Squeak-dev August 2004

squeak-dev@lists.squeakfoundation.org

154 participants
394 discussions

Shrinking
by Pavel Krivanek 29 Aug '04

29 Aug '04

I have modified shrinking process for image 3.0. I have prepared a patch which removes Morphic references from source codes. The result is cca 1.2 MB large compact image with Squeak kernel and basic MVC tools. It has NO OBSOLETE CLASSES!!! Resulting image includes full source codes and class orgranization so it is easy to use and it's prepared for next shrinking steps (abandonSources, zapOrganization...). Installation: Open - MVC project - enter into it - open file list - fileIn file shrink.st You will have to press 'yes' several times and enter your initials. Wait a while (Squeak recompiles all classes) Problems: - This image doesn't contain hundreds of useful methods - Some removed messages have still record in class organization - This is quite old image, but we have well documented sequence of update changes Maybe it's not the right starting point for making fully updated small image 'cos the amount of updates is formidable. Though it's clear we need very small reference image (maybe without Morphic) for making production images. Pavel

8 14

character ^ (caret) is missing in Squeak 3.6
by m r ameri 29 Aug '04

29 Aug '04

Hi every one , character ^ (caret) is missing in Squeak 3.6 windows full version. When i type it Up Arrow symbol is printed. I am using Win Xp. Any clues . kind regards m r ameri

2 1

Squeak Map server is down?
by danil a. osipchuk 29 Aug '04

29 Aug '04

Hi all. Can somebody open package loader in Squeak? It seems that something is going wrong with SM server - wget http://map1.squeakfoundation.org/sm/ping waits for response without result. May be somebody is aware of some manual search & download strategy for published items while Goran is unavaible? Danil

5 6

Re: Shrinking
by Pavel Krivanek 29 Aug '04

29 Aug '04

Yes and no. These classes aren't necessary for system functionality but if you remove them, there will be some wrong references in source codes. For example, try to modify (add space somewhere etc.) and recompile method ChangeSet>>checkForConversionMethods manually. I will get error messages from compiler. Pavel

2 1

[ENH] SelectionMorphEnh-dgd
by DiegoGomezDeck＠ConsultAr.com 29 Aug '04

29 Aug '04

from preamble: "Change Set: SelectionMorphEnh-dgd Date: 29 August 2004 Author: Diego Gomez Deck Assorted improvements for SelectionMorph. To try SelectionMorph you have to start a drag&drop operation with SHIFT pressed. May be be need an entry in Object Catalog for SelectionMorph? Improvements: - Some look&feel changes (color, border color and border width). - More convenient external name. - Two new options: distribute horizontally & distribute vertically. Redistribute the spaces between selected objects in a 'democratic' way. Enjoy it!! "!

