Re: [stack] Re: Cat article submission to Doctor Dobbs Journal

Hi John,

Thanks a lot for the comments.

On Feb 10, 2008 8:40 PM, John Nowak <[email protected]> wrote:
> I'll be quick here, as I haven't the time I wish I had.

Sorry about that.
> The introduction seems weak. Simplicity doesn't come from a low number
> of concepts.

I'm a little confused: a low number of concepts seems to be the very
definition of simplicity.

> If it did, the SK-calculus would be the language all new
> programmers would be starting with.

You seem to be saying here more that a low number of concepts doesn't
neccessarily make a language appropriate for beginners. Is this
correct? If so I do agree with you. The other reason, which I should
perhaps include is the fact that the model of computation closely
resembles what actually happens in a computer.

> It seems like offering a hook
> beyond how "simple" it is should be part of your introduction.

Fair enough, I'll see if I can do better.

> There
> is the discussion of efficiency, but at least for me, this is one of
> the least interesting parts about Cat.

What do you think is the most interesting part of Cat, if any? And I'm
not just fishing for compliments, honest :-)

>  That said, I'm not sure how I'd
> rework this...
>
> > In the Cat specification instructions are referred to as functions,
> > whether or not they have side-effects.
>
> Do functions actually have side-effects?

In Cat they do (according to my definition of function).

> If all functions are unary
> functions of stacks to stacks, it's easy to thread an implicit world
> state through the entire program.

It's possible, but I wouldn't call it easy. I struggled for a long
time to figure out how to do it in Cat. Then again "easy" is always
relative, I'm sure it was easier for you. :-)

>  I believe Backus's FL did something
> similar. As far as I'm concerned, Cat is purely functional. Functions
> of the type A ~> B are functions that read/write the world state on
> the top of the stack. All other operations work below it.
>
> > Functions can not be redefined, and are only visible after they are
> > defined.
>
> Does this mean recursive definitions are disallowed?

No. I see your point though, it is unclear. Thanks for pointing it out.

> > In Joy parlance this is called a quotation, but I prefer to think of
> > it as either an anonymous function or lambda expression.
>
> Lambda expression might not be the term to use here as there are no
> variables involved.

Good point.

> > A possible definition of bin_rec is shown Figure X.
>
> So bin_rec must be a primitive?

No, why do you ask? I am intending to show that it can be defined in a
library, so it doesn't have to be made a primitive.
I am probably not being clear. The Cat primitive set is somewhat
arbitrary, different implementations would choose different operations
to implement as predefined primitives.

> If so, that should probably be stated.
> I assume that's the case for reasons related to the type system, as I
> believe you can write it in Joy.

Yes you can.

> > For those who like big Greek words
>
> I don't know who the intended audience of this is (I'm not familiar
> with Doctor Dobbs), but I find this level of informality off-putting.

I can understand. I was on the fence here, I was just trying to add a
bit of levity, and to laugh a bit at the tendency of the community to
overuse greek-like words (I am not convinced that those are real
words). DDJ is a trade magazine intended for software development
professionals.

> > Metadata is a form of structured comment that can be associated with
> > a Cat program. I use it to document functions, and provide automatic
> > unit tests.
>
> Again, not sure on the audience, but "I use it" as opposed to "Its
> intended use is" strikes me as odd.
>
> > Static type systems are useful for documentation, static
> > verification of code, and optimization.
>
> And more!

I could use some inspiration here. ;-)
Any suggestions?

> > It is common practice in stack languages to document the stack
> > effects of each instruction, i.e. what type of values are remove
> > from the stack (the consumption), and what type of values are placed
> > on the stack (the production).
>
> That should be 'e.g.', not 'i.e.'.

Why is that? I thought it was more of an elaboration than an example.

> > If the type annotation is omitted, Cat is able to infer the type
> > automatically, using a variant of the Hindley-Milner algorithm [X]
>
> Should be: "If a type annotation is omitted, Cat is able to infer a
> type automatically using a variant of the Hindley-Milner algorithm [X]."

Okay.

> > Using the Cat interpreter "#t" will give you the type of any
> > function on the stack.
>
> I don't quite understand this sentence. Perhaps "Using '#t' in the Cat
> interpreter"?

Yes, thanks.

> > The most important properties are: whether the required input types
> > for a function are being supplied when it is called
>
> Perhaps "values of the correct types" is better.

Yep,

> > A somewhat novel feature of the Cat type system is that all
> > functions are row polymorphic [ref] (also called tail polymorphic)
>
> I don't have time to get into this now, but I think there are cases
> when you don't want functions to be row polymorphic. A trivial
> example: Supplying a function that takes two arguments for use in a
> callback. Simply requiring the function to unify with (A b c -> A)
> would be insufficient, as you'd be allowed to pass something like (A b
> c d e -> A b).

This is interesting, I will have to look at the issue carefully before
I comment on it.

>  I've had to introduce an additional concept to my type
> system to address this. First-class stacks (evaluated on via Joy's
> 'infra') also seem to require this.

I don't support first-class stacks in Cat.

> I'll post in more detail soon I'm
> sure...

I look forward to it.

> > define quadratic { [dupd [sqr] dip swap [*] dip] dipd [* +] dip + }
>
> If you take the arguments in reverse order, as you typically would,
> you can do this:
>
> 2dup [* * swap] dip * + +

That is interesting. Personally I wouldn't typically do that, because
I would like to write things like:

define quad234 { 2 3 4 quadratic }

> > Now we partially evaluate the "papply" functions:
>
> Better than 'curry'. :)

:-)

I can't help but think that maybe "partial curry" would be a more
accurate term, but history wins here.

> Anyway, I hope something in there helped. Sorry I don't have time for
> a more thorough reading.
>
> - John

I really appreciate your comments, thanks a lot!

- Christopher