Squeak-dev September 2002

squeak-dev@lists.squeakfoundation.org

165 participants
288 discussions

RE: [FIX] SortedCollectionFix-sr
by Richard A. O'Keefe 30 Sep '02

30 Sep '02

"Andreas Raab" <Andreas.Raab(a)gmx.de> wrote: So, for example, it would be trivial to rewrite Dictionary>>keysSortedSafely ^SortedCollection withAll: self keys sortBlock:[...] SortedCollection class>>withAll: aCollection sortBlock: aBlock "Answer an instance of me such that its elements are sorted according to the criterion specified in aBlock." ^(super new: aCollection size) sortBlock: aBlock; addAll: aCollection which is much more readable anyways There's no doubt about that. Can we please have this method added to SortedCollection class? Pretty please?

1 0

Re: [FIX] SortedCollectionFix-sr
by Richard A. O'Keefe 30 Sep '02

30 Sep '02

I should make it clear that I am in complete and total agreement with Stephan Rudlof <sr(a)evolgo.de> that clients of SortedCollection should not be using an *UNCHECKED* #addLast:. He found some code that was using #addLast: The first one is IndexFile>>readFrom: aStream messageFile: messageFile, its comment says: "Initialize myself from the given text stream. It is assumed that the .index file was written in order of ascending message timestamps, although this method is only less efficient, not incorrect, if this is not the case." A lazy implementation of SortedCollection with a sorting method that exploits existing order (and there are lots of those) would provide linear time building *without* any special bypass. Let me make this clear: there are several known sorting methods, including some that are quite efficient, which have the pleasant property that if you ask them to sort N items that are already in sorted order, they only do O(N) comparisons; N-1 in some cases. This should be an affordable overhead. Dictionary>>keysSortedSafely "Answer a SortedCollection containing the receiver's keys." | sortedKeys | sortedKeys _ SortedCollection new: self size. sortedKeys sortBlock: [:x :y | "Should really be use <obj, string, num> compareSafely..." ((x isString and: [y isString]) or: [x isNumber and: [y isNumber]]) ifTrue: [x < y] ifFalse: [x class == y class ifTrue: [x printString < y printString] ifFalse: [x class name < y class name]]]. self keysDo: [:each | sortedKeys addLast: each]. ^ sortedKeys reSort This would seem to be a case for #addAll:. SortedCollection>> addAll: aCollection aCollection size > (self size // 3) ifTrue: [aCollection do: [:each | super addLast: each]. self reSort] ifFalse: [aCollection do: [:each | self add: each]]. ^ aCollection This code is striking a balance between number of comparisons (ifFalse: case better) and number of moves (ifTrue: case usually better, but sort M then merge N+M would be better still, and for all but small N would do better than the ifTrue: code on comparisons as well). The odd thing about it is that the cross-over point is not at all well chosen. If we let M = N/4, then the cost of this method is O(N**2) even though O(NlgN) is achievable. Now I see a problem: the implementor (sma) of Dictionary>>keysSortedSafely possibly has seen, that in his case each element would be added one by one, instead of adding all unsorted followed by a reSort. And then he has tried to optimize it... How to deal with this case? Use a lazy implementation with an efficient order-exploiting sort and get #reSort out of the public interface. I see, I've opened up a can of worms... But my conviction stays, that my suggestion is - evolutionary - better than the current status. And an evolutionary change in the right direction - though admittedly not ideal - is better than no change. The "evolutionary" change here would be to allow #addLast: but make it check that the class invariant would not be broken. Everything can be made to go smoothly if we have *stronger* encapsulation than now (no addFirst:, no addLast:, and definitely no reSort) and a better (but still quite simple) implementation inside.

1 0

Re: Symbian platform port
by Chris Double 29 Sep '02

29 Sep '02

"Frank Shearar" <frank.shearar(a)rnid.org.uk> writes: > I've read talk here and there about Squeak being ported (or people wanting > to port it) to the Symbian platform (Nokia 9210, etc). > > I was just wanting to know if anyone knows the status of any > attempts to do this? I have the SDK for the 9210 and have ported a few languages to it (io [1], Ficl [2], and attempted Goo [3]). I'd be keen to see Squeak on the 9210 and would be happy to dig into a port. Some things that would need to be overcome would be: 1) No mouse or stylus. Only a keyboard is available. There is a 'virtual cursor' that can be controlled with a direction pad but that could be quite painful to use. 2) Only 3MB or so of memory to play with. Is it even possible to pare Squeak down to run in this with Morphic? How about MVC? 3) A Symbian application is really a DLL. It can contain no static variables at all. Are the many global statics in the Squeak source? 4) The Symbian application can only have an 8KB stack. Is this sufficient? 5) It's possible to do an .exe on Symbian rather than an app which works around 3 and 4. But an .exe doesn't play nicely with the OS. It won't appear in task lists, can't be switched to, etc. But I'd accept that if needed to get it going. [1] http://www.dekorte.com/Software/C/Io/ [2] http://ficl.sf.net [3] http://www.googoogaga.org Chris. -- http://radio.weblogs.com/0102385 http://www.double.co.nz

3 3

Re: Symbian platform port
by Chris Double 29 Sep '02

29 Sep '02

Thanks for the pointer to the code Tim, I've downloaded it and will take a look. Hopefully it'll help towards a Symbian port. Even if it uses too much memory on the 9210 there's the possibility of it running on future phones with more memory (Like the SonyEriccson P800). > One of the things that was/is a problem with the EPOC port is the > apparent need for the globals to be gathered into a struct of some > some sort and initialised. On Symbian (and probably Epoc) if the application is to play nicely with the OS it is really a DLL that is loaded by the OS. DLL's are not allowed to have static writable variables to allow them to be placed in ROM. So the big task in porting is usually to get rid of global variables. As you mention there is a hack to get around it. You have the application built as a .exe (which can have global variables) and a DLL 'front end' which spawns the .exe inside it's own address space. The technique is described in the write up about porting Mame to Epoc: http://www.symbian.com/developer/techlib/papers/MAME/MAMEDream.html Chris.

1 0

Bad Aspects of Morphic?
by Aaron J Reichow 29 Sep '02

29 Sep '02

Hello! Anyone with experience programming more average or normal GUI toolkits knows that Morphic is different. I think that most people, if they learn how to use it, find Morphic better in many ways. At least I do. I find myself wanting to change my environment, be it Mac OS X, Linux or Win98 (only at work, and not for long! :), but then I realize that Cocoa/GTK+/Win98 are a PIA compared to Morphic for a lot of things. But surely, Morphic has to have issues. Speed is the biggest one I've heard and observed myself. What else is wrong with Morphic and how can we fix it? I'd say we have one of the coolest UI toolkits around, but it can 1. Speed * Solution: ? 2. No a consistent and comprehensive 'regular' set of widgets for Morphic. * Solution: Commuinity could decide on Prefab, BobsUI or something new- but a partial consensus would be good, so newbies (or oldies) know where to put work when they want a new widget part of this consistent set 3. I'm told events in Morphic suck? Regards, Aaron Aaron Reichow :: UMD ACM Pres :: http://www.d.umn.edu/~reic0024/ "the profit system follows the path of least resistance and following the path of least resistance is what makes a river crooked." :: u. utah phillips

4 4

Re: Symbian platform port
by Chris Double 29 Sep '02

29 Sep '02

"Frank Shearar" <frank(a)angband.za.org> writes: > Is it then not feasible to write a mouse simulator that wraps the > direction pad? I've only done a couple hours' work with the Symbian > SDK, so I'm pretty much clueless as to its workings, unfortunately. Yes that's possible. It's the approach I'd take first if I was doing it. There is an 'epoc' port floating around that could perhaps be used as the basis for a Symbian port. I did some quick experimentation and the base Squeak spource code compiles fine under the Symbian SDK. It's just a matter of implementing the various platform defined functions and hopefully getting it to work in less than 3MB. > Would one then be able to craft a VM that didn't use a mouse? I'm > thinking here of being able to write applications in Squeak to run > on the phone, much as one writes Java ME apps... It's certainly possible. I've got Io [1] working on the phone for example which allows me to code in that language. The next release of Io will include my Symbian support hopefully. [1] http://www.dekorte.com/Software/C/Io/ Chris. -- http://www.double.co.nz

2 1

ThingLab Won't Filein
by Mark4Flies＠aol.com 29 Sep '02

29 Sep '02

I am trying to use the ThingLab but I cannot get past installing it. I downloaded ThingLab v2 today. I am running Squeak 3.2.7 beta 6 (Carbon) updated to 4956 under Mac OS X 10.2.1 on an iMac. I open a new MVC project and file in the Predefs.cs without any trouble at all. As soon as I try to file in the other part, I get the message: "You may not create any more undefined objects - use nil." Any suggestions? Thanks! -Mark

1 0

collaborative
by Niko Schwarz 28 Sep '02

28 Sep '02

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 A few Requests for Enhancements (dont feel cool enough to do it myself): The local IP address should be detected properly (it says my internet IP would be 127.0.0.1 -- hmm, that not cool). There should be a squeak server storing the "badged" users, for im too lazy to always seek for my friends IPs over different messangers. There should be a way to see all online squeakers, who wish to be seen. regards, nick - -- No committee could ever come up with anything as revolutionary as a camel -- anything as practical and as perfectly designed to perform effectively under such difficult conditions. -- Laurence J. Peter -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.0.6 (GNU/Linux) Comment: For info see http://www.gnupg.org iD8DBQE88Nj6c9uyCiO8RNQRAlgEAJkBNVqOC89nAPiStNsq8hCaMmAXXwCfWnaJ 4CH+hMTEko50bBmZNBCWs/E= =BfwJ -----END PGP SIGNATURE-----

14 33

[pre-ANN] SqueakNIC 30.11
by Marcus Denker 28 Sep '02

28 Sep '02

Hi! Just a short pre-Announcement. SqueakNIC, the Squeak Event that was first organised at the University of Bern, has started to travel north: WHAT ? SQUEAK EVENT WHERE ? UNIVERSITY OF KARLSRUHE - ATIS Karlsruhe, Germany WHEN ? Saturday 30th of November 2002 -- 12pm until ... CONTACT ? email : marcus(a)ira.uka.de Karlsruhe is 3h away from Bern, 1h from Frankfurt. What's SqueakNIC? Squeak can be regarded as a Great Big Toy. Morphics are for something. We can see each squeaker has his/her own hints and experience. Why not to share this knowledge ? The goal of SqueakNic is try to give an anwser... It consists basically of an meeting of squeakers (without regarding level) around computers (and pizza ? :-) The purpose is to show what you want about Squeak, these could be Morph manipulation, explanations about existing projects, new cooperation/interaction between research project (or simply independent developers), ... A official announcement will be posted end of October. Marcus -- Marcus Denker marcus(a)ira.uka.de -- Squeak! http://squeakland.org

1 0

None
by None 28 Sep '02

28 Sep '02

1 0

← Newer
1
2
3
4
5
6
7
...
29
Older →

Jump to page: