- Pharo-dev - lists.pharo.org

Re: [Pharo-project] [squeak-dev] FileSystem (was: About (backwards) Compatibility)

by Yanni Chiu

On 13/12/12 10:24 PM, Chris Muller wrote: > > Igor, a very good explanation for our individual views here is that > your journey is building a great new Smalltalk, mine is something > else. I don't see that as a good explanation. I see different paths to similar goals. The Igor/Pharo path is to first improve the Smalltalk, and from that platform, scale even greater heights than could have been imagined from the "random walk" along "safe" path.

Dec. 14, 2012

Re: [Pharo-project] [squeak-dev] FileSystem (was: About (backwards) Compatibility)

by Chris Muller

>> Yes, of course those capabilities could be added simply with factoring >> work and little disruption to the API. That's the promise object >> technology, encapsulation. > > The main problem with FileDirectory that there's no API. > I can hardly call a single class as API... > as to me, it is clear that file system domain is a bit more complex > and in order to reflect it properly in a language you will need more > that a single role (FileDirectory). FileDirectory is actually made up of 10 classes providing a rich but utilitarian API for traversal, navigation, stream creation, copying and renaming, filename parsing utilities. Platform-specific implementations all share the same API, including one for remote FTP access. For you to say "no API" is silly. > So, i really doubt that you end up with 'little disruption to API'. It > will be substantial. > > To me, design comes first. ALWAYS. > All of the rest is secondary. Design should be a means to some end, not an end in itself. You can design for 20 years but if real-world results are never harvested then it was only ever just playing around with a cool toy. My goal is to get as far along my software journey as my limited lifespan will allow.. > FileSystem establishing a clear > separation between conceptual model > ... snip ... > won't require serious changes and will serve you much much longer > without need to change anything in userland code. (Take a Collection > protocols for example). .. and so when I become faced with having to make "serious changes" in my "userland code," (e.g., a complex ODBMS system) just to hold position along my journey; naturally I'm resistant. You can try to justify it with design theory but, at the end of the day, the two frameworks do basically the same thing, converting to it is mostly pure cost. > Also, once you have such set of rules, it is much easier for people to > maintain implementation code, > as well as adding extra functionality if needed. > > Clearly, VM interfaces (such as primitives) should be not exposed to > ... Igor, a very good explanation for our individual views here is that your journey is building a great new Smalltalk, mine is something else. Cheers, Chris

Dec. 14, 2012

Re: [Pharo-project] How to load Phoseydon and Neptuno?

by Mariano Martinez Peck

On Thu, Dec 13, 2012 at 6:40 PM, Torsten Bergmann <astares(a)gmx.de> wrote: > Are there plans for SqueakDBX regarding porting it to Pharo 2.0 > or moving it from FFI to NativeBoost? > > It would be a nice idea. And not hard to do since the functions of OpenDBX are quite easy to wrap: easy types and no need of callbacks. Only one external structure. Unfortunately, right now my plate is full. If someone wants to take a look, I can provide more details on how that should be done. Cheers, -- Mariano http://marianopeck.wordpress.com

Dec. 14, 2012

Re: [Pharo-project] Helping the noobs help out - i.e. fixing the debugger

by Michael Roberts

Indeed we spent some time in Edinburgh looking at it :-) that was too long ago :-( The problem i see with the original debugger inherited from Squeak, in the Pharo context, is that it is very sensitive to a lot of the core code in the image. What this means is that the accelerated changes in Pharo code base had unintended side effects on original debugger machinery as it diverged from its ancestry. It goes all the way back to pushing Eliot's closure implementation in 1.0 which we were desperate and excited for. Since we relied on taking core compiler/closure/debugger code from Squeak it became more and more important to track the difference between squeak and Pharo. Anyone who has tried looking at the diffs will know how hard that is. I forget which class it was but we found this obscure bug in one of the collection classes IIRC that threw one tiny but annoying aspect of debugger highlighting off. I only found that by single stepping both images through known code snippets with this debugger 'oscilloscope' I had hacked up for the purpose. What that experience showed was it was really involved how the instruction machinery hangs together. As squeak trunk is where most fixes get pushed in this area it requires huge resources and diligence to track every change to see if Pharo needs it. For ages Stef would post every interesting looking trunk change to the pharo bug tracker but there were not many folks looking at them all. And also it is not nice work. We didn't have the tools or modularity to cherry pick changes in this area. So the new debugger model in Glamorous showed an interesting direction to go in and this comment from Marcus is also interesting on seemingly building a new debugger architecture which we have discussed before. I was trying to do was to figure out a way you could regression test the debugger by recording and replaying examples of it's operation and checking each release it hadn't been broken unexpectedly. I think there is still mileage in that area if it has not been done already. Also, historically, I am not convinced it was ever properly working in the sapphire build or even in 3.9. The bugs and effects were so subtle that you just got used to working around them. I.e. I have 20 mins to do some coding do I add a bit more to my cool seaside app or do I struggle with fixing the debugger? Last I looked at squeak trunk it was looking pretty good. But the code base is hard to track. What I was last thinking about in this area was trying to live 'trace' in some way all the code required by the debugger into a filed out and renamed set of classes like :SqDebugger SqArray SqCompiler SqInstructionStream and so on and then load them into Pharo. The idea being you would have an identical implementation that you would use to operate on all the Pharo code but entirely independent from it and maintained in squeak trunk. It is an unrealistic idea but the example i was thinking about from the electronic world is using one oscilloscope to test or observe the internals of another. You could do that image to image over the network of course but I am not sure if you just vary the complexity in a different direction. Anyway just my 2p, I care a lot I about the tooling and look forward to seeing what comes out! Cheers, Mike (a bit absent, but still enjoying the progress) On 13 Dec 2012, at 09:40, Igor Stasenko <siguctua(a)gmail.com> wrote: > On 13 December 2012 10:30, StÃ©phane Ducasse <stephane.ducasse(a)inria.fr> wrote: >> Now adrian I imagine that you saw that people worked on this bug and this is a rather complex one. >> So I would suggest to you to avoid to draw conclusions too fast. >> > yes, if i remember, we tried to approach it at least once.. > but unfortunately it requires a lot deeper knowledge about bytecode > and (de)optimizations to fix highlight. > >> Stef >> >> On Dec 12, 2012, at 11:28 PM, adrians wrote: >> >>> >>> Well, it just seems that I'm whining here instead of contributing, but if >>> the debugger is indeed a very (if not the most) useful tool which is used in >>> pretty much every bit of fix-up work, fixing it if it is broken would have >>> to come before all else, no? Otherwise, any work that needs to be done until >>> it is looked at is just compounded. >>> >>> Maybe it would make sense to correct (in the current code) to a small degree >>> just one of the things that is currently broken. No frills, but just getting >>> a more accurate indication of where the PC is - at least have it on the >>> correct line of code if it can't be pinpointed more accurately without too >>> much hassle. I'm curious what derails the location highlighting as half of >>> the time it is close, if not correct. Is it blocks that give it a hard time? >>> It wouldn't be too bad if other things were going to take a while if you >>> could at least keep track of where you were. >>> >>> If the idea is to wait until a whole big refactoring can be done on this, I >>> fear a fully debugger might be another year away. >>> >>> -- Adrian >>> >>> >>> >>> -- >>> View this message in context: http://forum.world.st/Helping-the-noobs-help-out-i-e-fixing-the-debugger-tp… >>> Sent from the Pharo Smalltalk mailing list archive at Nabble.com. >>> >> >> > > > > -- > Best regards, > Igor Stasenko. >

Dec. 13, 2012

[Pharo-project] How to load Phoseydon and Neptuno?

by Torsten Bergmann

Are there plans for SqueakDBX regarding porting it to Pharo 2.0 or moving it from FFI to NativeBoost?

Dec. 13, 2012

Re: [Pharo-project] another kid born: NBExternalEnumerationType

by Ciprian Teodorov

On Thu, Dec 13, 2012 at 8:47 AM, Igor Stasenko <siguctua(a)gmail.com> wrote: > On 13 December 2012 08:16, Ciprian Teodorov <ciprian.teodorov(a)gmail.com> > wrote: > > > > > > On Thu, Dec 13, 2012 at 3:11 AM, Igor Stasenko <siguctua(a)gmail.com> > wrote: > >> > >> On 13 December 2012 03:08, Igor Stasenko <siguctua(a)gmail.com> wrote: > >> > On 11 December 2012 20:33, Ciprian Teodorov < > ciprian.teodorov(a)gmail.com> > >> > wrote: > >> >> Hi guys, > >> >> > >> >> today I have commited to the NativeBoost repository a new core class, > >> >> the > >> >> NBExternalEnumerationType, > >> >> which reifies the enum declarations in C and offers a nice interface > >> >> (especially in terms of debug/inspect). > >> >> > >> >> To use it just subclass the NBExternalEnumeration and add an > #enumDecl > >> >> method to the class side such as: > >> >> > >> >> NBExternalEnumeration subclass: #NBTestEnumeration > >> >> instanceVariableNames: '' > >> >> classVariableNames: '' > >> >> poolDictionaries: '' > >> >> category: 'NativeBoost-Tests' > >> >> > >> >> NBTestEnumeration class>>enumDecl > >> >> ^ {(#AAA -> 1). > >> >> (#BBB -> 2). > >> >> (#CCC -> 3). > >> >> (#DDD -> 2400)} asDictionary > >> >> > >> > just one comment: imo, it would be better to not force users to > >> > remember that they should use > >> > complicated syntax in this method. > >> > > >> > i propose following syntax: > >> > > >> > NBTestEnumeration class>>enumDecl > >> > ^ #( > >> > #AAA 1 > >> > #BBB -> 2 > >> > #CCC -> 3 > >> > #DDD -> 2400 > >> > ) > >> > > >> oops, of course it should be just pairs, e.g.: > >> #( A 1 B 2 C 3) > >> Yep this is a good idea for most cases so I'll integrate it. But there > are > >> those cases where your enum has many items (more than the number of > literals > >> we can have in a method), and I think in those cases having the user > give us > >> a prebuilt dictionary is better. > > and where you see many literals? > #( ..... no matter how many stuff here ... ) > is 1 (single) literal! > .. > (i wonder, why people keep confusing literals in method body versus > stuff in a single literal array :) > > And you don't have to trust my word, you can check it by yourself, > just evaluate: > > | x | > > x := #(A B C). > ^ thisContext method literals size > > versus: > > | x | > x := #(A B C D E F G H I J K L M N). > ^ thisContext method literals size > > > My bet, the other day I was trying to do something like x := {#A. #B. #C. #D .......} which did not work, so I wrongly supposed we will have the same issue with #(A B C D). > >> > >> > then you can have > >> > > >> > NBTestEnumeration class>>initializeEnum > >> > > >> > dict := Dictionary newFromPairs: self enumDecl > >> > ..... > >> > > >> > then #initialize should just do 'self initializeEnum' > >> > > >> > > >> > > >> >> DO NOT FORGET to call the #initialize method on your class. The rest > is > >> >> automatically done ;) > >> >> > >> >> You can use your new enum in two ways: > >> >> > >> >> add it to a client class poolDictionaries list (see > >> >> #NBExternalEnumTests > >> >> for an example), and then just write CCC in your code -- CCC here is > an > >> >> item > >> >> of your enum > >> >> or you can send the name of an item to your class --- > NBTestEnumeration > >> >> DDD > >> >> > >> >> The NBExternalEnumeration implements (instance and class) some more > >> >> API > >> >> methods like: > >> >> > >> >> #itemAt: retrieves the item having a specific value --- > >> >> NBTestEnumeration > >> >> itemAt: 2 > >> >> #includes: checks the existence of a specific item in the enum --- > >> >> NBTestEnumeration includes: #AAA > >> >> > >> >> Having this already gives us a nice interface to working with enums. > >> >> Now I > >> >> plan to improve a little bit on performance and usage scenarios: > >> >> > >> >> the possibility to accept smalltalk integers as valid input/output > for > >> >> functions(or callbacks) that use enums as arguments/return. > >> >> support the different representation types for the enum values > >> >> (int/uint) > >> >> > >> > > >> > to my thinking, it not worth implementing separate type converter for > >> > enums.. > >> > just put a note that if people would like to deal nicely with enums > >> > returned by function(s), > >> > they should use a converter , something like that: > >> > > >> > enumValue := MyEnumeration fromInteger: (self > >> > callSomeFunctionReturningInt) > > > > This is a good idea too. > >> > >> > > >> > then this method can implement converting from int->enum object, > >> > as well as do a range checking etc. > >> > in same way, if people want to use 'enumValue' for passing it as > >> > parameter to external function, > >> > they should not pass it directly , but use 'enumValue value' > >> > > >> > In general, it would be cool to have good 'enum type' converter.. > >> > (despite that i will never use it, > >> > because to my taste its not really worth it, and i can live with > >> > integers and in most cases you hide such low-level details from > >> > end-user anyways ;)) > >> > but for those, who want more clarity and better structure in their > >> > code, of course, they can use it . > >> > > > > > I do not completely agree with you on this one (but you already now it). > The > > thing is that if we use integers then the user will have to constantly > ask > > which enum value should I pass here... > > The idea is to not expose enum values to user at all. > Sometimes it easily doable, sometimes not.. it depends , of course.. > But ideally, there should be only messages. > > > If you want that you can always have > > it by declaring your function signature as #(void fct(int anEnumValue)) > and > > then call it with MyEnum DDD value (that's fine). Having an enum type > > converter is very good for documenting the argument types (and > automatically > > checking them). > > True. But everything has its cost: in this case - performance. > > > So I think this should be the user choice. Moreover, I'm for > > incremental development (and learning) so the way I see it is: When you > > first implement the ffi wrapper just try to be as precise as you can be, > > then after your library binding is done you can change some declarations > to > > use more 'low level types' (the underlying representation) for speed > reasons > > mainly (if it ever comes to that). > > > Yes. > > > But at the end I think we should give our users options, which will help > > them make decisions based on their particular needs, not based on our > "gut" > > feeling ;) > > > I agree that this will be nice addition: for better coverage alien (C) > data structures. > I am totally with you on that. > > I just wanted to say that i never thought about it as a necessary part > (simply because > i can live without it). > > -- > Best regards, > Igor Stasenko. > > -- Dr. Ciprian TEODOROV IngÃ©nieur DÃ©veloppement CAO tÃ©l : 06 08 54 73 48 mail : ciprian.teodorov(a)gmail.com www.teodorov.ro

Dec. 13, 2012

Re: [Pharo-project] Helping the noobs help out - i.e. fixing the debugger

by Marcus Denker

On Thu, Dec 13, 2012 at 10:40 AM, Igor Stasenko <siguctua(a)gmail.com> wrote: > On 13 December 2012 10:30, StÃ©phane Ducasse <stephane.ducasse(a)inria.fr> wrote: >> Now adrian I imagine that you saw that people worked on this bug and this is a rather complex one. >> So I would suggest to you to avoid to draw conclusions too fast. >> > yes, if i remember, we tried to approach it at least once.. > but unfortunately it requires a lot deeper knowledge about bytecode > and (de)optimizations to fix highlight. > I looked at it a bit on the train... With the new infrastructure, we create an intermediate representation which is (nearly) a 1:1 mapping to bytecode (some are mapped to multiple bytecode, multiple IR nodes can result in just one BC), but mostly: 1:1 mapping. The nice thing is that the IRnode records the offset of the bytecode when bytecode is emitted. And the IR knows the AST Node it was created from. And the AST node knows everything (down to the Tokens and the offset in the source text). after adding #instructionForPC: to IRMethod (which just searches for the node), the following code just works: | method astNode| method := Integer>>#+. astNode := (method parseTree ir instructionForPC: 58) sourceNode. method sourceCode copyFrom: astNode debugHighlightStart to: astNode debugHighlightStop returns 'ifTrue: [^ (self digitAdd: aNumber) normalize] ifFalse: [^ self digitSubtract: aNumber]' This needs to be tested and enhanced a bit (e.g. for the case that one IR created multiple bytecodes, e.g. block pre-amble and postscript...). But it should not be that hard to do, I think. The good thing is that it does not rely on some table that is created somewhere, but instead you have the data (IR and AST) directly there to look at. Tweaking special cases should be quite easy with that. Marcus -- Marcus Denker -- denker(a)acm.org http://www.marcusdenker.de

Dec. 13, 2012

Re: [Pharo-project] About (backwards) Compatibility

by StÃ©phane Ducasse

> > I hate to keep repeating myself, but the Pharo manifesto is very clear, and > makes these types of arguments moot: > - Better for the better > - Beauty to learn from > - Not backward compatible > - Clean, lean and fast And we pay attention about change we will not change filesystem every release or announcement. Stef

Dec. 13, 2012

Re: [Pharo-project] who is "dfgs"?

by Camillo Bruni

haha, so these are commits from the pre-full-name aera. Anyway, dfgs was a student in bern ;), but since I put the code in the system I should be the owner ;). VTerm*Driver abstract over the terminal, and yes they are crucial for any decent terminal support. - provide colors for terminal - provide cursor navigation - provide terminal capability access due to lack of NativeBoost in the main image, so far I could not use tcap, but that will change. So far I did not port the input handler (but there is a fully fledged smalltalk-based readline implementation ready). Once that is in the image there should be no more undefined calls. And using #perform: is a very bad habit in my opinion. It breaks all tools, since you make everything implicit (and quite slow as well).. On 2012-12-11, at 09:05, Esteban Lorenzano <estebanlm(a)gmail.com> wrote: > ... and please, please, please, use real names, not initials when committing packages! > > Esteban > > On Dec 11, 2012, at 11:33 AM, Pavel Krivanek <pavel.krivanek(a)gmail.com> wrote: > >> Hi, >> >> I would like to ask who has the initials "dfgs". And I have a question for him/her. The VTermInputDriver initialization methods generate a lot of unimplemented calls because they expect a handler object with protocol that is not defined in the image. Do we need the VTerm classes at all? Can you provide an abstract handler class? Or what about to switch this calls to perform:/perform:with: form. >> >> Cheers, >> -- Pavel >> >> > >

Dec. 13, 2012

Re: [Pharo-project] Zinc-Resource-Meta-[Core|Tests]

by StÃ©phane Ducasse

I added the entry. stef

Dec. 13, 2012