1 0

Regex review
by danil a. osipchuk 29 Aug '04

29 Aug '04

Hi all I always was frustated a bit by absence of regex capabilites in Squeak and I hope that finally one of libraries will finaly get into the image. So, I've read related discussions and tried to figure out what folks are most interested in and what I can do myself. It seems I can play with libraries a little and evaluate 1)Speed 2)General ease of use 3)Feature lists for some extent as first iteration input to the list. I'm not Smalltalk or Regex expert, but anyway - little better than nothing. I've loaded both libs (they collided in String a bit, but stayed usable) and started writing methods for simple tasks in parallel. System: Celeron 1.7MHz 384DRAM running Unix VM 3.6-3, Squeak 3.7 #5976 on FreeBSD 4.10, Input file: August-2004 squeak-dev mail archive size ~1M. Both suits are extensivly documented. At first glance it seems that RX (Vassily's version) is more friendly, RE (A.Greenberg's one) - less, maybe as it faced to external library. Both suits seem to be 'feature complete' and common in regexp sense - RX claims that it passes H.Spencer test suit and RE is based on standard PCRE (and as such has some search flags, importance of which i didn't investigate too hard). Converting between two should't be difficult in most cases. I'll present and comment essential blocks of code I've come up solving simple tasks (#millisecondsToRun: #result): 1) counting words in file (pattern: '\w+') RX: self testInFileDo: [:fstream | matcher matchesOnStream: fstream do: [:s | num := num + 1]]. #(23296 277863) RE: self testInFileLinesDo: [:str | num := num + (matcher collectFrom: str) size]. #(11166 277909) Well, first impressions. RX has very nice enumerating protocol with stream support, I solved task in a minute or two. RE support of streams is limited (line-by-line grep) and I didn't even bother with it. Iterating through patterns in RE is much less powerfull - RE can only collect subgroups. I've spent quite a bit time to find the shortest way to count words. It turned out that with choosen file my #testFileLinesDo: gives only one line due to inproper line endings (the same with Re>>grepFrom:to:onMatch:onNonMatch: btw). Thus the whole thing is slow. Ok, further I'll switch to custom version: RE: self testInFileLinesLFDo: [:str | num := num + (matcher collectFrom: str) size]. Walkback. Failed on empty line - it returned nil instead collection. Sorry, and now: RE: self testInFileLinesLFDo: [:str | (matcher collectFrom: str) ifNotNilDo: [:col | num := num + col size]]. #(5715 277909) New version is four times faster than native Smalltalk implementation (not so much IMHO). There is some difference in result, it seems that one should tweak that mysterious search engine flags (and may be I was wrong stating that code between libs is easy to port). 2) Let's take slightly meaningful example - collecting email-like strings from file (pattern: '\w+(a)(\w+\.)+\w+'): RX: self testInFileDo: [:fstream | matcher matchesOnStream: fstream do: [:s | mails add: s]]. #(98460 aSet(483)) RE: self testInFileLinesLFDo: [:str | (matcher collectFrom: str) ifNotNilDo: [:col | mails addAll: col]]. #(2029 aSet(483)) RE violently outperforms RX on such task (50 times). It tends that RE wins in processing itself and looses a bit during massive exchanges from/in primitives (just guess). Result set is mainly garbage - if one needs to collect addresses (to offer something invaluable to all dwellers) correct pattern looks like '\w+ at (\w+\.)+\w+' - RX: #(113588 aSet(143)) RE: #(1961 aSet(143)) 3) Lets change this allready well recognized pattern to something else say 'squeaker at squeak.org' -> 'squeaker dog squeak point org' ('dog' is Russian alias for @). Sorry for a bit crazy engineering style: RX: | matcher | matcher := RxMatcher forString: '(\w+)( at )(\w+\.)+(\w+)'. self testInOutFilesDo: [:in :out | matcher copyStream: in to: out translatingMatchesUsing: [:weWillUseMatcherCacheInstead | String streamContents: [:stream| stream nextPutAll: (matcher subexpression: 2); nextPutAll: ' dog '; nextPutAll: ((matcher subexpression: 4) copyReplaceAll: '.' with: ' point '); nextPutAll: (matcher subexpression: 5) ]]]. Result: 157837ms md5:c33f76ca533ccaba3c486add18331b9e RE: | matcher str| matcher := Re on: '(\w+)( at )(\w+\.)+(\w+)'. self testInOutFilesDo: [:in :out | [in atEnd] whileFalse: [ str := in upTo: Character lf. out nextPutAll: (matcher search: str andReplace: [:m | String streamContents: [:stream| stream nextPutAll: (m matchAt: 1); nextPutAll: ' dog '; nextPutAll: ((m matchAt: 3) copyReplaceAll: '.' with: ' point '); nextPutAll: (m matchAt: 4) ]]). out nextPut: Character lf.]]. Result: 3056ms md5:c33f76ca533ccaba3c486add18331b9e ************************************************************************************************* Conclusion. I like Vassily's library a lot - it is very, very nice and pleasant in use (comments everywhere, carefully designed interface). But it is slow and maybe it can slow down Squeak if heavily used (although I don't know how effective string processing in Squeak core is now). It is unsuitable for extensive processing and good for episodical use (validate user input for instance). A.Greenberg's library of course is not bad at all - it simply follows external library specifics. For string manipulation it appears to be as handy as Vassily's. And it is very, very fast. Danil

3 3

Lessons learnt from been an integrator
by stéphane ducasse 29 Aug '04

29 Aug '04

Hi all here are the lessons we learnt while building the 3.7 image. We wanted to have shout, rb, connector morph in 3.7 but there were too much patches. We are sorry about that. - We had to decide if we add balloon 3d with pooh broken or not, same for the language editor. We decided to add manually the fixes because we do not have access to the code. But this cannot continue like that. For the next release here is a list of properties that packages should have: - be maintained - code access by one of the integrators - minimizing the overriding of methods - we should harvest the changes in the kernel - put in places some mechanisms such as services Conclusion: I will do 3.7 full but do not count on me in the future. Apparently you are all too busy to pay attention to people that spend their time to build stuff for everybody. Stef

3 2

3.8 Release Plan
by Doug Way 29 Aug '04

29 Aug '04

I created a "Release Plan for 3.8" page on the swiki, similar to the To-Do list page we had for 3.7, to help organize what we've discussed so far for the 3.8 release. Here 'tis: http://minnow.cc.gatech.edu/squeak/3832 Hopefully this is in the right ballpark of what we want to do for 3.8. While I was at it, I also did a major overhaul of the "Squeak Versions" page... http://minnow.cc.gatech.edu/squeak/275 - Doug

1 0

Re: project "squeak" on sourceforge.net
by lex＠cc.gatech.edu 29 Aug '04

29 Aug '04

Pupeno <pupeno(a)pupeno.com> wrote: > I see, I didn't know about it, but even then, the fact that USA prohibits > exporting Squeak from USA to Cuba, it doesn't mean that the license must > aprove it (that is, if I make some software under the same license as > Squeak's, here in Argentina,I still wouldn't be able to send it to Cuba, even > when those USA laws doesn't apply to me). > Here RMS explains it: http://www.gnu.org/philosophy/free-sw.html > "Sometimes government export control regulations and trade sanctions can > constrain your freedom to distribute copies of programs internationally. > Software developers do not have the power to eliminate or override these > restrictions, but what they can and must do is refuse to impose them as > conditions of use of the program. In this way, the restrictions will not > affect activities and people outside the jurisdictions of these governments." This analysis overlooks that indirect exports might also be disallowed. That is, if Apple gives software to someone in Argentina, and that person exports it to Cuba, then Apple may well be breaking the law. It seems rather daring for a US entity *not* to insist that recipients conform to US embargos. Incidentally, the USA is not unusual in this area. Here is an overview of the EU's trade sanctions: http://www.eurunion.org/legislat/Sanctions.htm Substitute any EU country for "USA" in these discussions and you run into the same issues with approximately the same list of embargoed countries. One rarely mentioned argument for removing the embargo clause from Squeak-L is that, possibly, the software is not subject to embargo anyway. I did a quick reading of some of the relevant law, and it appeared that completely free and widely-available stuff is not subject to embargo. I don't know whether that applies to Squeak or not, though. I certainly am not a lawyer! About the only lawyers involved in these debates are the guys who wrote Squeak-L, and since they are on our side, I am inclined to presume they did a decent job. Lex

1 0

Text to speech complaint and Hello everyone
by Joshua Scholar 29 Aug '04

29 Aug '04

I notice that the text to speech system is so incomplete that it blows up when handed any text that has punctuation in it. While it would be nice if it knew enough to pause for commas and periods etc. it's amazing that it won't even ignore a character it doesn't know... While I am a programmer, I'm new to Smalltalk, so it may be a while before I'm submitting fixes myself... How does that work anyway? Is Squeak really open source? I am impressed with squeak. My first impression is that Smalltalk is a vastly superior system for building GUIs than everything that came after it. For rapid development, you can't beat a typeless language who's classes can be queried and added to in real time. It looks to me like everyone copied bits of message passing and then copied some OO concepts and missed the importance of having: 1. an interactive language 2. a built in database describing the system 3. a typeless language to make rapid and flexible prototyping possible Joshua Scholar

5 6

← Newer
1
...
5
6
7
8
9
10
11
...
40
Older →

Jump to page: