October 2011 - Pharo-users - lists.pharo.org

Re: [Pharo-users] Goods db

by Jonathan Kelly

On 23/10/2011 7:45 PM, Lukas Renggli wrote: >>> I do not see how BTreePageOuter is used? Is that your own class? What >>> does that have to do with BTree? >> I wrote my own BTree, mainly as a learning exercise. I had a look at the >> other version, but it made my head hurt. Turns out mine is about 20% faster >> than the version on squeaksource for adds, and nearly twice as fast for >> deletes. I > Sounds cool, but IMHO correctness and customizability to different > database layers is more important than speed. > >>> Generally it is a bad idea to subclass collections. I am not sure, but >> Is that a general smalltalk principle, or just in regards to Goods?? > This is general to Smalltalk. Subclassing from concrete collections > inherits you way too much behavior that exposes way too much of the > internal structure. Subclassing introduces subtle bugs and hand hard > to detect dependencies. > > For example, did you think of overriding #removeAllSuchThat: in your > OrderedCollection subclass? If not, I bet that soon somebody starts to > use it and completely breaks all constraints of a B-Tree. > Um, sub-classed or in a variable, what's to stop them doing that either way? J

Oct. 25, 2011

Re: [Pharo-users] Hosted GLASS made easy?

by Norbert Hartl

Am 25.10.2011 um 03:38 schrieb Larry White: > Hi, > > I've been watching a thread on the Seaside mailing list about hosting > GLASS on Amazon's EC2. It hit 60 messages (try this, doesn't work, try > that...). I'm no sys admin, so I'd like to find cloud hosting for > pharo and Gemstone that is easier to manage. I got Gemstone & Pharo > running on my mac in just a few hours, but i'm sure that's a lot > easier than a manual install on someone else's remote linux box with > Apache etc. > > Is there a reliable cloud hosting environment where the GLASS virtual > appliance might 'just work' out of the box? > Don't be misled by a 60 messages thread :) There are always things that should work but in a single case they don't. That doesn't mean the whole thing is that complicated. I never tried Nicks stuff but I'm sure in nearly all cases it will take 10 minutes to get going. If you just want to host a gemstone application and you are not in a hurry it might be good to wait a little bit. The gemstone guys are in the progress to integrate GLASS services into cloud foundry [1]. There you will be able to deploy your application and only that. You won't have access to the underlying operating system but that means you don't need to care. If you need to do more integration stuff on the machine than there are a few options. Nicks machine is the best right now IMHO. I myself have developed some scripts that help dealing with some setup things of GLASS [2]. And we are building a bootstrapp chain for a GLASS appliance where you could build your own machine from scratch [3] (which is in very early state and could cause more than 60 messages :) ). [1] http://www.cloudfoundry.com/ [2] https://github.com/noha/stone-creator [3] https://github.com/noha/virtual-gemstone Norbert

Oct. 25, 2011

Re: [Pharo-users] [pharo-users] Can't Load Magma in Pharo 1.3

by Ankur Sethi

Hi Stef, On Tue, Oct 25, 2011 at 12:33 AM, StÃ©phane Ducasse <stephane.ducasse(a)inria.fr> wrote: > Welcome ankhur > > we are sorry that you encounter problems. > Even with these issues, Pharo is probably the most fun I've had with a programming language (environment?) in a long time :) I'll try asking about the syntax highlighting thing on the Seaside ML. (PS: is "Ankhur" a common name somewhere? You're the second person to call me that this month.) -- Ankur Sethi (GeneralMaximus on IRC and elsewhere)

Oct. 25, 2011

Hosted GLASS made easy?

by Larry White

Hi, I've been watching a thread on the Seaside mailing list about hosting GLASS on Amazon's EC2. It hit 60 messages (try this, doesn't work, try that...). I'm no sys admin, so I'd like to find cloud hosting for pharo and Gemstone that is easier to manage. I got Gemstone & Pharo running on my mac in just a few hours, but i'm sure that's a lot easier than a manual install on someone else's remote linux box with Apache etc. Is there a reliable cloud hosting environment where the GLASS virtual appliance might 'just work' out of the box? thanks. Larry

Oct. 25, 2011

Re: [Pharo-users] [pharo-users] Can't Load Magma in Pharo 1.3

by StÃ©phane Ducasse

Welcome ankhur we are sorry that you encounter problems. Stef On Oct 24, 2011, at 4:54 PM, Ankur Sethi wrote: > Hi, > > I'm new to Pharo and Smalltalk in general. I'm trying to build a small > blogging application using Pharo and Seaside. I've decided to use > Magma for persistence, but I'm having some issues loading it in Pharo > 1.3. > > I loaded Magma the usual way: > > Gofer new > squeaksource: 'MetacelloRepository'; > package: 'ConfigurationOfMagma'; > load. > > (Smalltalk at:#ConfigurationOfMagma) > project latestVersion > load: 'Server'; > load: 'Client'. > > ... and I was presented with a backtrace. Since didn't know how to get > a full backtrace out of Pharo in plain text, I took a screenshot: > http://imagebin.org/180559 > > I know this is not adequate, but I'm hoping someone can help me get > Pharo to give me information about this issue in a more convenient > format. > > When I told Pharo to abandon loading the package and then re-evaluated this: > > (Smalltalk at:#ConfigurationOfMagma) > project latestVersion > load: 'Server'; > load: 'Client'. > > ... then everything went just fine. I even tried creating a Magma > repository, which worked. However, some things were totally borked in > my Pharo image. I couldn't get the System Browser to do syntax > highlighting, for one. > > I'm using Pharo1.3 + CogVM 13307 on Mac OS X Lion. > > I hope I made sense. > > Thanks. > > -- > Ankur Sethi > (GeneralMaximus on IRC and elsewhere) >

Oct. 24, 2011

Re: [Pharo-users] [pharo-users] Can't Load Magma in Pharo 1.3

by Ankur Sethi

Hi, On Mon, Oct 24, 2011 at 9:59 PM, Mariano Martinez Peck > > Yes. Go to ThisOSProcess class >> initialize > and replace the addToStartUpList:after:Â by a simple addToStartUpList: I actually loaded the newest version of OSProcess before loading Magma in a fresh image. That works great. Now I'm noticing something else: loading Magma works, but as soon as I load Seaside, I lose all syntax highlighting in the System Browser! I mentioned this issue in my original query, but I assumed Magma was causing it. Turns out it wasn't Magma after all, it was Seaside. Syntax highlighting still works in the workspace. It's only the System Browser that's affected. What do I do? -- Ankur Sethi (GeneralMaximus on IRC and elsewhere)

Oct. 24, 2011

Re: [Pharo-users] [pharo-users] Can't Load Magma in Pharo 1.3

by Mariano Martinez Peck

On Mon, Oct 24, 2011 at 6:24 PM, Ankur Sethi <get.me.ankur(a)gmail.com> wrote: > Hi, > > On Mon, Oct 24, 2011 at 8:37 PM, Mariano Martinez Peck > <marianopeck(a)gmail.com> wrote: > > Hence, > > ConfigurationOfMagma should be updated to use the new version of > OSProcess. > > Is there a way I can fix this manually in my own image? > > Yes. Go to ThisOSProcess class >> initialize and replace the addToStartUpList:after: by a simple addToStartUpList: cheers > -- > Ankur Sethi > (GeneralMaximus on IRC and elsewhere) > > -- Mariano http://marianopeck.wordpress.com

Oct. 24, 2011

Re: [Pharo-users] [pharo-users] Can't Load Magma in Pharo 1.3

by Ankur Sethi

Hi, On Mon, Oct 24, 2011 at 8:37 PM, Mariano Martinez Peck <marianopeck(a)gmail.com> wrote: > Hence, > ConfigurationOfMagma should be updated to use the new version of OSProcess. Is there a way I can fix this manually in my own image? -- Ankur Sethi (GeneralMaximus on IRC and elsewhere)

Oct. 24, 2011

Re: [Pharo-users] [pharo-users] Can't Load Magma in Pharo 1.3

by Mariano Martinez Peck

Hi. It looks like OSProcess (a dependency of Magma) is registering a class in the startUp list explicitly before ImageSegment (which was removed). This was already fixed in new version of OSProcess. Hence, ConfigurationOfMagma should be updated to use the new version of OSProcess. Cheers On Mon, Oct 24, 2011 at 4:54 PM, Ankur Sethi <get.me.ankur(a)gmail.com> wrote: > Hi, > > I'm new to Pharo and Smalltalk in general. I'm trying to build a small > blogging application using Pharo and Seaside. I've decided to use > Magma for persistence, but I'm having some issues loading it in Pharo > 1.3. > > I loaded Magma the usual way: > > Gofer new > squeaksource: 'MetacelloRepository'; > package: 'ConfigurationOfMagma'; > load. > > (Smalltalk at:#ConfigurationOfMagma) > project latestVersion > load: 'Server'; > load: 'Client'. > > ... and I was presented with a backtrace. Since didn't know how to get > a full backtrace out of Pharo in plain text, I took a screenshot: > http://imagebin.org/180559 > > I know this is not adequate, but I'm hoping someone can help me get > Pharo to give me information about this issue in a more convenient > format. > > When I told Pharo to abandon loading the package and then re-evaluated > this: > > (Smalltalk at:#ConfigurationOfMagma) > project latestVersion > load: 'Server'; > load: 'Client'. > > ... then everything went just fine. I even tried creating a Magma > repository, which worked. However, some things were totally borked in > my Pharo image. I couldn't get the System Browser to do syntax > highlighting, for one. > > I'm using Pharo1.3 + CogVM 13307 on Mac OS X Lion. > > I hope I made sense. > > Thanks. > > -- > Ankur Sethi > (GeneralMaximus on IRC and elsewhere) > > -- Mariano http://marianopeck.wordpress.com

Oct. 24, 2011

[pharo-users] Can't Load Magma in Pharo 1.3

by Ankur Sethi

Hi, I'm new to Pharo and Smalltalk in general. I'm trying to build a small blogging application using Pharo and Seaside. I've decided to use Magma for persistence, but I'm having some issues loading it in Pharo 1.3. I loaded Magma the usual way: Gofer new squeaksource: 'MetacelloRepository'; package: 'ConfigurationOfMagma'; load. (Smalltalk at:#ConfigurationOfMagma) project latestVersion load: 'Server'; load: 'Client'. ... and I was presented with a backtrace. Since didn't know how to get a full backtrace out of Pharo in plain text, I took a screenshot: http://imagebin.org/180559 I know this is not adequate, but I'm hoping someone can help me get Pharo to give me information about this issue in a more convenient format. When I told Pharo to abandon loading the package and then re-evaluated this: (Smalltalk at:#ConfigurationOfMagma) project latestVersion load: 'Server'; load: 'Client'. ... then everything went just fine. I even tried creating a Magma repository, which worked. However, some things were totally borked in my Pharo image. I couldn't get the System Browser to do syntax highlighting, for one. I'm using Pharo1.3 + CogVM 13307 on Mac OS X Lion. I hope I made sense. Thanks. -- Ankur Sethi (GeneralMaximus on IRC and elsewhere)

Oct. 24, 2011

Re: [Pharo-users] [Pharo-project] Planning a nice pharo conference :)

by eMko

Thank you :) Martin MadÄra -- View this message in context: http://forum.world.st/Planning-a-nice-pharo-conference-tp3902751p3932872.ht… Sent from the Pharo Smalltalk Users mailing list archive at Nabble.com.

Oct. 24, 2011

Re: [Pharo-users] How to get temp names of BlockClosure

by Igor Stasenko

On 24 October 2011 12:44, Marcus Denker <marcus.denker(a)inria.fr> wrote: > > On Oct 24, 2011, at 12:40 PM, Marcus Denker wrote: >> >>> While looking I found >>> >>> [:a :b :c| ] method tempNames Â #('[a' 'b' 'c]') >>> >>> For me there are two errors. >> >> yes, this is wrong. >> >> The real problem is that the CompiledMethod does not store any info about temp names, >> so you need to either go to the source text or (better) AST... and than there is Dan's clever hack >> that can save temp names in the method trailer to allow decompiling without sources, complicating >> the logic in addition. >> > > Ah, and the temp-names embedded in the trailer is used excactly for DoIts... because normally the > #tempNames should always return an empty list. > > It only has meaning when using Dan's embedded temp-names in the trailer hack, which for the whole > image nobody ever uses since years. > yes, i think this is too much effort(s) for displaying a do-it temp names in debugger. As you said, it is possible to embed temps into compiled method, but it is largely unused in system, and used only for doits. Definitely, the way how (often) we using it in our system(s) was not worth spending manhours implementing it. > Â Â Â Â Marcus > > > -- > Marcus Denker -- http://marcusdenker.de > > > -- Best regards, Igor Stasenko.

Oct. 24, 2011

Re: [Pharo-users] How to get temp names of BlockClosure

by Marcus Denker

On Oct 24, 2011, at 12:40 PM, Marcus Denker wrote: > >> While looking I found >> >> [:a :b :c| ] method tempNames #('[a' 'b' 'c]') >> >> For me there are two errors. > > yes, this is wrong. > > The real problem is that the CompiledMethod does not store any info about temp names, > so you need to either go to the source text or (better) AST... and than there is Dan's clever hack > that can save temp names in the method trailer to allow decompiling without sources, complicating > the logic in addition. > Ah, and the temp-names embedded in the trailer is used excactly for DoIts... because normally the #tempNames should always return an empty list. It only has meaning when using Dan's embedded temp-names in the trailer hack, which for the whole image nobody ever uses since years. Marcus -- Marcus Denker -- http://marcusdenker.de

Oct. 24, 2011

Re: [Pharo-users] How to get temp names of BlockClosure

by Marcus Denker

On Oct 17, 2011, at 4:57 PM, Norbert Hartl wrote: > I'm trying to learn how to get temp names from a BlockClosure. Well, I try to be able to treat BlockClosures the same way than CompileMethod when it comes to arguments. I can do > > [:a :b :c| ] asContext tempNames #('a' 'b' 'c') > > but I'm not sure if this is the best to go. Yes, this uses the DebuggerMethodMap and thus should hide the temps that are there just to store the tempVecor. (with the closure design, there are temps that are just implementation artefacts, not temps that where defined). > While looking I found > > [:a :b :c| ] method tempNames #('[a' 'b' 'c]') > > For me there are two errors. yes, this is wrong. The real problem is that the CompiledMethod does not store any info about temp names, so you need to either go to the source text or (better) AST... and than there is Dan's clever hack that can save temp names in the method trailer to allow decompiling without sources, complicating the logic in addition. I always think that all this is an epic mess... but other people like it. Marcus -- Marcus Denker -- http://marcusdenker.de

Oct. 24, 2011

Re: [Pharo-users] GlorpDBX

by Mariano Martinez Peck

On Mon, Oct 24, 2011 at 9:28 AM, Jonathan Kelly <jonkelly(a)fastmail.fm>wrote: > Hi, > > > feeling particularly stupid right now. Been trying to get Glorp DBX > working with Postgresql for the last 2 hours and I'm busted. It took me > about an out just to find out how to load and configure things. ugh. It > took me a while to work out I had to actually install OpenDBX (doh). So > I have downloaded that and added the directory to my PATH. > That's good. > > So I'm trying to run the tests and I get DBXFatalError: FATAL OpenDBX: > Loading backend library failed. I know it's finding OpenDBX, Now, it is not that it is not finding OpenDBX, but PostgreSQL client library > because > when I move the OpenDBX folder, I get a different error - "Error: > External Module Not Found" from apiInitialize (and I can see it's > trying to load libopendbx-1.dll). > > Part of my problem is squeakdbx.org domain expired on 09/27/11 and is > now a godaddy parking page. double ugh. > > We have announced in all possible mailing lists that it was temporaly moved to http://dbxtalk.smallworks.com.ar/squeakdbx Check this links: http://dbxtalk.smallworks.com.ar/Compiling%20and%20installing%20OpenDBX http://dbxtalk.smallworks.com.ar/Installation > Anyone got any ideas? > Install PostgreSQL client library (32 bits) and make sure its lib is findable by windows. Cheers I'm running a fresh 1.3 using CogVM-Win-13307 install on Windows XP. > Jonathan. > > -- Mariano http://marianopeck.wordpress.com

Oct. 24, 2011

Re: [Pharo-users] GlorpDBX

by Sven Van Caekenberghe

Jonothan, On 24 Oct 2011, at 09:28, Jonathan Kelly wrote: > Hi, > > > feeling particularly stupid right now. Been trying to get Glorp DBX > working with Postgresql for the last 2 hours and I'm busted. It took me > about an out just to find out how to load and configure things. ugh. It > took me a while to work out I had to actually install OpenDBX (doh). So > I have downloaded that and added the directory to my PATH. > > So I'm trying to run the tests and I get DBXFatalError: FATAL OpenDBX: > Loading backend library failed. I know it's finding OpenDBX, because > when I move the OpenDBX folder, I get a different error - "Error: > External Module Not Found" from apiInitialize (and I can see it's > trying to load libopendbx-1.dll). > > Part of my problem is squeakdbx.org domain expired on 09/27/11 and is > now a godaddy parking page. double ugh. > > Anyone got any ideas? > I'm running a fresh 1.3 using CogVM-Win-13307 install on Windows XP. > Jonathan. A couple of days ago I got it working in 1.3 with the native (non-dbx) postgresql driver. I basically did: Gofer new squeaksource: 'MetacelloRepository'; package: 'ConfigurationOfGlorpDBX'; load. ConfigurationOfGlorpDBX project latestVersion load: 'GlorpPostgresV2NativeWithTests'. I can't remember exactly, but it more or less worked out of the box. I did some patches for timestamp decoding and blobs, but these won't hit you right away. Sven

Oct. 24, 2011

GlorpDBX

by Jonathan Kelly

Hi, feeling particularly stupid right now. Been trying to get Glorp DBX working with Postgresql for the last 2 hours and I'm busted. It took me about an out just to find out how to load and configure things. ugh. It took me a while to work out I had to actually install OpenDBX (doh). So I have downloaded that and added the directory to my PATH. So I'm trying to run the tests and I get DBXFatalError: FATAL OpenDBX: Loading backend library failed. I know it's finding OpenDBX, because when I move the OpenDBX folder, I get a different error - "Error: External Module Not Found" from apiInitialize (and I can see it's trying to load libopendbx-1.dll). Part of my problem is squeakdbx.org domain expired on 09/27/11 and is now a godaddy parking page. double ugh. Anyone got any ideas? I'm running a fresh 1.3 using CogVM-Win-13307 install on Windows XP. Jonathan.

Oct. 24, 2011

Re: [Pharo-users] Goods db

by StÃ©phane Ducasse

this is why we need a solid proxy. Stef On Oct 23, 2011, at 12:13 PM, Mariano Martinez Peck wrote: > > > > > Um, I'm not upto optimization tricks yet, so no. :) > > This has nothing to do with *your* optimization tricks. These are > optimization tricks that the VM and the compiler play on you. Below a > list of potentially dangerous selectors that might not cause the GOODS > proxy objects to load the object. You need to be aware of these > details to be able to write code that works with proxy objects: > > ifTrue: ifFalse: ifTrue:ifFalse: ifFalse:ifTrue: and: or: whileFalse: > whileTrue: whileFalse whileTrue to:do: to:by:do: caseOf: > caseOf:otherwise: ifNil: ifNotNil: ifNil:ifNotNil: ifNotNil:ifNil: > repeat + - < > <= >= = ~= * / \\ @ bitShift: // bitAnd: bitOr: at: > at:put: size next nextPut: atEnd == class blockCopy: value value: do: > new new: x y yourself > > > Just a small comment: not ALL those methods will be *never* executed. > > All those that are optimized by the compiler: MessageNode >> #initialize (instVar MacroSelectors) > ifTrue: ifFalse: ifTrue:ifFalse: ifFalse:ifTrue: > and: or: > whileFalse: whileTrue: whileFalse whileTrue > to:do: to:by:do: > caseOf: caseOf:otherwise: > ifNil: ifNotNil: ifNil:ifNotNil: ifNotNil:ifNil: > repeat > > > Those are never executed, hence you cannot intercept them. > > Now, those special selectors that have associated bytecodes: Smalltalk specialSelectors > -> #(#+ 1 #- 1 #< 1 #> 1 #'<=' 1 #'>=' 1 #= 1 #'~=' 1 #* 1 #/ 1 #'\\' 1 #@ 1 #bitShift: 1 #'//' 1 #bitAnd: 1 #bitOr: 1 #at: 1 #at:put: 2 #size 0 #next 0 #nextPut: 1 #atEnd 0 #'==' 1 #class 0 #blockCopy: 1 #value 0 #value: 1 #do: 1 #new 0 #new: 1 #x 0 #y 0) > > SOME of them are not executed. > > 1) Some are ALWAYS executed > For example, the bytecode for #atEnd does nothing more than a small optimization, but it is always executed at the end: > StackInterpreter >> bytecodePrimAtEnd > messageSelector := self specialSelector: 21. > argumentCount := 0. > self normalSend. > > 2) Some are SOMETIMES executed. > Example, #+, #>=, etc. All those math bytecode usually work if the receiver and argument are numbers. But if they are not, like in the case of a proxy, then the bytecode fails and hence, the method is executed. Example: > StackInterpreter >> bytecodePrimAdd > | rcvr arg result | > rcvr := self internalStackValue: 1. > arg := self internalStackValue: 0. > (self areIntegers: rcvr and: arg) > ifTrue: [result := (objectMemory integerValueOf: rcvr) + (objectMemory integerValueOf: arg). > (objectMemory isIntegerValue: result) ifTrue: > [self internalPop: 2 thenPush: (objectMemory integerObjectOf: result). > ^ self fetchNextBytecode "success"]] > ifFalse: [self initPrimCall. > self externalizeIPandSP. > self primitiveFloatAdd: rcvr toArg: arg. > self internalizeIPandSP. > self successful ifTrue: [^ self fetchNextBytecode "success"]]. > > messageSelector := self specialSelector: 0. > argumentCount := 1. > self normalSend > > 3) Some are NEVER executed. For example, #class and #==. > StackInterpreter >> bytecodePrimEquivalent > > | rcvr arg | > rcvr := self internalStackValue: 1. > arg := self internalStackValue: 0. > self booleanCheat: rcvr = arg. > > > So, for a proxy I think the only problem is (apart from the already mentioned compiler inlining) point 3), that is #== and #class > > Anyway, I have wirtten a paper about writting proxies in Smalltalk which you may find interesting. > http://rmod.lille.inria.fr/web/pier/software/Marea/GhostProxies > You can get the paper from there. > > Cheers > > > > > Lukas > > -- > Lukas Renggli > www.lukas-renggli.ch > > > > > -- > Mariano > http://marianopeck.wordpress.com >

Oct. 23, 2011

Re: [Pharo-users] Goods db

by Lukas Renggli

>> > Um, I'm not upto optimization tricks yet, so no. :) >> >> This has nothing to do with *your* optimization tricks. These are >> optimization tricks that the VM and the compiler play on you. Below a >> list of potentially dangerous selectors that might not cause the GOODS >> proxy objects to load the object. You need to be aware of these >> details to be able to write code that works with proxy objects: >> >> ifTrue: ifFalse: ifTrue:ifFalse: ifFalse:ifTrue: and: or: whileFalse: >> whileTrue: whileFalse whileTrue to:do: to:by:do: caseOf: >> caseOf:otherwise: ifNil: ifNotNil: ifNil:ifNotNil: ifNotNil:ifNil: >> repeat + - < > <= >= = ~= * / \\ @ bitShift: // bitAnd: bitOr: at: >> at:put: size next nextPut: atEnd == class blockCopy: value value: do: >> new new: x y yourself > > > Just a small comment: not ALL those methods will be *never* executed. That's why I used the word "might" ;-) Lukas > > All those that are optimized by the compiler: MessageNode >> #initialize (instVar MacroSelectors) > ifTrue: ifFalse: ifTrue:ifFalse: ifFalse:ifTrue: > and: or: > whileFalse: whileTrue: whileFalse whileTrue > to:do: to:by:do: > caseOf: caseOf:otherwise: > ifNil: ifNotNil: ifNil:ifNotNil: ifNotNil:ifNil: > repeat > > > Those are never executed, hence you cannot intercept them. > > Now, those special selectors that have associated bytecodes: Smalltalk specialSelectors > -> #(#+ 1 #- 1 #< 1 #> 1 #'<=' 1 #'>=' 1 #= 1 #'~=' 1 #* 1 #/ 1 #'\\' 1 #@ 1 #bitShift: 1 #'//' 1 #bitAnd: 1 #bitOr: 1 #at: 1 #at:put: 2 #size 0 #next 0 #nextPut: 1 #atEnd 0 #'==' 1 #class 0 #blockCopy: 1 #value 0 #value: 1 #do: 1 #new 0 #new: 1 #x 0 #y 0) > > SOME of them are not executed. > > 1) Some are ALWAYS executed > For example, the bytecode for #atEnd does nothing more than a small optimization, but it is always executed at the end: > StackInterpreter >> bytecodePrimAtEnd > messageSelector := self specialSelector: 21. > argumentCount := 0. > self normalSend. > > 2) Some are SOMETIMES executed. > Example, #+, #>=, etc. All those math bytecode usually work if the receiver and argument are numbers. But if they are not, like in the case of a proxy, then the bytecode fails and hence, the method is executed. Example: > StackInterpreter >> bytecodePrimAdd > | rcvr arg result | > rcvr := self internalStackValue: 1. > arg := self internalStackValue: 0. > (self areIntegers: rcvr and: arg) > ifTrue: [result := (objectMemory integerValueOf: rcvr) + (objectMemory integerValueOf: arg). > (objectMemory isIntegerValue: result) ifTrue: > [self internalPop: 2 thenPush: (objectMemory integerObjectOf: result). > ^ self fetchNextBytecode "success"]] > ifFalse: [self initPrimCall. > self externalizeIPandSP. > self primitiveFloatAdd: rcvr toArg: arg. > self internalizeIPandSP. > self successful ifTrue: [^ self fetchNextBytecode "success"]]. > > messageSelector := self specialSelector: 0. > argumentCount := 1. > self normalSend > > 3) Some are NEVER executed. For example, #class and #==. > StackInterpreter >> bytecodePrimEquivalent > > | rcvr arg | > rcvr := self internalStackValue: 1. > arg := self internalStackValue: 0. > self booleanCheat: rcvr = arg. > > > So, for a proxy I think the only problem is (apart from the already mentioned compiler inlining) point 3), that is #== and #class > > Anyway, I have wirtten a paper about writting proxies in Smalltalk which you may find interesting. > http://rmod.lille.inria.fr/web/pier/software/Marea/GhostProxies > You can get the paper from there. > > Cheers > > > >> >> Lukas >> >> -- >> Lukas Renggli >> www.lukas-renggli.ch >> > > > > -- > Mariano > http://marianopeck.wordpress.com > > -- Lukas Renggli www.lukas-renggli.ch

Oct. 23, 2011

Re: [Pharo-users] Goods db

by Mariano Martinez Peck

> > > > Um, I'm not upto optimization tricks yet, so no. :) > > This has nothing to do with *your* optimization tricks. These are > optimization tricks that the VM and the compiler play on you. Below a > list of potentially dangerous selectors that might not cause the GOODS > proxy objects to load the object. You need to be aware of these > details to be able to write code that works with proxy objects: > > ifTrue: ifFalse: ifTrue:ifFalse: ifFalse:ifTrue: and: or: whileFalse: > whileTrue: whileFalse whileTrue to:do: to:by:do: caseOf: > caseOf:otherwise: ifNil: ifNotNil: ifNil:ifNotNil: ifNotNil:ifNil: > repeat + - < > <= >= = ~= * / \\ @ bitShift: // bitAnd: bitOr: at: > at:put: size next nextPut: atEnd == class blockCopy: value value: do: > new new: x y yourself > Just a small comment: not ALL those methods will be *never* executed. All those that are optimized by the compiler: MessageNode >> #initialize (instVar MacroSelectors) ifTrue: ifFalse: ifTrue:ifFalse: ifFalse:ifTrue: and: or: whileFalse: whileTrue: whileFalse whileTrue to:do: to:by:do: caseOf: caseOf:otherwise: ifNil: ifNotNil: ifNil:ifNotNil: ifNotNil:ifNil: repeat Those are never executed, hence you cannot intercept them. Now, those special selectors that have associated bytecodes: Smalltalk specialSelectors -> #(#+ 1 #- 1 #< 1 #> 1 #'<=' 1 #'>=' 1 #= 1 #'~=' 1 #* 1 #/ 1 #'\\' 1 #@ 1 #bitShift: 1 #'//' 1 #bitAnd: 1 #bitOr: 1 #at: 1 #at:put: 2 #size 0 #next 0 #nextPut: 1 #atEnd 0 #'==' 1 #class 0 #blockCopy: 1 #value 0 #value: 1 #do: 1 #new 0 #new: 1 #x 0 #y 0) SOME of them are not executed. 1) Some are ALWAYS executed For example, the bytecode for #atEnd does nothing more than a small optimization, but it is always executed at the end: StackInterpreter >> bytecodePrimAtEnd messageSelector := self specialSelector: 21. argumentCount := 0. self normalSend. 2) Some are SOMETIMES executed. Example, #+, #>=, etc. All those math bytecode usually work if the receiver and argument are numbers. But if they are not, like in the case of a proxy, then the bytecode fails and hence, the method is executed. Example: StackInterpreter >> bytecodePrimAdd | rcvr arg result | rcvr := self internalStackValue: 1. arg := self internalStackValue: 0. (self areIntegers: rcvr and: arg) ifTrue: [result := (objectMemory integerValueOf: rcvr) + (objectMemory integerValueOf: arg). (objectMemory isIntegerValue: result) ifTrue: [self internalPop: 2 thenPush: (objectMemory integerObjectOf: result). ^ self fetchNextBytecode "success"]] ifFalse: [self initPrimCall. self externalizeIPandSP. self primitiveFloatAdd: rcvr toArg: arg. self internalizeIPandSP. self successful ifTrue: [^ self fetchNextBytecode "success"]]. messageSelector := self specialSelector: 0. argumentCount := 1. self normalSend 3) Some are NEVER executed. For example, #class and #==. StackInterpreter >> bytecodePrimEquivalent | rcvr arg | rcvr := self internalStackValue: 1. arg := self internalStackValue: 0. self booleanCheat: rcvr = arg. So, for a proxy I think the only problem is (apart from the already mentioned compiler inlining) point 3), that is #== and #class Anyway, I have wirtten a paper about writting proxies in Smalltalk which you may find interesting. http://rmod.lille.inria.fr/web/pier/software/Marea/GhostProxies You can get the paper from there. Cheers > > Lukas > > -- > Lukas Renggli > www.lukas-renggli.ch > > -- Mariano http://marianopeck.wordpress.com

Oct. 23, 2011

Re: [Pharo-users] Inconsistent links for Pharo 1.3 versions

by Marcus Denker

Thanks, I fixes it. Marcus On Oct 22, 2011, at 2:48 PM, Marco A. Gonzalez wrote: > FYI for whomever is maintaining the pharo pages... > On http://www.pharo-project.org/home and all other pages, the upper right icon badge to download the Pharo 1.3 One Click installer downloads version 13315 (1) whereas on the download page, the rest of the links refer to version 13307 (2). > > Summary: > (1) The icon badge has.... https://gforge.inria.fr/frs/download.php/29274/Pharo-1.3-13315-OneClick.zip > (2) The Pharo 1.3. one-click distribution section has.... http://gforge.inria.fr/frs/download.php/29054/Pharo-1.3-13307-OneClick.zip > (3) In the Custom downloads section, the Pharo 1.3 link has.... http://gforge.inria.fr/frs/download.php/29053/Pharo-1.3-13307.zip > > The above said... thank you for all you do. > > - Marco A. > > > -- Marcus Denker -- http://marcusdenker.de

Oct. 23, 2011

Re: [Pharo-users] Goods db

by Lukas Renggli

>> I do not see how BTreePageOuter is used? Is that your own class? What >> does that have to do with BTree? > > I wrote my own BTree, mainly as a learning exercise. I had a look at the > other version, but it made my head hurt. Turns out mine is about 20% faster > than the version on squeaksource for adds, and nearly twice as fast for > deletes. I Sounds cool, but IMHO correctness and customizability to different database layers is more important than speed. >> Generally it is a bad idea to subclass collections. I am not sure, but > > Is that a general smalltalk principle, or just in regards to Goods?? This is general to Smalltalk. Subclassing from concrete collections inherits you way too much behavior that exposes way too much of the internal structure. Subclassing introduces subtle bugs and hand hard to detect dependencies. For example, did you think of overriding #removeAllSuchThat: in your OrderedCollection subclass? If not, I bet that soon somebody starts to use it and completely breaks all constraints of a B-Tree. >> - I am not aware that anybody has used GOODS in the past year on Pharo >> or even on a Â Cog VM. So there might be (subtle) bugs. Do the tests >> pass? > > Up until now. :) There were a couple of test failures > 1.testFlushAllRemovesAllUnreferenced - this was failure > 2.testTwoDifferentShapesInDatabase - this was error. I haven't used GOODS for a long time, but it seems to me that fixing all tests is the first thing you would need to do. I doubt that any of you own code works reliably if there are things broken in the core code. >> - GOODS loads things lazily. So it might be that when you look at >> things through reflection (like the inspector), they do not load >> properly. Also you need to make sure not to not access your objects >> with optimized messages like #== or #yourself, because these do not >> trigger the loading of the object. > > Um, I'm not upto optimization tricks yet, so no. :) This has nothing to do with *your* optimization tricks. These are optimization tricks that the VM and the compiler play on you. Below a list of potentially dangerous selectors that might not cause the GOODS proxy objects to load the object. You need to be aware of these details to be able to write code that works with proxy objects: ifTrue: ifFalse: ifTrue:ifFalse: ifFalse:ifTrue: and: or: whileFalse: whileTrue: whileFalse whileTrue to:do: to:by:do: caseOf: caseOf:otherwise: ifNil: ifNotNil: ifNil:ifNotNil: ifNotNil:ifNil: repeat + - < > <= >= = ~= * / \\ @ bitShift: // bitAnd: bitOr: at: at:put: size next nextPut: atEnd == class blockCopy: value value: do: new new: x y yourself Lukas -- Lukas Renggli www.lukas-renggli.ch

