- Pharo-dev - lists.pharo.org

[Pharo-project] about method lookup
by StÃ©phane Ducasse March 4, 2013

March 4, 2013

Hi I'm rethinking about my objvlisp lecture. In this lecture I'm talking about building the objVlisp kernel in Smalltalk. Now everything is working well except the error handling when a method is not found with different arguments than the original. Methods are represented with blocks having different number of arguments. I have several send messages (I should simplify this in the future) ------------------------------------------ binarySend: selector with: argument "send the message whose selector is <selector> to the receiver. The arguments of the messages are an <argument>. The method is lookep up in the class of the receiver. self is an objObject or a objClass." | ans | ans := (self objClass lookup: selector for: self) value: self value: argument. ^ ans send: selector withArguments: arguments "send the message whose selector is <selector> to the receiver. The arguments of the messages are an array <arguments>. The method is lookep up in the class of the receiver. self is an objObject or a objClass." | ans | ans := (self objClass lookup: selector for: self) valueWithArguments: (Array with: self) , arguments. ^ ans lookup is as simple as ------------------------------ lookup: selector for: anObjObject "look for the method named <selector> starting in the receiver. The lookup is done for a message sent to <anObjObject>. self is an objClass" ^(self doesUnderstand: selector) ifTrue: [ self bodyOfMethod: selector] ifFalse: [self objName = #ObjObject ifFalse: [ (Obj giveClassNamed: self objSuperclassId) lookup: selector for: anObjObject.] ifTrue: [ anObjObject objClass lookup: #error for: anObjObject]] Now the problem is that error redefinition only works if the same number of arguments are used and this is not nice. For example error redefinition works: --------------------------------------------------- testErrorRedefinition "(self run: #testErrorRedefinition)" self should: [aPoint unarySend: #zork] raise: Error. (pointClass at: pointClass offsetForMethodDict) at: #error put: ([ :superClassOfClassDefiningTheMethod | [:objself | 33]] value: objectClass). self assert: (aPoint unarySend: #zork) = 33. I could use cull: or something like that to be able to execute the block representing the error message But I have the impression that I need a way to say that the lookup failed and that the arguments should be packed in a message object. I looked in the StackVM to see how the arguments were handled: how do we go some lookupMethodInClass: class | currentClass dictionary found | ... currentClass := class. [currentClass ~= objectMemory nilObject] whileTrue: [dictionary := objectMemory fetchPointer: MethodDictionaryIndex ofObject: currentClass. ... found := self lookupMethodInDictionary: dictionary. found ifTrue: [^currentClass]. currentClass := self superclassOf: currentClass]. ... "Cound not find a normal message -- raise exception #doesNotUnderstand:" self createActualMessageTo: class. messageSelector := objectMemory splObj: SelectorDoesNotUnderstand. ^self lookupMethodInClass: class Am I correct to believe that createActualMessageTo: eats up the call args and does basically what I would like to do Stef

2 1

[Pharo-project] Pastell is now migrated to SmalltalkHub and has a nice CI job
by Torsten Bergmann March 3, 2013

March 3, 2013

Stef, this is not a problem of SqS. I copied the two new versions manually today back from MetacelloRepository to the SqS/Pastell repo since it looks like they were missing there. This is historic and from the time when usually sqS/MetacelloRepository was used for all configurations. It's not a real problem that the numbers have changed - I dont think many people reference Pastell. I already updated my other configs. If you have a script it should take care in the future if a newer ConfigurationOfXXX is either - in sqS/MetacelloRepository - or the projects own directory Regarding version numbers: I would vote for two digits - Major and minor version. So usually you start with 0.1, then 0.2, 0.3, ... When you make a patch to an old (problematic) version then you branch with a patch number. So for instance 0.2.1 means it is the first patch for Version 0.2. Bye T.

2 1

[Pharo-project] Pastell is now migrated to SmalltalkHub and has a nice CI job
by Torsten Bergmann March 3, 2013

March 3, 2013

Stef wrote: > Pastell is now migrated to SmalltalkHub and has a nice CI job Please note that previously ConfigurationOfPastell had the versions: 1.0, 1.0.2, 1.0.3, 1.0.4 and now has the versions: 1.0, 1.1., 1.2, 1.3 That broke other configurations that depend for instance on 1.0.4 and it adds a lot of confusion! Why didnt you just continue with 1.0.5 or 2.0? Thanks T.

3 4

[Pharo-project] about OrderedCollection new replaceFrom: 1 to: 5 with: #(6 7 8 9 10) startingAt: 1
by StÃ©phane Ducasse March 3, 2013

March 3, 2013

Hi guys in 1.2.1 OrderedCollection new replaceFrom: 1 to: 5 with: #(6 7 8 9 10) startingAt: 1 gave an OrderedCollection(6 7 8 9 10) while in 1.4 and 2.0 OrderedCollection new replaceFrom: 1 to: 5 with: #(6 7 8 9 10) startingAt: 1 raised an error Do you have an idea how I can handle that? Because XMLObservableList and XMLOrderedList implements replaceFrom: to: with: startingAt: because there is no self makeRoomFor: (stop - self capacity) invocation. in 1.4>> replaceFrom: start to: stop with: replacement startingAt: repStart "This destructively replaces elements from start to stop in the receiver starting at index, repStart, in the sequenceable collection, replacementCollection. Answer the receiver. No range checks are performed." | index repOff | repOff := repStart - start. index := start - 1. [(index := index + 1) <= stop] whileTrue: [self at: index put: (replacement at: repOff + index)] at: anInteger put: anObject "Put anObject at element index anInteger. at:put: cannot be used to append, front or back, to an ordered collection; it is used by a knowledgeable client to replace an element." self ensureBoundsFrom: anInteger to: anInteger. ^array at: anInteger + firstIndex - 1 put: anObject in 1.2.1 replaceFrom: start to: stop with: replacement startingAt: repStart "An ordered collection is growable, so a call to this can make the collection grow." self makeRoomFor: (stop - self capacity). array replaceFrom: start + firstIndex - 1 to: stop + firstIndex - 1 with: replacement startingAt: repStart. lastIndex := lastIndex max: stop + firstIndex - 1.

5 10

[Pharo-project] strange autocomplete behavior
by Ciprian Teodorov March 3, 2013

March 3, 2013

Hi guys, I don't know if any of you experience this problem, but with last versions of Pharo I have a strange behavior with code completion. The autocomplete list shows as I should, however after selecting an item hitting tab does not always completes the code as it should. For example try something like while typing Configuration... Do someone have an idea why this happens? Cheers, ciprian -- Dr. Ciprian TEODOROV IngÃ©nieur DÃ©veloppement CAO tÃ©l : 06 08 54 73 48 mail : ciprian.teodorov(a)gmail.com www.teodorov.ro

4 3

[Pharo-project] Spreadsheet widget
by Hilaire Fernandes March 3, 2013

March 3, 2013

Hello, Do we have such widget? Thanks Hilaire

7 16

[Pharo-project] XMLParser green bar on ci.inria.fr/pharo-contribution
by StÃ©phane Ducasse March 3, 2013

March 3, 2013

Hi guys so we have a first version of XMLParser migrated to SmalltalkHub and with tests running on our cool integration server. http://ci.inria.fr/pharo-contribution I will continue with OPAX/Pastell and migrate SOUP to use this new version of XMLParser. Stef

1 0

March 3, 2013

Hi I saw that some change in the XML parser were commited on the gemstone repository. Does it means that this is only for gemstone? Should I take take of them in the new ConfigurationOfXMLParser I'm building? I cannot test gemstone so I'm dealing with it right now. still I was wondering if some of the changes should be push in the main branch or not. Stef

1 0

March 3, 2013

Hi guys this is at least two times that I reimplement this behavior why cmd-b on a symbol is not equal to cmd-m why cmd-m on a class name is not equal to cmd-b so boringâ¦. this idea that there is a difference between implementor and browse is plain bad. Stef

1 0

[Pharo-project] About replace:with/copyReplace:with:
by Benjamin March 3, 2013

March 3, 2013

Hello guys, I just encounter these methods and got a bit disappointed First there is SequenceableCollection>>#replaceAll: oldObject with: newObject "Replace all occurences of oldObject with newObject" On String it something like: string := 'lapin'. string replaceAll: $a with: $b so now string is 'lbpin' Looking for a non destructive version, I looked at: #copyReplaceAll:with: SequenceableCollection>>#copyReplaceAll: oldSubstring with: newSubstring "Default is not to do token matching. See also String copyReplaceTokens:with:" So here it's not destructive, but 1) The argument names are wrong (or the method is not defined on the correct class) 2) why one take an element as argument and the other one a collection of elements ? Am I the only one disturb by this ? Ben

2 1