November 2011 - Pharo-dev - lists.pharo.org

Re: [Pharo-project] RenameMethodRefactoring Comment

by Sean P. DeNigris

Sean P. DeNigris wrote > > permutation: (0 to: oldSelector numArgs). > should be: permutation: (1 to: oldSelector numArgs). -- View this message in context: http://forum.world.st/RenameMethodRefactoring-Comment-tp4117939p4117958.html Sent from the Pharo Smalltalk mailing list archive at Nabble.com.

Nov. 29, 2011

[Pharo-project] RenameMethodRefactoring Comment

by Sean P. DeNigris

Here's a free bonus class comment: I do just what my name says. Example Usage: refactoring := RenameMethodRefactoring renameMethod: oldSelector in: class to: newSelector permutation: (0 to: oldSelector numArgs). refactoring execute. A few things to keep in mind: - even though a class is specified, my action is system-wide. The only thing I do with the class is make sure it defines oldSelector - the permutation should be a collection that specifies what order the arguments from the oldSelector should appear in the newSelector. If the argument order doesn't change, you can just use the interval shortcut shown above in the example. Example when arguement order changes: oldSelector: arg1 with: arg2 with: arg3 newSelector: arg2 with: arg3 with: arg1 permutation := { 2. 3. 1 } -- View this message in context: http://forum.world.st/RenameMethodRefactoring-Comment-tp4117939p4117939.html Sent from the Pharo Smalltalk mailing list archive at Nabble.com.

Nov. 29, 2011

Re: [Pharo-project] Problem with NonInteractiveUIManager

by Igor Stasenko

On 28 November 2011 20:01, Norbert Hartl <norbert(a)hartl.name> wrote: > > 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? > ok, seems like someone extended protocol without implementing analog in non-interactive manager. please file a bug report we will fix that. > Norbert > > > > -- Best regards, Igor Stasenko.

Nov. 29, 2011

Re: [Pharo-project] Problem with NonInteractiveUIManager

by Igor Stasenko

On 28 November 2011 23:36, Sean P. DeNigris <sean(a)clipperadams.com> wrote: > > Igor Stasenko wrote >> >> 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. >> > > Igor, I added your response to the class comment, but could you expand the > above paragraph (maybe some periods, lol). I don't understand it. > here what happens: [ some error signal ] -> attempt to open debugger -> [signal ErrorNonInteractive] -> leave to OS > -- > View this message in context: http://forum.world.st/Problem-with-NonInteractiveUIManager-tp4115412p411698… > Sent from the Pharo Smalltalk mailing list archive at Nabble.com. > -- Best regards, Igor Stasenko.

Nov. 29, 2011

Re: [Pharo-project] Problem with NonInteractiveUIManager

by Sean P. DeNigris

Igor Stasenko wrote > > 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. > Igor, I added your response to the class comment, but could you expand the above paragraph (maybe some periods, lol). I don't understand it. -- View this message in context: http://forum.world.st/Problem-with-NonInteractiveUIManager-tp4115412p411698… Sent from the Pharo Smalltalk mailing list archive at Nabble.com.

Nov. 28, 2011

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