Self-interest March 2014

self-interest@lists.selflanguage.org

10 participants
31 discussions

Linux Makefile
by ardeujho 27 Mar '14

27 Mar '14

I've written a Makefile to build Self using GNU Make without needing CMake. I started with the Tup support that was posted recently and converted it to Make. It's available in my 'linux_makefile' branch: <https://github.com/ardeujho/self/tree/linux_makefile> Builds can then be done with: make -f Makefile.linux Or parallel builds: make -f Makefile.linux -j8 I did this so I could easily hack on it to cross compile and add other libraries. I'm a bit lost with CMake in that regard. I made it public in case others find it useful. Ideally I'd like to remove the need to call 'makeDepends' as this would make cross compilation easier. This would require stripping out the generated 'incls' stuff and changing the source to include things directly. I'm not sure how feasible or how much work that is.

1 1

Re: [self-interest] Creating objects from the shell
by Stephen De Gabrielle 15 Mar '14

15 Mar '14

Thats brilliant, lobby _AddSlotsIfAbsent: ( | person = ( | name. age. test = ( 'test' printLine ) | ) | ) In english, I am sending the lobby object a message to add a new slot called person, and in that slot put an object '( | name. age. test = ( 'test' printLine ) | ) ' which consists of three different slots; name, age and test. the test slot contains a method that prints 'test'. I tried the following in the terminal and in the shell "Self 1" > "Self 1" lobby _AddSlotsIfAbsent: ( | person3 = ( | parent* = traits > clonable . name . age | 'test' printLine ) | ) > lobby > "Self 1" I'm trying to make a method object and it seems to succeed,am I right about the following; 1. defining a method object doesn't seem to run it's code? (I don't get 'test' returned). Instead if I do 'lobby person5'. - Am I right that method objects only 'run' when called, and their slots only exist for the duration of their action? 2. method objects don't get an outliner - because they only exist as objects while they are running. This seems to indicate I can't create an object that can do some setup code when it is created? Am I missing a trick? Thanks for your help. Stephen -- Stephen De Gabrielle stephen.degabrielle(a)acm.org Telephone +44 (0)20 85670911 Mobile +44 (0)79 85189045 http://www.degabrielle.name/stephen On Fri, Mar 14, 2014 at 12:02 AM, Jecel Assumpcao Jr. <jecel(a)merlintec.com>wrote: > > > Stephen, > > > > I've had some success creating objects from Morphic and using the > > outliner, but I can't seem to create objects using the shell. > > I have tried typing in some of the early examples in the handbook, > > but I only get errors. > > > > Person = (|name. age | 'test' printLn ) > > > > Example from handbook > > > > > > ( | slot1. slot2 | ?here is some code? printLine ) > > That is an example of a method, not a normal object. A normal object > would look like: > > ( | slot1. slot2 | ) > > The second problem is that you are defining a slot at the prompt, but > slots can only be defined inside some object. Before we had a GUI (Self > 4) we did stuff like: > > lobby _AddSlotsIfAbsent: ( | person = ( | name. age. test = ( 'test' > printLn ) | ) | ) > > If that works, you can then try this at the prompt: > > person test > > This example has at least two problems even if it works: the new person > object doesn't have any traits, like clonable or oddball and its data > slots don't have any interesting values, only nil. > > -- Jecel > > >

2 2

Loading Shared Libraries
by ardeujho 14 Mar '14

14 Mar '14

In 'objects/applications/serverDemo' there is an example of loading a shared library and using it from Self. The code has bitrot. The makefile no longer works and the binaries in there are old. I've put a patch that gets this working on Mac OS X at: <http://pastebin.ca/2662490> Also required is a fix to a core file: <http://pastebin.ca/2662493> Once that patch is applied you'll need to reload that file into your image (or build a new one). That patch fixes an issue where Self prepends an underscore to foreign function symbol names before lookup. I haven't done a git pull request yet as I want to get this working on Linux as well first but I'm traveling for a couple of days and won't get around to that. I'm mentioning it here to get feedback before I add other platform support. To do the demo, after doing a Mac OS X build of Self, change to the 'objects/applications/serverDemo' directory and run 'make'. This will build the shared library and C client program. To run, start Self from the 'objects' directory: $ cd objects $ ../vm/Self.app/Contents/MacOS/Self -s myimage.snap Then load the server.self file: Self> bootstrap read: 'server' From: 'applications/serverDemo'. Run the server: Self> (message copy receiver: serverDemo server Selector: start) fork. This will block as it runs a C 'accept' call exposed by the shared library. In another terminal window run the client to send Self comands to the server: $ ./toself "'hello' print" $ ./toself _Quit Should this demo be kept the same? It uses the lower level Glue functions directly? Or would it be better to modify it to use primitiveMaker? Or maybe do a 'serverDemo2' that uses primitiveMaker to compare the approaches? Is there a better way to 'fork' the server? The example given in a server.self comment uses a block but this fails due to non-LIFO blocks.

1 0

Creating objects from the shell
by Stephen De Gabrielle 14 Mar '14

14 Mar '14

Hi, I've had some success creating objects from Morphic and using the outliner, but I can't seem to create objects using the shell. I have tried typing in some of the early examples in the handbook, but I only get errors. Person = (|name. age | 'test' printLn ) Example from handbook ( | slot1. slot2 | ’here is some code’ printLine ) -- Sent from Gmail Mobile

2 1

FFI callbacks
by ardeujho 13 Mar '14

13 Mar '14

Is it possible to call Self methods from C glue code? I know Self can call C code but I'd like to be able to pass a Self block or object to a C function and have it call it. Is this possible?

3 4

Copy down slots
by ardeujho 12 Mar '14

12 Mar '14

I'm having trouble getting copy down slots working and I'm wondering if I'm doing something wrong. These are the steps I took: 1) copy a new object (| parent*=traits clonable. x. |) 2) Set that as 'myfoo' in globals and set the creator path to 'myfoo'. 3) Do a "Subclass me" on the myfoo object. Use defaults for the prompts. ...the new object appears with the copy down'd 'x' slot... 4) Add a new slot 'y' to 'myfoo' Here I expected 'y' to appear in my subclassed object but it did not. This also does not work if I manually create an object and use the copy down parent annotations. I see the copy down slots but adding a new slot in the parent doesn't make it appear in the child. If I add a field to 'set' though I do see it propagate to 'dictionary'. Is my own created objects missing some annotation to make this work?

1 1

Re: [self-interest] A question about method slots
by David Ungar 11 Mar '14

11 Mar '14

Methods are immutable, BTW. When you edit one in the debugger it does a "become". On Mar 8, 2014, at 12:52 PM, Jecel Assumpcao Jr. <jecel(a)merlintec.com> wrote: > Michael, > > > But what about a method slot? If I edit a method in a clone, the > > original is /not/ affected. > > This is a GUI issue, not a language one. The idea is that the most > common use for editing a method in an object that also exists in other > objects in the same "clone family" is that you now want it to be > different. You want to start a new clone family. > > Note that if you edit the exact same method in a debugger instead of an > outliner then your change *will* be applied to all the objects in the > clone family. The idea here is that you found a bug and so all copies of > the method should be changed. So these two tools act very differently > but the underlying language is the same. > > "Read-only" means for normal use. Besides methods, we also have > immutable strings (what would be Symbols in Smalltalk-80). The > programming tools (and the mirror objects they are built on) are not > normal use. If we didn't have this "escape hatch" it would be impossible > to create these read-only objects in the first place. > > -- Jecel > >

4 10

Re: [self-interest] A question about method slots
by David Ungar 10 Mar '14

10 Mar '14

Jecel, Many thanks! - David Sent from my iPhone, tap tap > On Mar 10, 2014, at 3:20 PM, "Jecel Assumpcao Jr." <jecel(a)merlintec.com> wrote: > > Michael, > > > If you want to say, that the statement, > > > > "if you clone an object, the original becomes the parent of the clone." > > > > is not just "rumours", but right in some way, I would like to understand > > this better. > > I am pretty sure David is saying that the statement is wrong. It isn't > very hard to see where the idea came from. If you first read "Using > Prototypical Objects to Implement Shared Behavior in Object-Oriented > Systems" presented by Henry Lieberman at OOPSLA 1986: > > http://web.media.mit.edu/~lieber/Lieberary/OOP/Delegation/Delegation.htm > l > > and then you have that very much in mind as you read "Self: The Power of > Simplicity" presented by David Ungar and Randall Smith at the next > OOPSLA: > > http://bibliography.selflanguage.org/self-power.html > > it is *very* easy to see stuff in the paper that really isn't there. You > have to remember that this is the extent of the experience that most > people who discuss Self have had with the language. On top of that we > had "SELF and the Origins of NewtonScript" by Walter Smith (Originally > published in PIE Developers magazine, July 1994): > > http://waltersmith.us/newton/SELF%20and%20the%20Origins%20of%20NewtonScr > ipt.pdf > > which reinforced the idea that the Lieberman model in NewtonScript had > come from Self. I know that Steve Dekorte thought so for quite a while > after having created Io based on NewtonScript. And people often look at > the title "Parents are Shared Parts: Inheritance and Encapsulation in > Self" by Craig Chambers, David Ungar, Bay-Wei Chang and Urs Hölzle (Lisp > and Symbolic Computation 4(3), Kluwer Academic Publishers, June, 1991): > > http://bibliography.selflanguage.org/parents-shared-parts.html > > without reading the actual paper (and often not even the abstract) and > it confirms their idea that Self uses the Lieberman differential > prototype model. But Self's model comes from ARK (The Alternate Reality > Kit by Randall Smith) instead where new objects are perfect clones of > their prototypes instead of empty children of the same: > > http://www.open-video.org/details.php?videoid=8050 > > In short: it is reasonable that people think that in Self "if you clone > an object, the original becomes the parent of the clone." but they are > wrong. > > -- Jecel > >

1 0

What does 'Do ``userDefinedOperation'' do?
by Stephen De Gabrielle 10 Mar '14

10 Mar '14

Hi, What does the menu item [Do ``userDefinedOperation''] do? It is found in the middle-button(yellow) menu when clicked on an outliner. Kind regards, Stephen -- Stephen De Gabrielle

2 2

Re: [self-interest] A question about method slots
by David Ungar 09 Mar '14

09 Mar '14

Interesting! Is this the right link? Was hoping for a drawing of the idea. Thank you, - David Sent from my iPhone, tap tap > On Mar 9, 2014, at 1:18 PM, "Jecel Assumpcao Jr." <jecel(a)merlintec.com> wrote: > > David Ungar wrote: > > Methods are immutable, BTW. When you edit one in the debugger > > it does a "become". > > Indeed. Which is why an interesting alternative to the red and green > buttons in the method editor for the outliners would be to pop up a > dedicated editor hovering just above the original method. You could make > some changes and then either just throw it away (the equivalent of the > red button) or you could drop it down on the outliner below, were it > would atomically become the new version of the method (like pressing the > green button). > > This might seem a little awkward, but would visually more closely > present the story we want to tell. With the red and green buttons you > see a method in the outliner which doesn't correspond to what will run > if you send the message to it right now. But I had already made this > suggestion in 1995: > > http://www.merlintec.com/old-self-interest/msg00718.html > > -- Jecel > >

2 1

2024

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 March 2014