain the lot. I suppose. — Reply to this email directly, view it on GitHub <#1159 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABSVZNFQBVZAFD2URAFP6SDW2FP5DANCNFSM6AAAAAAVNYKM2I>. You are receiving this because you commented.Message ID: ***@***.***>

[dukester]

Thanks! I have a look ...