Smalltalk › Squeak › Squeak - Dev

Traits in use (was Re: Re: Squeak and Namespaces)

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

18 messages Options

Avi Bryant

Traits in use (was Re: Re: Squeak and Namespaces)

On 12/1/06, J J <[hidden email]> wrote:

> Yes, the inheritance model without them has some real problems. And finding
> a solution is tough. C++'s multiple inheritance? Too complicated. Java's
> interfaces? Too much work and code duplication. So now smalltalk has a
> solution, and IMO the best one.

Is that opinion based on real-world use of Traits? Now that they've
been in the base image for a decent length of time, I'm curious to
hear of any war stories from using Traits outside of a research
context. I've never used them in anger myself, so it would be very
interesting to hear from those who have.

Avi

timrowledge

Re: Traits in use (was Re: Re: Squeak and Namespaces)

On 1-Dec-06, at 11:05 AM, Avi Bryant wrote:

> On 12/1/06, J J <[hidden email]> wrote:
>
>> Yes, the inheritance model without them has some real problems.
>> And finding
>> a solution is tough. C++'s multiple inheritance? Too
>> complicated. Java's
>> interfaces? Too much work and code duplication. So now smalltalk
>> has a
>> solution, and IMO the best one.
>
> Is that opinion based on real-world use of Traits? Now that they've
> been in the base image for a decent length of time, I'm curious to
> hear of any war stories from using Traits outside of a research
> context. I've never used them in anger myself, so it would be very
> interesting to hear from those who have.

I've not tried to use them but I did trip over them when working on
cleaning up the compiled method source pointer mess (which I still
haven't finished, rats). The problem is that the tool support appears
to be missing, which is a real pain. For example, I edit a trait-
method in one place that it is incorporated (imported? linked?
whatever) and there is nothing that tells me it is a trait, or asks
me if I want to make my change a local version or a global one or
anything at all. No matter how good an idea is, without proper
integration it ends up causing problems.

tim
--
tim Rowledge; [hidden email]; http://www.rowledge.org/tim
Useful random insult:- Pedaling real fast, but not getting anywhere.

Lukas Renggli

Re: Re: Traits in use (was Re: Re: Squeak and Namespaces)

> I've not tried to use them but I did trip over them when working on
> cleaning up the compiled method source pointer mess (which I still
> haven't finished, rats). The problem is that the tool support appears
> to be missing, which is a real pain. For example, I edit a trait-
> method in one place that it is incorporated (imported? linked?
> whatever) and there is nothing that tells me it is a trait, or asks
> me if I want to make my change a local version or a global one or
> anything at all.

OmniBrowser has a small extension in 3.9 that shows imported methods
in italic, redefined and locally defined ones with normal fonts. This
helps a little bit, but I agree it does not help more than for toy
examples: very soon you end up not understanding anymore what is going
on.

I wonder if the virtual method-categories of Damien Cassou would help
here? In the traits model all the info is there, it just needs to be
displayed in the browser nicely. Should be a matter of minutes to
define those method-categories (requires, supplies, overrides) to get
a useable trait browser, no?

And to answer the initial question: no, unfortunately not.

Cheers,
Lukas

--
Lukas Renggli
http://www.lukas-renggli.ch

timrowledge

Re: Traits in use (was Re: Re: Squeak and Namespaces)

On 1-Dec-06, at 11:32 AM, Lukas Renggli wrote:

>
> OmniBrowser has a small extension in 3.9 that shows imported methods
> in italic, redefined and locally defined ones with normal fonts. This
> helps a little bit, but I agree it does not help more than for toy
> examples: very soon you end up not understanding anymore what is going
> on.

Anything relying on extremely minor cues like that is not going to
work for many people. It has to be clear and informative and include
added/changed menus etc. Anything much less than full integration
with documentation to assist will just mean that at the very best a
new capability like this will get ignored. Most likely it will be
misused through ignorance and cause bugs down the line. Surely nobody
wants that?

tim
--
tim Rowledge; [hidden email]; http://www.rowledge.org/tim
Logic: The art of being wrong with confidence...

Philippe Marschall

Re: Re: Traits in use (was Re: Re: Squeak and Namespaces)

In reply to this post by Lukas Renggli

2006/12/1, Lukas Renggli <[hidden email]>:

> > I've not tried to use them but I did trip over them when working on
> > cleaning up the compiled method source pointer mess (which I still
> > haven't finished, rats). The problem is that the tool support appears
> > to be missing, which is a real pain. For example, I edit a trait-
> > method in one place that it is incorporated (imported? linked?
> > whatever) and there is nothing that tells me it is a trait, or asks
> > me if I want to make my change a local version or a global one or
> > anything at all.
>
> OmniBrowser has a small extension in 3.9 that shows imported methods
> in italic, redefined and locally defined ones with normal fonts. This
> helps a little bit, but I agree it does not help more than for toy
> examples: very soon you end up not understanding anymore what is going
> on.
>
> I wonder if the virtual method-categories of Damien Cassou would help
> here? In the traits model all the info is there, it just needs to be
> displayed in the browser nicely. Should be a matter of minutes to
> define those method-categories (requires, supplies, overrides) to get
> a useable trait browser, no?
>
> And to answer the initial question: no, unfortunately not.

For the n-th time:

TraitsOmniBrowser (not the one in the image): that's what virtual
categories come from in the first place anyway.

Philipe

Klaus D. Witzel

Re: Traits in use (was Re: Re: Squeak and Namespaces)

In reply to this post by Avi Bryant

Hi Avi,

I do not believe that you haven't used the trait concept for real-world
problems (no offense).

Instead, whenever you factored some methods such that instance variables
and class variables are on accessors, you have traitified these methods.
Every other class, which provides the respective accessors, can (re-)use
these traitified methods (modulo that that makes sense for an application).

Stateless traits is nothing much more. But the Trait system in Squeak
comes with Trait *composition* and composition is something which perhaps
was not used before (perhaps not intentionally). Doing Trait composition
manually, without using the supportive code in Squeak, is like gcc -S
followed by cut&paste, IMHO ;-)

/Klaus

On Fri, 01 Dec 2006 20:05:38 +0100, Avi Bryant wrote:

> On 12/1/06, J J <[hidden email]> wrote:
>
>> Yes, the inheritance model without them has some real problems. And
>> finding
>> a solution is tough. C++'s multiple inheritance? Too complicated.
>> Java's
>> interfaces? Too much work and code duplication. So now smalltalk has a
>> solution, and IMO the best one.
>
> Is that opinion based on real-world use of Traits? Now that they've
> been in the base image for a decent length of time, I'm curious to
> hear of any war stories from using Traits outside of a research
> context. I've never used them in anger myself, so it would be very
> interesting to hear from those who have.
>
> Avi
>
>

J J-6

RE: Traits in use (was Re: Re: Squeak and Namespaces)

In reply to this post by Avi Bryant

Well honestly no. I have been working on smaller things at the moment so I
haven't yet built anything big enough to run into the problems one has with
single inheritance. My oppinion is purely based on the papers, but it did
seem like the best fit I have seen yet.

In anger?

>From: "Avi Bryant" <[hidden email]>
>Reply-To: [hidden email], The general-purpose Squeak developers
>list<[hidden email]>
>To: "The general-purpose Squeak developers
>list"<[hidden email]>
>Subject: Traits in use (was Re: Re: Squeak and Namespaces)
>Date: Fri, 1 Dec 2006 11:05:38 -0800
>
>On 12/1/06, J J <[hidden email]> wrote:
>
>>Yes, the inheritance model without them has some real problems. And
>>finding
>>a solution is tough. C++'s multiple inheritance? Too complicated.
>>Java's
>>interfaces? Too much work and code duplication. So now smalltalk has a
>>solution, and IMO the best one.
>
>Is that opinion based on real-world use of Traits? Now that they've
>been in the base image for a decent length of time, I'm curious to
>hear of any war stories from using Traits outside of a research
>context. I've never used them in anger myself, so it would be very
>interesting to hear from those who have.
>
>Avi
>

_________________________________________________________________
View Athletes Collections with Live Search
http://sportmaps.live.com/index.html?source=hmemailtaglinenov06&FORM=MGAC01

J J-6

Re: Traits in use (was Re: Re: Squeak and Namespaces)

In reply to this post by timrowledge

>From: tim Rowledge <[hidden email]>
>Reply-To: The general-purpose Squeak developers
>list<[hidden email]>
>To: [hidden email], The general-purpose Squeak developers
>list<[hidden email]>
>Subject: Re: Traits in use (was Re: Re: Squeak and Namespaces)
>Date: Fri, 1 Dec 2006 11:16:28 -0800
>
>
>On 1-Dec-06, at 11:05 AM, Avi Bryant wrote:
>
>>On 12/1/06, J J <[hidden email]> wrote:
>>
>>>Yes, the inheritance model without them has some real problems. And
>>>finding
>>>a solution is tough. C++'s multiple inheritance? Too complicated.
>>>Java's
>>>interfaces? Too much work and code duplication. So now smalltalk has a
>>>solution, and IMO the best one.
>>
>>Is that opinion based on real-world use of Traits? Now that they've
>>been in the base image for a decent length of time, I'm curious to
>>hear of any war stories from using Traits outside of a research
>>context. I've never used them in anger myself, so it would be very
>>interesting to hear from those who have.
>
>I've not tried to use them but I did trip over them when working on
>cleaning up the compiled method source pointer mess (which I still haven't
>finished, rats). The problem is that the tool support appears to be
>missing, which is a real pain. For example, I edit a trait- method in one
>place that it is incorporated (imported? linked? whatever) and there is
>nothing that tells me it is a trait, or asks me if I want to make my
>change a local version or a global one or anything at all. No matter how
>good an idea is, without proper integration it ends up causing problems.
>
>
>tim
>--
>tim Rowledge; [hidden email]; http://www.rowledge.org/tim
>Useful random insult:- Pedaling real fast, but not getting anywhere.

I was under the impression that there was a "trait's aware" browser in the
image, or at least loadable from squeak map or something. Am I wrong?

_________________________________________________________________
Share your latest news with your friends with the Windows Live Spaces
friends module.
http://clk.atdmt.com/MSN/go/msnnkwsp0070000001msn/direct/01/?href=http://spaces.live.com/spacesapi.aspx?wx_action=create&wx_url=/friends.aspx&mk

Avi Bryant

Re: Re: Traits in use (was Re: Re: Squeak and Namespaces)

In reply to this post by Klaus D. Witzel

On 12/1/06, Klaus D. Witzel <[hidden email]> wrote:
> Hi Avi,
>
> I do not believe that you haven't used the trait concept for real-world
> problems (no offense).

Klaus,

No offense, but I was asking for practice and you're giving me more
theory. My question wasn't about traits. It was about Traits.
There's a big difference between using a concept/pattern, and using a
particular implementation, and I was asking about the latter. Yes,
given that we use traits all the time, Traits ought to be a big win,
but I'd love to see some evidence to confirm that.

Avi

Avi Bryant

Re: RE: Traits in use (was Re: Re: Squeak and Namespaces)

In reply to this post by J J-6

On 12/1/06, J J <[hidden email]> wrote:

> In anger?

I assume that expression comes from weapons - the difference between
shooting at targets and shooting at people. If you'll forgive the
bloodthirsty metaphor, Traits makes some nice holes in paper
silhouettes but I'm looking for the bodies.

Avi

J J-6

Re: RE: Traits in use (was Re: Re: Squeak and Namespaces)

Haha. Comming from the USA, the land of South Park, etc., I don't get
offended by bloodthirsty or even shocking metaphors. :)

Yes, at some point I hope to get in a project big enough to test drive them,
but it could be a while. The closest thing I have seen (and what convinced
me they were a good idea) was the description of how they are used with the
Collection classes. And I was under the impression this was in on 3.9?

_________________________________________________________________
Get the latest Windows Live Messenger 8.1 Beta version. Join now.
http://ideas.live.com

Klaus D. Witzel

Re: Re: Traits in use (was Re: Re: Squeak and Namespaces)

In reply to this post by Avi Bryant

Avi, what about 3.9's

ProtoObject #()
Object #()

Behavior #('superclass' 'methodDict' 'format')
ClassDescription #('instanceVariables' 'organization')
Class #('subclasses' 'name' 'classPool' 'sharedPools' 'environment'
'category' 'traitComposition' 'localSelectors')
[ ... all the Metaclasses ... ]
Metaclass #('thisClass' 'traitComposition' 'localSelectors')

There it is, a particular non-trivial implementation, in form of open
heart surgery ;-)

Okay the work on Collection (in the paper) doesn't count that much *but*
the above is real-world.

/Klaus

On Fri, 01 Dec 2006 22:53:30 +0100, Av wrote:

> On 12/1/06, Klaus D. Witzel <[hidden email]> wrote:
>> Hi Avi,
>>
>> I do not believe that you haven't used the trait concept for real-world
>> problems (no offense).
>
> Klaus,
>
> No offense, but I was asking for practice and you're giving me more
> theory. My question wasn't about traits. It was about Traits.
> There's a big difference between using a concept/pattern, and using a
> particular implementation, and I was asking about the latter. Yes,
> given that we use traits all the time, Traits ought to be a big win,
> but I'd love to see some evidence to confirm that.
>
> Avi
>
>

Avi Bryant

Re: Re: Re: Traits in use (was Re: Re: Squeak and Namespaces)

On 12/1/06, Klaus D. Witzel <[hidden email]> wrote:

> There it is, a particular non-trivial implementation, in form of open
> heart surgery ;-)

Did you implement it? I'm looking for experiences, not artifacts.
And since this was done as part of the implementation of Traits
themselves, it's not the ideal example.

Avi

Klaus D. Witzel

Re: Traits in use (was Re: Re: Squeak and Namespaces)

On Fri, 01 Dec 2006 23:59:15 +0100, Avi Bryant wrote:
> On 12/1/06, Klaus D. Witzel wrote:
>
>> There it is, a particular non-trivial implementation, in form of open
>> heart surgery ;-)
>
> Did you implement it? I'm looking for experiences, not artifacts.

Stupid me. I should either better understand the difference or go to sleep
:)

Sorry. But I'm shure the Traits developers will speak up.

> And since this was done as part of the implementation of Traits
> themselves, it's not the ideal example.

IC.

/Klaus

> Avi
>
>

stephane ducasse

Re: Traits in use (was Re: Re: Squeak and Namespaces)

In reply to this post by Avi Bryant

Hi avi

People could not test traits before the implementation was not
totally good. Now we have it, there is a first browser
that can help, but if nobody build systems with them we will not
know. I did not want to refactor squeak to use traits
so that people do not complain that we are ....

After the conclusions that we can draw are at the level of our
experimentations.
So I would love to see traits used (successfully or not). Adrian has
a cs that allow you to flatten your code so
the risk is quite low.

Stef

>> Hi Avi,
>>
>> I do not believe that you haven't used the trait concept for real-
>> world
>> problems (no offense).
>
> Klaus,
>
> No offense, but I was asking for practice and you're giving me more
> theory. My question wasn't about traits. It was about Traits.
> There's a big difference between using a concept/pattern, and using a
> particular implementation, and I was asking about the latter. Yes,
> given that we use traits all the time, Traits ought to be a big win,
> but I'd love to see some evidence to confirm that.
>
> Avi
>
>

stephane ducasse

Re: Traits in use (was Re: Re: Squeak and Namespaces)

In reply to this post by Klaus D. Witzel

The real problem is bootstrapping (not the language but the use).
We could not use traits before 3.9 (if you want to have real
experiences and not toys)
And to get traits benefices you need to have multiple inheritance
problems, ie not trivial
cases.

Lukas you mentioned to me once a cool place in seaside where traits
would really help.
But again we cannot change seaside.
So bootstrap.

Stef

Lukas Renggli

Re: Re: Traits in use (was Re: Re: Squeak and Namespaces)

> Lukas you mentioned to me once a cool place in seaside where traits
> would really help. But again we cannot change seaside. So bootstrap.

The canvas framework has a lot of duplicated code and/or code that is
defined too high in the hierarchy. The framework provides an
implementation of the different XHTML tags as specified in [1]. The
abstract element (WATagBrush) should only implement the relevant
methods (#attributeAt:, #attributeAt:put:, etc.) that are specified in
the W3C DOM [2] implementation.

Now XHTML has the so called generic attributes [2] (coreattrs, events,
focus, ...). It also has a notion of what can be nested into each
other. Moreover there is Seaside that defines different kinds of
callbacks, unfortunately not all kinds of callbacks are valid on all
elements. To keep it short, there is a lot of orthogonal things going
on. Tough it would be easy to put these things into traits and compose
them with the right elements. Like this you wouldn't get 11 mostly
identical implementations of #callback:, 7 identical implementations
of #on:of:, 4 identical implementations of #form:, #extnet:, #width:,
#height:, ...

Cheers,
Lukas

[1] http://www.w3.org/TR/2002/REC-xhtml1-20020801/dtds.html
[2] http://www.w3.org/TR/REC-DOM-Level-1/idl-definitions.html

--
Lukas Renggli
http://www.lukas-renggli.ch

Klaus D. Witzel

Re: Traits in use (was Re: Squeak and Namespaces)

Now that you mention xhtml: Todd Blanchard's HTMLCSS is a very
professional framework

- http://www.squeaksource.com/htmlcssparser

and to his HTML*Node class hierarchy I've added Traits, along the lines of

- http://www.w3.org/TR/xhtml-modularization/abstract_modules.html
- http://www.w3.org/TR/html4/sgml/dtd.html

Not much to show, yet. Mainly because I wait for stronger tool support
(lazy me ;-)

But send me an email if you want the source code.

/Klaus

On Sat, 02 Dec 2006 10:00:50 +0100, Lukas Renggli wrote:

>> Lukas you mentioned to me once a cool place in seaside where traits
>> would really help. But again we cannot change seaside. So bootstrap.
>
> The canvas framework has a lot of duplicated code and/or code that is
> defined too high in the hierarchy. The framework provides an
> implementation of the different XHTML tags as specified in [1]. The
> abstract element (WATagBrush) should only implement the relevant
> methods (#attributeAt:, #attributeAt:put:, etc.) that are specified in
> the W3C DOM [2] implementation.
>
> Now XHTML has the so called generic attributes [2] (coreattrs, events,
> focus, ...). It also has a notion of what can be nested into each
> other. Moreover there is Seaside that defines different kinds of
> callbacks, unfortunately not all kinds of callbacks are valid on all
> elements. To keep it short, there is a lot of orthogonal things going
> on. Tough it would be easy to put these things into traits and compose
> them with the right elements. Like this you wouldn't get 11 mostly
> identical implementations of #callback:, 7 identical implementations
> of #on:of:, 4 identical implementations of #form:, #extnet:, #width:,
> #height:, ...
>
> Cheers,
> Lukas
>
> [1] http://www.w3.org/TR/2002/REC-xhtml1-20020801/dtds.html
> [2] http://www.w3.org/TR/REC-DOM-Level-1/idl-definitions.html
>