Squeak-dev December 2009

squeak-dev@lists.squeakfoundation.org

97 participants
553 discussions

Daily Commit Log
by commits＠source.squeak.org 04 Dec '09

04 Dec '09

Changes to Trunk (http://source.squeak.org/trunk.html) in the last 24 hours: http://lists.squeakfoundation.org/pipermail/packages/2009-December/001942.h… Name: ST80-dtl.74 Ancestors: ST80-dtl.73 Move folder selection dialogs from Project to MVCProject and MorphicProject to eliminate MVC/Morphic dependency in Project. ============================================= http://lists.squeakfoundation.org/pipermail/packages/2009-December/001943.h… Name: Morphic-dtl.260 Ancestors: Morphic-dtl.259 Move folder selection dialogs from Project to MVCProject and MorphicProject to eliminate MVC/Morphic dependency in Project. ============================================= http://lists.squeakfoundation.org/pipermail/packages/2009-December/001944.h… Name: System-dtl.189 Ancestors: System-ar.188 Move folder selection dialogs from Project to MVCProject and MorphicProject to eliminate MVC/Morphic dependency in Project. But note that PluggableFileList used for MVC is incorrectly categorized in package Morphic. Change Imports>>viewImages to use "Project current isMorphic" rather than "self couldOpenInMorphic". ============================================= http://lists.squeakfoundation.org/pipermail/packages/2009-December/001945.h… Name: Kernel-dtl.318 Ancestors: Kernel-nice.317 Remove Object>>couldOpenInMorphic and eliminate "world hacking" method category. ============================================= http://lists.squeakfoundation.org/pipermail/packages/2009-December/001946.h… Name: Collections-ul.238 Ancestors: Collections-ul.237 - added #lf and #crlf to WriteStream (like #cr) - updated String >> #withInternetLineEndings ============================================= http://lists.squeakfoundation.org/pipermail/packages/2009-December/001947.h… Name: CollectionsTests-ul.121 Ancestors: CollectionsTests-ar.120 - added tests for WriteStream's new lf and crlf ============================================= http://lists.squeakfoundation.org/pipermail/packages/2009-December/001948.h… Name: Network-ul.41 Ancestors: Network-nice.40 - merged http://bugs.squeak.org/view.php?id=7291 ============================================= http://lists.squeakfoundation.org/pipermail/packages/2009-December/001949.h… Name: KernelTests-ul.115 Ancestors: KernelTests-nice.114 - tests for Behavior >> #allSelectorsBelow: and Behavior >> #allSelectors ============================================= http://lists.squeakfoundation.org/pipermail/packages/2009-December/001950.h… Name: Kernel-ul.319 Ancestors: Kernel-dtl.318 From: http://bugs.squeak.org/view.php?id=7166 - improved preformance for Behavior >> #allSelectors - added Behavior >> #allSelectorsBelow: ============================================= http://lists.squeakfoundation.org/pipermail/packages/2009-December/001951.h… Name: KernelTests-ul.116 Ancestors: KernelTests-ul.115 - fix BehaviorTest >> #testAllSelectorsBelow ============================================= http://lists.squeakfoundation.org/pipermail/packages/2009-December/001952.h… Name: Compiler-ul.102 Ancestors: Compiler-ul.101 - lifted a missing change from Pharo for ByteArray literals - ScannerTest >> #testLiteralSymbols and a test in DecompilerTestFailuresCollector pass now =============================================

1 0

UndefinedObject(Object)>>doesNotUnderstand: #asSymbol trying to load Shout
by Germán Arduino 03 Dec '09

03 Dec '09

Hi: Trying to load Shout (and others dev tools) on an updated trunk image I get: UndefinedObject(Object)>>doesNotUnderstand: #asSymbol Cheers.

4 13

Faster FileStream experiments
by Nicolas Cellier 03 Dec '09

03 Dec '09

I just gave a try to the BufferedFileStream. As usual, code is MIT. Implementation is rough, readOnly, partial (no support for basicNext crap & al), untested (certainly has bugs). Early timing experiments have shown a 5x to 7x speed up on [stream nextLine] and [stream next] micro benchmarks See class comment of attachment Reminder: This bench is versus StandardFileStream. StandardFileStream is the "fast" version, CrLf anf MultiByte are far worse! This still let some more room... Integrating and testing a read/write version is a lot harder than this experiment, but we should really do it. Nicolas

11 42

progress
by Colin Putney 03 Dec '09

03 Dec '09

Hi all, Just now I had occasion to use #collect:into:. I realized what kind of method I needed, remembered some discussion on the list about it, realized I'm using a fairly recent trunk image, and tried it out. It worked! Immensely satisfying. It's so good to be part of a living project again, and see Squeak growing and improving. Thanks everyone. Colin

2 1

Re: [squeak-dev] Embracing Projects
by Igor Stasenko 03 Dec '09

03 Dec '09

2009/12/3 Dan Ingalls <Dan(a)squeakland.org>: > FWIW - > > I, too, am very happy to hear of projects being revisited. It always seemed to me that a Project was *just* a sub-Squeak. It's so with regard to Display area, it's so with regard to changes and (when the planets are lined up right) it can even be so with regard to an ImageSegment as part of the image. > > As Tansel points out, global variables and similar referential issues were the bane of the image segments. I (and Ted K) worked hard to deal with this problem back in the day, and I tried redefining Smalltalk as a Project-local Dictionary ("Environment") linked to the dictionary of its parent project. I don't know if this kind of inheritance structure is best in general, but at one time the reorganizeEverything method actually restructured the whole system this way, introduced a name-space approach to cross-module references, and kept running afterwards. > > The motivation (which began with the need to support Alan's mondo 200M images with 100 projects) was to keep the "kernel" image small, and to load and unload projects as swapping units. But the same unit served well to divest the image of big categories like VM construction, EToys, etc (though this was just two levels, not deep). The one thing I did differently from most of the other namespace proposals was to make global references be (capitalized) messages to the module dictionaries. This had the nice property that, if you worked on a new Stream package called MyStreams, then simply executing, eg, > OriginalStreams := KernelStreams. > KernelStreams := MyStreams. > Would cause all code to start using the new streams, and it could all be reverted atomically with > KernelStreams := OriginalStreams > One could do this on a project-local basis too, allowing for multiple coexisting versions. > [There are other reasons for having global references be messages] > > Henrik Gedenryd picked up this work and carried it quite a bit farther into the system and sadly things got a bit too incompatible. It was a big frustration for me (and for him++) that we never managed to pull this off to completion. > > I wouldn't want to interfere with Andreas's nice suggestion. But I do think there's much more possible if anyone gets interested. Don't know if Spoon uses the synergy of projects as modules. > My very first impression, when i started discovering Squeak & Projects was exactly as you describing: - a separate, isolated object & name space, which can be easily plugged in/out from system. I expected that any changes i doing in one project can't be seen by others. Sure , later i found that its not quite as it supposed to be. To my perception, a Project should be a wrapper or something built on top of 'module' in modularized system. What i mean, that modules should be more basic building brick & should serve towards automation (loading/unloading/interfacing with other parts of system), but not UI/Morphic at first place as Project(s) do. > - Dan > ----------------------------- > >>This is great news! >> >>I have always liked projects but my earlier attempts of trying to separate them from images didn't go smooth and I had difficulty moving projects between images. (Anyone remember trying to save projects from within SqueakNews issues?) In hindsight most of them were due to lousy global variables or some complex relationships but even a smarter extraction process that can inform potential problems will go a long way. Mobilization of projects a la Squeakland but on a grander scale will greatly enhance the ability to make "the experience" of being in an object world much more attractive. >> >>Tansel >> >> >>On 26/11/2009, at 3:25 PM, Andreas Raab wrote: >> >>>Folks - >>> >>>I always felt that one of the least used "big ideas" in Squeak is that of projects. Projects are incredibly powerful but have often been severely misunderstood. A Project in Squeak can define an entire environment (MVC, Morphic, Tweak), it can be a modification to an existing environment, it can identify a set of tools to be used in some context and more. >>> >>>The progress that we've made with refactoring of class Project makes it feasible to illustrate the power of projects. If you install the attached change set in an updated trunk image, you will find a new type of project available - a GamesProject. When you choose this type of project you will notice that it downloads and installs the MorphicGames package that Edgar just announced (thanks Edgar!), sets up the environment with a set of games before launching into it. (note that I didn't go as far as setting up the menu bar with games, removing the tools, sealing the programmer facilities etc. but all of this is possible and should be considered for the various types of projects) >>> >>>And of course, when we ship the next image, we could actually include a number of such "proto projects" like the GamesProject - it takes a tiny amount of space but makes it very accessible and very visible from an user point of view. As a conseqence, we can hopefully ship the next Squeak version with (proto-)Games, Etoys, VMMaker, Seaside and whatnot projects. >>> >>>Thoughts? >>> >>>Cheers, >>> - Andreas >>> >>><GamesProject.st> > > > -- Best regards, Igor Stasenko AKA sig.

1 0

The Trunk: Compiler-ul.102.mcz
by commits＠source.squeak.org 03 Dec '09

03 Dec '09

Levente Uzonyi uploaded a new version of Compiler to project The Trunk: http://source.squeak.org/trunk/Compiler-ul.102.mcz ==================== Summary ==================== Name: Compiler-ul.102 Author: ul Time: 3 December 2009, 6:04:35 am UUID: b4de7fcc-4713-aa42-ac62-f7f1de2b6c16 Ancestors: Compiler-ul.101 - lifted a missing change from Pharo for ByteArray literals - ScannerTest >> #testLiteralSymbols and a test in DecompilerTestFailuresCollector pass now =============== Diff against Compiler-nice.100 =============== Item was changed: ----- Method: Scanner class>>isLiteralSymbol: (in category 'testing') ----- isLiteralSymbol: aSymbol "Test whether a symbol can be stored as # followed by its characters. Symbols created internally with asSymbol may not have this property, e.g. '3' asSymbol." + + | i ascii type next last | - | i ascii type next | i := aSymbol size. i = 0 ifTrue: [^ false]. + - i = 1 ifTrue: [^('$''"()#0123456789_' includes: (aSymbol at: 1)) not - and: ["should be (TypeTable at: (aSymbol at: 1) charCode ifAbsent: [#xLetter]) ~~ #xDelimiter" - (aSymbol at: 1) isSeparator not]]. - ascii := (aSymbol at: 1) charCode. "TypeTable should have been origined at 0 rather than 1 ..." + ascii := (aSymbol at: 1) asciiValue. + type := TypeTable at: ascii ifAbsent: [^false]. + type == #xLetter ifTrue: [ + next := last := nil. + [i > 1] + whileTrue: + [ascii := (aSymbol at: i) asciiValue. + type := TypeTable at: ascii ifAbsent: [^false]. + (type == #xLetter or: [type == #xDigit or: [type == #xColon + and: [ + next == nil + ifTrue: [last := #xColon. true] + ifFalse: [last == #xColon and: [next ~~ #xDigit and: [next ~~ #xColon]]]]]]) + ifFalse: [^ false]. + next := type. + i := i - 1]. + ^ true]. + type == #xBinary ifTrue: [^i = 1]. "Here we could extend to + ^(2 to: i) allSatisfy: [:j | + ascii := (aSymbol at: j) asciiValue. + (TypeTable at: ascii ifAbsent: []) == #xBinary]" + type == #verticalBar ifTrue: [^i = 1]. - ascii = 0 ifTrue: [^ false]. - type := TypeTable at: ascii ifAbsent: [#xLetter]. - (type == #xColon or: [type == #verticalBar or: [type == #xBinary]]) ifTrue: - [^i = 1]. - type == #xLetter ifTrue: - [next := nil. - [i > 1] whileTrue: - [ascii := (aSymbol at: i) charCode. - ascii = 0 ifTrue: [^false]. - type := TypeTable at: ascii ifAbsent: [#xLetter]. - (type == #xLetter or: [type == #xDigit or: [type == #xColon]]) ifFalse: - [^false]. - (next == #xDigit and: [type == #xColon]) ifTrue: [^false]. - next := type. - i := i - 1]. - ^true]. ^false!

1 0

The Trunk: KernelTests-ul.116.mcz
by commits＠source.squeak.org 03 Dec '09

03 Dec '09

Levente Uzonyi uploaded a new version of KernelTests to project The Trunk: http://source.squeak.org/trunk/KernelTests-ul.116.mcz ==================== Summary ==================== Name: KernelTests-ul.116 Author: ul Time: 3 December 2009, 5:46:36 am UUID: a4eed34e-28b7-c14b-b2c5-0323756ef7c4 Ancestors: KernelTests-ul.115 - fix BehaviorTest >> #testAllSelectorsBelow =============== Diff against KernelTests-ul.115 =============== Item was changed: ----- Method: BehaviorTest>>testAllSelectorsBelow (in category 'tests') ----- testAllSelectorsBelow self assert: (Object allSelectorsBelow: ProtoObject) = Object selectors asIdentitySet. + self assert: (Object allSelectorsBelow: nil) = (Object selectors union: ProtoObject selectors) asIdentitySet! - self assert: (Object allSelectorsBelow: nil) = Object selectors asIdentitySet!

1 0

The Trunk: Kernel-ul.319.mcz
by commits＠source.squeak.org 03 Dec '09

03 Dec '09

Levente Uzonyi uploaded a new version of Kernel to project The Trunk: http://source.squeak.org/trunk/Kernel-ul.319.mcz ==================== Summary ==================== Name: Kernel-ul.319 Author: ul Time: 3 December 2009, 5:12:50 am UUID: b8275b69-ee08-9f45-b8c2-7e020776e941 Ancestors: Kernel-dtl.318 From: http://bugs.squeak.org/view.php?id=7166 - improved preformance for Behavior >> #allSelectors - added Behavior >> #allSelectorsBelow: =============== Diff against Kernel-dtl.318 =============== Item was added: + ----- Method: Behavior>>allSelectorsBelow: (in category 'accessing method dictionary') ----- + allSelectorsBelow: topClass + | coll | + coll := IdentitySet new. + self withAllSuperclassesDo: + [:aClass | + aClass = topClass + ifTrue: [^ coll ] + ifFalse: [aClass selectorsDo: [ :sel | coll add: sel ]]]. + ^ coll + + + ! Item was changed: ----- Method: Behavior>>allSelectors (in category 'accessing method dictionary') ----- allSelectors + "Answer all selectors understood by instances of the receiver" - "Answer all selectors understood by instances of the receiver" + ^(Array streamContents: [ :stream | + self withAllSuperclassesDo: [ :aClass | + aClass selectorsDo: [ :each | + stream nextPut: each ] ] ]) asIdentitySet! - | coll | - coll := OrderedCollection new. - self withAllSuperclasses do: - [:aClass | coll addAll: aClass selectors]. - ^ coll asIdentitySet!

1 0

The Trunk: KernelTests-ul.115.mcz
by commits＠source.squeak.org 03 Dec '09

03 Dec '09

Levente Uzonyi uploaded a new version of KernelTests to project The Trunk: http://source.squeak.org/trunk/KernelTests-ul.115.mcz ==================== Summary ==================== Name: KernelTests-ul.115 Author: ul Time: 3 December 2009, 5:11:22 am UUID: b7d613d4-6eba-7a43-be09-23f76c5856e0 Ancestors: KernelTests-nice.114 - tests for Behavior >> #allSelectorsBelow: and Behavior >> #allSelectors =============== Diff against KernelTests-nice.114 =============== Item was added: + ----- Method: BehaviorTest>>testAllSelectors (in category 'tests') ----- + testAllSelectors + + self assert: ProtoObject allSelectors = ProtoObject selectors asIdentitySet. + self assert: Object allSelectors = (Object selectors union: ProtoObject selectors) asIdentitySet.! Item was added: + ----- Method: BehaviorTest>>testAllSelectorsBelow (in category 'tests') ----- + testAllSelectorsBelow + + self assert: (Object allSelectorsBelow: ProtoObject) = Object selectors asIdentitySet. + self assert: (Object allSelectorsBelow: nil) = Object selectors asIdentitySet!

1 0

PluggableFileList in the wrong package? (was: [squeak-dev] The Trunk: System-dtl.189.mcz)
by David T. Lewis 03 Dec '09

03 Dec '09

On Thu, Dec 03, 2009 at 02:22:01AM +0000, commits(a)source.squeak.org wrote: > > Name: System-dtl.189 > Author: dtl > Time: 2 December 2009, 9:17:54 am > UUID: 3785764e-2a0f-4582-b416-f2903c553df7 > Ancestors: System-ar.188 > > Move folder selection dialogs from Project to MVCProject and MorphicProject > to eliminate MVC/Morphic dependency in Project. But note that PluggableFileList > used for MVC is incorrectly categorized in package Morphic. While reorganizing some Project methods into the proper MVC and Morphic subclasses, I notice that PluggableFileList is used for opening a file dialog in MVC, so it now it is referenced by MVCProject. PluggableFileList works in both MVC and Morphic, but is categorized in package Morphic. Am I right in thinking that PluggableFileList (and probably some other classes) needs to move out of the Morphic package? If so, where should it go (perhaps under System-* ?). Dave

1 0

← Newer
1
...
48
49
50
51
52
53
54
55
56
Older →

Jump to page: