Scratch

coolhogs wrote:

I meant when it gets done.

Ah, okay.  I still don't really understand all these apps that are really just web pages -- although if it were totally local to your computer and had access to its filesystem and hardware, that'd be worth doing.

bharvey wrote:

Hmm, maybe you could let others submit their ideas?

You don't like it?      Sure, by all means.

...

No offense intended; but I did notice you simply cropped the image in your sig. 

I was hoping for Alonzo the Gobo to make a comeback...

Ah. So I do have to code a special form for lambda, I thought so.

I found out one big mistake I was making—I was first evaluating arguments, than the actual procedure. I should have been evaluating the procedure, and the procedure's evaluation script called "evaluate" recursively on arguments which need evaluation. Does that make sense?

The first thing you do is look at the expression's first element, without evaluating it!  If it's a keyword "lambda" "define" "if" etc, then you go off and evaluate each kind of special form with its own rules.  No part of a lambda expression is evaluated, but (e.g.) IF immediately evaluates the second element, then decides whether to evaluate the third or the fourth, depending on the value of the second.  If that first element isn't a keyword, then you evaluate everything -- doesn't matter what order.

The value of a lambda expression is essentially (list exp env), combining the expression itself with the defining environment, but in practice it's usual to add a procedure type-tag to that, so you can't later confuse it with a user-generated list.

Note: After typing this out, I noticed the next paragraph is a rather brutal rant, so skim it over if you don't have the mental energy. 

So how do I go about representing an environment? As a dictionary? If so, when evaluating a lambda I'd have to pass in a copy of the environment to make sure input variables don't leak into the global variables... but I'm not sure how to copy dictionaries so that they remain mutable (Objective C distinguished between mutable and immutable data, which is really silly in my opinion; everything should be mutable and copyable) and all members are copied, too. It's one of those things I hate about Objective-C. It seems to be specially designed to send you to the developer documentation, where the answer is hopelessly buried in optimizations, exceptions, compiler directives, subclassing notes, and deprecations...    Why do we, for example, have to declare the return type of a method (along with the types of inputs)? The compiler should just assume we know what we're doing and pass along whatever value we input/output. There is a "general" value type called "id" but it only works for pointers to instances of classes, so you cannot return a number. Not to mention we have to also copy method declarations into the header for no apparently good reason. They reside outside the instance variable declarations, which is silly because a method is like an instance variable with a lambda value. But of course Obj-C doesn't support lambda, it supports "blocks" whose return type is inferred from return statements, so if you try to return an item of an array somewhere, you won't be able to return an integer somewhere else, even if the item of the array was a number (why do we distinguish between int and float anyway?). And there is no Block type, so returning a block is close to impossible.

Okay, rant over. I'm doing my interpreter over in JS, which is so much nicer to me. 

P.S. You can make a long dash with shift-alt-dash, I notice you're putting in double-dashes (--). 

Hardmath123 wrote:

Block request:
imgpaste.com/v1kt.gif

probably not going to be implemented - what would you do with a lambda script that broadcasts in the hat block → infinite loop 
and it's just so easy to workaround, so probably not...

also, it requires evaluating of all broadcast hat blocks while broadcasting. think of speed consequences
just use if/else and a variable 

roijac wrote:

I suggest the chromeless project

Oh, that looks wonderful!  GPL, cross-platform, the works.  Thanks!

Hardmath123 wrote:

Block request:
imgpaste.com/v1kt.gif

I just remark that BROADCAST , contrarily to its sister instruction WHEN I RECEIVED accepts a variable as input. This is convenient if you need to trigger an action to a group of sprites, whose number is variable You just have to read a list and trigger the BROADCAST of each item of the list.

Hardmath123 wrote:

That thing on the image I pasted is an upvar... sorry about the bad graphics, I'm not an artist. 

Ah, sorry, roijac and I both misinterpreted it as a variable reference.  In that case we are in agreement.

bharvey wrote:

Hardmath123 wrote:

(which reminds me, you need a favicon urgently). 

http://snap.berkeley.edu/favicon.png

Uh, you know favicons are supposed to be 32x32, preferably 16x16?

bharvey wrote:

scimonster wrote:

Uh, you know favicons are supposed to be 32x32, preferably 16x16?

Oh, fussy, fussy....

http://snap.berkeley.edu/favicon.png

or maybe

http://snap.berkeley.edu/fav2.png

I like the second. 

i'm going to write the explicit control evaluator in scratch!

roijac wrote:

Hardmath123 wrote:

I like the second. 

+1 

I'd pick the second if it weren't so aliased.

Hey, Bharvey, about the quines, didn't you write this one?

invoke [[x] print (list "invoke :x :x)] [[x] print (list "invoke :x :x)]

Also, this ought to work, but doesn't:

((lambda (a) 
   (lambda () (a a))) 
 (lambda (a) 
   (lambda () (a a))))

Last edited by Hardmath123 (2012-06-05 09:30:59)

dreamod wrote:

I just started modifying BYOB, and I've noticed a lot of stuff I haven't noticed before, and I'm VERY impressed by it. I'm adding Image support which can be exchanged through variables, watchers etc, and even an arg morph which I call ImageArgMorph. It looks like a miniature palette and when clicked let's you edit/view the image which is stored in the arg morph. If you want to have this, tell me. I have no problem with letting you use this for BYOB. I love BYOB.

well, it probably won't work with javascript 

bharvey wrote:

joefarebrother wrote:

i'm going to write the explicit control evaluator in scratch!

In Snap!, you mean, I trust.  Scratch doesn't have goto.

I'm going to use broadcast as goto.