Oct. 23, 2011

Re: [Pharo-users] Goods db

by Jonathan Kelly

Hi, On 22/10/2011 7:09 PM, Lukas Renggli wrote: > I do not see how BTreePageOuter is used? Is that your own class? What > does that have to do with BTree? I wrote my own BTree, mainly as a learning exercise. I had a look at the other version, but it made my head hurt. Turns out mine is about 20% faster than the version on squeaksource for adds, and nearly twice as fast for deletes. I > Generally it is a bad idea to subclass collections. I am not sure, but Is that a general smalltalk principle, or just in regards to Goods?? > it could be that GOODS optimizes OrderedCollection access (look for > class extension from GOODS in OrderedCollection and superclasses). I couldn't find anything - but I'm quite new to smalltalk, so that's no guarantee. > Some other random remarks: > > - I am not aware that anybody has used GOODS in the past year on Pharo > or even on a Cog VM. So there might be (subtle) bugs. Do the tests > pass? Up until now. :) There were a couple of test failures 1.testFlushAllRemovesAllUnreferenced - this was failure 2.testTwoDifferentShapesInDatabase - this was error. These seemed somewhat esoteric errors to me, so i kept going. All the DB sort of tests passed, which is what i was interested in. So ... I massaged my implementation to move from subclass to component (another learning experience :) and now GOODS is working. I'll do some more testing/playing/learning and report back. I picked GOODS because I wanted to get my brain into OO thinking, and it seemed more OO that other alternatives, but I'll look at GLORP as well. Another learning experience. > - GOODS loads things lazily. So it might be that when you look at > things through reflection (like the inspector), they do not load > properly. Also you need to make sure not to not access your objects > with optimized messages like #== or #yourself, because these do not > trigger the loading of the object. Um, I'm not upto optimization tricks yet, so no. :) Jonathan.

Oct. 22, 2011

Re: [Pharo-users] Goods db

by Johan Fabry

I think you mean the Chilean mirror http://dsal.cl/squeaksource , unless there is a second mirror that I am not aware of? (And in that case, where is it and who is in charge? We should cooperate ...) On 22 Oct 2011, at 04:43, Jonathan Kelly wrote: > PS... squeaksource was down, so I downloaded the packages from the spanish mirror into my local package cache, and loaded them from there. ---> Save our in-boxes! http://emailcharter.org <--- Johan Fabry jfabry(a)dcc.uchile.cl - http://dcc.uchile.cl/~jfabry PLEIAD Lab - Computer Science Department (DCC) - University of Chile

Oct. 22, 2011

Re: [Pharo-users] Goods db

by StÃ©phane Ducasse

I know that andy and johan fixed some bugs in goods and that david shaffer too. Now I do not know if andy code has been published. Andy? Stef On Oct 22, 2011, at 9:43 AM, Jonathan Kelly wrote: > Hi, > > I thought I'd ask here first ... I got my BTree code working, so I thought I'd try to use it with the Goods DB, but it seems it's only storing and recovering the instance variables. My page class is > > OrderedCollection variableSubclass: #BTreePageOuter > instanceVariableNames: 'order count' > classVariableNames: '' > poolDictionaries: '' > category: 'BTree' > > db := KKDatabase onHost: 'localhost' port: 6543. > db root: (BTree withOrder: 2). > db root add: 'O1'. > db commit. > db logout. > > When I reconnect, instance variables order and count are set, but array (from the parent OrderedCollection) is nil. > > Jonathan. > PS... squeaksource was down, so I downloaded the packages from the spanish mirror into my local package cache, and loaded them from there. I'm running 1.3 on COG (that came from the 1.2.1 single click). > Package versions are: > GOODS-cds.137.mcz > GOODSCollections-cds.1.mcz > GOODSPlatformSqueak-cds.6.mcz >

Oct. 22, 2011