a simple question (I think)

I have a strong dislike for incompletely initialised objects.

So I would do it like this:

   myInstVar := TheDependentClass newFor: self.

(together with any other essential information as parameters).

In my Smalltalk system it would look like this:

  Whatever subclass: #TheDependentClass

    instanceConstantNames: 'owner'

    ...

    class methods for: 'instance creation'

      newFor: owner

        ^(self basicNew) pvtOwner: owner; yourself

    methods for: 'initialising'

      pvtOwner: anObject

        super pvtPostNew.

        owner := anObject.

Private access to methods with a 'pvt' prefix is enforced.

Instance constants can only be set in private methods.

There's an additional wrinkle that makes it a run time

error to call TheDependentClass new, so that you cannot get

an incompletely initialised object.

In any Smalltalk you can follow similar conventions even if

they are not enforced.

Of course if you don't mind incompletely initialised objects,

nothing stops you writing code that allows them.  The question

is whether you want to be able to trust the 'owner' field or

not.

On Thu, 18 Jun 2020 at 09:46, Esteban Maringolo <[hidden email]> wrote:

Objects interact with each other by sending messages, although you can
get the sender by using reflection (thisContext sender), it is not a
common practice to do so.
So if you need another object to know the sender (or any other object)
you simply pass it as a parameter.

So modifying your example, and making the syntax actually work it would be:

self property: ObjectClass new.
self property caller: self.

and then when you need to deal with that you can simply get a
reference to that object and send it a message to do something:
self property caller doSomething.

You can also have something like

self property: (ObjectClass for: self)


In that case the #for: message is received by "ObjectClass class",
which you can think of as if it were a Factory or a "static" method
(it is not! it is much more powerful than that).

ObjectClass class>>#for: callerObject
  ^self new
     setCaller: callerObject;
     yourself

And that's it, you have an instance of ObjectClass that you
initialized by sending #for: instead of #new (internally it ended up
calling #new, but from the outside it was just a message send).

And for the most part that's Smalltalk: Objects and messages.

Regards!

Esteban A. Maringolo

On Wed, Jun 17, 2020 at 6:19 PM Russ Whaley <[hidden email]> wrote:
>
> I find myself, over and over, creating a & storing q new object in an instVar, then storing (my)self on that object - so I can have easy access to the 'object that created me.'
>
> <what I do now...>
>
> self instVar := NewClass new.
>
> self instVar callingObject: self.
>
> "where these call the same messages"
>
> self myMessage.
>
> self instVar callingObject myMessage.
>
> <is there a way to...>
>
> self instVar := NewClass new.
>
> self instVar objectWhoCreatedMe myMessage.
>
> "I know there is a 'thisContext sender' - but is there a 'who created me' concept?"
>
>
> Ultimately, there is no real overhead to storing (my)self on a new object, I guess, but as I'm developing and testing, I tend to have a whole lot of 'stranded objects' I have to remove manually.
>
> Thoughts?
>
> Thanks!
> --
> Russ Whaley
> [hidden email]

as a general principle do not use thisContext. 

S. 

On 20 Jun 2020, at 01:12, Russ Whaley <[hidden email]> wrote:

Thank you both for your insights. I’ve done the newFor: style many times, but where it falls short is when I read in a STON object. No worries, I was just wondering if there was something cool like “what object am I sitting on?”... I looked through Context and the client message will be helpful to me in other areas. I’m always eager to learn. Thank you very much for being willing to teach. 

Thanks!

Russ

On Wed, Jun 17, 2020 at 9:05 PM Richard O'Keefe <[hidden email]> wrote:

I have a strong dislike for incompletely initialised objects.

So I would do it like this:

   myInstVar := TheDependentClass newFor: self.

(together with any other essential information as parameters).

In my Smalltalk system it would look like this:

  Whatever subclass: #TheDependentClass

    instanceConstantNames: 'owner'

    ...

    class methods for: 'instance creation'

      newFor: owner

        ^(self basicNew) pvtOwner: owner; yourself

    methods for: 'initialising'

      pvtOwner: anObject

        super pvtPostNew.

        owner := anObject.

Private access to methods with a 'pvt' prefix is enforced.

Instance constants can only be set in private methods.

There's an additional wrinkle that makes it a run time

error to call TheDependentClass new, so that you cannot get

an incompletely initialised object.

In any Smalltalk you can follow similar conventions even if

they are not enforced.

Of course if you don't mind incompletely initialised objects,

nothing stops you writing code that allows them.  The question

is whether you want to be able to trust the 'owner' field or

not.

On Thu, 18 Jun 2020 at 09:46, Esteban Maringolo <[hidden email]> wrote:

Objects interact with each other by sending messages, although you can
get the sender by using reflection (thisContext sender), it is not a
common practice to do so.
So if you need another object to know the sender (or any other object)
you simply pass it as a parameter.

So modifying your example, and making the syntax actually work it would be:

self property: ObjectClass new.
self property caller: self.

and then when you need to deal with that you can simply get a
reference to that object and send it a message to do something:
self property caller doSomething.

You can also have something like

self property: (ObjectClass for: self)


In that case the #for: message is received by "ObjectClass class",
which you can think of as if it were a Factory or a "static" method
(it is not! it is much more powerful than that).

ObjectClass class>>#for: callerObject
  ^self new
     setCaller: callerObject;
     yourself

And that's it, you have an instance of ObjectClass that you
initialized by sending #for: instead of #new (internally it ended up
calling #new, but from the outside it was just a message send).

And for the most part that's Smalltalk: Objects and messages.

Regards!

Esteban A. Maringolo

On Wed, Jun 17, 2020 at 6:19 PM Russ Whaley <[hidden email]> wrote:
>
> I find myself, over and over, creating a & storing q new object in an instVar, then storing (my)self on that object - so I can have easy access to the 'object that created me.'
>
> <what I do now...>
>
> self instVar := NewClass new.
>
> self instVar callingObject: self.
>
> "where these call the same messages"
>
> self myMessage.
>
> self instVar callingObject myMessage.
>
> <is there a way to...>
>
> self instVar := NewClass new.
>
> self instVar objectWhoCreatedMe myMessage.
>
> "I know there is a 'thisContext sender' - but is there a 'who created me' concept?"
>
>
> Ultimately, there is no real overhead to storing (my)self on a new object, I guess, but as I'm developing and testing, I tend to have a whole lot of 'stranded objects' I have to remove manually.
>
> Thoughts?
>
> Thanks!
> --
> Russ Whaley
> [hidden email]

--

Russ Whaley
[hidden email]

--------------------------------------------

Stéphane Ducasse

http://stephane.ducasse.free.fr / http://www.pharo.org 

03 59 35 87 52

Assistant: Aurore Dalle 

FAX 03 59 57 78 50

TEL 03 59 35 86 16

S. Ducasse - Inria

40, avenue Halley, 

Parc Scientifique de la Haute Borne, Bât.A, Park Plaza

Villeneuve d'Ascq 59650

France