Squeak-dev March 1999

squeak-dev@lists.squeakfoundation.org

153 participants
338 discussions

re: external objects
by Craig Latta 30 Mar '99

30 Mar '99

Hi Marcel-- > I've looked at the code a little bit, and was intrigued by the > references to EXTERNAL objects, but not exactly sure how this works > or even if it really is what I think it is, a way of referencing user > settable Squeak objects from the outside. In fact, I would *love* > to see a small architectural overview if that is possible, because it > looks like there's a lot of really interesting stuff there. Sure... You can register any object in the "external objects array" with SystemDictionary>>registerExternalObject:. That method answers an integer. You can use that number to access the object in virtual machine code, because the exteral objects array is an element in the "special objects array". The special objects array is a global virtual machine variable called specialObjectsOop. I don't think there is generic external object access support in the virtual machine, but we could easily define a macro for it. It would be something like: #define externalObjectAt(index) longAt(((((char *) (longAt(((((char *) specialObjectsOop)) + 4) + (38 << 2))))) + 4) + ((index - 1) << 2)); All of the external objects in my networking framework happen to be semaphores. There's already a convenient mechanism for signalling such semaphores from the virtual machine: synchronizedSignalSemaphoreWithIndex(). The parameter is the external object index of the desired semaphore. When you're finished using an external object, you unregister it with SystemDictionary>>unregisterExternalObject:. In my networking framework, I do this when external resources are closed. It would be nicer if the unregistering happened via finalization. There's no major penalty for not unregistering, just a slow memory leak. Even nicer than using "external objects" at all would be fixed-address objects. That's what I used in my previous implementation of the networking framework. The basic issue here is that the addresses of Smalltalk objects being used by C libraries shouldn't change (e.g., by the garbage collector). I hope that makes sense, -C -- Craig Latta composer and computer scientist craig.latta(a)netjam.org www.netjam.org latta(a)interval.com Smalltalkers do: [:it | All with: Class, (And love: it)]

1 0

Re: [BUG] Showing a large file in File List (.changes for example)
by Bob Arning 29 Mar '99

29 Mar '99

Reiner, My guess would be the amount of free memory needed to load the entire file into the PluggableTextController. A little experimentation (in 2.3 MVC) on a 6meg change file showed that I needed about 31meg free memory to read the whole thing. After it was successfully loaded, the memory used dropped to 25meg and afer positioning the insertion point somewhere in the text, usage dropped to around 19meg. This final value (3 times the size of the changes file) seems to be the minimum as the PluggableTextController held a copy of the string in "initialText", the Paragraph also held a copy (actually several copies up to the time I made a selection) and the Paragraph requires an Array of TextLineInterval objects at one per line. This last is an overhead of 40+ bytes per line and in my case consumed more memory that a single copy of the string itself. So, I would recommend increasing Squeak memory so that you have 18 to 20 meg free and try again. Cheers, Bob On Mon, 29 Mar 1999 00:24:35 +0200 "Reinier van Loon" <R.L.J.M.W.van.Loon(a)inter.nl.net> wrote: >Does anyone know why Squeak2.3 has a problem showing large files in a File >List? > >I have no problem loading (3-4 seconds for 3.5 Mb) the .changes file in a >workspace of Visual Smalltalk but when I try to view the file in File List >(Squeak) everything comes to a halt.

1 0

[BUG] Showing a large file in File List (.changes for example)
by Reinier van Loon 29 Mar '99

29 Mar '99

Does anyone know why Squeak2.3 has a problem showing large files in a File List? I have no problem loading (3-4 seconds for 3.5 Mb) the .changes file in a workspace of Visual Smalltalk but when I try to view the file in File List (Squeak) everything comes to a halt. I haven't looked into it yet, but if anyone knows something... Reinier.

2 1

Squeak mentioned on 32bitsonline
by "Gran" Hultgren 29 Mar '99

29 Mar '99

Just saw this little blurb: http://www.32bitsonline.com/article.php3?file=issues/199903/lj-19990325&pag… Very nice to see that the Linux community is taking note of Squeak. regards, Gran === Gran Hultgren, gohu(a)rocketmail.com, icq#:6136722 GSM: +46 709 472152, http://195.22.65.4 "First they ignore you. Then they laugh at you. Than they fight you. Then you win." -- Gandhi _________________________________________________________ DO YOU YAHOO!? Get your free @yahoo.com address at http://mail.yahoo.com

1 0

Async primitives?
by Duane Maxwell 28 Mar '99

28 Mar '99

All - Since I'm new here, I'm not sure if this has been discussed, so please bear with me. I've been giving some thought to the external primitive mechanism and was wondering if any thought had been given to somehow integrating some of the primitive support with the Smalltalk process model and native OS threading. At the moment, calling an external primitive occurs in the same native thread as the rest of VM, which means everything grinds to a halt until the primitive is complete. The question is this - is it possible to add another parameter to the <primitive> call which tells the VM to attempt to run the (thread-safe) primitive in another native thread, and suspend the current Smalltalk Process, then resume it when the primitive completes. An alternative is to provide VM callbacks that allow the plug-in developer to initiate that behavior in sections that aren't otherwise requiring support of the VM. I expect there will be at some point be primitives that are time-consuming for some reason, but aren't critical enough to wait for. An example might be some sort of compression or encryption which might be difficult to factor for a polling model. =================================================== Duane Maxwell dmaxwell (at) launchpados.com CTO http://www.launchpados.com Launchpad, Inc. (619) 578-8500 x226 Information contained herein is my personal opinion and not necessarily that of Launchpad, Inc. ===================================================

2 1

Background on that crazy dot idea
by Marcel Weiher 28 Mar '99

28 Mar '99

Wow, my crazy little idea generated quite a bit of discussion. The funny thing is that I didn't really consider the 'dot' part all that crazy, in fact it is actually not new to Smalltalk. GemStone uses it, and my post was actually triggered when I found out that ThingLab uses it to enable change management. My central idea was that messaging consists of naming the participants and then sending the messages, and that naming is really just a special case of selection (compare 'Design Principles Behind Smalltalk'). Generalizing naming to selection by allowing 'complex names' would not really be a change to Smalltalk, but more the lifting of a current restriction on what can be accessed by name. The idea is definitely not just to provide convenient shortcuts for certain message sends, although that is a nice effect. See 'Adaptive Programming' (http://www.ccs.neu.edu/research/demeter) for a more thorough treatment of the problems of graph traversal and an interesting solution. Having a name for 'the contents of this collection' or other dynamic entities is something that appeals greatly to me, especially if the mechanism is simple and can be the core for quite a few other applications. However, this is really a blue-plane idea, and definitely something that needs to be thought about a little more, although as Stefan demonstrated, a simple version isn't really that hard to add. A more comprehensive approach that also takes name-spaces into account would be to treat (complex) names as entities that are resolved recursively in the dynamic execution context, with only simple + local names being resolved at compile time. Marcel

1 0

[BUG] - MVC - Selector Finder (2.3+)
by Norton, Chris 28 Mar '99

28 Mar '99

Hi Folks. I have to thank Maurice Rabb for getting me to play with the Selector Finder. Apparently, the Morphic Selector Finder is working as designed. If I type a selector name, like "printOn:" in the upper left hand box and then type <Enter>, the lower pane fills up with selectors whose name contains this substring. The MVC Selector finder does not do this. I'm not sure if this is the only way to use the tool, as I just recently started playing around with it. I can't seem to figure out how to use the "find...", "find again" or "set search string" popup items. They don't appear to do anything in this dialog. In any event, the two dialogs do not seem to work the same, so I'm calling this a bug. Thanks. ---==> Chris

3 2

RE: VM cleanup
by Raab, Andreas 28 Mar '99

28 Mar '99

> I think a naming convention must address the following concerns : > > 1) At the moment, it seems all named primitive method names must be unique > across all classes, with the possibility of collisions higher than the > normal case with Smalltalk. I'm not sure if there is a collision between > names across modules. > At the moment there isn't - the function looked up is only searched in the given module. There is, however, a potential collision problem when stuff gets simply put into the VM itself and not into separate modules but I don't consider this to be a really big problem. The current naming convention for most of the primitives is 'primitiveSubsystemAction' (e.g., primitiveMIDIPortOpen, primitiveAsyncFileClose, primitiveSocketError) and to me this looks pretty sufficient. > This is a side effect of the C translation - there's > really no reason why the class name can't be prepended to the method name. > The class name can of course be prepended to the primitive name - it may however be confusing at certain points since the same primitive may be invoked from different classes. And ... it's more stuff to type ;-> > 2) There should be a way to handle the named primitive call where some VMs > have them external, and others internal, with the Smalltalk code glue > indifferent. As I understand the Mac code, the search is done in either > one or the other depending upon the call. It seems like a minor fix to > have the function searched in the supplied module name, then in the VM on > failure. This would also allow the overriding of malfunctioning or > "in-the-process-of-enhancement" primitives. Currently, moving the Balloon > classes to an external module requires modification all of the primitive > calls. > There are two problems with the approach: #1: If you have a brand-new VM and an older plugin you would actually use the old stuff rather than the newer, VM-builtin versions. Given that plugins may not only be in the image or VM directory this could cause a lot of *very* confusing problems. #2: Assume your VM contains an older Plugin version. You start your image and notice that something goes wrong (the ST code requires the newer version). If you then get the newest plugin and drop it into your VM directory you'll mixup calls to outdated internal primitives (basically all the ones you did before you dropped the Plugin in there) with newer external ones. If (such as in the balloon case) these primitives require certain behavior of a set of primitives you may get a really nice crash here. Later on, you may have an updated VM which now contains the newest version and you'll get the same problem just the other way around (e.g., using older external versions and newer internal versions). > 3) Mac people in particular might be turned off to the Squeak directory > containing piles of plug-ins - there should also be a > (platform-dependent?) > convention for collecting plug-ins that will be searched for. > On Unix/Win32 there are conventions for finding these DSOs/DLLs even if they are not in your Squeak directory - which is part of the above mentioned problems ;-( Andreas -- PLEASE NOTE NEW EMAIL ADRESS! +===== Andreas Raab ========= (andreasr(a)wdi.disney.com) ==+ | Walt Disney Imagineering Phone: +1 818 544 5016 I I Glendale, CA Fax: +1 818 544 4544 I +======< http://isgwww.cs.uni-magdeburg.de/~raab >========+

2 1

Messaging vs. Methoding
by agree＠carltonfields.com 27 Mar '99

27 Mar '99

I have found the discussions quite informative, and have surprised myself as I find me leaning away from changes to syntax and convention of Smalltalk-80 in favor of less invasive solutions. To that end, I found myself wishing for a better solution, but was unable to find it. I would appreciate the advice of Smalltalkers with more experience than I. In building a module, I find myself wishing to provide broad functionality with a range of options, but also desiring to have the morass of options mostly hidden by use of highly common defaults. For the simplest example, consider this message from the rePlugin package: aString reMatch: pattern opt: oString from: begin to: end This directs a search of the receiver for matches of patterns from the specified substring indices, using certain options specified by oString. Assume that this full functionality is all warranted and necessary. The problem is that users most frequently will call this by using: aString reMatch: pattern opt: ' ' from: 1 to: (aString size) tending to make us want to provide: aString reMatch: pattern with defaults for the opt:, from: and to: portions of the message. On the other hand, the other calls are not infrequent, leading us to want to provide message for EVERY combination, in this case a total of 8 messages. Worse, every new degree of freedom requires another doubling of the message count. In many cases, each of these messages simply call the "most heavily specified" version, using the appropriate defaults. This leads to lots of messages, and IMHO, aesthetically too many. Practically, managing new factorings becomes unwieldy and error-prone. In languages such as Python, methods can be defined to have a plurality of default keywords, such as: reMatch(self,pattern,opt='',from=1,to=self.size()) which has the effect of the combinatorial expansion of independent messages, but the virtue of having the code handled all in one place, and requires no "pure glue" messages. (Actually, I'm not sure that the default for to can be specified in this way. You may need to default to nil and put the logic in the preamble of the method.) I think this is a good thing. I note that from a messaging standpoint, this approach requires no change in syntax (as would the Python keyword parameters, for example). I presume that the reason this isn't done in Smalltalk is that dynamic lookups of methods would be complicated and slowed thereby. Still, it seems to me that the a method lookup process needn't be impeded by default keywords -- the compiler can still generate each method selector independently, having them all point to the same code, or perhaps code that performs the "glue assignments," followed by a message to the "fully specified" procedure. Of course, browsers would need to know that removing or modifying such a message would require removal or modification of ALL of the "derivative" methods, but that doesn't seem too complex. It would be "nice," IMHO, to be able to specify every one of the messages above with a single message (from a browsing perspective) having a declaration, perhaps, of the form: reMatch: pattern opt: oString := ' ' from: begin := 1 to: to := (self size) or equivalent (and perhaps better-looking) syntax? I presume that this has all been considered and weighed before. What am I missing? Should I have better factored the design? Is the keyword defaults approach practical (or even desirable) in Squeak? Would it be straightforward to implement? How would you go about it?

5 4

Complex...
by SLAWICK CHARLIER 27 Mar '99

27 Mar '99

Hi, I've continue to build my own complex class (as so said, it's for my own education so even if a such things already exist, I will continue this work anyway...I've surch on the UIUC server but did not find a such class, are you sure there is one there?) I've just a little problem to optimize my code: I've created a methode i in the Number class, so 5 i is sending a complex with a real value egal to 0. But when writing 2 + 5 i , it send a complex as parameter of +. Do I need to change the + methode (and -, /, *...) for the number class and for EACH of its subclasses or is there a way for me to avoid that dirty work? Exept for that pb, the latest version is working well and allow you to do much of the basic operetions you can do with a a+bi form. Next step will be to implement a exp. form (c*ecp(iT))...a hard week-end to come... Slawick, university of Troyes (France) PS:I don't really know what I am doing on this ML. All of you seam to be hard programmer, all searchers in lab. all around the world... just at higher level of knowledge than me!

1 0

← Newer
1
2
3
4
5
6
7
...
34
Older →

Jump to page: