Squeak-dev December 1999

squeak-dev@lists.squeakfoundation.org

146 participants
451 discussions

Re: I should have known better: method categories and file-outs
by JArchibald＠aol.com 30 Dec '99

30 Dec '99

=> 12/29/99 6:47:04 PM, dnsmith(a)watson.ibm.com => << I then filed out the change set >> Dave, Sorry about previous; I see you are attacking a more subtle problem. Well, we have an entirely new millenium to address all of the problems (or not). Jerry.

1 0

Re: I should have known better: method categories and file-outs
by JArchibald＠aol.com 30 Dec '99

30 Dec '99

=> 12/29/99 6:47:04 PM, dnsmith(a)watson.ibm.com => << But is sure isn't what I wanted and I don't see any way around. I want to categorize my new stuff. I want someone else to be able to file it into some version of Squeak that is not identical to mine. In theory the file-in could run across a wide spectrum of versions. >> Dave, ChangeSets are the way to capture a collection of changes. I thought you taught me that, or was it the other way around. In any case, I am working on a generalized 'difference' facility between two source descriptions (sound familiar). At the rate I am going it should be ready about mid-Spring. Jerry.

1 0

Re: I should have known better: method categories and file-outs
by Bob Arning 30 Dec '99

30 Dec '99

On Wed, 29 Dec 1999 17:41:37 -0500 "David N. Smith \(IBM\)" <dnsmith(a)watson.ibm.com> wrote: >But is sure isn't what I wanted and I don't see any way around. I >want to categorize my new stuff. I want someone else to be able to >file it into some version of Squeak that is not identical to mine. In >theory the file-in could run across a wide spectrum of versions. > >There seems to be no way to do that. Am I missing something >(obvious)? or is this a possible requirement for an improved change >management system? It could be handled better, but you could manually edit out the !SomeClass reorganize! etc stuff. The methods you are filing in will categorize themselves. Cheers, Bob

1 0

Re: [Q] fullBounds and Bounds in Morphic
by Bob Arning 30 Dec '99

30 Dec '99

On Wed, 29 Dec 1999 13:24:32 -0800 "Andrew P. Black" <black(a)cse.ogi.edu> wrote: >I'm still not very happy with my initialExtent method, though. > >!AccessorPreferences methodsFor: 'displaying' stamp: 'apb 12/29/1999 11:35'! >initialExtent > > ^ 340 @ ((rows min: 12) * 19 + 35)! ! > >Too many magic numbers! The rows min: 12 calculation is intended to >leave room for at most 11 instance variables and the header; more >than that and the user uses the scrollbar. But I would like to be >able to ask the scrollpane what the width of its submorphs are, and >what the height of the window frame is, rather than putting in 340 >and 35. I couldn't even get the height of a row by asking it (I got >40 as the answer); the magic number 19 came from the explorer. If the scrollPane actually has some contents, you can resort to the (admittedly ugly) scrollPane scroller firstSubmorph height "or width" > >I would have thought that shrinkwrap would do something like what I want. > Well, it does, but only for AlignmentMorphs. The problem here is that there are a couple of other classes that have different ideas: ScrollPane expects to be given a extent and then show whatever can be shown in that area while SystemWindow expects to start at some size (resizable by the user) and divide its space up porportionally. You could fix the first by defining a subclass of ScrollPane that expanded in size (up to some limit) as items were added and then acted like the present ScrollPane. The second problem (important only after the first is fixed) could be dealt with by creating a SystemWindow-ish thing that is really an AlignmentMorph and thus able to #shrinkWrap. Or you could define a subclass of SystemWindow that chose its initial extent from its subpanes rather than from the model. > >So, if I understand this right, fullBounds is the space that I want >to occupy on the screen, or more generally in the enclosing morph. My >fullBounds might be > (1) the same as my bounds (often) > (2) bigger than my bounds, if some submorphs extend beyond > my bounds. > (3) smaller than bounds (never?) > >When I'm asked for my fullBounds, I would normally ask all of my >submorphs what their fullBounds are, and answer the smallest >enclosing rectangle. But if, for example, I'm a vertical scrollPane, >I might give a smaller (in height) answer, because I can scroll to >make the cropped morphs visible. Yup. You can think of fullbounds as the outer limit on a) what parts of the screen might need to be redrawn if you were changed/removed and b) where you or your submorphs might conceivably be interested in mouse clicks. >If this is right, it seems that bounds should be private. I might, >or might not, use my bounds in calculating my fullBounds, but no >other object should ask for my bounds. Well, we could probably argue for privacy for all sorts of things, but sometimes the expedient access (e.g. allowing one morph to reposition/resize other morphs) seems to be worth the price. >Please correct me where I'm wrong! No, I think you've got it. Cheers, Bob

1 0

Re: [FIX] Integer #+, #-, #/
by Andres Valloud 30 Dec '99

30 Dec '99

Hi Stephan. > But here we could go the same path and move the normalization to > #bitAnd:, #bitOr: and #bitXor. Note that #bitOr: with normalized numbers does not need #normalize. Incidentally, some time ago I had produced faster versions of bitOr:, bitAnd: and bitXor: for LargeIntegers that were between 33 and 40 times faster or so. Unfortunately, they do not work with the new bitwise behavior standard for negative integers. Another place that I think would be very nice to revamp is the digitLogic methods in Integer (suggestion by Dan Ingalls)... I thought about doing it but then run out of time and then fell out of my stack... so now you have reminded me of that. Perhaps after I finish some small tasks remaining... > Then we'd have the unique convention that the private methods don't > normalize and the callers are responsible for that. I like this very much. Andres.

1 0

Squeak/Siren vs Csound et al.
by Michal Seta 30 Dec '99

30 Dec '99

Hello, I was just browsing the archives of the squeak mailing list and noticed this thread about squeak and music and all. I have used Csound extensively in my music and still I am quite satisfied with it (at least the synthesis side of it). The real pain is score generation. I am not a programmer and I have been looking for all kinds of tools to help/speed up/customize the composition process for a while. In the meantime I have tried various tools/progamming languages. I even bought Max- and it's still being used in a variety of fun ways. Anyways, in my recent search of best tool for Csound score generation I started learning python (well, also for some cgi apps) and almost started developing some tools for my csound scores. And suddenly I happened to read about squeal (followed some stray link I guess). And then I discovered that Mr Pope developed a music system based on it. So I just spent 3 nights playing around with Squeal (and Siren) trying to understand how it all works. Having said all that, I'm planning to dive a bit more into squeal and explore the music/multimedia side of it. I'm just starting with it so I guess it's going to take a while (is the learning curved considered to be steep ???) before I can contribute anything (or even read code) but i'd be interested in following the development. Is there anyone else using Siren? ...... .... ...... Michal Seta ...... ...... ....

3 2

I should have known better: method categories and file-outs
by David N. Smith (IBM) 29 Dec '99

29 Dec '99

I just carefully categorized the changes for additions to a bunch of basic classes so that all my changes are in a category with the same name. The changes go in a bunch of system classes including Object, Collection, Boolean, etc. (These are purely additions; there are no replacement methods.) I then filed out the change set, which files out updates to some 20 classes, and of course it filed out code to completely replace all categories for all methods in each of the classes rather than just adding my new one. As soon as I saw what it did I knew that I knew it would do that! But is sure isn't what I wanted and I don't see any way around. I want to categorize my new stuff. I want someone else to be able to file it into some version of Squeak that is not identical to mine. In theory the file-in could run across a wide spectrum of versions. There seems to be no way to do that. Am I missing something (obvious)? or is this a possible requirement for an improved change management system? 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

[BUG] Compiler bug
by David N. Smith (IBM) 29 Dec '99

29 Dec '99

I found a way to get a walkback in the compiler; Try: (#1 2 3) instead of the intended: #(1 2 3) 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: Display speed on large monitor
by Raab, Andreas 29 Dec '99

29 Dec '99

Jan, > Folks on the list can correct me if I'm wrong, but Squeak heavily > depends on doing bitblt's from memory to the screen buffer. > The bitblt from memory to screen performance of a modern AGP > based PC video card is as high as 200-300 megabytes/sec Correct. There is one thing that's interesting as well and that's that blitting between VRAM directly is again quite a bit faster (since it doesn't need to go to the AGP bus - compared to the memory speed of a modern graphics card AGP is slow as hell ;-) > Some older display adapters had no way to map the screen buffer into > memory and had to execute graphics processor commands that would > take a stream of data written to an I/O port to get the bits to > the screen. This is exactly what an X-Server does (and the reason why using the XSHM extension for local display make such a difference on Unix Squeak). Andreas

1 0

Re: [Q] fullBounds and Bounds in Morphic
by Bob Arning 29 Dec '99

29 Dec '99

On Tue, 28 Dec 1999 22:59:08 -0800 "Andrew P. Black" <apb(a)cse.ogi.edu> wrote: >I'm trying to build a control panel. The Morphic containment >hierarchy looks like the following (extracted by playing with an >explorer) > > aSystemWindow (fullBounds = 263@340 corner: 863@712) > aScrollPane (fullBounds = 264@357 corner: 863@711) > aTransformMorph (fullBounds 267@345 corner: 849@697) > anAlignmentMorph (fullBounds 0@0 corner: 320@61) > >and the AllignmentMorphs contains more stuff which looks as I expect. >What I don't expect is that the enclosing objects (ATransformMorph >and outwards) have bounds that are so much bigger than the enclosed >AlignmentMorph. I don't give any of these objects explicit sizes, >and I would expect them to re-size around the AllignmentMorph that >they contain. [snip] >What am I doing wrong? As the comment indicates, I stole a lot of >this from Preferences class, where it seems to work just fine. The driving force here is the SystemWindow which has been told to be a certain size and adjusts its suppanes (the ScrollPane) accordingly. How it got told (and something you didn't copy from Preferences) is via #initialExtent. If you add: ========= !AccessorPreferences methodsFor: 'as yet unclassified' stamp: 'RAA 12/29/1999 07:18'! initialExtent ^330@200! ! ========= then you will get more of the behavior you expect. >If there is anything written on how bounds and fullBounds interact, I >would appreciate a pointer. I'm not sure what you mean by interact, but the simple difference is this: - bounds is the rectangle enclosing a morph - fullBounds is generally the rectangle enclosing the morph PLUS all its submorphs. Some morphs (those related to scrolling, e.g.) apply clipping and ignore anything outside their own bounds so they will implement this a bit differently. Hope this is what you were asking. Cheers, Bob

2 1

Jump to page: