Scratch

MathWizz wrote:

bharvey wrote:

Are you kidding about that?

Didn't you see the smiley?              

Ref : bharvey Yesterday 10:03

1- I've changed the path, now :

www.xleroy.net/Byobtuto/Thumbnails.html

2- Drawing of the shuttle is now completed...but it is very, very slow.
Consequently, I've tried to compile the project, to see the difference if any. Impossible, due to one sever crash, see screencopy :

www.xleroy.net/Byobtuto/Forum/XL-shuttle.gif

bharvey wrote:

shadow_7283 wrote:

Will BYOB 3 include any improvments in speed?

We have some ideas about how to make BYOB 4 a lot faster, but otoh we're hoping that there won't have to be a BYOB 4 because the Scratch Team will adopt the idea of first class data.

To me, speed is not a major concern. After all, the agenda is not to make a language to run big companies or make astronomy number-crunching.

I see Scratch-Byop as an excellent tool to familiarize young (and less young) persons to learn the basis of algorithmic science, to stimulate the creativity, to provide a proactive occupation to kids (instead of playing passively video games, create them !). Scratch already aims at this objective, but BYOP extension will help to fully achieve this achievement.

Bharbey, I fully support your agenda. At this stage the objective is not to add to BYOB all the possible bells and whistles that anybody can dream of, but add to Scratch all the nice features of BYOP, missing today in Scratch.

I agree with you that the "buy your own blocks" philosophy is more or less incompatible with "I want to find every control structure that I have met in a previous programmer's life". One good example of this is the MAP function which is not a basic procedure of Logo, but can easily be written by a beginner (idem for FOR ...EACH etc).

To come back to the "speed" concern :

1- I have noticed that according  the method used to draw a shape the speed varies a lot. I'll investigate.
2 - Refering to your experience, is the speed of an application improved, or not, after compilation ?

3 - Probably that many applications of an educational language do not need high speed, and if absolutely needed, the programmer will have the choice to skip the "byop" extensions of Scratch.

Amongst the wishes and desires, one can hope that in a near future, ScratchByop will be adapted to benefit from the Nvidia-CUDA technology. I'm not kidding !!

xly wrote:

To come back to the "speed" concern :

1- I have noticed that according  the method used to draw a shape the speed varies a lot. I'll investigate.
2 - Refering to your experience, is the speed of an application improved, or not, after compilation ?

3 - Probably that many applications of an educational language do not need high speed, and if absolutely needed, the programmer will have the choice to skip the "byop" extensions of Scratch.

Amongst the wishes and desires, one can hope that in a near future, ScratchByop will be adapted to benefit from the Nvidia-CUDA technology. I'm not kidding !!

I'll asume you were adressing me. I think that speed is a really important factor for more amibitious projects, which BYOB 3 does a great job of allowing for. If you take a look at my Web Wizard project, you'll notice it moves at a snail's pace. A BYOB 3 version would greatly improve it, but I don't want the project moving any slower.

I think that BYOB can be used for much more than just an "educational tool". True, it may never reach the standards of text based programming languages, but I feel that BYOB is the first step in a long walk to making programming easy. Scratch is about as simple as it gets so far, but with future (BYOB) custom blocks that I am sure will be made, Scratch/BYOB can make it easy for anyone to create complex games and environmenets.

But in order for any of this to happen, we can't let speed be an obstacle. I'm sure that BYOB won't reach lightning speed any time soon because of the way it evaluates code (as bharvey explained earlier), but once again, the idea that lag will eliminate great ideas and concepts for projects is really frustrating.

bharvey wrote:

shadow_7283 wrote:

But in order for any of this to happen, we can't let speed be an obstacle. I'm sure that BYOB won't reach lightning speed any time soon because of the way it evaluates code (as bharvey explained earlier), but once again, the idea that lag will eliminate great ideas and concepts for projects is really frustrating.

Yes, precisely.  I don't expect BYOB to compete with C speeds, but at least it should compete with Logo speeds.  I rewrote my Logo tic-tac-toe game in BYOB and it takes several minutes per move; it's unplayable.  There's no reason it should be that bad!

WOO! Someone actually read my lectures. 

Bad news. I have written the traditional procedure of Hilbert curve (see H. Abelson Turtle geometry   p 98 -combine form). It blocks hardly and definitively Byob, after several tests and dozens of verification of the code  ? I continue my investigations.

bharvey wrote:

xly wrote:

Bad news. I have written the traditional procedure of Hilbert curve (see H. Abelson Turtle geometry   p 98 -combine form). It blocks hardly and definitively Byob, after several tests and dozens of verification of the code  ? I continue my investigations.

Try letting it run overnight -- it might just be really slow.

Running overnight  = hard to sleep (If the computer is in your room) and big electricity bills. Plus if you live with your parents. Inquiring parents to why your leaving your computer running through the night.

Coming back (seriously - Hilbert blocks even with level 2 !) to the speed concern... It is well known that the functional languages with one all-in-memory management are very,very,very limited for the file management. If I take the example of "drawing a shuttle", with coordinates, several intermediary files, a lot of functions useful or not, there comes a problem of memory management, garbage collecting etc For example, at this stage it is impossible with Scratch/Byob to import/export correctly a structured list. At the opposite way Logo/Scratch/Byob are reasonably fast to build small games, scientific or musical applications, which need few data.

Inversedly it is ideal to read SICP or Turtle geometry books with Byob "in hand" !

bharvey wrote:

shadow_7283 wrote:

Will BYOB 3 include any improvments in speed?

BYOB 2.99 will probably (there is a last-minute speed improvement coming, but since it's not here yet I don't know how much it'll help) be slow as molasses.  The reason is that Scratch lists are stored in a way that's tuned for iteration rather than recursion, which is the cleanest way to think about many problems once you can define procedures.  We should be back to at least traditional Scratch speed in the real release in August.  We have some ideas about how to make BYOB 4 a lot faster, but otoh we're hoping that there won't have to be a BYOB 4 because the Scratch Team will adopt the idea of first class data.

2.99.007 isn't slow at all.    It's faster than Scratch!

Just asking, but does BYOB 2.99.013 support connections with computers that aren't in your network?

Thanks,
Whiz

soupoftomato wrote:

I have made a forum community for BYOB!

Wow -- that's quite an infrastructure for a program with probably 30 or 40 users.     Maybe we'll grow into needing it after August.  Thanks.

soupoftomato wrote:

... Harry Scratcher ...

Fwiw, some of us set our browsers to always use a black background no matter what the page says.  Your sig doesn't look very legible against black!

@bharvey

I decided to not chase the ghost of Hilbert during the night, happily because during my deep sleep I've had a dream. The aim of Byobers should not be to develop anything else than the hard kernel allowing Scratch to offer the full capabilities of functional programming + a few number of essential primitives.

But, on top of that, Byob would develop the idea (or the concept , or the technology ? ) of function-sharing. Byob would provide an easy way to import/export "third-party" functions, like it already exists for sprites, thus allowing each scratcher to extent his tool-box by picking around available functions. Just to take an example, I can offer one function to project 3d coordinates to 2d ( a much more sophisticated one ...) . It's something similar to the Firefox extensions (why not comparable to the Appstore for Iphone).

Each Scratcher specialist in any domain, could propose to all scrtachers  new functions in accordance with his own skills : computer science, music, arts, mathematics, chemistry, 3d, medecine, astronomy , games,etc

Some of these functions could be permanently installed/uninstalled into the user Scratch-Byob kernel according his needs. (this function exists in Logo: PPS=protected functions).

In the meantime I am starting now to chase the "Von Koch" and to have a look to your streams.