Squeak-dev January 1998

squeak-dev@lists.squeakfoundation.org

81 participants
196 discussions

Re: Squeak 1.3 problems
by Hans-Martin Mosner 31 Jan '98

31 Jan '98

Chris Reuter wrote: > > ... > > Many of the methods have mangled source code, specifically that they > have no newlines. The changes file is, I notice, in Unix format, so > it appears that Squeak is failing to properly do end-of-line > conversions. I've tried filing in CrLfFileStream and making > FileStream class >> concreteStream but that doesn't seem to work. > > Does anyone know what's going on. You've most likely downloaded the changes file as 'text', not 'binary'. Both sources and changes should have CR (ASCII 13) as the line end character, nothing else. Using UNIX tools usch as 'tr', you can fix the file (or you can do it with some Squeak statements). Hans-Martin

1 0

Re: Smalltalk Design Question
by Hans-Martin Mosner 31 Jan '98

31 Jan '98

Mike Klein wrote: > It certainly would be nice for the corrector to change > > iszero -> isZero > > but it would be nicer, still for it to correct: > > expandCollection -> growCollection > > or > > emitFourBytes -> emit4Bytes > > sort of a thesaures based corrector. Yeah, exactly what I wanted all those years: A Do-What-I-Mean-Not-What-I-Say approach to programming. The ideal thing would be a syntax checker and corrector which would generate working programs even if I just rolled my head on the keyboard. :-) I'm not saying that those little helpful tricks such as spell checking, automatic declaration of variables etc. are bad. They're actually quite useful, and whenever I'm forced to use VisualAge at a customer project I yearn for the helpfulness of Smalltalk-80 derived system such as VisualWorks or Squeak. However, there's a limit. Software that tries to be overly helpful and intelligent soon gets in the way of getting real work done. Like the MS Word spelling checker which has bitten me more than once when it helpfully corrected an 'its' into 'ist' because it thought I were writing a german text. Giving me hints is ok, but trying to guess (incorrectly) what I could have meant is not. So please, if you think you have a great idea that would make programming a breeze, just implement it in Squeak, and see whether it really helps. It's not that difficult. But please stop the endless discussion... nuf said Hans-Martin

1 0

double dispatching versus coercion
by kyma＠SymbolicSound.com 31 Jan '98

31 Jan '98

I think that the number of double dispatching methods needed is far fewer than n^2 for n arithmetic classes. In Kyma, we currently have a hierarchy of 94 classes of "arithmetic" objects, of which about 18 have substantially different behavior (and are not abstract). To implement double dispatched addition, for example, we needed about 90 methods: 17 implementors of #+, and 71 implementors of #sumFromXXXX:, where XXXX is some class name. You can greatly reduce the number of methods by taking advantage of the "receiver" superclass hierarchy using "forwarding" double dispatching methods. For example, #sumFromSmallInteger: is defined in ArithmeticObject (the abstract superclass of all of these objects) to resend the message as #sumFromInteger:; that message is defined to resend as #sumFromNumber:; that message resends #sumFromScalar:, etc. working up the class hierarchy of the receiver. Only in those classes that need to distinguish between adding themselves to SmallIntegers and Numbers, for instance, need to implement both #sumFromSmallInteger: and #sumFromNumber:. Kurt Hebel

1 0

Radical interface changes (was Windoze-style 3D borders and titlebar etc)...
by Kevin R. Powell 31 Jan '98

31 Jan '98

Well the recent changes made by Sam S. Adams, et al got me to coding up my favorite style of interaction with windows. I have always loved Self's Morphic, and the current work of course, and its visual and interaction style. So I have tried to capture some of it by using the windoze junta as a starting point. So I made a few screendumps to see if any one is interested in the changes. Take a look at: http://www.canis.uiuc.edu/~powell/squeak-dump.gif http://www.canis.uiuc.edu/~powell/squeak-dump-2.gif If there is any interest I will package up the code as best I can and give it out, Kevin Powell

1 0

Re: Transcendental Numbers
by Hamish (DP) Harvey 31 Jan '98

31 Jan '98

A minor reorganisation and some small changes let it file in. If it works properly I still don't know, though matrix multiplication looks OK. One thing which has come up, though: several methods use local variables in blocks thus: aNumber isComplex ifTrue: [| u v | u := ...snip... which squeak doesn't like. It reckons there should be a right bracket where the | opening the local variable section is. Is there a reason for this? I haven't used Smalltalk for a couple of years, and have just come back after finding Squeak on a web crawl, so I don't yet feel up to delving too deeply for the answer. > Double-dispatching was faster in VisualWorks than coercion was. It > seems a lot more elegant to me, as well. Further, it works It does indeed. Smalltalks handling of arithmetic always struck me as a weak point. Cheers, Hamish

1 0

Re: Makefile quesion: #include "sq.h"
by Ian Piumarta 30 Jan '98

30 Jan '98

Brian, > Does anyone know why sqFilePrims.o, sqSoundPrims.o, sqUnixNetwork.o, and > sqXWindow.c apparently aren't reading in their header file "sq.h" [...] > The sq.h file definitions don't seem to be getting into the above files, > though the #include "sq.h" is at the top of each file. Here are some things to try. Compile with "-H" to see if the preprocessor is including the right "sq.h". Compile with "-E" to check the output from the preprocessor to see if it looks reasonable -- you should be able to see the definitions for "int success(int);", and so on, in the pre-processed output of sqFilePrims.c and friends. Run "nm" on "sqFilePrims.o" and "interp.o" and pipe it into "grep _success": the former should have a symbol table entry of type "U" for _success, and the latter an entry of type "T". If all of that checks out okay, then I'm baffled. You might want to double-check that the "make" program is linking all the required ".o" files, or try linking them manually: gcc -o Squeak *.o -lX11 -lXext -lm and maybe with the "-v" flag to see precisely which options the driver is passing to the various compilation stages. > A Squeak Makefile is needed to work on a NetBSD-1.3 unix system which > has not yet got a port of Squeak. [...] > gcc generated system info through gcc -E -dM c.c This will print a bunch of symbols, amongst which you should choose one (or more) to identify the platform. (You could equally well discover these by running "gcc -v" which will print the path to the "specs" file, in which you will find a list of the platform-specific definitions that gcc is using on your platform, immediately after the line "*cpp:".) Once you have rectified the linking problems I suggest you copy the iX86/FreeBSD portion of the config file, change the symbols in the #if defined(...) lines to check for your platform, and alter the #defines and #undefs to suite the M68K instead of the Pentium. My guess is that the OS stuff is all the same, and the CPU stuff is all the opposite way round except for DOUBLE_WORD_ALIGNMENT (if this is incorrect then you'll get a SEGV the first time the Squeak tries to read or write a Float): #if defined(__BSD__) && defined(__m68k) /* 68k/BSD */ # if defined(SQ_CONFIG_DONE) # error configuration conflict # endif # undef HAS_D_NAMLEN # undef HAS_TIMEZONE # undef HAS_ON_EXIT # define HAS_MSB_FIRST /* changed */ # undef DOUBLE_WORD_ALIGNMENT # undef DOUBLE_WORD_ORDER /* changed */ # define JUMP_ALIGN_STRICT /* changed */ # define SQ_CONFIG_DONE #endif I'm not 100% certain about DOUBLE_WORD_ORDER -- you'll have to try it both ways and see which one works. (The image might even come up, and restore the display, with the wrong setting. The first thing you should try is "print it" on "Float pi", and if the value is right then then flag is right.) Good luck! Ian

3 2

RE: multithreading support
by Alejandro F. Reimondo 30 Jan '98

30 Jan '98

Are you thinking in implementing threads as : 1.- "Smalltalk Spaces" (implementation of connectable Virtual Machines each with its own Interpreter+GC+ClassHierarchy+ObjectArea+...) ? 2.- processes attachables to native threads of execution ? First approach can be more powerful, and it can be interesting for defining a new way to break our monolithic image. Ale. Alejandro F. Reimondo Feel free to visit Smalltalk User Group of Argentina (SUGAR) at http://www.sugarWeb.com Evolution make Smalltalk gurus. Mutation make Java funs. ---------- Desde: Sehyo Chang Enviado el: Miércoles 28 de Enero de 1998 20:52 Para: squeak Asunto: multithreading support <<Archivo: ATT00000.htm>> I am starting prototyping effort of add native multi-threading support to Squeak. Any suggestions, hints, advice would be appreciated. -- sehyo chang

3 4

Re: Transcendental Numbers
by John Brant 30 Jan '98

30 Jan '98

It should be ftp://st.cs.uiuc.edu, not http://st.cs.uiuc.edu. John Brant ----------------------------------- At 09:52 PM 1/29/98 -0600, James Rupert wrote: >Ralph, > >I tried to access the paper mentioned below, but the site is inaccessible. >Is this the right location? Is this server site down? Any ideas? > >Regards, >James >-----Original Message----- >From: Ralph E. Johnson <johnson(a)cs.uiuc.edu> >To: squeak(a)cs.uiuc.edu <squeak(a)cs.uiuc.edu> >Date: Friday, January 30, 1998 8:37 AM >Subject: Re: Transcendental Numbers > > >>Double-dispatching was faster in VisualWorks than coercion was. >>It seems a lot more elegant to me, as well. Further, it works >>better for matrices. Kurt Hebel made an implementation for >>ParcPlace Smalltalk before ParcPlace switched over, and we wrote >>a paper about it that you can find at >>http://st.cs.uiuc.edu/pub/papers/patterns/double-dispatch.ps >>The code is at http://st.cs.uiuc.edu/pub/Smalltalk/st80_r4/arithmetic.st >>It includes all sorts of wierd arithmetic objects, such as matrices, >>infininities and infinitessimals, polynomials and functions. It shouldn't >>be that hard to port to Squeak. >> >>-Ralph >> >> >> > > >

1 0

Re: Transcendental Numbers
by johnson＠cs.uiuc.edu 30 Jan '98

30 Jan '98

One of the things that the double-dispatching paper describes is how careful use of forwarding functions can eliminate the N^2 problem. It also describes a browser that generates a lot of the code for you. I've wanted to have somebody reimplement that browser for years. -Ralph

1 0

Re: Three beginner's questions
by guzdial＠cc.gatech.edu 30 Jan '98

30 Jan '98

At 7:40 PM 1/29/98, Ward Cunningham wrote: >> > 1) What is the easiest way to draw a graphical object (e.g., a line)? > >Phillippe -- Andreas offers some nicely general solutions to your >question. I'd like to add that Smalltalk has always supported turtle >graphics which is probably the EASIEST way to draw a line: > > Pen new > turn: 30; > go: 100 I can get one level easier, if you're willing to interact with the code: Line example Click at the starting point, click at the ending point. POOF! A line! This is the example that I use in class to shock students by drawing lines across windows and scrollbars. (This quarter, there was an audible "Cool!" :-) BTW, I just wrote a position paper for a CHI workshop on Learner Centered Design where I talked about Squeak and being able to draw on your windows. Comments welcome! http://guzdial.cc.gatech.edu/Commentary/userAsLearner.html Mark -------------------------- Mark Guzdial : Georgia Tech : College of Computing : Atlanta, GA 30332-0280 (404) 894-5618 : Fax (404) 894-0673 : guzdial(a)cc.gatech.edu http://www.cc.gatech.edu/gvu/people/Faculty/Mark.Guzdial.html

1 0

Jump to page: