On 26 Jan 2015, at 07:52, Hernán Morales Durand <hernan.morales@gmail.com> wrote:

Marcus, I don't see it. Why an UI tool defect should impact Behavior?

my problem is a bit that I have so much to do (there are 672 other open issues…) that I can’t even think about this, so I will ignore it for now.

2015-01-26 3:49 GMT-03:00 Marcus Denker <marcus.denker@inria.fr <mailto:marcus.denker@inria.fr>>:

...

On 26 Jan 2015, at 07:44, Hernán Morales Durand <hernan.morales@gmail.com <mailto:hernan.morales@gmail.com>> wrote:

I want to discuss case 14789 (https://pharo.fogbugz.com/f/cases/14789/Nautilus-assumes-a-class-always-unde... <https://pharo.fogbugz.com/f/cases/14789/Nautilus-assumes-a-class-always-unde...>) because a class should be browseable even without a definition. Anonymous classes are still classes.

If Nautilus project cannot deal with the whole Smalltalk power then it is poorly designed and I wonder what's the advantage over the old OmniBrowser...

It means pushing a lot of API down to Behaviour. Do we want that?

Marcus

Hi Hernàn, I'm curious. How are the anonymous behaviours found so they can be shown in the browser? On Mon, Jan 26, 2015 at 12:20 AM, Hernán Morales Durand < hernan.morales@gmail.com> wrote:

2015-01-26 5:04 GMT-03:00 Nicolai Hess <nicolaihess@web.de>:

...

2015-01-26 7:44 GMT+01:00 Hernán Morales Durand <hernan.morales@gmail.com

...

:

...

I want to discuss case 14789 ( https://pharo.fogbugz.com/f/cases/14789/Nautilus-assumes-a-class-always-unde...) because a class should be browseable even without a definition. Anonymous classes are still classes.

If Nautilus project cannot deal with the whole Smalltalk power then it is poorly designed and I wonder what's the advantage over the old OmniBrowser...

Hernán

Can you show what the old OmniBrowser is displaying for Behavior new new browse

See attach (Image is Pharo 1.2.2 #12353). The Behavior definition is not updated but one can browse the behaviors opening first the Instances Browser and then finding the class. Code is here: http://www.smalltalkhub.com/#!/~hernan/LightweightClasses

...

I can not check this in Pharo, as I could not find a way to load OmniBrowser in that image.

I doubt it could be ported back without significan effort :(

...

nicolai

-- best, Eliot

On Mon, Jan 26, 2015 at 12:29 PM, Hernán Morales Durand < hernan.morales@gmail.com> wrote:

Hi Eliot.

I iterated using #allInstances in LCBLightweightInstancesClassNode>>classInstances The are conditionals there because instances of Array and ContextPart were created on the fly and then selecting them in the Browser created infinite recursion.

Depending on how many anonymous behaviours there are and how frequent they're browsed you might think of keeping them in a weak set, and enumerating that.

Hernán

2015-01-26 15:33 GMT-03:00 Eliot Miranda <eliot.miranda@gmail.com>:

Hi Hernàn,

...

I'm curious. How are the anonymous behaviours found so they can be shown in the browser?

On Mon, Jan 26, 2015 at 12:20 AM, Hernán Morales Durand < hernan.morales@gmail.com> wrote:

...

2015-01-26 5:04 GMT-03:00 Nicolai Hess <nicolaihess@web.de>:

...

2015-01-26 7:44 GMT+01:00 Hernán Morales Durand < hernan.morales@gmail.com>:

...

I want to discuss case 14789 ( https://pharo.fogbugz.com/f/cases/14789/Nautilus-assumes-a-class-always-unde...) because a class should be browseable even without a definition. Anonymous classes are still classes.

If Nautilus project cannot deal with the whole Smalltalk power then it is poorly designed and I wonder what's the advantage over the old OmniBrowser...

Hernán

Can you show what the old OmniBrowser is displaying for Behavior new new browse

See attach (Image is Pharo 1.2.2 #12353). The Behavior definition is not updated but one can browse the behaviors opening first the Instances Browser and then finding the class. Code is here: http://www.smalltalkhub.com/#!/~hernan/LightweightClasses

...

I can not check this in Pharo, as I could not find a way to load OmniBrowser in that image.

I doubt it could be ported back without significan effort :(

...

nicolai

-- best, Eliot

-- best, Eliot

Hi Stef 2015-01-31 6:13 GMT-03:00 stepharo <stepharo@free.fr>:

Hernan

I would really love that you think twice before pressing the button sent mail. Else I will finish to put you on my kill list and I will be sad but so far I see from you often frustration. Could you try to be positive? You will see as a state of mind this is so much more pleasant.

I will reply to this by private e-mail.

There are many aspects of Nautilus that I do not like but there are even more aspect about the old browser and OB that I hate! And as you know doing is difficult.

You know I wrote several OB browsers so I know OB had problems. OB was not *that* bad. The old browser was. But I just spent too much energy on OB.

So what is the point. How many anonymous classes did I created over the last five year: none.

We have different POV and use cases.

What do we do with class side? packages?

Why do you expect that a tools supporting the Smalltalk metaclass model should suport Anonymous classes?

Because tools should be flexible enough to support other similar artifacts (like anonymous classes) at least without too much modification. How we can evolve if tools are not adaptable?

BTW I'm the guy in Squeak that fixed long long long time ago Behavior new new (So I know a bit on anonymous classes). When I used anonymous classes I did it to support instance based programming. So should I expect my hack to work with normal browser?

When I think about halting SystemWindow>>delete automatically I miss the instance-based behavior. I do not consider it a hack. Now Glamour could be a solution? It would support building tools for all the other experiments (MethodWrappers, Ghosts, Multi-Methods, Wide Classes, etc)? Now may be the anonymous classes you create should return a dummy package.

I think that would be a good solution.

Stef

Le 26/1/15 07:44, Hernán Morales Durand a écrit :

I want to discuss case 14789 (https://pharo.fogbugz.com/f/

...

cases/14789/Nautilus-assumes-a-class-always-understand-package) because a class should be browseable even without a definition. Anonymous classes are still classes.

If Nautilus project cannot deal with the whole Smalltalk power then it is poorly designed and I wonder what's the advantage over the old OmniBrowser...

Hernán

I will reply to this by private e-mail.

There are many aspects of Nautilus that I do not like but there are even more aspect about the old browser and OB that I hate! And as you know doing is difficult.

You know I wrote several OB browsers so I know OB had problems. OB was not *that* bad. The old browser was. But I just spent too much energy on OB.

:)

What do we do with class side? packages?

Why do you expect that a tools supporting the Smalltalk metaclass model should suport Anonymous classes?

Because tools should be flexible enough to support other similar artifacts (like anonymous classes) at least without too much modification. How we can evolve if tools are not adaptable?

now classes belongs to a package so the class browser assume that your class must belong to a package.

BTW I'm the guy in Squeak that fixed long long long time ago Behavior new new (So I know a bit on anonymous classes). When I used anonymous classes I did it to support instance based programming. So should I expect my hack to work with normal browser?

When I think about halting SystemWindow>>delete automatically I miss the instance-based behavior. I do not consider it a hack.

Now Glamour could be a solution? It would support building tools for all the other experiments (MethodWrappers, Ghosts, Multi-Methods, Wide Classes, etc)?

May be. :)