- Pharo-dev - lists.pharo.org

Re: [Pharo-project] [COTDC] 85 - HashedCollection

by StÃ©phane Ducasse

thanks this is a good one! Stef On Nov 28, 2011, at 5:33 PM, Nicolas Cellier wrote: > The Squeak trunk version is: > > I am an abstract collection of objects that implement hash and > equality in a consitent way. This means that whenever two objects are > equal, their hashes have to be equal too. If two objects are equal > then I can only store one of them. Hashes are expected to be integers > (preferably SmallIntegers). I also expect that the objects contained > by me do not change their hashes. If that happens, hash invariants > have to be re-established, which can be done by #rehash. > > Since I'm abstract, no instances of me should exist. My subclasses > should implement #scanFor:, #fixCollisionsFrom: and > #noCheckNoGrowFillFrom:. > > Instance Variables > array: <ArrayedCollection> (typically Array or WeakArray) > tally: <Integer> (non-negative) > > array > - An array whose size is a prime number, it's non-nil elements are > the elements of the collection, and whose nil elements are empty > slots. There is always at least one nil. In fact I try to keep my > "load" at 75% or less so that hashing will work well. > > tally > - The number of elements in the collection. The array size is always > greater than this. > > Implementation details: > I implement a hash table which uses open addressing with linear > probing as the method of collision resolution. Searching for an > element or a free slot for an element is done by #scanFor: which > should return the index of the slot in array corresponding to it's > argument. When an element is removed #fixCollisionsFrom: should rehash > all elements in array between the original index of the removed > element, wrapping around after the last slot until reaching an empty > slot. My maximum load factor (75%) is hardcoded in #atNewIndex:put:, > so it can only be changed by overriding that method. When my load > factor reaches this limit I replace my array with a larger one (see > #grow) ensuring that my load factor will be less than or equal to 50%. > The new array is filled by #noCheckNoGrowFillFrom: which should use > #scanForEmptySlotFor: instead of #scanFor: for better performance. I > do not shrink. > > > 2011/11/28 StÃ©phane Ducasse <stephane.ducasse(a)inria.fr>: >> I'm an abstract class supporting dictionary based data structure. I provide access to my elements based on their hash. >> >> Stef >> >> On Nov 28, 2011, at 4:29 PM, Sean P. DeNigris wrote: >> >>> Today: HashedCollection >>> >>> >>> Comment Of The Day Contest - One Day One Comment >>> Rules: >>> #1: Each day a not commented class is elected. Each day the best comment >>> will be integrated with name of the author(s). >>> #2: If you cannot comment it, deprecate it. >>> Results: http://code.google.com/p/pharo/wiki/CommentOfTheDayContest >>> >>> -- >>> View this message in context: http://forum.world.st/COTDC-85-HashedCollection-tp4115592p4115592.html >>> Sent from the Pharo Smalltalk mailing list archive at Nabble.com. >>> >> >> >> >

Nov. 28, 2011

Re: [Pharo-project] Morphic replacement

by Douglas Brebner

On 27/11/2011 21:08, Fernando Olivero wrote: > When the vm, starts announcing multitouch events, it's a matter of > adding that particular peripheral device to the model. For now, the vm > only announces mouse and keyboard events, therefore i only modeled > those. > > Thanks for the question, maybe we should ask the vm guys to add this > to their todo list. > > I believe Bert Freudenberg made an iPad VM with multitouch support, though I don't know its current status. Good luck with your work, we could use a better Morphic :)

