Squeak-dev November 1998

squeak-dev@lists.squeakfoundation.org

128 participants
265 discussions

Re: PROPOSAL: create the squeak-help list
by Eric Ulevik 05 Nov '98

05 Nov '98

-----Original Message----- From: Dwight Hughes <dwighth(a)ipa.net> To: squeak(a)cs.uiuc.edu <squeak(a)cs.uiuc.edu> Date: Thursday, 05 November 1998 10:56 pm Subject: Re: PROPOSAL: create the squeak-help list >russell.allen(a)firebirdmedia.com wrote: >> >> Hi >> >> Would there be some way of moving answers to questions on this new list over to >> the Squeak Swiki so that people are not forced to trawl through thousands of >> old emails to find that answer they are seeking? (Either something automated >> or a rule such as "The person who asks the question updates the Swiki"?). >> >> That sort of rule could make the swiki into a self-updating FAQ... > >I like the "the person who asks the question updates the Swiki" rule -- >it would have to be pasted into the Swiki by hand though. Perhaps we >could have a special "append" form or something like the PARC Sparrow >lightweight editable webpage to simplify the process. It would make a >great resource for everyone. > >> Russell >> >> P.S. As as aside, 1 and 2 cent coins are still legal tender in Australia >> (although not used), so $0.02 must be worth *something* ;-) > >Hmmm, what is your smallest value coin then? > >-- Dwight > > 5 cents. 1 and 2 cent coins are no longer legal tender.

1 0

Re: PROPOSAL: create the squeak-help list
by russell.allen＠firebirdmedia.com 05 Nov '98

05 Nov '98

Hi Would there be some way of moving answers to questions on this new list over to the Squeak Swiki so that people are not forced to trawl through thousands of old emails to find that answer they are seeking? (Either something automated or a rule such as "The person who asks the question updates the Swiki"?). That sort of rule could make the swiki into a self-updating FAQ... Russell P.S. As as aside, 1 and 2 cent coins are still legal tender in Australia (although not used), so $0.02 must be worth *something* ;-) ---------------------------------- Dwight Hughes wrote: > I would like to semiformally propose that we create the squeak-help > list. > > This list would be for how-to questions about Squeak and its environment > ("how do I turn the sound off?", "how do I get a list of all the > classes?", "how do I store something in a file?",...); for learning how > to start programming in Squeak/Smalltalk; for "why doesn't this code > work?" questions; and the like. > > A few reasons for a squeak-help list: > > (1) The advanced topics often discussed in the Squeak list can be a > rather overwhelming to those just learning Smalltalk/Squeak. > > (2) Those who know Smalltalk but are new to Squeak may be embarrassed to > ask simple "stupid" questions that they need answers to. > > (3) The shear volume of email to the list can bury simple questions for > help, which may never get answered. > > (4) Newbies may just not feel comfortable discussing simple basic > problems when everyone else on the list is splitting quarks. > > (5) With the upcoming version 2.3, I expect the level of discussion to > get even more advanced (and that the email volume will increase): > pluggable primitives, weak arrays and finalization, 2-D > device-independent anti-aliasing vector graphics, .... > > (6) The number of newbies coming to Squeak through things like the > Pluggable Web Server and through Squeak becoming more and more widely > known on the net seems to be increasing -- it would be good for them to > have a forum that more directly addresses their needs. > > If you would like to see such a list created, DO NOT REPLY TO THE SQUEAK > LIST - email ME directly at mailto:dwighth@ipa.net and I will tally the > responses and send the results back to the list. I will not be the one > setting up this list: I do not have any webserver resources available to > me -- this will probably fall to John Brant, should he be willing. > However, I might be persuaded to compile a Squeak FAQ as my > contribution. > > -- Dwight

2 1

Double dispatch
by Jim Menard 05 Nov '98

05 Nov '98

Gang, I've seen quite a few references to double dispatching. What is that? URL's, pointers to docs, or a simple RTFM (but *which* FM) would be appreciated. Jim -- Jim Menard jimm(a)io.com http://www.io.com/~jimm/ BeOS developer #1283 "An object at rest cannot be stopped!" -- The Evil Midnight Bomber What Bombs At Midnight

6 6

PROPOSAL: create the squeak-help list
by Dwight Hughes 05 Nov '98

05 Nov '98

I would like to semiformally propose that we create the squeak-help list. This list would be for how-to questions about Squeak and its environment ("how do I turn the sound off?", "how do I get a list of all the classes?", "how do I store something in a file?",...); for learning how to start programming in Squeak/Smalltalk; for "why doesn't this code work?" questions; and the like. A few reasons for a squeak-help list: (1) The advanced topics often discussed in the Squeak list can be a rather overwhelming to those just learning Smalltalk/Squeak. (2) Those who know Smalltalk but are new to Squeak may be embarrassed to ask simple "stupid" questions that they need answers to. (3) The shear volume of email to the list can bury simple questions for help, which may never get answered. (4) Newbies may just not feel comfortable discussing simple basic problems when everyone else on the list is splitting quarks. (5) With the upcoming version 2.3, I expect the level of discussion to get even more advanced (and that the email volume will increase): pluggable primitives, weak arrays and finalization, 2-D device-independent anti-aliasing vector graphics, .... (6) The number of newbies coming to Squeak through things like the Pluggable Web Server and through Squeak becoming more and more widely known on the net seems to be increasing -- it would be good for them to have a forum that more directly addresses their needs. If you would like to see such a list created, DO NOT REPLY TO THE SQUEAK LIST - email ME directly at mailto:dwighth@ipa.net and I will tally the responses and send the results back to the list. I will not be the one setting up this list: I do not have any webserver resources available to me -- this will probably fall to John Brant, should he be willing. However, I might be persuaded to compile a Squeak FAQ as my contribution. -- Dwight

2 1

Re: Where can I get some etoys?
by scottw＠wdi.disney.com 04 Nov '98

04 Nov '98

Bob, It will be much more straightforward in the forthcoming Squeak 2.3 release to create "etoy-like" things than it has been in 2.2. I advise waiting a couple of weeks. However, if you wish to push forward on this in the meantime, a good place to look is in the "ReadMe.text" file shipped with Squeak 2.2; the relevant section starts with the heading "Programming in Morphic". Or, more conveniently, you can point a web browser directly at: http://minnow.cc.gatech.edu/SqueakDoc.11 Finally, if none of the above satisfies your immediate need, here are a few steps that will get you on the road to reconstructing Alan's "race car" demo in the current release: * Start out in a fresh Morphic Construction Project. * From the "Scripting" submenu of the world menu, get a "detachable scripting space". * Drag out a blank playfield (pale green rect) from the parts bin and deposit it in the scripting space. This will be the place where you draw your car and where it will travel around. * Paint your car by dragging a "paint palette" from the parts bin and dropping it into Playfield, then painting the car, then hitting "Keep". (It's also possible to script a "Paint" tab of the palette so that it invokes painting automatically whenever it's hit, as in Alan's etoy demo, but drag-n-drop of the palette icon should do to get you started.) * Obtain a "Viewer" for the car by dragging from its cyan-colored halo handle; place the viewer where you wish. (It is also possible to have viewers all show up at a single, reusable place, as in Alan's etoy demo, but it's rather strenuous right now to set this up. In 2.3 Squeak this will be *much* easier.) * Rename the car from "Sketch" to "Car" by clicking on its name at the top of the Viewer (the name-editing area will turn yellow), backspacing to get rid of the unwanted text, typing "Car", and then hitting ENTER. * Get a nascent script for the car by dragging from the phrase in the Viewer that reads "Car turn by 5" and dropping somewhere into your scripting space. * Now proceed to modify and test your script, by adding new phrases, changing parameters, etc. * Later on, paint your steering wheel, bring up its viewer, and drag tiles from the relevant places in that viewer, dropping them atop the appropriate places in your script, to achieve the effect you want. For example, if you want the amount by which the car turns to be determined by the heading of the steering wheel, get a tile for "steeringWheel's heading" anddrop it over the "5" parameter in the "car turn by 5" phrase of the car's script. Hope this helps, -- Scott -------------------------------------------------------------- At 11:32 AM -0800 11/4/98, Houston, Bob wrote: >I had the opportunity to see Alan do his etoy demo last week. Wow! This >stuff is really cool!! > >The etoy authoring tools and scripting environment have come a long way in >the past few months. I'd really like to play around with this stuff, but I >seem to be missing something. I remember playing with the authoring >tools/scripting support in a previous release of squeak, but I can't seem to >get it to work in the version 2.2. I'm sure it's just pilot error on my >part. ;-) > >Could someone send me a pointer to get me going? If this info is available >on the SqueakDoc site please send me the page number. Here's what I' trying >to do: > >I'd like to re-create the "race car" demo, where you draw two sketch morphs >(a race car and a steering wheel) and "wire" them together so that rotating >the steering wheel causes the car to be driven around the screen. Pretty >simple, right? Question: How exactly do I invoke the authoring tools and >connect the two morphs in this way? > >Thanks in advance, > >Bob Houston, Tech Rep bobh(a)rational.com ><mailto:bobh@rational.com> >Rational Software Corporation 703 761-4466 >8000 Westpark Drive, Fifth Floor >McLean, VA 22102 USA

1 0

Re: Units package (was Method names)
by David N. Smith 04 Nov '98

04 Nov '98

At 4:09 -0500 11/4/98, JArchibald(a)aol.com wrote: >Dave Smith (of IBM Research) did a units package of some completeness many >years ago. I was surprised when he did it how much it clarified things in >certain cases. This stuff may be in his book (early one). > >Jerry. Jerry: I still have that code someplace, probably. Since I wrote it I have run into two guys who wrote a much better, open ended (ie, define your own units), and much faster package for the engineering department of a large aerospace(?) company. The technique used to make it really fast should have been patented but the company they worked for didn't want to do it. They were going to publish, and at one point we agreed I'd even write the paper, but somehow it never got done. Anyhow there is a truly neat trick which can let units accumulate (ie, feet cubed * seconds squared * meters) and rapidly figure out later what's included. I'll send some mail and see what the current state of publicity is. Dave _______________________________ David N. Smith IBM T J Watson Research Center Hawthorne, NY _______________________________ Any opinions or recommendations herein are those of the author and not of his employer.

1 0

Re: Double dispatch
by Ralph Johnson 04 Nov '98

04 Nov '98

The best paper on double-dispatching was the one that Dan Ingalls wrote for OOPSLA'86. But if you don't have the OOPSLA'86 proceedings, or the CD with the first ten years of papers on it, you might read the paper that Kurt Hebel and I wrote on it at ftp://st.cs.uiuc.edu/pub/papers/patterns/double-dispatch.ps Our paper focuses on how to implement arithmetic, and on a tool that we built to make double-dispatching easier. -Ralph

3 4

Squeak memory allocation
by lex＠cc.gatech.edu 04 Nov '98

04 Nov '98

Okay, since my words have been unconvincing, I hacked a little code to see how this really works. First, my gloom and doom about static memory allocations was worse than the true case. I claimed that if you allocate a larger heap (via command line flags or whatever) than you have actual RAM, that you will definately get swapping. In fact this is not quite true. The current ObjectMemory does an incremental collection after every 2000 allocations, and so in the typical case the upper parts of a large but mostly empty heap will never get touched. However, that's only the typical case. If you start allocating very large objects, then you can allocate tons of space before the 2000-allocations trigger goes off. To test this, I ran an image with 50 megs of heap, and executed: 10000 timesRepat: [ Array new: 1000000 ] The Squeak process grew to 40 megs resident, and all my other running apps started getting swapped out. Next, I modified the ObjectMemory to also trigger a GC if more than 3 megs have been allocated since the last GC. When I executed the above line with the new VM, the Squeak process only went up to 17 megs. The patch is below, for anyone who wants to play with it. It's a bit of a hack as is, but simply adding a "maxAllocationBetweenGCs" instance variable, and finding a better name than "lastFreeSpace" would make it pretty clean. Also, maybe the max allocation should be smaller for smaller images; for a 2 meg image, for instance, you might want to do a GC after just 1 meg of allocation. Lex 'From Squeak 2.2 of Sept 23, 1998 on 3 November 1998 at 4:52:29 pm'! "Change Set: FrequentGCs Date: 3 November 1998 Author: Lex Spoon trigger a GC after every 3 megabytes of allocation, so that Squeak doesn't use memory high in its heap unless it really needs it."! Object subclass: #ObjectMemory instanceVariableNames: 'memory youngStart endOfMemory memoryLimit lastFreeStart nilObj falseObj trueObj specialObjectsOop rootTable rootTableCount child field parentField freeBlock lastHash freeLargeContexts freeSmallContexts allocationCount lowSpaceThreshold signalLowSpace compStart compEnd fwdTableNext fwdTableLast remapBuffer remapBufferCount interruptCheckCounter checkAssertions allocationsBetweenGCs tenuringThreshold statFullGCs statFullGCMSecs statIncrGCs statIncrGCMSecs statTenures statRootTableOverflows displayBits ' classVariableNames: 'AllButHashBits AllButMarkBit AllButMarkBitAndTypeMask AllButRootBit AllButTypeMask BaseHeaderSize CharacterTable ClassArray ClassBitmap ClassBlockContext ClassByteArray ClassCharacter ClassCompiledMethod ClassFloat ClassInteger ClassLargePositiveInteger ClassMessage ClassMethodContext ClassPoint ClassProcess ClassPseudoContext ClassSemaphore ClassString ClassTranslatedMethod CompactClasses ConstMinusOne ConstOne ConstTwo ConstZero Done ExternalObjectsArray FalseObject FreeSizeMask GCTopMarker HashBits HashBitsOffset HeaderTypeClass HeaderTypeFree HeaderTypeGC HeaderTypeShort HeaderTypeSizeAndClass LargeContextSize MarkBit MinimumForwardTableBytes NilContext NilObject RemapBufferSize RootBit RootTableSize SchedulerAssociation SelectorCannotReturn SelectorDoesNotUnderstand SelectorMustBeBoolean SmallBlockContext SmallContextSize SmallMethodContext SpecialSelectors StackStart StartField StartObj TheDisplay TheFinalizationSemaphore TheInputSemaphore TheInterr! ! ! uptSemaphore TheLowSpaceSemaphore TheTimerSemaphore TrueObject TypeMask Upward ' poolDictionaries: '' category: 'Squeak-Interpreter'! !ObjectMemory methodsFor: 'initialization' stamp: 'ls 11/3/1998 15:43'! initializeMemoryFirstFree: firstFree "Initialize endOfMemory to the top of oop storage space, reserving some space for forwarding blocks, and create the freeBlock from which space is allocated. Also create a fake free chunk at endOfMemory to act as a sentinal for memory scans." "Note: The amount of space reserved for forwarding blocks should be chosen to ensure that incremental compactions can usually be done in a single pass. However, there should be enough forwarding blocks so a full compaction can be done in a reasonable number of passes, say ten. (A full compaction requires N object-moving passes, where N = number of non-garbage objects / number of forwarding blocks)." | fwdBlockBytes | "reserve space for forwarding blocks" fwdBlockBytes _ MinimumForwardTableBytes. (memoryLimit - fwdBlockBytes) >= (firstFree + BaseHeaderSize) ifFalse: [ "reserve enough space for a minimal free block of BaseHeaderSize bytes" fwdBlockBytes _ memoryLimit - (firstFree + BaseHeaderSize). ]. "set endOfMemory and initialize freeBlock" endOfMemory _ memoryLimit - fwdBlockBytes. freeBlock _ firstFree. self setSizeOfFree: freeBlock to: (endOfMemory - firstFree). "bytes available for oops" "make a fake free chunk at endOfMemory for use as a sentinal in memory scans" self setSizeOfFree: endOfMemory to: BaseHeaderSize. checkAssertions ifTrue: [ ((freeBlock < endOfMemory) and: [endOfMemory < memoryLimit]) ifFalse: [ self error: 'error in free space computation' ]. (self oopFromChunk: endOfMemory) = endOfMemory ifFalse: [ self error: 'header format must have changed' ]. (self objectAfter: freeBlock) = endOfMemory ifFalse: [ self error: 'free block not properly initialized' ]. ]. lastFreeStart _ freeBlock.! ! !ObjectMemory methodsFor: 'allocation' stamp: 'ls 11/3/1998 15:57'! allocateChunk: byteSize "Allocate a chunk of the given size. Sender must be sure that the requested size includes enough space for the header word(s)." "Details: To limit the time per incremental GC, do one every so many allocations." | enoughSpace newFreeSize newChunk | self inline: true. allocationCount >= allocationsBetweenGCs ifTrue: [ "do an incremental GC every so many allocations to keep pauses short" self incrementalGC. ]. freeBlock - lastFreeStart > (3*1024 * 1024) ifTrue: [ "do a GC if we've allocadet more than 3 megs" self incrementalGC. ]. enoughSpace _ self sufficientSpaceToAllocate: byteSize. enoughSpace ifFalse: [ "signal that space is running low, put proceed with allocation if possible" signalLowSpace _ true. lowSpaceThreshold _ 0. "disable additional interrupts until lowSpaceThreshold is reset by image" interruptCheckCounter _ 0. ]. (self sizeOfFree: freeBlock) < (byteSize + BaseHeaderSize) ifTrue: [ self error: 'out of memory'. ]. "if we get here, there is enough space for allocation to succeed" newFreeSize _ (self sizeOfFree: freeBlock) - byteSize. newChunk _ freeBlock. freeBlock _ freeBlock + byteSize. "Assume: client will initialize object header of free chunk, so following is not needed:" "self setSizeOfFree: newChunk to: byteSize." self setSizeOfFree: freeBlock to: newFreeSize. allocationCount _ allocationCount + 1. ^ newChunk! ! !Interpreter methodsFor: 'object memory support' stamp: 'ls 11/3/1998 15:42'! postGCAction "Mark the active and home contexts as roots if old. This allows the interpreter to use storePointerUnchecked to store into them." (activeContext < youngStart) ifTrue: [ self beRootIfOld: activeContext ]. (theHomeContext < youngStart) ifTrue: [ self beRootIfOld: theHomeContext ]. lastFreeStart _ freeBlock.! !

5 5

Re: Double dispatch
by Vassili Bykov 04 Nov '98

04 Nov '98

At 09:35 AM 11/4/98 -0600, Lyn A Headley <laheadle(a)cs.uchicago.edu> wrote: > >I'm afraid I must speak up here. I believe you are confusing >delegation with double dispatch. The way I understand it is that >double-dispatch is a *language* feature which looks up the method to >call based on the runtime type/class of *two* of its arguments rather >than simply the first as is the case with most OO languages. when I >say argument I really mean the object to which the message is sent, >but multi-dispatch languages usually use generic messages so the >object really *is* an argument to the method. > >confusing, huh? see Common lisp's CLOS for an example. This is just a mix-up with terminology. "Multiple dispatch" is a language feature of looking up a method depending on the types of more than one argument (so, in fact, there are no messages any more, the "driving force" are generic functions, methods (that are parts of GFs), and method combinations). "Double dispatch" is an implementation technique used in languages with single dispatch to implement multiple dispatch for the case of two discriminating arguments. Not confusing at all. -- Vassili Bykov vassili(a)objectpeople.com The Object People http://www.objectpeople.com +1(613)225-8812 "Any sufficiently complicated C or Fortran program contains an ad hoc informally-specified bug-ridden slow implementation of half of Common Lisp." -- Greenspun's Tenth Rule of Programming

1 0

Re: Regenerating the VM
by lex＠cc.gatech.edu 04 Nov '98

04 Nov '98

Ian Piumarta <Ian.Piumarta(a)inria.fr> wrote: > > > The quick fix is simple: in sq.h, redefine ioLowResMSecs to just be ioMSecs: > > > > #define ioLowResMSecs() ioMSecs() > > It should already be like in the 2.2 sq.h! Maybe there's an erroneous > #define USE_CLOCK_MSECS in sqConfig.h someplace? Which platform were > you running on when you saw delays taking 1000 times too long??? > > Ian Linux 2.0.34, using libc5, with a Cyrix 6x86 133Mhz, for what it's worth. Does the Unix port use a customized sq.h? Looking in (InterpreterSupportCode class>>squeakHeaderFile), the default definition of ioLowRes..() is not in an #ifdef (even though the regular ioMSecs() is). Or does one of the other Unix headers #undef this, which would mean my headers are out of date? Lex

1 0

← Newer
1
...
19
20
21
22
23
24
25
26
27
Older →

Jump to page: