- Pharo-dev - lists.pharo.org

Re: [Pharo-project] Startup Preferences

by Benjamin

Guys, I am ok for any proposition but please try to find only one together :) some of you are for ~/.config/pharo/bla others for ~/.pharo/bla some of you are for multi location, other for only one. Personally I am for ~/.pharo and for multi locations :) Waiting for your arguments ;) Ben On Dec 13, 2011, at 5:57 PM, Camillo Bruni wrote: > Cool, > > For me the default config dir is still in "~/.pharo/" so that would be cool if added... > > cami > > On 2011-12-13, at 11:37, Benjamin wrote: > >> New version: >> >> Now there is a hierarchy for the setting file >> >> - first it lloks in the image diretory for st files >> - if none is found, it looks into >> ~/.config/pharo/${VERSION} >> - if none is found it looks into >> ~/.config/pharo/general >> >> We could also add a look into >> ~/.pharo/ >> >> >> >> Gofer new >> squeaksource: 'StartupPreferences'; >> package: 'StartupPreferences'; >> load. >> >> Ben >> >> On Dec 1, 2011, at 6:50 PM, StÃ©phane Ducasse wrote: >> >>> me too. >>> >>> Stef >>> >>> On Dec 1, 2011, at 6:38 PM, Camillo Bruni wrote: >>> >>>> Yeah I would really love to see this. >>>> >>>> maybe with a fixed lookup chain >>>> >>>> 1. Settings from the Image directory/startup.st >>>> 2. Settings from ~/.pharo/$VERSION/startup.st >>>> 3. Settings from ~/.pharo/startup.st >>>> >>>> cami >>>> >>> >>> >> > >

Dec. 13, 2011

Re: [Pharo-project] IdentitySet but using #hash rather than #identityHash ?

by Eliot Miranda

Hi Mariano, On Tue, Dec 13, 2011 at 12:37 AM, Mariano Martinez Peck < marianopeck(a)gmail.com> wrote: > > > On Tue, Dec 13, 2011 at 8:43 AM, Michael Roberts <mike(a)mjr104.co.uk>wrote: > >> Hi Mariano, when I read this thread I was a bit confused that you wanted >> an IdentitySet that used #hash. From that statement it sounded like you >> just wanted a Set. This would allow any object to define its own hash and >> importantly what equality means with #=. So if you want to delegate that to >> the object use a Set. >> >> > No, I cannot use a Set because I cannot have repetitions. This that I am > asking is what we do in Fuel serializer while traversing the object graph. > Each analyzed object is put in an IdentityDictionary (but the question is > the same for IdentitySet) as key. To avoid cycles, I need to put each > object only once. Since graphs can be very big, such dict could have lots > of objects. > > >> However as the thread has gone on perhaps you want the identity >> relationship but you just wanted a bigger identity hash space? >> > > Yes, exactly. I were thinking if there could be a way (maybe...that's why > I am asking) of improve its performance considering that I could have much > more objects than 2^13. In other words, I wanted to see if I could avoid > colisions. > You can assume that certain properties of objects will not change during serialization, for example the class of objects, the basic size of objects. So you can construct a valid extended identity hash from these properties. For example fuelSerializationHash ^self identityHash + (self class identityHash bitShift: 12) + (self basicSize bitShift: 24) In general this idea may not work because of meta-primitives like changeClassTo:, which would change the hash. But in Fuel's case I think it's safe to assume that objects won't change class or size during serialization. HTH > > >> >> IdentitySets are fast because they bypass any delegation. Once you have >> seen the object in your traversal (common usage pattern) that's it. You >> grab it's identity which is a pretty low level thing to do. >> > > Ok...it is a tradeoff. If I use #identityHash it is fast because there is > no delegation and it is almost an immediate primitive. But I gues it will > be slow if there are lots of colisions. Not using #identityHash but > something else maybe could decrease maybe the amount of colisions, but > maybe with the delegation it will gets slower. > > I will try with Levente idea of what it is done in SystemTracer: use as a > hash the identityHash of the object mixed with the identityHash of its > class. Maybe that decreases the colisions and at the same time I don't pay > delegation (#class is special bytecode, so nothing, and ok..there are 2 > sends to #identityhash but I don't thinnk it is that much). > > Anyway, thanks for the interesting post, I always learn :) > > >> >> As for what happens with collections who knows. Depends. Relying on >> identity set semantics for a collection is easy. Set semantics is not so. >> Remember that both hash and equals are important to know if the set already >> contains the element. Depending on the collection implementation both of >> these could be composite in terms of the parts. Who knows where you end or >> how long it takes. I.e. if it is a function of the size of the collection >> and further collections are composite.... >> >> > yes... > > >> Cheers >> Mike >> >> >> >> On Tuesday, December 13, 2011, Carlo <snoobabk(a)yahoo.ie> wrote: >> > Hi Mariano >> > I'm no expert either ;) >> > Without having access to the exact code it would look like either you >> have a collection that references itself (which would break all collection >> implementations) or maybe the tests have just slowed down to the point >> where you think it's 'crashed'. >> > Do you have anymore info or perhaps which methods you changed on >> IdentitySet? >> > Cheers >> > Carlo >> > On 13 Dec 2011, at 1:57 AM, Mariano Martinez Peck wrote: >> > >> > >> > On Tue, Dec 13, 2011 at 12:32 AM, Mariano Martinez Peck < >> marianopeck(a)gmail.com> wrote: >> >> >> >> >> >> On Mon, Dec 12, 2011 at 1:56 PM, Carlo <snoobabk(a)yahoo.ie> wrote: >> >>> >> >>> Hi >> >>> Wouldn't the fact that you use hash cause potential loops now? e.g. >> collection refers to another object that refers to first collection. --> >> aCollection>>hash references an item which causes this current collection's >> hash to be called again? >> >> >> >> Hi Carlo. I am still newbie with Collections but I think I am having >> exactly that problem. During my tests, it loops in Collection >> #hash >> when sending #hash to its elements. >> >> Sorry, but I couldn't undertand what is the cause of the problem? why >> it doesn't work while it does using #identityHash? could you elaborate? >> >> >> > >> > Well, now I understood, and I understand also why it doesn't happen >> with #identityHash. But what happens then with regular Dictionaries using >> #hash? why it doesn't happen there? >> > >> >> >> >> thanks >> >> >> >>> >> >>> identityHash is deterministic in this case. >> >>> Does this help? >> >>> Cheers >> >>> Carlo >> >>> On 12 Dec 2011, at 10:58 AM, Mariano Martinez Peck wrote: >> >>> Hi guys. I hope this is not a very stupid question. Background: in >> Fuel we have a IdentityDictionary where we put each of the objects we find >> while traversing the graph. We need to use IdentitySet because we cannot >> have repetitions (and to avoid loops) so we NEED to use #==. In such >> dictionary we put ALL objects of the graph, so it can be very big. Since >> IdentitySet uses #identityHash, it means it will be using those ONLY 12 >> bits in the object header. It means that we have 2^12 = 4096 different >> values. >> >>> >> >>> Question: having explained the previous, I wanted to be able to use >> #hash rather than #identityHash since several classes implement #hash and >> hence I thought that using #hash I could have less colisions and hence a >> better performance. I tried to make a subclass of IdentitySet that uses >> #hash rather than #identityHash but my image freezes. I also tried >> something like: >> >>> >> >>> set := PluggableSet new. >> >>> set hashBlock: [ :elem | elem hash ]. >> >>> set equalBlock: [ :a :b | a == b ]. >> >>> >> >>> But it doesn't work either. I works with simple tests in a workspace >> but when I run the full tests of Fuel, it enters in a loop in the method >> #hash of Collection.. >> >>> >> >>> Anyway, my question is, should that work? if not, what is the exact >> reason? >> >>> >> >>> Thanks in advance, >> >>> >> >>> -- >> >>> Mariano >> >>> http://marianopeck.wordpress.com >> >>> >> >>> >> >> >> >> >> >> >> >> -- >> >> Mariano >> >> http://marianopeck.wordpress.com >> >> >> > >> > >> > >> > -- >> > Mariano >> > http://marianopeck.wordpress.com >> > >> > >> > >> > > > > -- > Mariano > http://marianopeck.wordpress.com > > -- best, Eliot

Dec. 13, 2011

Re: [Pharo-project] [Vm-dev] OSS sound plugin.

by Krishsmalltalk

Yes I can have it working albeit with some hacks... For ubuntu base Can post some notes on it thru when I can present it sensibly... Hacks are to do with some flags that I just plain commented out... Will need to chk more if I can fix whatever it is more reliably Works with vm sound OSS plugin, Sudhakar krishnamachari Extn 91-40403012 Cell 9902104814 On Dec 13, 2011, at 5:14 PM, Bernat Romagosa <tibabenfortlapalanca(a)gmail.com> wrote: > It doesn't seem like it was done... I can't manage to get sound to work in > Linux and the OSS plugin is not there. > > Anyone succeeded? > > -- > View this message in context: http://forum.world.st/OSS-sound-plugin-tp3357110p4190128.html > Sent from the Pharo Smalltalk mailing list archive at Nabble.com. >

Dec. 13, 2011

[Pharo-project] Old smalltalker is slightly lost

by Sean P. DeNigris

>From KasperOsterbye: > > This post has NOT been accepted by the mailing list yet. > 15 years ago I did a lot in Smalltalk, and have done some recreational > meta programming along the way. However, I now wanted to do something > where I was in need of drawing a diagram with x and y axes on what I > presume might be called a Canvas. > > I lack a simple tutorial to get me started. I have looked at many, but > they seem to presume I want to draw on the world "openInWorld". I should > like not to scatter my world with line fragments and points, but keep it > inside a "normal" window with borders, a background color (i.e. not a see > through window). > I have gone dead in two approaches. > > First attempt is to use a SystemWindow, and then draw on the canvas I > presume is there somewhere. I have just not been able to figure out how to > find the canvas of that SystemWindow, nor what method presumeable does the > painting. > > Second attempt is to use a SystemWindow, and then add (addMorph:) line and > circle morphs for my diagram. But it seems something is missing in my > understanding there as well. > > While it is fun that I can open circles and other stuff in "world", it is > not what I need. > > Small hits or direct solutions more than welcome. > > -- Kasper > -- View this message in context: http://forum.world.st/Old-smalltalker-is-slightly-lost-tp4191432p4191432.ht… Sent from the Pharo Smalltalk mailing list archive at Nabble.com.

Dec. 13, 2011

[Pharo-project] Old smalltalker is slightly lost

by KasperOsterbye

15 years ago I did a lot in Smalltalk, and have done some recreational meta programming along the way. However, I now wanted to do something where I was in need of drawing a diagram with x and y axes on what I presume might be called a Canvas. I lack a simple tutorial to get me started. I have looked at many, but they seem to presume I want to draw on the world "openInWorld". I should like not to scatter my world with line fragments and points, but keep it inside a "normal" window with borders, a background color (i.e. not a see through window). I have gone dead in two approaches. First attempt is to use a SystemWindow, and then draw on the canvas I presume is there somewhere. I have just not been able to figure out how to find the canvas of that SystemWindow, nor what method presumeable does the painting. Second attempt is to use a SystemWindow, and then add (addMorph:) line and circle morphs for my diagram. But it seems something is missing in my understanding there as well. While it is fun that I can open circles and other stuff in "world", it is not what I need. Small hits or direct solutions more than welcome. -- Kasper -- View this message in context: http://forum.world.st/Old-smalltalker-is-slightly-lost-tp4191137p4191137.ht… Sent from the Pharo Smalltalk mailing list archive at Nabble.com.

Dec. 13, 2011

Re: [Pharo-project] Startup Preferences

by Camillo Bruni

Cool, For me the default config dir is still in "~/.pharo/" so that would be cool if added... cami On 2011-12-13, at 11:37, Benjamin wrote: > New version: > > Now there is a hierarchy for the setting file > > - first it lloks in the image diretory for st files > - if none is found, it looks into > ~/.config/pharo/${VERSION} > - if none is found it looks into > ~/.config/pharo/general > > We could also add a look into > ~/.pharo/ > > > > Gofer new > squeaksource: 'StartupPreferences'; > package: 'StartupPreferences'; > load. > > Ben > > On Dec 1, 2011, at 6:50 PM, StÃ©phane Ducasse wrote: > >> me too. >> >> Stef >> >> On Dec 1, 2011, at 6:38 PM, Camillo Bruni wrote: >> >>> Yeah I would really love to see this. >>> >>> maybe with a fixed lookup chain >>> >>> 1. Settings from the Image directory/startup.st >>> 2. Settings from ~/.pharo/$VERSION/startup.st >>> 3. Settings from ~/.pharo/startup.st >>> >>> cami >>> >> >> >

Dec. 13, 2011

[Pharo-project] hasPoolVarNamed: semantics :(

by StÃ©phane Ducasse

Hi guys Class>>hasPoolVarNamed: aString "Return whether the receiver has a pool variable named: aString" self hasSharedPools ifTrue: [ self sharedPools do: [:each | (each usesClassVarNamed: aString) ifTrue: [ ^true ]]] ifFalse: [ ^false ]. ^false So this means to me that a subclass of a shared pools user does not inherit the fact that its superclass effectively use a sharedPool. I would prefer that this would be the inverse. self assert: (RootClassPoolUser hasPoolVarNamed: 'Author'). "a subclass " self assert: (SubclassPoolUser hasPoolVarNamed: 'Author'). What do you think? may be we should rename it hasLocalPoolvarNamed: especially since we have usesPoolVarNamed: which takes into account the hierarchy only when the class starts to use a shared pool (oh boy) and is bogus (I wrote probably that code of course without tests. Now I have them.). Class>>usesPoolVarNamed: aString "Return whether the receiver has a pool variable named: aString" self hasSharedPools ifTrue: [ self allSharedPools do: [:each | (each usesClassVarNamed: aString) ifTrue: [^true]]] ifFalse: [^false]. ^false I wrote it now with tests as follows: Class>>usesPoolVarNamed: aString "Return whether the receiver has a pool variable named: aString" self allSharedPools do: [:each | (each usesClassVarNamed: aString) ifTrue: [^true]]. ^ false Stef

Dec. 13, 2011

Re: [Pharo-project] Startup Preferences

by Benjamin

On Dec 13, 2011, at 2:51 PM, Sean P. DeNigris wrote: > > Benjamin Van Ryseghem-2 wrote >> >> - first it lloks in the image diretory for st files >> > > Any .st file (not good) or just startup.st? right now it's any, but easily good be changed ;) Ben > > -- > View this message in context: http://forum.world.st/Startup-Preferences-tp4129591p4190520.html > Sent from the Pharo Smalltalk mailing list archive at Nabble.com. >

Dec. 13, 2011

[Pharo-project] allSharedPools

by StÃ©phane Ducasse

This is a pity that allSharedPools does not return an ordered unique collection. Class>>allSharedPools "Answer a Set of the pools the receiver shares, including those defined in the superclasses of the receiver." | aSet | ^self superclass == nil ifTrue: [self sharedPools copy] ifFalse: [aSet := self superclass allSharedPools. aSet addAll: self sharedPools. aSet] since sharedPools returns now an orderedCollection I doubt that the uniqueness will the guaranteed. Behavior>>allSharedPools "Answer a Set of the names of the pools (Dictionaries or SharedPool subclasses) that the receiver and the receiver's ancestors share." ^superclass allSharedPools Now on a metaclass we will climb up the inheritance chain for nothing so I would define allSharedPools as follows ClassDescription>>allSharedPools "Answer a Set of the pools the receiver shares, including those defined in the superclasses of the receiver." ^ OrderedCollection new Tell me what you think? Stef

Dec. 13, 2011

[Pharo-project] sharedPools

by StÃ©phane Ducasse

Hi guys I would like to see how we can clean the protocols. Behavior>>sharedPools "Answer a Set of the names of the pools (Dictionaries) that the receiver shares. 9/12/96 tk sharedPools have an order now" ^ OrderedCollection new => I will update the comment (this is not a set nor with names!) => I would like to move down to classDescription such methods because all the other methods are defined there or in Class. Class>>sharedPools "Answer a Set of the pool dictionaries declared in the receiver." sharedPools == nil ifTrue: [^OrderedCollection new] ifFalse: [^sharedPools] => I will update the comment TraitBehavior>>sharedPools ^ Dictionary new oops! I would be good to have polymorphic results! I will return ^ OrderedCollection new Now I'm not sure if it makes sense (but this is probably better) that RBClass returns also a polymorphic results. I have the impression that we get an orderedCollection too. RBClass>>sharedPools ^ self allPoolDictionaryNames collect: [ :each | Smalltalk globals at: each asSymbol ifAbsent: [ Dictionary new ] ] Ring seems to miss sharedPools. Stef

Dec. 13, 2011