Nov. 28, 2011

Re: [Pharo-project] Problem with NonInteractiveUIManager

by Norbert Hartl

Am 28.11.2011 um 18:53 schrieb Igor Stasenko: > On 28 November 2011 17:44, Norbert Hartl <norbert(a)hartl.name> wrote: >> Ok, >> >> this is the behaviour with an image #13320 and the newest cog vm. I don't need to do anything because the NonInteractiveUIManager is set to default automatically regardless if -nodisplay or -headless is specified. And in my case this is broken. However if I hack away the noninteractive stuff by doing >> >> UIManager default: (UIManager default instVarAt: 1). >> >> than it works. A debugger pops up in the image and can see the real exception with VNC. So to me it seems that there is something broken. Igor, can you add anything. I think Marcus is not available at the moment. >> > > a non-interactive manager takes ownership when you either: > - run image with -headless param > - during image startup (StartupUIManager). > > the default behavior of it, when any interaction with user is > requested, it signals ErrorNonInteractive error, > and default action for this error is to print stack trace to log file > and leave to OS. > signaling any other unhandled error leading to opening a debugger > window, which is considered a user interaction request, and > consequently bails out to OS as well by signaling ErrorNonInteractive. > > if you run headless and don't want to use this manager, > you can disable it, just by setting a morphic ui manager: > > UIManager default: MorphicUIManager new. > Yes, that's the same as what I did with restoring the original manager. Using the non-interactive manager does not output the stack trace of the exception that occured but the exception that is triggered because openDebuggerOn:context:label:contents:fullView: is not implemented in NonInteractiveUIManager. And even if I impement the method with ^ self nonInteractive I don't get the exception that originally ocurred. So you cannot see the exception that caused the image to exit. Shouldn't that be different? Norbert

Nov. 28, 2011

Re: [Pharo-project] Problem with NonInteractiveUIManager

by Igor Stasenko

On 28 November 2011 17:44, Norbert Hartl <norbert(a)hartl.name> wrote: > Ok, > > this is the behaviour with an image #13320 and the newest cog vm. I don't need to do anything because the NonInteractiveUIManager is set to default automatically regardless if -nodisplay or -headless is specified. And in my case this is broken. However if I hack away the noninteractive stuff by doing > > UIManager default: (UIManager default instVarAt: 1). > > than it works. A debugger pops up in the image and can see the real exception with VNC. So to me it seems that there is something broken. Igor, can you add anything. I think Marcus is not available at the moment. > a non-interactive manager takes ownership when you either: - run image with -headless param - during image startup (StartupUIManager). the default behavior of it, when any interaction with user is requested, it signals ErrorNonInteractive error, and default action for this error is to print stack trace to log file and leave to OS. signaling any other unhandled error leading to opening a debugger window, which is considered a user interaction request, and consequently bails out to OS as well by signaling ErrorNonInteractive. if you run headless and don't want to use this manager, you can disable it, just by setting a morphic ui manager: UIManager default: MorphicUIManager new. > Norbert > -- Best regards, Igor Stasenko.

Nov. 28, 2011

Re: [Pharo-project] Athens work in progress: couple screenshots

by Igor Stasenko

On 28 November 2011 16:04, Fernando Olivero <fernando.olivero(a)usi.ch> wrote: > Don't know, but in case we cant modify Athens, we can always rename > the project and start fresh, given that Igor rewrote most of it > right?. > > Fernando > > On Fri, Nov 25, 2011 at 12:58 PM, StÃ©phane Ducasse > <stephane.ducasse(a)inria.fr> wrote: >> >> On Nov 25, 2011, at 10:35 AM, Fernando Olivero wrote: >> >>> Ok, i will try to do so from now on! >>> >>> Igor, thanks for detailed response, i see now there's no real need for >>> caching the glyphs, even though i didn't consider rotation as an >>> available affine transformation. >>> >>> Just a question on the protocol of the canvas, i found it better to >>> use the following convenience methods, instead of exposing the >>> transformation (or canvas state) directly to the sender: >>> >>> drawOn: aCanvas >>> Â aCanvas pathTransform â¦ >>> >>> drawOn: aCanvas >>> Â Â aCanvas scaleBy: 0.5 @0.5 ; translateBy: 10@10 . >>> Â Â â¦.. the canvas #pathTransform actually should answer a delegate, which should allow you to manipulate coordinate system transformation. There's no intent to expose the state directly. You can modify transformation matrix only via corresponding protocol, but don't manipulate its state directly. now about canvas scaleBy: ... versus canvas pathTransform scaleBy:... why i did it like that? because there are multiple transformation matrices responsible for different parts of computations. for instance, in OpenVG model there are four: - path - image - stroke - (forgot) :) or take an openGL for example there are 3 (or 4) matrices again: - model - projection - texture - color space if you expose all of them via canvas, then you have to bloat canvas protocol with duplicate messages like: canvas pathScale: canvas pathRotate: canvas imageScale: canvas imageRotate: ... etc that's why there are separate delegates, which represent concrete matrix, and common protocol which you can use for all of them: canvas pathTransform scaleBy: canvas imageTransform scaleBy: etc -- Best regards, Igor Stasenko.

Nov. 28, 2011

Re: [Pharo-project] Building 32-bit Cairo library on Macs

by Igor Stasenko

On 28 November 2011 16:26, Fernando Olivero <fernando.olivero(a)usi.ch> wrote: > Igor, i've downloaded the latest native boots vm, loaded the NB and > NBCairo, only missing Cairo now. > Could you provide the library you built for the Mac. if you just copy cairo library it will fail to load because of missing dependencies. as i told, there are multiple dependencies, so it is better to install it instead of copying. you shall install it from ports or homebrew. just make sure you build either universal binaries or 32-bit, because 64 bit ones won't work. > Thanks, > Fernando > On Sun, Nov 27, 2011 at 10:12 PM, fernando.olivero(a)usi.ch > <fernando.olivero(a)usi.ch> wrote: >> nice! thanks a lot for dealing with the ugliness outside the smalltalk >> world, so we mortals can deal with higher level languages. >> >> Looking forward to see an Athens binding to Cairo. >> Fernando >> >> On Fri, Nov 25, 2011 at 7:33 PM, StÃ©phane Ducasse >> <stephane.ducasse(a)inria.fr> wrote: >>> cool >>> We want more :) >>> >>> Stef >>> >>> On Nov 25, 2011, at 5:17 PM, Igor Stasenko wrote: >>> >>>> Just to let you know: >>>> >>>> it works on Mac! >>>> I just loaded the code which Javier did for linux, changed path to >>>> library and it worked out of the box. >>>> all examples in [1] are working out of the box! >>>> >>>> So, now i can start wiring Cairo backend for Athens :) >>>> >>>> [1] http://squeaksource.com/NBCairo/ >>>> >>>> -- >>>> Best regards, >>>> Igor Stasenko. >>>> >>> >>> >>> >> >> > > -- Best regards, Igor Stasenko.

Nov. 28, 2011

[Pharo-project] Class Comments from Squeak (was Re: [COTDC] 85 - HashedCollection)

by DeNigris Sean

<quote author="Nicolas Cellier"> The Squeak trunk version is: </quote> Great idea - low hanging fruit: Issue 5034 in pharo: Harvest class comments from Squeak http://code.google.com/p/pharo/issues/detail?id=5034

