The Inbox: Traits-pre.307.mcz

A new version of Traits was added to project The Inbox:
http://source.squeak.org/inbox/Traits-pre.307.mcz

==================== Summary ====================

Name: Traits-pre.307
Author: pre
Time: 20 November 2015, 2:58:26.562 pm
UUID: f0955b2d-775c-4862-85b9-5d6e616cd2e4
Ancestors: Traits-eem.306

This fixes http://bugs.squeak.org/view.php?id=7767 which is about the - and @ operator of Trait compositions ignoring brackets. This implementation uses the normal left to right evaluation order of Smalltalk to avoid confusions about the way trait composition expressions are evaluated.

=============== Diff against Traits-eem.306 ===============

Item was changed:
  ----- Method: TraitAlias>>printOn: (in category 'operations') -----
  printOn: s
  "Answer the trait composition string (used for class definitions)"
+ s nextPutAll: '('.
  s nextPutAll: subject asString.
  s nextPutAll: ' @ {'.
  aliases do:[:assoc| s print: assoc] separatedBy:[s nextPutAll:'. '].
  s nextPutAll: '}'.
+ s nextPutAll: ')'.!
- !

Item was changed:
  ----- Method: TraitComposition>>- (in category 'converting') -----
  - anArray
- "the modifier operators #@ and #- bind stronger than +.
- Thus, #@ or #- sent to a sum will only affect the most right summand"
 
+ self updateTraits: (self traitsCollect: [:aTrait | aTrait - anArray ])!
- self addLast: (self removeLast - anArray)!

Item was changed:
  ----- Method: TraitComposition>>@ (in category 'converting') -----
  @ anArrayOfAssociations
+
+ self updateTraits: (self traitsCollect: [:aTrait | aTrait @ anArrayOfAssociations ])!
- "the modifier operators #@ and #- bind stronger than +.
- Thus, #@ or #- sent to a sum will only affect the most right summand"
-
- self addLast: (self removeLast @ anArrayOfAssociations)!

Item was added:
+ ----- Method: TraitComposition>>traitsCollect: (in category 'accessing') -----
+ traitsCollect: aBlock
+ ^self collect: [:each| each traitsDo: aBlock]!

Item was changed:
  ----- Method: TraitComposition>>traitsDo: (in category 'accessing') -----
  traitsDo: aBlock
+ ^self do: [:each| each traitsDo: aBlock]!
- ^self do:[:each| each traitsDo: aBlock]!

Item was added:
+ ----- Method: TraitComposition>>updateTraits: (in category 'converting') -----
+ updateTraits: aCollection
+
+ self
+ removeAll;
+ addAll: aCollection!

Item was changed:
  ----- Method: TraitExclusion>>printOn: (in category 'composition') -----
  printOn: aStream
  "Answer the trait composition string (used for class definitions)"
+ aStream nextPutAll: '('.
  aStream nextPutAll: subject asString.
  aStream nextPutAll: ' - {'.
  exclusions asArray sort do:[:exc| aStream store: exc] separatedBy:[aStream nextPutAll: '. '].
+ aStream nextPutAll: '}'.
+ aStream nextPutAll: ')'.!
- aStream nextPutAll: '}'.!

I want to add some more reasoning to this commit. The commit is intended to fix this issue: http://bugs.squeak.org/view.php?id=7767

My implementation is based on the corresponding Pharo Traits implementation, except for the TraitsComposition>>#printOn: implementation. Do you think it is worth also recreating that behavior?

In general, the most important thing regarding this change is, that this might break older .st or .cs files containing Traits definitions based on the old evaluation order.

Bests
Patrick
________________________________________
Von: [hidden email] <[hidden email]> im Auftrag von [hidden email] <[hidden email]>
Gesendet: Freitag, 20. November 2015 09:52
An: [hidden email]
Betreff: [squeak-dev] The Inbox: Traits-pre.307.mcz

A new version of Traits was added to project The Inbox:
http://source.squeak.org/inbox/Traits-pre.307.mcz

==================== Summary ====================

Name: Traits-pre.307
Author: pre
Time: 20 November 2015, 2:58:26.562 pm
UUID: f0955b2d-775c-4862-85b9-5d6e616cd2e4
Ancestors: Traits-eem.306

This fixes http://bugs.squeak.org/view.php?id=7767 which is about the - and @ operator of Trait compositions ignoring brackets. This implementation uses the normal left to right evaluation order of Smalltalk to avoid confusions about the way trait composition expressions are evaluated.

=============== Diff against Traits-eem.306 ===============

Item was changed:
  ----- Method: TraitAlias>>printOn: (in category 'operations') -----
  printOn: s
        "Answer the trait composition string (used for class definitions)"
+       s nextPutAll: '('.
        s nextPutAll: subject asString.
        s nextPutAll: ' @ {'.
        aliases do:[:assoc| s print: assoc] separatedBy:[s nextPutAll:'. '].
        s nextPutAll: '}'.
+       s nextPutAll: ')'.!
- !

Item was changed:
  ----- Method: TraitComposition>>- (in category 'converting') -----
  - anArray
-       "the modifier operators #@ and #- bind stronger than +.
-       Thus, #@ or #- sent to a sum will only affect the most right summand"

+       self updateTraits: (self traitsCollect: [:aTrait | aTrait - anArray ])!
-       self addLast: (self removeLast - anArray)!

Item was changed:
  ----- Method: TraitComposition>>@ (in category 'converting') -----
  @ anArrayOfAssociations
+
+       self updateTraits: (self traitsCollect: [:aTrait | aTrait @ anArrayOfAssociations ])!
-       "the modifier operators #@ and #- bind stronger than +.
-       Thus, #@ or #- sent to a sum will only affect the most right summand"
-
-       self addLast: (self removeLast @ anArrayOfAssociations)!

Item was added:
+ ----- Method: TraitComposition>>traitsCollect: (in category 'accessing') -----
+ traitsCollect: aBlock
+       ^self collect: [:each| each traitsDo: aBlock]!

Item was changed:
  ----- Method: TraitComposition>>traitsDo: (in category 'accessing') -----
  traitsDo: aBlock
+       ^self do: [:each| each traitsDo: aBlock]!
-       ^self do:[:each| each traitsDo: aBlock]!

Item was added:
+ ----- Method: TraitComposition>>updateTraits: (in category 'converting') -----
+ updateTraits: aCollection
+
+       self
+               removeAll;
+               addAll: aCollection!

Item was changed:
  ----- Method: TraitExclusion>>printOn: (in category 'composition') -----
  printOn: aStream
        "Answer the trait composition string (used for class definitions)"
+       aStream nextPutAll: '('.
        aStream nextPutAll: subject asString.
        aStream nextPutAll: ' - {'.
        exclusions asArray sort do:[:exc| aStream store: exc] separatedBy:[aStream nextPutAll: '. '].
+       aStream nextPutAll: '}'.
+       aStream nextPutAll: ')'.!
-       aStream nextPutAll: '}'.!

Should we keep Traits? It was a neat idea that I was happy to support but it got left unfinished. Where are tools to develop & manage Traits? Where is the usage?

Unless there is a compelling reason - and subsequent effort to fill out support - I suggest we should remove them. Along with Islands. And Universes. And probably Environments too, since that has stalled without becoming a proper part of the system.

tim
--
tim Rowledge; [hidden email]; http://www.rowledge.org/tim
Strange OpCodes: IG: Insert Garbage

On Fri, 20 Nov 2015, tim Rowledge wrote:

> Should we keep Traits? It was a neat idea that I was happy to support but it got left unfinished. Where are tools to develop & manage Traits? Where is the usage?

I always considered Traits to be a weak concept. IMHO with less effort,
one could have created a stateful deterministic mixin system which would
have been superior to Traits.

Andreas has replaced the original Traits implementation with his own
"NanoKernel" variant. It's probably the simplest thing that can be
considered as Traits.

If you need tools, then you'll have to look somewhere else. The Pharo guys
might have come up with something over the years.

I think only Pharo and some related projects use Traits, but even there
the use-cases are extremely limited.

>
> Unless there is a compelling reason - and subsequent effort to fill out support - I suggest we should remove them. Along with Islands. And Universes. And probably Environments too, since that has stalled without becoming a proper part of the system.

Backwards compatibility is the only reason for it to be in the image. In
the 4.1 release, it was unloadable. We should check whether it still is.
Sadly I have to agree with you on the rest, too.


Levente

>
> tim
> --
> tim Rowledge; [hidden email]; http://www.rowledge.org/tim
> Strange OpCodes: IG: Insert Garbage
>
>
>
>

I like traits. I use them for documenting extension points/interfaces:

https://github.com/hpi-swa/vivide/tree/master/repository/Vivide.package/TViObjectView.trait
https://github.com/hpi-swa/vivide/tree/master/repository/Vivide.package/TViMemento.trait

Please keep them in the image.

Best,
Marcel

On Fri, Nov 20, 2015 at 12:40 PM, tim Rowledge <[hidden email]> wrote:
> Should we keep Traits? It was a neat idea that I was happy to support but it got left unfinished. Where are tools to develop & manage Traits? Where is the usage?
>
> Unless there is a compelling reason - and subsequent effort to fill out support - I suggest we should remove them. Along with Islands. And Universes. And probably Environments too, since that has stalled without becoming a proper part of the system.

+1.  +1.  +1 and, +1.  Doing the most with the least is the most
beautiful and admirable aspect of Squeak.  Six reserved words,
assignment, and sending messages to objects.  That's pretty much it.
Those concepts alone form the building blocks of the entire IDE, and
still able to push the language with clever hacks like Mixins,
Generators, Promises, Futures, WriteBarriers, and Continuations.

Traits, Slots, Islands, Environments and Pragmas never convinced me
that they deserve to be part of a language this wonderfully sparse.

The then trade-off would be to make them unloadable, right? ;) Just like anyone can unload *-Deprecated packages. This simplifies maintenance.

Best,
Marcel

I would like island to stay also. To me, they are one of the most interesting additions to our base over the recent years. (We use them in Archipelago [*]...)

Best,
Robert

[*] http://www.hpi.uni-potsdam.de/swa/publications/media/SecklerHirschfeld_2014_ArchipelagoAResearchPlatformForComponentInteractionInDistributedApplications_AuthorsVersion.pdf




--
Robert Hirschfeld
[hidden email]
www.hirschfeld.org

On Sat, Nov 21, 2015 at 12:00:08AM -0800, marcel.taeumel wrote:
> The then trade-off would be to make them unloadable, right? ;) Just like
> anyone can unload *-Deprecated packages. This simplifies maintenance.
>

Right. Making these packages reloadable is a very good idea. The idea of
just deleting them seems rather unhelpful to me.

Dave
 

On Sat, Nov 21, 2015 at 06:43:00PM +0100, karl ramberg wrote:
I am not sure what is involved, but is it because of this?

  http://bugs.squeak.org/view.php?id=7814

If so we should try Colin's suggested workaround:

  http://lists.squeakfoundation.org/pipermail/squeak-dev/2014-September/180018.html

Dave

On 20 November 2015 at 18:40, tim Rowledge <[hidden email]> wrote:
> Should we keep Traits? It was a neat idea that I was happy to support but it got left unfinished. Where are tools to develop & manage Traits? Where is the usage?
>
> Unless there is a compelling reason - and subsequent effort to fill out support - I suggest we should remove them. Along with Islands. And Universes. And probably Environments too, since that has stalled without becoming a proper part of the system.

There's definitely a pattern there: someone has a great idea for a
fairly advanced capability, heroically tries to do all the work solo,
or with minimal help from the community, burns out and the work never
gets finished.

Traits, or things close enough to traits that you end up splitting
hairs to tell them apart, are a core feature of so many languages
nowadays (Ruby, Newspeak, Scala, Perl 6, Rust, off the top of my
head), while we let the idea die on the vine, for want of tooling
support. And I'm sure Environments will, too.

Sure, if it's not providing value, and no one's willing to do the
work, just kill the thing and be done. I'd rather see people pitch in
and help _make_ the dang thing a proper part of the system. ("Thing"
here applies mostly to Environments, but Islands and Traits too.) But
I'm also not going to run around pointing fingers: I'm too burned out
to do anything to help, so I'll just shut up now.

frank

> tim
> --
> tim Rowledge; [hidden email]; http://www.rowledge.org/tim
> Strange OpCodes: IG: Insert Garbage
>
>
>

Hi Karl,

On 23.11.2015, at 16:45, karl ramberg <[hidden email]> wrote:

> Software is hard to make. Even harder to make right.
> Picking up other peoples projects can be daunting.
>
> Most of these projects are quite difficult and involve almost as much political as technical skills to do. (Political in the sense as to get the community to agree to changes and adapt)
>
> I'm not sure how to proceed.
>

I have to concur, for every very line.

We should have at least a list of things that people/community think are orphaned… so that we can bail if they are actually not.

Best regards
        -Tobias

On Mon, 23 Nov 2015, Frank Shearar wrote:

Traits are complex: they introduce a new kind of Behavior, one which
doesn't do anything on its own. If Traits were all Classes and all Classes
were Traits, the whole implementation would probably be a lot simpler.
Traits are weak: they only provide a way to share stateless methods
between classes (methods which can't reference variables). If you Traits
were Classes, you could share stateful methods by providing a mapping for
variables, the same way you can do it for methods.

> support. And I'm sure Environments will, too.

Environments is at a point where it needs complex things to be found out
to move forward. The current implementation is incomplete and broken. The
reason why we don't face the problems (so often) is that we only use one
environment.

Levente