Self-interest December 1990

self-interest@lists.selflanguage.org

13 participants
20 discussions

self blocks
by kuba%starcom%navtech＠sj.ATE.SLB.COM 13 Dec '90

13 Dec '90

It is said that blocks in self are close to closures in scheme, so I have tried to implement the following: (define (cond-gen par1) (define (parametrized-test par2) (= par1 par2)) parametrized-test) (define test-if-equals-4 (cond-gen 4)) (test-if-equals-4 4) and after couple failures gave up. Does anybody have a solution to this problem? Kuba

1 0

Catch & Throw
by dje＠scs.carleton.ca 13 Dec '90

13 Dec '90

As Craig Chambers mentioned, SELF has exit and exitValue to provide "catch & throw" capability. Parc Place Smalltalk versions 2.5 and later provide exception handling capabilities. This is a superset of catch & throw (essentially named catch and throw). One problem I had with SELF's exit and exitValue was that it doesn't seem to work when an _Eval is on the stack. I was trying to implement an evaluator (which I will be sending to self-interest as soon as I've prettied it up a bit) which reads user expressions and evaluates them. The problem is that returning from a block whose context is down the stack a ways (which is what exit and exitValue do) doesn't work when there is an _Eval on the stack. It has no effect. It seems that the _Eval is done on a separate stack. This makes it difficult to handle compile errors. The only approach I can think of is to actually parse the expression! Run-time errors shouldn't be a problem. The _ReturnFromSelf primitive didn't seem to work either, although the documentation mentions that the _AbortFromSelf primitive also aborts any running primitives (such as _Eval). If anyone has a way around this problem, I would like to incorporate a solution into the evaluator. Thanks, Danny Epstein dje(a)scs.carleton.ca

1 0

Re: catch & throw (ATTN: Dave Ungar & co.)
by craig 12 Dec '90

12 Dec '90

There's already a similar capability defined in Self. What you call "catch" we call "exit" and "exitValue" (depending on whether nil or a particular value is returned from the catch block). We don't create any special throwBlock objects to send "throw:" to, instead using normal blocks to send "value:" to to exit the catch block. So instead of writing [|:errorTag| message: errorTag ] catch. message: errorTag = ( errorTag throw: 'oops!'. 'this code never happens!' print. ). as you did, you would write [|:errorTag| message: errorTag ] exitValue. "or exit" message: errorTag = ( errorTag value: 'oops!'. 'this code never happens!' print. ). BTW, you can't make the "block" data slot in your throw block object private, since the "block:" message in the "with:" method of throw blocks would not have access; the sending method holder (in this case, the prototypical throwBlock) is not an ancestor of the receiver of the "block:" message, so private slots are invisible. So just make the "block" data slot public (or unspecified) to get your example to work. -- Craig Chambers

1 0

catch & throw (ATTN: Dave Ungar & co.)
by burdick＠mentor.cc.purdue.edu 12 Dec '90

12 Dec '90

Dave & co.: I didn't see catch and throw in the docs, so if you guys haven't put catch & throw in the standard world, you might want to just put this code in there (at least I don't know of any rules of good programming it violates or anything)... Maybe this has been discussed in this group already, but just in case, here's a useful addition to the startup.self... ------------ traits block _AddSlots: (| ^ catch = (value: throwBlock with: [|:throwValue| ^throwValue]). _ throwBlock = (| ^ with: aBlock = (_Clone block: aBlock). ^ throw: aValue = (block value: aValue). ^ throw = (throw: nil). _ block. |). |). ------------ You can use it like this: [|:errorTag| message: errorTag ] catch. message: errorTag = ( errorTag throw: 'oops!'. 'this code never happens!' print. ). The reason I defined throwBlock instead of just putting throw in block along with catch is to help with errors; to make sure that only throwBlocks receive the throw[:] message and not just any old block. If this doesn't make sense, any CommonLISP book should have a good explanation. I could never figure out why ParcPlace never included catch and throw in their release; it's easy to add and it's great for handling errors. -- Bill burdick(a)mentor.cc.purdue.edu P.S. There may be a syntax error or two in this code -- I don't have self (I guess I'm trying to find my self :) ), yet, so I haven't been able to test this.

1 0

commercial/free version
by burdick＠mentor.cc.purdue.edu 11 Dec '90

11 Dec '90

Since it seems that self is truly meant to be a practical, pure OO language, are there any plans for a commercial version of self? or are there plans to release it free and without restrictions on its use? People in the business community are starting to pick up on Smalltalk-80 (it's now possible after only ten years!). I have heard that Objective-C isn't as widely accepted as C++ because there was no free version of it available. -- Bill

2 1

Re: inheriting data slots
by ungar＠thoreau 11 Dec '90

11 Dec '90

Sharing format information is harder in Self than in classical languages. What you describe seems like it's about the best way to me. David

1 0

inheriting data slots
by burdick＠mentor.cc.purdue.edu 11 Dec '90

11 Dec '90

A question I've had about self since I read the self intro paper in OOPSLA is how you inherit data slots. I read the Organizing Programs Without Classes paper, and that sort of answered it, but I want to make sure that I'm not missing anything... Using an extra parent slot (like dataParent in the example in the paper), you model data inheritance by copying a list of prototypes when you clone an object. That would be these slots in a traits object: dataParent: arg = (). copy = (| new. | new: resend.copy. new dataParent: dataParent copy. |). Then, for each object inheriting those slots that wanted to inherit data from an object 'o', you just define a slot: dataParent** <- o. By doing this, I'm trying to share some format information as well as behavior. Is there a better way to do this, or is this 'the right way?' -- Bill burdick(a)mentor.cc.purdue.edu

1 0

Re: X & Self
by ungar＠thoreau 05 Dec '90

05 Dec '90

Well, I am finally catching up with my mail, and it's great to hear that you are going to try some Self hacking! I think that we may have to start a Self-exchange to let pioneers share their objects! David

1 0

None
by Benjamin E. Tucker 02 Dec '90

02 Dec '90

To: Sean.Levy(a)CS.CMU.EDU, self-interest(a)self.stanford.edu Subject: Re: Am I missing something?

1 0

Re: Continuing woes, X11, UIs
by craig 01 Dec '90

01 Dec '90

I just copied the sun4/Self.Z binary, ran uncompress, and did ls -l: -r-xr-x--- 1 craig self 884736 Nov 30 16:37 Self* This size is different from your 735076 result. Are you using uncompress? Seems like you just aren't getting a real copy of Self, for some reason. -- Craig Chambers

1 0

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

Self-interest December 1990