At 10:34 AM 8/26/98, Travis Griggs wrote:
Yup. One might call what Squeak has a form of double-coercion. The generality scale is built right into the various #adaptToType: messages. So that negates the use of the older coercion scheme found as fallback in VW. It also limits having to reproduce the tree for each of the basic four arithmetic ops (+, -, /, *). OTOH, it requires extra message sends to get some work done, and can create intermediate objects.
So beyond needing to create less methods in total, what if any is the real advantage over DD?
Yes, I still use the DD changeSet. Note that Stephen Travis Pope found a bug or two in the LargeInteger stuff, which he generously posted to the archive.
I was planning to use DD for the Rectangle arithmetic as well. When I am done how about we unify this with STP's bug fixes?
I haven't done any actual real work as of late, just poked around. I can't even figure out where to download the 2.1 stuff for the Mac. The UIUC link downloads a 2.0 binhex file.
I got my 2.1 in st.cs.uiuc.edu Smalltalk/Squeak/2.0/Squeak2.1sea.bin. I seem to remember that after downloading the archive I was surprised because it was name 2.0. Ignore this because inside it is actually 2.1.
--Maurice
--------------------------------------------------------------------------- Maurice Rabb 773.281.6003 Stono Technologies, LLC Chicago, USA
Yes, but where is the 2.1J image? Or has the JIT implementation been abandoned on the Mac? The only one I can find is the same one Travis mentioned, which seems to be 2.0 - which I was under the impression meant there'd be problems playing with Morphic using the 2.1 image (of course, I'm under a lot of impressions, and I'm sure many of them are quite wildly wrong - perhaps this is one of those :-)).
Maurice Rabb wrote:
I got my 2.1 in st.cs.uiuc.edu Smalltalk/Squeak/2.0/Squeak2.1sea.bin. I seem to remember that after downloading the archive I was surprised because it was name 2.0. Ignore this because inside it is actually 2.1.
-- .. Donald Major 677-8000, x6937 "I don't remember any dtm System Developer sasdtm @ unx sas com explosions in 'Bambi'..." SAS Instutitute Inc Lexington - 'Gargoyles' http://www.unx.sas.com/~sasdtm/
This was posted on http://minnow.cc.gatech.edu/squeak.370 concerning the present status of the JitterVM:
"The Jitter VM was released for Squeak 1.31. Users should note that it has a serious bug in the way it caches the structure of classes. If you write code that redefines a class with many subclasses, you may crash shortly afterwards. I have written test code that crashes it every time. Ian has not released a Jitter for 2.0 or 2.1. He is working on a from-scratch new one. Presumably it will not have the cache bug. -- Ted Kaehler"
-- Dwight
Donald T. Major, II wrote:
Yes, but where is the 2.1J image? Or has the JIT implementation been abandoned on the Mac? The only one I can find is the same one Travis mentioned, which seems to be 2.0 - which I was under the impression meant there'd be problems playing with Morphic using the 2.1 image (of course, I'm under a lot of impressions, and I'm sure many of them are quite wildly wrong - perhaps this is one of those :-)).
Maurice Rabb wrote:
I got my 2.1 in st.cs.uiuc.edu Smalltalk/Squeak/2.0/Squeak2.1sea.bin. I seem to remember that after downloading the archive I was surprised because it was name 2.0. Ignore this because inside it is actually 2.1.
-- .. Donald Major 677-8000, x6937 "I don't remember any dtm System Developer sasdtm @ unx sas com explosions in 'Bambi'..." SAS Instutitute Inc Lexington - 'Gargoyles' http://www.unx.sas.com/~sasdtm/
Thanks - I'd not heard this!
GO TED!
(Hoping every bit of encouragement will make it happen that much sooner :-)).
Dwight Hughes wrote:
This was posted on http://minnow.cc.gatech.edu/squeak.370 concerning the present status of the JitterVM:
"The Jitter VM was released for Squeak 1.31. Users should note that it has a serious bug in the way it caches the structure of classes. If you write code that redefines a class with many subclasses, you may crash shortly afterwards. I have written test code that crashes it every time. Ian has not released a Jitter for 2.0 or 2.1. He is working on a from-scratch new one. Presumably it will not have the cache bug. -- Ted Kaehler"
-- .. Donald Major 677-8000, x6937 "I don't remember any dtm System Developer sasdtm @ unx sas com explosions in 'Bambi'..." SAS Instutitute Inc Lexington - 'Gargoyles' http://www.unx.sas.com/~sasdtm/
Maurice Rabb wrote:
At 10:34 AM 8/26/98, Travis Griggs wrote:
Yup. One might call what Squeak has a form of double-coercion. The generality scale is built right into the various #adaptToType: messages. So that negates the use of the older coercion scheme found as fallback in VW. It also limits having to reproduce the tree for each of the basic four arithmetic ops (+, -, /, *). OTOH, it requires extra message sends to get some work done, and can create intermediate objects.
So beyond needing to create less methods in total, what if any is the real advantage over DD?
When I posed this question a couple o' months back, Dan Ingalls indicated that he felt the #adaptToType: approach might be a little more tractible to a Squeak newcomer. I don't really feel qualified to argue that point one way or the other, since I've been talkin' Small for a while now. I will give that double dispatching, at least for me, was a piece of the Smalltalk enlightenment puzzle. An #adaptToType: type message is basically a polymorphic application of a type cast, something that we're all familiar with from more arcane systems. OTOH, the only people that have really delved into the core number stuff so far, have been people comfortable with 100% messaging. Ultimately, the whole thing would best be solved by adding multi-methods to Squeak, but I'm not that good of a VM hacker. :)
I was planning to use DD for the Rectangle arithmetic as well. When I am done how about we unify this with STP's bug fixes?
Sounds great. I'll go ahead and move the numeric collection stuff in as well. I also intend to have a go at adding strings to numbers and vice versa (aka Perl).
I haven't done any actual real work as of late, just poked around. I can't even figure out where to download the 2.1 stuff for the Mac. The UIUC link downloads a 2.0 binhex file.
I got my 2.1 in st.cs.uiuc.edu Smalltalk/Squeak/2.0/Squeak2.1sea.bin. I seem to remember that after downloading the archive I was surprised because it was name 2.0. Ignore this because inside it is actually 2.1.
Dang. I should've let it finish downloading. I saw the name and slapped the cancel button.
-- Travis Griggs Key Technology tgriggs@keyww.com Smalltalk - 100% Pure Objects, Always had 'em, Always will!
Rolf Wilms has written a MIDI package for Smalltalk MT.
http://www.ruhr.de/home/kottan/
Those attempting to get Siren up and running on Squeak might want to take a look at it to see how he interfaces to the Windows MIDI system.
adam hill...
squeak-dev@lists.squeakfoundation.org