Scratch

technoboy10 wrote:

what should I learn in my last two years of high school? I'm homeschooled, so I can pretty much pick and choose what to learn. 

You're so lucky!  (Provided you don't have we-don't-want-our-kid-learning-evolution parents.  Funny how homeschooling attracts the most and least intellectual parents.)

Disclaimer:  There's no official answer to your question; someone else might have very different ideas.

1.  SICP!  SICP!  SICP!

2.  As much math as you can.  Discrete math (logic, set theory, probability, mainly) is most directly applicable, plus linear algebra if you're interested in graphics.  But the starting point for a mathematical understanding of fractals is point set topology.  If you get interested in the related field of electrical engineering, you'll need complex analysis.  Of course you can do all that in college, but that'll be way easier if it's second time around.  And thinking like a mathematician, as distinct from knowledge about specific bits of math, is the most important skill you can have.  And, math is more fun than anything, even programming.  (Edit:  If you study calculus, use Tom Apostol's textbook.)

3.  Don't collect programming languages.  Learning a programming language, after the first time, is a one-weekend activity.  Instead, collect programming paradigms.  Understand functional programming; understand OOP; understand logic programming.  (To that end, learning other languages can help if they're languages that emphasize a paradigm you don't know yet.  So, instead of learning Java and Javascript and C++ and Python and Perl, which are really all the same language, learn ML (functional), Smalltalk (OOP), and Prolog (logic).)

4.  Don't spend all your time on technical stuff.  Read Freud.  Read Marx.  Read Dickens.  And find a DVD called "The TAMI Show" and watch it.

Last edited by bharvey (2013-02-06 13:33:08)

bharvey wrote:

Hardmath123 wrote:

Now I can import everything except scripts and sounds! 

Wow!  Great start.

Thanks! It's not too hard; the formats are quite similar.

Scripts are pretty important.   

Good point, that's next on my list. The thing is, I need to translate each block and separate out the procedures into custom blocks.

When you get this done, what do you want to build next?  The debugger?

Maybe. I was thinking about writing a fast Snap! interpreter in JS, since making my own language is clearly a pain, and I did write a nice little Scheme interpreter (with Lambda!).

@Jens: aside from the SVG glitch, I also (only once!) ran into a glitch where the whole UI just fell apart. There were hazy circles floating around (!!!) and all the panes were displaced.    I'll let you know if I can reproduce it.

Another issue I have is retrieving the project data (notes and creator's name). In Scratch 2.0 the API is beta.scratch.mit.edu/api/v1/project/<id>/?format=json. This only seems to work if you're looking up your project, and you're logged in. That's really strange. How do I access this from Python? It treats me as not-logged-in from a urllib request and gives me a blank page.

@Cloning: bharvey, have you been to the Scratch 2.0 Beta site? Check it out, it's really well made. But I want to point out specifically how cloning works there, because I think that's a good way to do it and I'm 80% sure that's  similar to what Jens did on Snap!. Basically, you only have 3 clone blocks: clone[myself|sprite list], delete this clone, and a {clone startup} hat block, which is triggered whenever the clone is initialized. On cloning, you retain all the features (position, direction, etc.).

Hardmath123 wrote:

the API is beta.scratch.mit.edu/api/v1/project/<id>/?format=json. This only seems to work if you're looking up your project, and you're logged in. That's really strange. How do I access this from Python? It treats me as not-logged-in from a urllib request and gives me a blank page.

I've mentioned it before, but try httplib2 — it handles forms and cookies better than urllib, so you can login using the main site login (which I can't find right now!).

If everything else fails, I can actually load the project page and parse the HTML I get to find the notes and creator. But I'm not too convinced that that's a good idea, and would rather use that as a last resort.

@bharvey: It it alright if it's a command-line tool?

cd ~/path/to/my/tool
./snapin8r.py 100203

(100203 being the ID of the project on the main site). It will take two parameters, one for an output file, the other being a boolean flag telling it whether to open the project instantly or not—I use the #open: trick and Python's webbrowser module to open projects automatically for debugging, and I figured it would be useful for users, too. 

EDIT: Of course, it also takes --help for help.

Last edited by Hardmath123 (2013-02-07 07:32:52)

Aha, hmm. Maybe you have to export the Scratch project using Scratch's FILE menu, first? Or we have to ask the Scratch Team to add a CORS header for Snap...

Hardmath123 wrote:

Is it alright if it's a command-line tool?

Aww, we can't just drag the project from their page to ours?
I'll let Jens answer that; he's the drag-and-drop fan.

I think we should have more control over threads in snap!. To accomplish this, I think the LAUNCH block should return a procedure that is the thread's control procedure. When this is called with no arguments, it will stop the tread and return the thread's continuation (a command one with no arguments) which can be called to resume the thread. When the thread finishes, it's thread control procedure will return a function which when called, returns the value if it exited via REPORT, or returns "No Value" or something if it exited via STOP BLOCK. The way to distinguish between these "finished thread functions" and a thread continuation would be that when it's finished, the control procedure will return the same function (in terms of IS IDENTICAL TO?) but successive callings will return different continuations of a running thread even if called while it is stopped. The thread control procedure could also be called with one argument, a procedure, that replaces the thread's continuation with that procedure, resuming it if it is currently stopped by the control procedure of no args. This is so you could make the thread do something extra when it's finished or something with the thread control procedure being able to keep up. I think that is all you need to implement thread control tools like pausing and resuming it, making something happen when the thread finishes, waiting for a thread to finish, ect.
(there could also be a block called THIS THREAD'S CONTROL PROCEDURE or something for even more control)

@Hardmath: Okay, now I'm getting this Django response page:

Forbidden (403)
CSRF verification failed. Request aborted.

Request:

headers = {'Content-type': 'application/x-www-form-urlencoded'}
data = urlencode({"username": "blob8108", "password": "whatever"})
(response, content) = http.request("http://beta.scratch.mit.edu/login/", "POST", data, headers=headers)

EDIT: I'm now convinced the API only allows access to your own projects. 

Last edited by blob8108 (2013-02-07 12:11:35)

MathWizz wrote:

http://beta.scratch.mit.edu/internalapi/project/<id>/get

Oh, clever! And this for the images and media:
http://beta.scratch.mit.edu/internalapi/asset/<filename.png>/get/

nXIII wrote:

blob8108 wrote:

MathWizz wrote:

http://beta.scratch.mit.edu/internalapi/project/<id>/get

Oh, clever! And this for the images and media:
http://beta.scratch.mit.edu/internalapi/asset/<filename.png>/get/

Yeah, but most Scratch 2.0 assets are .svg, not .png.

I assumed the other media files could be retrieved in the same way...

you need to GET the main site or any other page and get a CSRF token.

Oh, will any page do? I assumed you'd need to get the login form. Cool!
But you still need to parse the HTML, don't you? Meh.

nXIII wrote:

just pointing out that you probably shouldn't hard-code ".png" in your script.

Sure 

You can just search for the string:

Code:

name='csrfmiddlewaretoken' value='

and go to the next quotation mark.

...ew. 

blob8108 wrote:

MathWizz wrote:

http://beta.scratch.mit.edu/internalapi/project/<id>/get

Oh, clever! And this for the images and media:
http://beta.scratch.mit.edu/internalapi/asset/<filename.png>/get/

I'm already doing that to get the data (images, sounds and scripts), I just need a way to get the project notes, creator name, and other junk like that.

@Jens I got sounds and scripts working (I still need to make a library to translate Scratch 2.0 blockspecs to Snap! specs… but I'm procrastinating that.  ). Are the 'id' attributes required for lists of costumes/sounds? It works without it, so it can't be vital. 

@bharvey The best way would be for you to be able to type the project's URL under "open…" (open>Scratch 2.0 project>[dialog box]). You would send the URL to a Python server running my converter, and I would return the XML from a GET request. Happy?

Also, has Jens started on the server yet, and if so, is it in Python? I'd suggest in the end using Google App Engine to get the final release at snap.edu or something.

@bharvey "About cloning etc."
May I give you the layman viewpoint.
OOP features of old Byob3.1 are very powerful and full of capabilities to write kids projects with bunches of targets, mobiles, monsters, bricks etc But these capabilities have been poorly documented and suffer of no project-sharing to offer examples of applications. IMHO you could in a first stage simply port Byob 3.1 OOP system in Snap!
But one may regret that Byob OOP does not expose the traditionnal CLASS instruction as used by Java,JS and others. Byob 3.1 can mimick this (see page 30 of your documentation). But for "compliance" reasons, it should be wise to provide one CLASS instruction not related to a SPRITE with its own graphical & visual characteristics. Something similar to your Dispatch/Method procedures as documented into page 27 to 30 of your documentation. For example you create a CLASS of circles and then you can use it to display a lot of circles, or move dozens of simple shapes without needing to create a lot of real cloned sprites.

Last edited by xly (2013-02-08 08:47:48)