Comments on Moserware: Meta-FizzBuzz

WildWil: Meta-programming is its own game for me :...

2009-05-31T13:59:47.085-04:00

WildWil: Meta-programming is its own game for me :)

Interesting notion indeed. We didn't go that far, ...

2009-05-29T09:34:22.260-04:00

Interesting notion indeed.
We didn't go that far, just made it an iPhone app with a score attack mode. Beer not included :) And certainly can be used as a weed out tool too. Though I have to admit we hadn't thought of that one....

Feel free to check it out here:

http://buzzyteam.wordpress.com/fizzbuzz/

Mark Miller: I found your post to be a good review...

2008-10-18T11:25:00.000-04:00

Mark Miller: I found your post to be a good review of the philosophy behind The Computer Revolution Hasn't Happened Yet. I also liked Crawford's analogy to the need for an alphabet to democratize what Kay might call "computering" for all.

Meta-Fizzbuzz was just a very basic attempt to demonstrate what might happen if we allow scoping rules to be a bit less conventional (e.g. a global state that remembers variable references).

To be fair, I don't think "it" would be useful in large blocks of code as you'd lose contexts, but I think the design would force "programmers" to be brief, which would be a good feature.

The larger overall context is that bootstrapping to a world where you can experiment with these ideas keeps getting cheaper. This trend just might eventually allow "Revolution."

After a recent post I did I feel like I understand your post a little better now. I came upon an article by Chris Crawford that I think relates real well to your use of the term "it" in your language. In his essay he talks about abstraction, both in natural language and in programming. He said that abstraction exists in natural language, but it's "loosey-goosey". It's arbitrary, but our brains are sophisticated enough to figure it out (usually). He said that programming introduces precise notions of abstraction, enabling us to "think about issues of abstraction more clearly and powerfully, just as a strongly conjugated language allows one to express ideas of action with greater clarity."

alexey: You're right; arguments go onto an argumen...

2008-09-02T22:26:00.000-04:00

alexey: You're right; arguments go onto an argument stack that is difficult to type. I mentioned this in my last post. OMeta/JS does this and is fine since types sort of melt away in a dynamic way. I could achieve the same in C# by using Reflection everywhere at the expense of some speed. I think that with C# 4.0, I could put all of the semantic actions in a dynamic { ... } block and let the compiler worry about emitting optimal DLR expressions to make it most efficient.

The difficulty I've run into is typing ApplyWithAr...

2008-09-02T09:47:00.000-04:00

The difficulty I've run into is typing ApplyWithArgs and SuperApplyWithArgs. And I've no idea how I could get around it without a) making a lot of generic rule types for different number of arguments, which breaks down once someone needs one argument more than I've supplied; or b) making them generic in output argument only and taking an array of objects as arguments, which gets us right back to doing a lot of casting.

alexey: Thanks for looking at the code. Let me kno...

2008-09-02T09:11:00.000-04:00

alexey: Thanks for looking at the code. Let me know if you need help understanding something. I haven't yet done a pass to make the code more readable yet. Unfortunately the code is still in a "braindump" state. I'm hoping that I'll be able to actually add syntax sugar so that you don't have to explicitly say "Sugar" in the host actions.

You're right in calling out that defaults and variable capture type definitions are hacks. They help, but aren't perfect. I think the long term solution might be to accept static typing where possible and then allow for reflection-based (e.g. dynamic) where you don't want to do type annotations.

At least.. that seems easier than a foolproof duck type-inference algorithm.

I've looked at the latest commit (just after leavi...

2008-09-02T01:06:00.000-04:00

I've looked at the latest commit (just after leaving my comment), and it is much nicer in this respect. But consider a grammar where several output types are widely used, say, a calculator which should work on ints, doubles, and decimals. One of them may be used as a default, and variable capture helps too.

But all the casting is still going on under the surface. I've tried to make HostExpression generic in the m_value type, and Rule in the output type, but haven't had much progress so far.

alexey: Thanks for taking the time to leave a comm...

2008-09-01T15:41:00.000-04:00

alexey: Thanks for taking the time to leave a comment!

The latest commit of OMeta# is much closer to what you're talking about because of the notion of being able to specify a default type as well as specifying the type at point of variable declaration.

Here's a line from the latest Calculator grammar that exploits this feature:

Number ^= Number:n Digit:d -> { n * 10 + d } | Digit,

This is made possible because of this declaration: "ometa Calculator<char, int> : Parser<char> {" which states that by default it is mapping char streams to integer objects.

Alternatively, you could have been explicit and said

Number ^= Number:n<int> Digit:d<int> -> { n * 10 + d } | Digit,

You're absolutely right that it's a struggle against the static type system. My goal is to keep doing as much as I can to make things feel more dynamic while keeping the goodness properties that static typing allows.

I'm just making gut feel guesses as to what seems best based on my own usage of OMeta#. I love any and all feedback on these decisions.

One thing that I really don't like about the c...

2008-08-30T11:25:00.000-04:00

One thing that I really don't like about the current incarnation of OMeta# is the constant struggle with the static type system. Wouldn't

int Number = Number:int n Digit:int d -> { n * 10 + d }

or something like that fit .NET better than

Number = Number:n Digit:d -> { n.As<int>() * 10 + d.As<int>() }

?

maysonicwrites:

I thought about doing a Met...

2008-08-25T19:00:00.000-04:00

maysonicwrites:

I thought about doing a Meta-Lisp, but it's already been done in OMeta, so I went with Meta-FizzBuzz instead.

As it turned out, Meta-FizzBuzz turned out to be Pascal-ish. The fact that I have statements and global variables makes it a very imperative language. However, these were just arbitrary decisions, you could have gone with a functional approach instead.

The Lisp Way - on steroids.

2008-08-25T11:55:00.000-04:00

The Lisp Way - on steroids.

zack: I don't think I'd be mean enough to actually...

2008-08-25T09:43:00.000-04:00

zack: I don't think I'd be mean enough to actually ask the Meta-FizzBuzz problem in an interview... yet :)

The current state of language-oriented thinking is seems a bit too niche right now for people to take it seriously.

However, I keep coming back to the thoughts that Alan Kay said to the effect of "if you really understand something, you should be able to craft a language for it" going off the Plato idea of being able to "divid[e a problem] at the joints, as nature directs, not breaking any limb in half as a bad carver might"

It just seems that a Moore's Law increase in software expressiveness is going to spread the notion that languages are cheap and can be crafted to ideally suit whatever business or inquisitive problem you want to solve.

Definitely an interesting problem. I can see this ...

2008-08-25T09:17:00.000-04:00

Definitely an interesting problem. I can see this sending some candidates running screaming from the interview, but I guess if that happens you don't really want them anyhow :-)

Comments on Moserware: Meta-FizzBuzz

WildWil: Meta-programming is its own game for me :...

Interesting notion indeed. We didn't go that far, ...

Mark Miller: I found your post to be a good review...

alexey: You're right; arguments go onto an argumen...

The difficulty I've run into is typing ApplyWithAr...

alexey: Thanks for looking at the code. Let me kno...

I've looked at the latest commit (just after leavi...

alexey: Thanks for taking the time to leave a comm...

One thing that I really don't like about the c...

maysonicwrites:I thought about doing a Met...

The Lisp Way - on steroids.

zack: I don't think I'd be mean enough to actually...

Definitely an interesting problem. I can see this ...

maysonicwrites:

I thought about doing a Met...