Nov. 28, 2011

Re: [Pharo-project] Problem with NonInteractiveUIManager

by Norbert Hartl

Ok, this is the behaviour with an image #13320 and the newest cog vm. I don't need to do anything because the NonInteractiveUIManager is set to default automatically regardless if -nodisplay or -headless is specified. And in my case this is broken. However if I hack away the noninteractive stuff by doing UIManager default: (UIManager default instVarAt: 1). than it works. A debugger pops up in the image and can see the real exception with VNC. So to me it seems that there is something broken. Igor, can you add anything. I think Marcus is not available at the moment. Norbert Am 28.11.2011 um 15:42 schrieb Norbert Hartl: > I'm trying to do some Katas in deploying pharo images on my server. After finding my image crashing all the time I tried to research what I need to take care of. I came to the conclusion that > > UIManager default: UIManager default nonInteractiveManager. > NonInteractiveUIManager snapshotErrorImage: true. > > would be the right thing to do in a server environment. But I only get the stack trace attached. So I'm asking myself if I need to do something different in order to get the stack trace of my own application in case something goes wrong. > > Norbert > > ---- > > 28 November 2011 3:35:54 pm > > VM: Mac OS - intel - 1072 - Croquet Closure Cog VM [CoInterpreter VMMaker-oscog.52] Cog VM 4.0.0 > Image: Pharo1.3 [Latest update: #13320] > > NonInteractiveUIManager(Object)>>doesNotUnderstand: #openDebuggerOn:context:label:contents:fullView: > Receiver: a NonInteractiveUIManager > Arguments and temporary variables: > aMessage: openDebuggerOn: a Process in GRPharoPlatform>>openDebuggerOn: context...etc... > exception: MessageNotUnderstood: NonInteractiveUIManager>>openDebuggerOn:contex...etc... > resumeValue: nil > Receiver's instance variables: > uiManager: a MorphicUIManager > doNotQuitOnRestart: false > > Debugger class>>openOn:context:label:contents:fullView: > Receiver: Debugger > Arguments and temporary variables: > process: a Process in GRPharoPlatform>>openDebuggerOn: > context: OrderedCollection(Object)>>error: > title: 'Error: startIndex is out of bounds' > contentsStringOrNil: nil > bool: true > fullView: true > Receiver's instance variables: > superclass: CodeHolder > methodDict: a MethodDictionary(#addOptionalButtonsTo:at:plus:->(Debugger>>#addO...etc... > format: 166 > instanceVariables: #('interruptedProcess' 'interruptedController' 'contextStack...etc... > organization: ('*Ocompletion' guessTypeForName:) > ('*Polymorph-Widgets' addOptio...etc... > subclasses: nil > name: #Debugger > classPool: a Dictionary(#AlwaysOpenFullDebugger->false #ContextStackKeystrokes-...etc... > sharedPools: nil > environment: a SystemDictionary(lots of globals) > category: #'Tools-Debugger' > traitComposition: nil > localSelectors: nil > > [:tool | ^ tool > openOn: aProcess > context: aContext > label: aString > contents: contents > fullView: aBool] in ToolRegistry>>debug:context:label:contents:fullView: > Receiver: a ToolRegistry > Arguments and temporary variables: > aProcess: Debugger > aContext: a Process in GRPharoPlatform>>openDebuggerOn: > aString: OrderedCollection(Object)>>error: > contents: 'Error: startIndex is out of bounds' > aBool: nil > tool: true > Receiver's instance variables: > tools: an IdentityDictionary(#basicInspector->BasicInspector #browser->[self de...etc... > inspectorsMap: an IdentityDictionary(#CompiledMethod->#CompiledMethodInspector ...etc... > > ToolRegistry>>using:do: > Receiver: a ToolRegistry > Arguments and temporary variables: > aToolName: #debugger > aBlock: [:tool | ^ tool > openOn: aProcess > context: aContext > label: aString...etc... > tool: Debugger > Receiver's instance variables: > tools: an IdentityDictionary(#basicInspector->BasicInspector #browser->[self de...etc... > inspectorsMap: an IdentityDictionary(#CompiledMethod->#CompiledMethodInspector ...etc... > > ToolRegistry>>debug:context:label:contents:fullView: > Receiver: a ToolRegistry > Arguments and temporary variables: > aProcess: a Process in GRPharoPlatform>>openDebuggerOn: > aContext: OrderedCollection(Object)>>error: > aString: 'Error: startIndex is out of bounds' > contents: nil > aBool: true > Receiver's instance variables: > tools: an IdentityDictionary(#basicInspector->BasicInspector #browser->[self de...etc... > inspectorsMap: an IdentityDictionary(#CompiledMethod->#CompiledMethodInspector ...etc... > > Process>>debug:title:full: > Receiver: a Process in GRPharoPlatform>>openDebuggerOn: > Arguments and temporary variables: > context: OrderedCollection(Object)>>error: > title: 'Error: startIndex is out of bounds' > bool: true > topCtxt: GRPharoPlatform>>openDebuggerOn: > Receiver's instance variables: > nextLink: nil > suspendedContext: GRPharoPlatform>>openDebuggerOn: > priority: 60 > myList: nil > threadId: nil > errorHandler: nil > name: 'ZnMultiThreadedServer HTTP worker' > env: nil > > [process > debug: anError signalerContext > title: anError description > full: true] in GRPharoPlatform>>openDebuggerOn: > Receiver: a GRPharoPlatform > Arguments and temporary variables: > anError: Error: startIndex is out of bounds > process: a Process in GRPharoPlatform>>openDebuggerOn: > Receiver's instance variables: > a GRPharoPlatform > > WorldState>>runStepMethodsIn: > Receiver: a WorldState > Arguments and temporary variables: > aWorld: a PasteUpMorph(425197568) [world] > queue: SharedQueue with 0 items > nextInQueue: [process > debug: anError signalerContext > title: anError descrip...etc... > Receiver's instance variables: > hands: an Array(a HandMorph(843055104)) > viewBox: 0@0 corner: 1296@875 > canvas: a FormCanvas on: RFBDisplayScreen(1296x875x32) > damageRecorder: a DamageRecorder > stepList: a Heap(StepMessage(#stepAt: -> an OverflowRowMorph(210239488))(an Ove...etc... > lastStepTime: 19250 > lastStepMessage: nil > lastCycleTime: 19271 > alarms: a Heap() > lastAlarmTime: 19250 > menuBuilder: nil > > PasteUpMorph>>runStepMethods > Receiver: a PasteUpMorph(425197568) [world] > Arguments and temporary variables: > > Receiver's instance variables: > bounds: 0@0 corner: 1296@875 > owner: nil > submorphs: an Array(a TaskbarMorph(984088576) a SystemWindow(347865088) a Syste...etc... > fullBounds: nil > color: (Color r: 0.97 g: 0.98 b: 1.0) > extension: a MorphExtension (666632192) [eventHandler = an EventHandler] [othe...etc... > borderWidth: 0 > borderColor: (Color r: 0.03 g: 0.02 b: 0.0) > backgroundMorph: nil > worldState: a WorldState > griddingOn: nil > > WorldState>>doOneCycleNowFor: > Receiver: a WorldState > Arguments and temporary variables: > aWorld: a PasteUpMorph(425197568) [world] > Receiver's instance variables: > hands: an Array(a HandMorph(843055104)) > viewBox: 0@0 corner: 1296@875 > canvas: a FormCanvas on: RFBDisplayScreen(1296x875x32) > damageRecorder: a DamageRecorder > stepList: a Heap(StepMessage(#stepAt: -> an OverflowRowMorph(210239488))(an Ove...etc... > lastStepTime: 19250 > lastStepMessage: nil > lastCycleTime: 19271 > alarms: a Heap() > lastAlarmTime: 19250 > menuBuilder: nil > > WorldState>>doOneCycleFor: > Receiver: a WorldState > Arguments and temporary variables: > aWorld: a PasteUpMorph(425197568) [world] > Receiver's instance variables: > hands: an Array(a HandMorph(843055104)) > viewBox: 0@0 corner: 1296@875 > canvas: a FormCanvas on: RFBDisplayScreen(1296x875x32) > damageRecorder: a DamageRecorder > stepList: a Heap(StepMessage(#stepAt: -> an OverflowRowMorph(210239488))(an Ove...etc... > lastStepTime: 19250 > lastStepMessage: nil > lastCycleTime: 19271 > alarms: a Heap() > lastAlarmTime: 19250 > menuBuilder: nil > > PasteUpMorph>>doOneCycle > Receiver: a PasteUpMorph(425197568) [world] > Arguments and temporary variables: > > Receiver's instance variables: > bounds: 0@0 corner: 1296@875 > owner: nil > submorphs: an Array(a TaskbarMorph(984088576) a SystemWindow(347865088) a Syste...etc... > fullBounds: nil > color: (Color r: 0.97 g: 0.98 b: 1.0) > extension: a MorphExtension (666632192) [eventHandler = an EventHandler] [othe...etc... > borderWidth: 0 > borderColor: (Color r: 0.03 g: 0.02 b: 0.0) > backgroundMorph: nil > worldState: a WorldState > griddingOn: nil > > [[World doOneCycle. > Processor yield. > false] whileFalse. > nil] in Project class>>spawnNewProcess > Receiver: Project > Arguments and temporary variables: > > Receiver's instance variables: > superclass: Model > methodDict: a MethodDictionary(#beep->(Project>>#beep "a CompiledMethod(3169320...etc... > format: 132 > instanceVariables: nil > organization: ('*sound' beep) > > subclasses: nil > name: #Project > classPool: a Dictionary(#UIProcess->a Process in nil ) > sharedPools: nil > environment: a SystemDictionary(lots of globals) > category: #'System-Support' > traitComposition: {} > localSelectors: nil > > [self value. > Processor terminateActive] in BlockClosure>>newProcess > Receiver: [[World doOneCycle. > Processor yield. > false] whileFalse. > nil] > Arguments and temporary variables: > > Receiver's instance variables: > outerContext: Project class>>spawnNewProcess > startpc: 57 > numArgs: 0 > > > --- The full stack --- > NonInteractiveUIManager(Object)>>doesNotUnderstand: #openDebuggerOn:context:label:contents:fullView: > Debugger class>>openOn:context:label:contents:fullView: > [:tool | ^ tool > openOn: aProcess > context: aContext > label: aString > contents: contents > fullView: aBool] in ToolRegistry>>debug:context:label:contents:fullView: > ToolRegistry>>using:do: > ToolRegistry>>debug:context:label:contents:fullView: > Process>>debug:title:full: > [process > debug: anError signalerContext > title: anError description > full: true] in GRPharoPlatform>>openDebuggerOn: > WorldState>>runStepMethodsIn: > PasteUpMorph>>runStepMethods > WorldState>>doOneCycleNowFor: > WorldState>>doOneCycleFor: > PasteUpMorph>>doOneCycle > [[World doOneCycle. > Processor yield. > false] whileFalse. > nil] in Project class>>spawnNewProcess > [self value. > Processor terminateActive] in BlockClosure>>newProcess >

Nov. 28, 2011

Re: [Pharo-project] What happened to COTDC?

by Sean P. DeNigris

Sean P. DeNigris wrote > > We can't afford to take a break! > I didn't realize the script had been updated to exclude TestCases... (SystemNavigation default allClassesAndTraits reject: [:aClass| aClass hasComment or: [aClass inheritsFrom:TestCase]] ) size "782" -- View this message in context: http://forum.world.st/What-happened-to-COTDC-tp4110480p4115829.html Sent from the Pharo Smalltalk mailing list archive at Nabble.com.

Nov. 28, 2011

Re: [Pharo-project] [COTDC] 85 - HashedCollection

by Nicolas Cellier

The Squeak trunk version is: I am an abstract collection of objects that implement hash and equality in a consitent way. This means that whenever two objects are equal, their hashes have to be equal too. If two objects are equal then I can only store one of them. Hashes are expected to be integers (preferably SmallIntegers). I also expect that the objects contained by me do not change their hashes. If that happens, hash invariants have to be re-established, which can be done by #rehash. Since I'm abstract, no instances of me should exist. My subclasses should implement #scanFor:, #fixCollisionsFrom: and #noCheckNoGrowFillFrom:. Instance Variables array: <ArrayedCollection> (typically Array or WeakArray) tally: <Integer> (non-negative) array - An array whose size is a prime number, it's non-nil elements are the elements of the collection, and whose nil elements are empty slots. There is always at least one nil. In fact I try to keep my "load" at 75% or less so that hashing will work well. tally - The number of elements in the collection. The array size is always greater than this. Implementation details: I implement a hash table which uses open addressing with linear probing as the method of collision resolution. Searching for an element or a free slot for an element is done by #scanFor: which should return the index of the slot in array corresponding to it's argument. When an element is removed #fixCollisionsFrom: should rehash all elements in array between the original index of the removed element, wrapping around after the last slot until reaching an empty slot. My maximum load factor (75%) is hardcoded in #atNewIndex:put:, so it can only be changed by overriding that method. When my load factor reaches this limit I replace my array with a larger one (see #grow) ensuring that my load factor will be less than or equal to 50%. The new array is filled by #noCheckNoGrowFillFrom: which should use #scanForEmptySlotFor: instead of #scanFor: for better performance. I do not shrink. 2011/11/28 StÃ©phane Ducasse <stephane.ducasse(a)inria.fr>: > I'm an abstract class supporting dictionary based data structure. I provide access to my elements based on their hash. > > Stef > > On Nov 28, 2011, at 4:29 PM, Sean P. DeNigris wrote: > >> Today: Â HashedCollection >> >> >> Comment Of The Day Contest - One Day One Comment >> Rules: >> #1: Each day a not commented class is elected. Each day the best comment >> will be integrated with name of the author(s). >> #2: If you cannot comment it, deprecate it. >> Results: http://code.google.com/p/pharo/wiki/CommentOfTheDayContest >> >> -- >> View this message in context: http://forum.world.st/COTDC-85-HashedCollection-tp4115592p4115592.html >> Sent from the Pharo Smalltalk mailing list archive at Nabble.com. >> > > >

Nov. 28, 2011