Squeak-dev May 2003

squeak-dev@lists.squeakfoundation.org

204 participants
710 discussions

Re:ReadWriteStream Issues
by Richard A. O'Keefe 06 May '03

06 May '03

Time for a look at the ANSI standard. (Not for a full and final legal settlement, but for some guidance, is all.) (Part of) the protocol hierarchy looks like this, using a Javaish syntax (because protocols aren't classes, and because it's hard to draw a DAG in ASCII): protocol "sequenced stream" close // dissociate stream from backing store contents // return complete contents of stream. isEmpty // self contents isEmpty, NOT at end position // return current logical position position: n // set logical position reset // same as self position: 0 setToEnd // same as self position: (self contents size) (I must say that having position/position: in the Stream class in Squeak is a *major* pain. Nearly every stream I've ever written has been unable to inherit from Stream because of that. I also note that #contents is unimplementable for streams connnected to many kinds of devices.) protocol "collection stream" extends "sequenced stream" contents // returns same class as aColl select: [...] // where aColl is underlying collection; the // answer might or might not be aColl itself. protocol "gettable stream" atEnd // at end of readable data yet? do: aBlock // [self atEnd] whileFalse: [aBlock value: self next] next // return next item (undefined when atEnd) next: n // return collection of next n items nextLine // return next line nextMatchFor: x // return self next = x peek // return next item but don't move over it, nil atEnd peekFor: x // return self atEnd not and: [self peek = x] skip: n // skip next n items, stop early atEnd skipTo: x // [self atEnd] whileFalse: [self next = x ifTrue: // [^true]]. ^false upTo: x // collect items; stop collecting just before x; // stop reading just after x or AtEnd. (#nextLine is clearly the odd man out here. What does (ReadStream on: #(1 2 3 4)) nextLine mean?) protocol ReadStream extends "gettable stream", "collection stream" next: n // returns same class as self contents upTo: x // returns same class as self contents protocol "puttable stream" cr // self nextPut: Character cr. flush // "update backing store". nextPut: x // write x. If position is at end, append, else // replace & step over next element. nextPutAll: xs // xs do: [:x | self nextPut: x] space // self nextPut: Character space. tab // self nextPut: Character tab. protocol WriteStream extends "puttable stream", "collection stream" // no extra methods, no changes to semantics. protocol ReadWriteStream extends ReadStream, WriteStream // no extra methods, no changes to semantics. factory ReadStream on: aCollection // return a ReadStream reading from aCollection Past sequence values: none. Future sequence values: the elements of aCollection. In effect, the position is 0 and the read limit is aCollection size. factory WriteStream with: aCollection // return a WriteStream writing to aCollection Past sequence values: the elements of aCollection. Future sequnce values: none. In effect, the position and the write limit are both aCollection size. factory ReadWriteStream with: aCollection // return a ReadWriteStream based on aCollection Past sequence values: the elements of aCollection. Future sequnce values: none. In effect, the position, read limit, and write limit are all aCollection size. What we _don't_ find is a definition for ReadWriteStream>>on: Let's look at Squeak 2.8. Stream PositionableStream ReadStream WriteStream ReadWriteStream PositionableStream class>>on: aCollection ^self basicNew on: aCollection PositionableStream class>>on: aCollection from: first to: last "You were maybe expecting something clever, maybe?" ^self basicNew on: (aCollection copyFrom: first to: last) PositionableStream>>on: aCollection collection := aCollection. readLimit := aCollection size. position := 0. self reset "position := 0 again." WriteStream class>>on: aCollection "inherited from PositionableStream" WriteStream class>>on: aCollection from: first to: last ^self basicNew on: aCollection from: first to: last WriteStream class>>with: aCollection ^self basicNew with: aCollection WriteStream class>>with: aCollection from: first to: last ^self basicNew with: (aCollection copyFrom: first to: last) WriteStream>>on: aCollection "boils down to" collection := aCollection. position := 0. readLimit := 0. "OOPS" writeLimit := aCollection size. WriteStream>>on: aCollection from: first to: last "boils down to" collection := aCollection. readLimit := writeLimit := aCollection size min: last. position := first - 1 max: 0. WriteStream>>with: aCollection "boils down to" collection := aCollection. position := readLimit := writeLimit := aCollection size. We see a strange thing here. WriteStream on: aColl sets readLimit to 0, but WriteStream on: aColl from: 1 to: aColl size sets readLimit to aColl size. ReadWriteStream class "inherits #on:, #on:from:to:, #with:, #with:from:to:" ReadWriteStream "inherits #on:, #on:from:to:, #with:" It looks very much as if WriteStream>>on: aCollection should read self on: aCollection from: 1 to: aCollection size. This doesn't seem to be a new problem; I was looking at the code in 2.8.

1 0

Re: Squid Environments
by Richard A. O'Keefe 06 May '03

06 May '03

Anthony Hannan <ajh18(a)cornell.edu> wrote: When importing modules into a module, where do we get the imported modules from. Eiffel addresses this problem with LACE files (.ace extension). It's worth looking at how they do that to see what kind of functionality might be worth having. Of course Eiffel is static, and a .ace file basically describes a single application (or library). Erlang addresses this problem with SASL descriptors, and Erlang is nearly as dynamic as Smalltalk.

1 0

[BUG] DVS cross-platform file error
by Eric Scharff 06 May '03

06 May '03

For the curious, I tracked down why the changes button was broken for me in DVS. The problem seems to be how DVS uses CrLfFileStream. BUG: DVS fileOuts on a stock Mac OS X VM (3.5.0Beta8 VM, Squeak 3.5-5180) have DOS (CRLF) line endings, but when they read in, they expect Mac cr line endings. If you change the line endings from a DVS fileOut from CRLF to CR line endings, the changes button works properly. DISCUSSION: On my 3.5 Unix VM, things properly save and load with Unix line endings. I'd really want things to save with Unix line endings anyway because I want to use DVS with CVS. So, I ran CrLfFileStream defaultToLF but DVS doesn't seem to honor this. I'd like DVS to either properly use CRLF file stream (so that they save and load in the same format) or simply to use CVS-friendly LF format. -Eric

5 9

Re: Squid Environments
by Anthony Hannan 06 May '03

06 May '03

Well after thinking some more I noticed some opportunities for improvement on my environments design. Here they are (I hope you guys don't mind all my thinking out loud, and changing my mind). Method name prefixes (protocol part) aren't needed anymore since the environment can serve as the protocol. If two methods from different inherited environments have the same name then the inheriting environment must alias them. Calling methods will never need to provide selector prefixes anymore since aliasing will resolve conflicts. When an independent (protocol) method delegates to the receiver class, where does it look for the class's method? We can't just say in the class method dictionary because that dictionary changes depending on which environment you are looking from. So which environment do you search when you delegate? We can't use the execution environment since dynamic binding violates encapsulation and security. We can't use the environment the protocol method is defined because the receiver class may be defined in a higher environment not giving it an opportunity to override. We can't use the environment the class is defined in since the class extension may be defined in an higher environment with the protocol method or orginal sender. The solution is to look in both the original sender's environment and the class's environment. The original sender's environment will necessarily include the protocol environment since it is calling it. Both environments are searched as if they were both inherited by some environment, meaning an error is raised if a conflict exists (in a normal environment an alias would resolve the conflict). This is similar to ClassBoxes, and I have kind of come full circle to it. But I am still not sure which environment/classbox Alexandre searches when looking up methods for a class. I think he searches the execution environment/classbox, which although convenient, violates (I believe) environment modularity (encapsulation) and capability security. Cheers, Anthony

1 0

[BUG] Magma repository creation
by Derek Brans 06 May '03

06 May '03

Chris, In MaFilename>>indexOfDotDelimiter Since directories and filenames can have periods in them, you probably want to get the index of the last dot delimiter. Cheers, Derek Derek Brans Nerd on a Wire Web design that's anything but square http://www.nerdonawire.com phone: 604.874.6463 mailto: brans(a)nerdonawire.com

1 0

Re: ReadWriteStream Issues
by Anthony Adachi 06 May '03

06 May '03

Frank Sergeant wrote: > Also, rather than just executing your block of code in a workspace, > highlight it, right click, and choose 'debug' and then you can single step > through the code and easily examine the stream etc as you go. As, I indicated in my first post and which Bert Mayo also later pointed out the readLimit gets set to 0 in the on: method. This is likely to be the source of the problem. That is, assuming this really is unexpected behavior on the part of the ReadWriteStream class. Anthony __________________________________ Do you Yahoo!? The New Yahoo! Search - Faster. Easier. Bingo. http://search.yahoo.com

1 0

Re:ReadWriteStream Issues
by Frank Sergeant 06 May '03

06 May '03

> <<I've been reading the "Squeak: A Quick Trip to > ObjectLand" book and have come across some problems > with getting the examples using ReadWriteStream to > work (on page 174-175). I don't get the same results > as the book states as one should expect.>> Just a quick note or two in case it helps before someone posts the definitive answer: I don't think you need to modify any of Squeak. I think you just need to add the missing #reset after creating the ReadWriteStream and before trying to read from it. Otherwise, you are trying to ready past the end and, naturally, there is nothing there. Also, rather than just executing your block of code in a workspace, highlight it, right click, and choose 'debug' and then you can single step through the code and easily examine the stream etc as you go. There was the mention of the problem of writing to a ReadStream, but, of course, you shouldn't write to a ReadStream. Also, you have been opening the ReadStream or ReadWriteStream on a *literal*. I think this is ok for the ReadStream but I don't think it is a good idea in general for a stream that can be written to. Instead of s := ReadWriteStream on: #(a b c). I would would be more comfortable with opening it on a copy, e.g. s := ReadWriteStream on: #(a b c) copy. Also note #on: versus #with:. Isn't one of these preferred and the other obsolete? -- Frank

1 0

Re: ReadWriteStream Issues
by Anthony Adachi 06 May '03

06 May '03

Frank Sergeant wrote: > I don't think you need to modify any of Squeak. I think you just need > to add the missing #reset after creating the ReadWriteStream and before > trying to read from it. Otherwise, you are trying to ready past the end > and, naturally, there is nothing there. Sending a reset message to the ReadWriteStream instance does not resolve the unexpected behaviour. s := ReadWriteStream on: #(1 2). s reset. s next. "returns nil" s atEnd. "returns true" s position. "returns 0" s contents. "returns #()" > Also, you have been opening the ReadStream or ReadWriteStream on a > *literal*. I think this is ok for the ReadStream but I don't think it > is a good idea in general for a stream that can be written to. Instead > of > > s := ReadWriteStream on: #(a b c). > > I would would be more comfortable with opening it on a copy, e.g. > > s := ReadWriteStream on: #(a b c) copy. Making a copy of the literal does not alter the results either. s := ReadWriteStream on: #(a b c) copy. s reset. "a ReadWriteStream" s atEnd. " true" s position." 0" s next. " nil" s contents." #()" However, as I mentioned in my first post, if one sends the message "with:" instead of "on:" immediatly followed by "reset" then one can get the results the book expects. Whether this is the expected behaviour by the ReadWriteStream class' authors is the question? Anthony __________________________________ Do you Yahoo!? The New Yahoo! Search - Faster. Easier. Bingo. http://search.yahoo.com

1 0

[ANN] New Comanche Website
by Stephen Pair 05 May '03

05 May '03

I've put up a new website to organize the distribution and ongoing development of Comanche. I tried to follow the design of Ian's Unix dowload site (because I like it so much). http://people.advantive.com/~spair/comanche - Stephen

1 0

Re: Let's awake!
by Daniel Vainsencher 05 May '03

05 May '03

tblanchard(a)mac.com wrote: > I sort of agree - there is a lot of work going on and its difficult for > the core people to absorb it. Right. > My personal pet desire is to get some kind of date/time/timezone thing > in there so you could have a couple squeaks on machines in different > locations and still have coordinated time. The ansi fixes on squeak > map provide a fairly simple mechanism - I'd like to see them adopted. > The timezone database is probably a more complete solution, but the > current Timestamp that downloads with SqueakMap is not solution at all > because its naive about time zones. A review of all the different solutions by someone interested, explaining the trade offs about adding them to the image vs. keeping them as packages, would probably help. Note that we do have an add-to-the-image-aversion, since we'd rather be making it smaller, not bigger, but something might get in IF it is the simplest solution that could possibly work, and the current solutions is really just broken (whether most of us notice or not). Daniel > On Sunday, May 4, 2003, at 12:19 PM, <diegogomezdeck(a)consultar.com> > wrote: > > > Hi guys... > > > > The following is an email I'm thinking to write for weeks. > > > > To say it shortly: Squeak is dieing in front of us... > > > > Since SqC leaves the "dictatorship" role we don't find the way to > > continue. Despite the apparent organization and procedures we had > > created, > > simply in the last months we was not able to satisfy the expectations. > > > > To probe my point just take a look to the amount of fixes waiting for > > approval and the changes produced by MCP and KCP. > > > > Personally, I paused my work in MCP until something happens in our > > community. I invest most of my free time in Squeak and I don't feel > > this > > sacrifice is producing results. > > > >> From my point of view We're trying to organize so much, and in the > >> reality > > we have pages and pages describing a process (the harvesting process) > > that > > don't work. The bureaucracy is killing us. The steps to get a fix > > approved are, in most of the cases, more difficult and time consumer > > than > > the fix itself. We have lists of people with roles (guides, > > harvesters, > > etc) but in the reality only a few of them are active and working. > > > > We have big problems and I don't feel the community realize of it. > > > > Cheers, > > > > Diego Gomez Deck > > > > PS: I'm talking for my own and my thoughts are not the MCP-team > > opinions. > > > > > > > >

3 2

← Newer
1
...
60
61
62
63
64
65
66
...
71
Older →

Jump to page: