Squeak-dev May 2003

squeak-dev@lists.squeakfoundation.org

204 participants
710 discussions

Squeak and http proxy
by Giovanni Giorgi 02 May '03

02 May '03

Hi all, I'd like to use squeak behind a proxy. I need to forward all my http requests to a proxy server (with a ip address, a port and a login/password). There is something to deal with a proxy in the base image? I have found the class InternetConfiguration but it seems unused and has no comments. Someone can help me?! -- [ [ [ JJ ] ] ] | Temi cio' che conosci, non | cio' che non conosci http://www.siforge.org |

3 4

RE: [ANN]Draft rough plan for 3.6!
by Magistrello Alejandro (SFA) 02 May '03

02 May '03

Hi all, GamesTests is now at SqueakMap http://map2.squeakfoundation.org/sm/package/eb813272-250f-43a3-9ee8-c28b5e1e f565 regards, Alejandro -----Mensaje original----- De: Marcus Denker [SMTP:marcus@ira.uka.de] Enviado el: martes 15 de abril de 2003 11:05 Para: The general-purpose Squeak developers list Asunto: Re: [ANN]Draft rough plan for 3.6! On Tue, Apr 15, 2003 at 01:16:28PM +0100, goran.hultgren(a)bluefish.se wrote: > Hi Marcus and all! > > Marcus, what are your thoughts on how we should verify the (removed) > package installations? > You have been working with some of them and also with tests so it would > be interesting to hear. > > Do you think there is a point in creating a test-package for each that > at least does *some* kind of test - no matter how simple? > Yes, there are two reasons: 1. This will be a nice starting point for adding more tests. 2. even a simple test will catch some trivial errors... I think for many packages we could come up with some pretty simple tests. lets see... we have the folowing removals: BaseImage tests --> Ok. that's easy ;-) Celeste: we have a test that opens Celeste on a nil database and presses all buttons. not bad as a starter. Games: just start up all games. Shouldnt give an error. The only game were we have a real good test for is FreeCell: (freeCell _ FreeCell new) openInWorld. freeCell board pickGame: 1 VM generation: we could do the stuff that the macroBechmarks are doing: generating the Interp.c. Then maybe opening/closing of vmMaker. Macro benchmarks: just run them PWS: That's difficult. But nobody uses pws anyways. SUnit: comes with tests, these need to be moves into it's own package. Scamper: just open it: It renders the "about" page, so the whole rendering need to work for this. Speech: Run the demos (Possible with no sound-output ;-) ). Balloon3D: don't know. Maybe we could simply open the 3dExplorermorph. These are really only meant as first tests. In the future, I hope people will add more tests (and refactore the code to make this possible). But we need to start somewhere... Marcus -- Marcus Denker marcus(a)ira.uka.de -- Squeak! http://squeak.de

1 0

Re: [Q] Fiddling with system innards, was [Q] Newbie - Multiple Return Values
by Richard A. O'Keefe 02 May '03

02 May '03

There are at least three ways to return multiple values from a method, and one of them is not obvious. (1) Return an Array. SomeClass>>multivalueMethod ... ^{x. y. z} This made a lot of sense when Squeak let you do {a. b. c} := foo multivalueMethod. (2) Return an object. SomeClass>>multivalueMethod ... ^MvResult with: x with: y with: z This is accessed like r := foo multivalueMethod. a := r x. b := r y. c := r z. Plus side: the caller has to retrieve only what the caller is interested in. Minus side: if you add more information to the result, the caller may never find out that it needs updating. This is usually the best thing to do IF you can come up with some other things the new class can do. Can it be returned from another method? Is there result processing code that could be moved into it? (3) Do the Scheme thing, use a block. SomeClass>>multivalueMethod: resultBlock ... ^resultBlock value: x value: y value: z This is accessed like foo multivalueMethod: [:x :y :z | a := x. b := y. c := z]. The thing I like about this approach is that while there still _is_ an object being created (the new Block) it doesn't _feel_ like an object. I'm not _doing_ anything to pick the result apart, I'm _given_ the result pieces.

2 1

Aggregated removal script for 3.6
by goran.hultgren＠bluefish.se 02 May '03

02 May '03

Hi all! Just hacked up a load script for performing all registered removals on SM - 10 of them (don't forget Balloon3D - it isn't in the correct category). Read more on http://map2.squeakfoundation.org/sm/packagebyname/aggregated Tested by firing up a vanilla 3.6a-5179, "open package loader" and then install the above package. Answer yes twice and then you have an image somewhat smaller. NOTE: I haven't tried to bring stuff back in yet... :-) But so far so good. regards, Göran

5 9

Re: [ANN] [ENH] FillInTheBlankMorph Look
by Daniel Vainsencher 02 May '03

02 May '03

You raise exactly the sorts of issues we're talking about. As another example, some process will be driven by different "clients" - sometimes with a user attached, sometimes guided by a script. So an oracle answering, for example, installation questions without a GUI is also useful. To sharpen the different issues we have - While Views separate display from the model (allowing configurable, and multiple views on the model (a one way effect)), and Controllers allow different reactions to user requests (one way, user initiated requests), this discussion is about requests for information initiated by the running process, which is not supposed to know who it is asking (user, install script or whatever). Since most interactions between UI and application model indeed start at the UI, which is naturally configured for the model, and not the other way, this has not been so urgent to address until now. Daniel Anthony Adachi <adachipro(a)yahoo.com> wrote: > > Stephane Ducasse wrote: > > > I would not do this change. I imagine that we should > find a much clever > > way of notifying user. Avi suggested to me that we > could have exception > > that would be caught and letting the used decide if > he wants to put a UI or not > > for headless image for examples. > > Tim Rowledge then wrote: > > > The key item to separate progress notification from > > needing the UI > > classes is to change the defaultAction to do > nothing > > rather than display > > the progress bar (well, modulo the stuff to make > sure > > the block runs > > etc). Then one has to find all the place where the > > exception could need > > to be handled (which is actually pretty tricky and > > tedious - tools > > would be good here) and handle it the right way. > > Avi Bryant wrote: > > >The only way to do it right is per process, and the > cleanest way to do > >that is probably to do exactly what Tim suggests, > which is to have a > >Notification subclass signalled for each type of user > interaction - > >#inform:, #confirm:, progress, etc. > > First, I should note up front that I'm new to Squeak > and don't know much about it's inner workings so > please feel free to ignore my opinion. > > Wasn't the concepts behind MVC suppose to be about the > separation of the interface and the model? By the way, > I realize that Morphic isn't MVC (I'm speaking of the > motivations initially behind Smalltalk's MVC at a > conceptual level). Also, I have read somewhere that > MVC originally had 4 elements Model, View, Controller > and User. Of which User being the consideration of > whom is the end user and their needs. > > Many people seem to think of View and Controller being > some sort of GUI elements but originally was not this > concept meant to be whatever mechanisms were > appropriate to the end user? > > For instance, people with handicaps may have different > needs in how they control or view the computer. A > quadriplegic cannot control the computer with a mouse > or keyboard and a person with a visual disability will > have difficulties with a "View" which is a Graphical > UI. > > Also, what if someone is perfectly able visually but > doesn't need a GUI? Like perhaps if they desired to > run Squeak as a server? In such cases perhaps the > "view" they would want would be a record of some sort > on how long certain time consuming operations took. > Perhaps, in order to keep tabs on performance or to > monitor regular occurring events such as how long > periodic backups took. e.g.-perhaps in the form of > some kind of log file (either saved inside the image, > on disk or to be part of a summary of notable > activities emailed out on a regular basis)? They may > not even wish to have Progress dialog windows > popping-up at all. > > Again, what about people who are blind? Obviously, a > GUI progress bar is rather useless to them and would > be in need of some other form of progress notification > other than visual. i.e.- a different form of "View". > > I think the key words here are notification ... of > progress. > > Maybe messages intended for View like things should be > sent to a collection of Observers? Observers of > pertinence who can be attached or removed from the > notification mechanism. (i.e.- as in GoF Observer > pattern or similar). Observers who respond in a way > appropriate to the needs of the end User. > > Just a thought which came to mind. > > Take care, > > Anthony > > __________________________________ > Do you Yahoo!? > The New Yahoo! Search - Faster. Easier. Bingo. > http://search.yahoo.com

1 0

Re: [Q] Version Control with Squeak was [Q] Fiddling with system innards
by Daniel Vainsencher 02 May '03

02 May '03

BTW, the Swiki has lots of Squeak usage information. If you've looked at that and didn't find these kinds of questions answered, we want to know. There's a documentation team listening at the mailing list, and they try to make the swiki (more) perfect. Anyway, questions are welcome here, and even more welcome after one's looked at the Swiki... ;-) Daniel Anthony Adachi <adachipro(a)yahoo.com> wrote: > Avi Bryant wrote: > > > For more info on one way this gets done, see: > > http://beta4.com/squeak/aubergines/docs/dvs.html > > Thanks for the link, > > Anthony > > __________________________________ > Do you Yahoo!? > The New Yahoo! Search - Faster. Easier. Bingo. > http://search.yahoo.com

1 0

Newbie
by Tony Fugere 02 May '03

02 May '03

I am new to Squeak. I have created a little project, and I would like to save the source code for it. I know you can save the project, but the file is in binary. I would like to be able to share SmallTalk source code for a project. What am I missing? Thanks. Tony Fugere

2 1

Re: [Q] ProgressMorph cleanup
by Daniel Vainsencher 02 May '03

02 May '03

Sounds good to me. Daniel diegogomezdeck(a)consultar.com wrote: > Hi, > > Waiting for the changes for progress notification I'll start to change the > look of ProgressMorph. > > Reading the code I found thet ProgressMorph is using FlashProgressMorph. > > I propose to create a ProgressBarMorph (using the code on > FlashProgressMorph). > > Then I can change FlashProgressMorph to be a subclass of ProgressBarMorph > or just delete FlashProgressMorph (and use ProgressBarMorph). > > I'll produce 2 changesets: one with the refactoring I'm talking about and > the second one with the look changes. > > Do you agree? > > Diego

1 0

Re: [Q] Fiddling with system innards, was [Q] Newbie - Multiple Return Values
by Daniel Vainsencher 02 May '03

02 May '03

Building on what Cees said - the methods you will write initially will probably be geared to your specific application (String>>findComicMarker). The next step is to realize there's a general part to that method (some kind of searching in strings), and an application specific part, which String really should have no knowledge of (the pattern to be found). Then you'll make it into String>>findMarker: which is called from ComicParser>>findMarker content findMarker: self marker When you've cut down the functionality related to String to it's most general form, you're quite likely to come up with something that is generally useful, and then it will either - A. Be a useful or at least reasonable addition to the base class library, and you might want to submit it to the list members' scrutiny. B. Already be implemented. This is very very common - the basic classes are very very rich in functionality. This is not a waste of time, it's an opportunity to simplify. I differ with Cees comments on one point - changesets are a very easy way to start playing around (they are maintained automatically, and can be edited easily), but they stink as a way to manage core over time. This is especially if you have class extensions. If your application is more than a few classes, use DVS or something like it. Daniel Cees de Groot <cg(a)cdegroot.com> wrote: > > --Boundary_(ID_wFgDtfzrwXRzid8qV2dN5w) > Content-type: multipart/signed; boundary="=-zkE9eJEpApU0f0bzcQS+"; > protocol="application/pgp-signature"; micalg=pgp-sha1 > > > --=-zkE9eJEpApU0f0bzcQS+ > Content-Type: text/plain > Content-Transfer-Encoding: quoted-printable > > On Thu, 2003-05-01 at 18:53, Ingo Hohmann wrote: > > where I come from fiddling with the system innards is supposed to be a=20 > > 'black art'. > >=20 > In Smalltalk, it isn't supposed to be a black art. So get used to that. > > > Actually, my problem is that, everything I do is in the image to stay -=20 > > that's quite a difference to writing a script that changes whatever it=20 > > wants to in the system, when it does not interfere in any way with other=20 > > scripts/programs. > >=20 > However, this is not 'fiddling with system innards'. Everything you do > is logged in a changeset, and you can thus see what you changed, undo > your changes, file them out to a text file and throw them at a version > repository, etcetera. Yes, programming in an image-based Smalltalk is > different from programming most other languages, but don't let that stop > you from writing good code. > > > _That's_ why I am reluctant to create new classes. > >=20 > And that's completely the wrong reason :-) > > > Should I extend the String class in the system? > > Create a subclass of String that understands my additional messages? > > Create a StringHelper which understands #myMessage: aString? > >=20 > > The latter seems to be the cleanest to me, what do you think? > >=20 > Not clean at all, you'll create parallel hierarchies that serve no-one > except the vendors of expensive parallel-hierarchy-tracking tools (also > called 'UML modellers' or 'CASE tools' ;-)). > > First, you need to think about what the best place is for a method. When > you found it, you put it there. Of course, when you put it in an > existing class you'll take some extra care to make the thing either > completely non-obvious (give it a name prefixed with your initials so > that no-one else is likely to make the same 'class extension') or > completely obvious (say the method 'isSymbol', which would return > 'false' in its implementation in the Object class and 'true' in its > implementation in the Symbol class) so that if someone else makes a > class extension with the same name, it will have the same behavior so it > won't collide with your version. But there's no a priori reason you > shouldn't put your method in String if all signs are that it belongs in > String. > > Smalltalk has tools to keep these class extensions apart from the 'base > methods'. The default method in Squeak is that all your code is kept in > a changeset; a changeset basically is a sort of transaction log of your > changes, and does not care when you define a method whether the class > was 'yours' or 'someone else's'. A newer method in Squeak is a DVS > package, which relies on some naming conventions in the extended classes > to make them part of the package you want them to be in. On both topics > you'll find abundant information on the Squeak Swiki. > > Regards, > > Cees > > --=-zkE9eJEpApU0f0bzcQS+ > Content-Type: application/pgp-signature; name=signature.asc > Content-Description: This is a digitally signed message part > > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v1.2.1 (GNU/Linux) > > iD8DBQA+sVS28wOTf+CYnosRAuLyAJ9fpPG+5lmB1FwxoGswlxnwpNSPmgCfd2Z3 > 2CFsk9gy8EhBEiCV8VRY4cU= > =WKU2 > -----END PGP SIGNATURE----- > > --=-zkE9eJEpApU0f0bzcQS+-- > > > --Boundary_(ID_wFgDtfzrwXRzid8qV2dN5w) > MIME-version: 1.0 > Content-type: text/plain; charset=us-ascii > Content-transfer-encoding: 7BIT > > > > --Boundary_(ID_wFgDtfzrwXRzid8qV2dN5w)--

1 0

[Q}[CONNECTORS] Connecting submorphs
by Martin Drautzburg 02 May '03

02 May '03

I used to have a PasteUpMorph with a couple of embedded submorphs. The submorphs were connected via connectors. No problem so far. Then I wanted to connect some of the PasteUpMorphs (the containers) and realized, that they don't accept connectors. So I thought: lets turn them into plain rectangeles. Now I *could* connect the Containers but could no longer connect the Morphs inside the Container. Am I attempting a logically impossible thing ? Or is there an easy solution ?

2 4

Jump to page: