Squeak-dev May 2003

squeak-dev@lists.squeakfoundation.org

204 participants
710 discussions

[Q] Version Control with Squeak was [Q] Fiddling with system innards
by Anthony Adachi 01 May '03

01 May '03

On Cees de Groot wrote: > 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. On the subject of changes is it possible if you could you go into a little more detail about how people handle version control with Squeak? So, I gather, from what you're saying, that people file out the classes they've created (and the existing one's they've modified) out of Squeak and store them into their Version Control program on a periodic basis? Thus, providing the way to bring back individual classes or methods in an image to an earlier state (beyond what undo allows- especially after quitting and reopening Squeak). No? Thanks, Anthony __________________________________ Do you Yahoo!? The New Yahoo! Search - Faster. Easier. Bingo. http://search.yahoo.com

2 1

[Q] Newbie - Multiple Return Values
by Ingo Hohmann 01 May '03

01 May '03

Hi to all again, for a network protocol I want to return to values from a function - the payload of a message, and wether the message was tagged as being successful or not. What would be the best way to return those values? - I thought about an instance variable for one of these, but there might be multiple message pending. - Using Exceptions for the an unsuccessful message would be misusing the exception mechanism. - And overloading the returnvalue - either an 'this was an error' value _or_ the payload does not particularly please me either. - Creating a new Class for this on the other hand seems to be quite an overkill. What's the best SmallTalk way to handle this? Thanks, Ingo

3 2

Speech sar
by Frank Shearar 01 May '03

01 May '03

Hi all It's probably just a newbie problem, but I've installed the SpeechRemoval package followed by the Speech package in a 3.6 (5184) image. (This is the right thing to do to follow the trend of stripping down the base image, isn't it?) At any rate, when I install (or reinstall) the Speech package, nothing happens - the installation happens instantaneously and I can find no KlattFrame class, which means that the classes in the sar weren't filed in. What gives? Any help will be most gratefully received! frank

1 0

[Q] ProcessorSchedule>>activeProcess doesn't always return activeProcess??
by Phil_Hargett＠groove.net 01 May '03

01 May '03

I have been testing some code with SUnit and an upgraded TestRunner (e.g., the TestRunner Enhancements SM package), and I have noticed that when I explicitly put in my code a call to the debugger to interrupt execution and start debugging: Processor activeProcess debugWithTitle: 'Debug foo'. I typically receive a debugger but on a different process than the one that ran the above code! Quite often, it is the Morphic UI thread itself. I last noticed that perhaps since TestRunner executes tests on a background process (yeah!) I should be checking invoking #debugWithTitle: on Processor backgroundProcess instead, but that seems unclean: if I just want a reference to the process in which my code is executing, why do I know I have to know anything about what process I am? I'm sure I'm missing something. Any assistance appreciated. Thanks! Phil

3 2

Newbie speaks again
by Tony Fugere 01 May '03

01 May '03

Is it even possible to view a single morphic project just as SmallTalk source code? In or outside of Squeak? Tony Fugere

2 1

Re: Well factored Objects
by Richard A. O'Keefe 01 May '03

01 May '03

Jimmie Houchin <jhouchin(a)texoma.net> has now explained his problem a bit more. I wasn't going to mention such but it seems beneficial to clarify things. I am a licensed Funeral Director. I fill out legal certified documents. They have certain data criteria. AH *HAH*. Now all becomes clear. You aren't modelling a person at all. The thing you are REALLY modelling is a *FORM*. (I'm a little unclear on the concept of "legal name". My father (R.I.P.) was a lawyer, taught law in two countries, wrote many law textbooks, and told me on more than one occasion that someone can call themselves anything they want as long as it isn't with intent to defraud. Your birth certificate may say "Phillip Arthur George McDonald Fraser McKie" but you can call yourself "Bill McGonagle", even in contracts, without having to inform the government. I have no idea what USA law might be.) For all my forms the SSN or an 'unknown' or 'not available' must be provided. Empty is not an option, it will be rejected. I will address these real life issues. The forms I complete have spaces on them for *all* the above data. Right. You are not modelling social reality, you are modelling official forms. The legal first name will be the first name, and the legal last name will be the last name. All middle name(s) will go in the middle. I'm rather unhappy about the idea of a legal system defining what is a legal first name and/or a legal last name. (Yes, I'm aware of the practice in some Scandinavian countries of the government telling you what you can call a child/yourself, and I'm aware that some other countries do it too. It does prevent a certain about of silliness, but it can go too far.) For obituaries I can put anything they want. For death certificates I must put the legal name. So you have forms, defined by the government, and there are constraints and semantics for those things defined by the government. > * Do "first" and "last" really mean "written first" and "written last" > or are they code-words for "personal" and "family" name? Is there > an expectation that names should be sorted with "last" name as most > significant? In that case, what about Chinese and Hungarian names? Legal first, legal last. That tells me nothing, because I don't know what "legal first" and "legal last" are. The point about Chinese and Hungarian names is that they write family name first, personal name last. I'm _guessing_ that "legal first" means "personal" and "legal last" means "family". But again, the point is that the thing you are modelling is not a JHPerson but a UsGovtDeathCertificate. I think it will probably pay, long term, to regard the *person*'s name(s) and the *form*'s name fields as independent things and not try to derive one from the other, while making it easy in the user interface for people to copy one to the other for the common case where there is a simple relationship. Suffix is used to designate the legal designation as would be on a birth certificate. Senior, Sr. does not exist generally in the legal sense. But myself being Jimmie Lee Houchin, Jr., Jr. would be mine. I don't think we _have_ a "legal designation" in this country, which is why I didn't understand. Presumably someone's legal name is whatever it is the their birth certificate or their most recently accepted change of name form (if any). What happens if a resident non-citizen dies? How is the legal name defined then? This is true. But I do need somewhere to tie the persons UUID (SSN) to their name. I might have multiple 'John Smith's but their SSN should be unique. Yes, you need to tie the SSN to the person. No, not directly to "the" name. William Kent's "Data and Reality" really is an indispensable guide. IIRC, he advises to use your *own* "surrogate" (approx= oid) for entities. Still talkikng in data base terms here, There should be a Person relation with primary key PersonID generated by YOU, *not* an SSN. You have already mentioned that SSN might be 'unknown' or 'not available', which makes it unusable as a primary key (for you). On your account, a person can have at most one SSN when they die. The SSN is a property of the person, not of the person's name(s). So it makes sense to have Person(PersonID, SSN?, ...) There should be a DeathCertificate relation with primary key PersonID which is also a foreign key for Person. This describes the form. Does the form require the address and legal name for the informant, as well as for the deceased? In that case, it makes sense for legal name to be part of Person. It might then make sense to have a UsaLegalName domain. So you might have Person(PersonID, SSN?, UsaLegalName, Address, UsualName, ...) UsaDeathCertificate(PersonID, PersonID InformantID, Date, ...) In this situation, it makes sense to have a UsaLegalName object split into fields because the fields exist on the form, but it's really the form that's being modelled, not the person. But, I understand Richard. I quite concur with the frustration of arbitrary decisions and constraints. Just today I was unable to register on a web site to participate in a language standardisation effort I care about greatly, because it wouldn't let me spell my name correctly... One of the reasons why modelling business information is sometimes (sometimes!) easier than modelling "real life" is that what one is really modelling in the business situation is a set of forms. In effect, one is modelling an already simplified model. One of the reasons why this tends to make life hard for customers is that the computer is enforcing the simplified rules. When it's your own system, and you're in complete charge, you can always work around it. On a vaguely related note, the NZ government wants people to save 10% on electricity use this winter. The local newspaper pointed out a day or two ago that the town hall was still brightly lit at night. A council spokesman apologise, and explained that they had been searching for days trying to find the switch... Somewhere, a timer is still enforcing what seemed like a good idea at the time.

2 2

Re: [ANN] [ENH] FillInTheBlankMorph Look
by Anthony Adachi 01 May '03

01 May '03

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

What makes up a "User" in a Class Object? (was Well factored Objects)
by Darius 01 May '03

01 May '03

I enjoyed Richard A. O'Keefe's treatise on name representation in class objects: "People don't have names. Instead there is a *relation* between people and names." This touches a little on my concerns about Graphics User Interface representation. Please excuse any of my lack of insights for Ive a psychology degree and work as a programmer. [A *real* Computer Analyst] ;) Why is there no persistent class representing the user in total? Shouldnt there be something at the Kernel level? Why is such a person only represented by keyboard strokes and Hand Morphs? Isnt the most cherished person in our programming life, the "user", more than an I/O port? A person has memory, a unique place and time, a skill level, a purpose, a single focus or locus of attention at any moment, modes of communication across devices, learning styles, personality types, and limitations to the above (among a few other things...). Every sent communication and every stored data element has a purpose for a person or a group which is most often lost in todays data representational systems. A person is also a member of several groups for which that person shares an identity an which act as a proxy on their behalf. All this relates directly to the data use and functionality of any application. People dont _have_ fingers. Instead there is a *relation between what people want and what their fingers do. People dont _have_ eyes. Instead there is a *relation* between what they see, what they remember from what they see, what they know. They quickly shift their focus or locus of attention over time. They imagine half of what they see by filling in from memory. They focus on juxtapositions both of whats similar and what contrasts. They compare what they see to a cognitive map or mental heuristics of what they expect. They give attention to the information that has an emotional impact on themselves. Why dont we create class hierarchies of our GUIs to match the already researched rules of cognitive science? At this point I should say But there's also a simple solution." But I cant. The best I can do is refer you to a more detailed exploration of the topic in Jef Raskins The Humane Interface http://humane.sourceforge.net/humane_interface/hollands_review.html Certainly Squeak with its OOP methodology could get us there and gain notoriety in the process. Cheers, Darius

3 2

[Q] What draws the text insertion point?
by tblanchard＠mac.com 01 May '03

01 May '03

I'm playing around with TextMorph's appearance. The text looks squished up against the left side of the bounding box so I was trying to shift it to the right a bit. I've found that adding a margin offset to the paragraph's position moves the characters over nicely. But the text insertion point and selection drawings don't shift along with it. Digging through TextEditorMorph isn't getting me there. Anybody who knows this wanna give me a hint?

1 0

[ANN] [ENH] FillInTheBlankMorph Look
by diegogomezdeck＠consultar.com 01 May '03

01 May '03

Hi, Attached is a changeset with a renewed look for FillInTheBlankMorph "compatible" with the other look enhancements I made. Comments, as usual, are welcome. Diego PS: What is the next part to change the look?

9 13

Jump to page: