Norbert, I have also noted the "return nil on error" in parsers, and (my current mission) streams. IMHO, these decisions were questionable before structured exception handling, and are all the more unfortunate now. My preference would be to create something like #readFrom:onError: and then re-write #readFrom: in terms of it, raising an error inside the block. Failing to do that risks intermittent downstream failures due to "unexpected" garbage values. I like my bad news early, and in context. Skipping to the end of your message, I am hopeful that Pharo will not suffer from Squeak's barriers to adoption of good code, but I cannot promise same. It is probably best to start small and ask up-front whether people want what you propose. I suspect you will find that good work will be included in Pharo. Bill Wilhelm K. Schwab, Ph.D. University of Florida Department of Anesthesiology PO Box 100254 Gainesville, FL 32610-0254 Email: bschwab@anest.ufl.edu Tel: (352) 846-1285 FAX: (352) 392-7029
norbert@hartl.name 6/2/2008 2:28:15 PM >>> Hi,
I really appreciate the existence of the pharo project. For me I declared it as a good starting point for me to dive into the deeper waters of squeak/pharo. My biggest problem is that I do not know how exactly things have to be done. I don't want to just submit changes without to know if the worth it. A little story. I started last week to toy around with the newest pharo release. On the web site I read that all tests green is a primary goal. So I started to let the test runner run all tests. After having some trouble I found a failing test in NumberParsingTest. The test is failing because the test case expects Number>>readFrom: to throw an error if a number couldn't be parsed. Searchig the ST80 book I found nothing about Number>>readFrom:. Searching the web I found a thread on squeak-dev from january where the outcome was: - readFrom: returns 0 on unsuccesful read attempt to satisfy the parser (sounds like some history thing) - readExactlyFrom: does throw an error an unsuccesful read attempt. And it utilizes sqNumberParser So what to do? I tried to be least aggressive and fixed the test case to use readExactlyFrom: (and then splitting the test case into two because the test does different things). Not thinking first I would have asked a lot of questions on this list: - Do we want to preserve the current readFrom: ? - Should we integrate sqNumberParser completely avoiding mulitple implementations? - Is it only me having slightly problems having a class Number which uses a class SqNumberParser instead of integrating the code into the Number hierarchy? Is there a way to get this questions answered without generating a lot of noise on the list? Other questions: - I scanned the image for obsolete references and found some. I don't know how to deal with it in an update. Using ServerDirectory resetServers. FileServices removeObsolete. PackageOrganizer default unregisterPackageNamed: 'FlexibleVocabularies' I have removed all obsolete references but one. How is this integrated into the upstream? - Trying to solve the last obsolete references I found an obsolete class referenced in PasteUpMorph. The method which referenced it is PasteUpMorph>>isSafeToServe. The method is in category Nebraska. I thought Nebraska has been removed from the system. Thinking that I would have removed all methods of this category but one because they have no senders. One method transferRemoteServerFrom: is referenced from Project>>enter: returningFlag revert: revertFlag saveForRevert: saveForRevert Is it ok to rip this call out (by trying to understand what it does in class Project ;) ) Some words of advice would be helpful to get a feeling about those things. I don't have a problem preparing some stuff and put it into the Inbox. A problem would be to have a patch rejected without explanation :) Norbert _______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project