Unintuitive behavior of class-side initialize
Locked
Unintuitive behavior of class-side initialize
 
|
As class is just an object in Smalltalk, it would be reasonable to believe that #initialize message is always sent to class on creation.
However, that's not true. It is only sent to classes that redefine #initialize. Also sending super initialize may lead to problems. None of these is mentioned in documentation (I assume Pahro By Example is the official one). It only says that #initialize is sent when class is loaded into memory. I think either this behavior should be made more consistent, or excplicitly mentioned in the docs. |
Re: Unintuitive behavior of class-side initialize
|
|
On Wed, Aug 27, 2014 at 8:22 AM, webwarrior <[hidden email]> wrote: As class is just an object in Smalltalk, it would be reasonable to believe Ohhh yes. There is always a time for every Smalltalkers where you discover that doing super for class side initialize is not a good idea. Here was my time: http://forum.world.st/super-initialize-is-not-a-good-idea-for-Behaviors-td3086162.html
None of these is mentioned in documentation (I assume Pahro By Example is +1 to the documentation
Mariano http://marianopeck.wordpress.com |
Re: Unintuitive behavior of class-side initialize
|
I will be adding an "Introduction to Object Orientation" chapter to the new updated Pharo By Example online book, so I will add this to the chapter. The chapter will target people not familiar with Object Orientation and people new to coding. Its not an immediate target but it will happen till the end of the year.
On Wed, Aug 27, 2014 at 3:44 PM, Mariano Martinez Peck <[hidden email]> wrote:
|
Re: Unintuitive behavior of class-side initialize
|
|
This is not something fundamental to OOP, just specific to Pharo or
Smalltalk. Take for example Python. It has similar object model (objects, classes, metaclasses). Every time class is defined (either by class keyword, or by type() function, or by calling a metaclass), its metaclass' __init__ method is called. Usually you don't need to redefine metaclass' __init__ method, because class attributes are initialized in class definition. But when you do, it works as expected. 27.08.2014 15:50, kilon.alios [via Smalltalk] написав(ла): > I will be adding an "Introduction to Object Orientation" chapter to the > new updated Pharo By Example online book, so I will add this to the > chapter. The chapter will target people not familiar with Object > Orientation and people new to coding. Its not an immediate target but it > will happen till the end of the year. > > > On Wed, Aug 27, 2014 at 3:44 PM, Mariano Martinez Peck <[hidden email] > </user/SendEmail.jtp?type=node&node=4775051&i=0>> wrote: > > > > > On Wed, Aug 27, 2014 at 8:22 AM, webwarrior <[hidden email] > </user/SendEmail.jtp?type=node&node=4775051&i=1>> wrote: > > As class is just an object in Smalltalk, it would be reasonable > to believe > that #initialize message is always sent to class on creation. > > However, that's not true. It is only sent to classes that redefine > #initialize. Also sending super initialize may lead to problems. > > > Ohhh yes. There is always a time for every Smalltalkers where you > discover that doing super for class side initialize is not a good idea. > Here was my time: > http://forum.world.st/super-initialize-is-not-a-good-idea-for-Behaviors-td3086162.html > > None of these is mentioned in documentation (I assume Pahro By > Example is > the official one). It only says that #initialize is sent when > class is > loaded into memory. > > I think either this behavior should be made more consistent, or > excplicitly > mentioned in the docs. > > > > +1 to the documentation > > > > > -- > View this message in context: > http://forum.world.st/Unintuitive-behavior-of-class-side-initialize-tp4775042.html > Sent from the Pharo Smalltalk Users mailing list archive at > Nabble.com. > > > > > -- > Mariano > http://marianopeck.wordpress.com > > > > > ------------------------------------------------------------------------ > If you reply to this email, your message will be added to the discussion > below: > http://forum.world.st/Unintuitive-behavior-of-class-side-initialize-tp4775042p4775051.html > > To unsubscribe from Unintuitive behavior of class-side initialize, click > here > < > NAML > <http://forum.world.st/template/NamlServlet.jtp?macro=macro_viewer&id=instant_html%21nabble%3Aemail.naml&base=nabble.naml.namespaces.BasicNamespace-nabble.view.web.template.NabbleNamespace-nabble.view.web.template.NodeNamespace&breadcrumbs=notify_subscribers%21nabble%3Aemail.naml-instant_emails%21nabble%3Aemail.naml-send_instant_email%21nabble%3Aemail.naml> > |
Re: Unintuitive behavior of class-side initialize
|
|
ok officially I am confused are we talking about initialise at the instance side or class side (metaclass) side ? I am a python coder myself but I have not done any metaprogramming .
If it is a metaprogramming thing then yes it wont go inside my intro chapter since this is advanced topic and metaprogramming is already covered by PBE on a separate chapter so that can be added there. Since I am not experienced with metaprogramming it would be better if people experienced with Pharo and metaprogramming would write the documentation .
On Wed, Aug 27, 2014 at 6:16 PM, webwarrior <[hidden email]> wrote: This is not something fundamental to OOP, just specific to Pharo or |
Re: Unintuitive behavior of class-side initialize
|
|
class-side initialize.
I don't think it can be considered a metaprogramming in Smalltalk. In Python, yes, messing with metaclass is metaprogramming. But things are a bit different there and metaclasses are very rarely used [explicitly]. 27.08.2014 19:05, kilon.alios [via Smalltalk] написав(ла): > ok officially I am confused are we talking about initialise at the > instance side or class side (metaclass) side ? > > I am a python coder myself but I have not done any metaprogramming . > > If it is a metaprogramming thing then yes it wont go inside my intro > chapter since this is advanced topic and metaprogramming is already > covered by PBE on a separate chapter so that can be added there. Since I > am not experienced with metaprogramming it would be better if people > experienced with Pharo and metaprogramming would write the documentation . |
Re: Unintuitive behavior of class-side initialize
|
|
Yeah don't do super initialize unless you know why on the class side. That being said class side initialize is pretty cool when loading new packages and setting everything up automagically. Phil Le 27 août 2014 18:33, "webwarrior" <[hidden email]> a écrit :
class-side initialize. |
Re: Unintuitive behavior of class-side initialize
|
|
In reply to this post by webwarrior
yeah python does not need initialisation for class variables like it does for instance variables, but then same applies for instance variables, for example you can add instance variables and even instance method after the instance is created , pharo does something similar too. Well OOP is a deep subject and this is an advanced topic that I would not want to put in an introductionary chapter. I had issues with initialise on class side recently where i was told i had to initialise the class myself instead I chose to go down the ifNil route and initialise my class variables inside other methods. This way I avoided the creation of an initialise method, and probably i will avoid using it in the future too. From what I see most classes (metaclasses) dont use an initialise method either.
On Wed, Aug 27, 2014 at 7:33 PM, webwarrior <[hidden email]> wrote: class-side initialize. |
Re: Unintuitive behavior of class-side initialize
|
|
I've also chosen ifNil route.
However, in that case you either: - break encapsulation by providing accessors - have to invoke initializing method in every method that uses class instance variables or class variables 27.08.2014 20:31, kilon.alios [via Smalltalk] написав(ла): > yeah python does not need initialisation for class variables like it > does for instance variables, but then same applies for instance > variables, for example you can add instance variables and even instance > method after the instance is created , pharo does something similar too. > > Well OOP is a deep subject and this is an advanced topic that I would > not want to put in an introductionary chapter. > > I had issues with initialise on class side recently where i was told i > had to initialise the class myself instead I chose to go down the ifNil > route and initialise my class variables inside other methods. This way > I avoided the creation of an initialise method, and probably i will > avoid using it in the future too. From what I see most classes > (metaclasses) dont use an initialise method either. > > > On Wed, Aug 27, 2014 at 7:33 PM, webwarrior <[hidden email] > </user/SendEmail.jtp?type=node&node=4775110&i=0>> wrote: > > class-side initialize. > > I don't think it can be considered a metaprogramming in Smalltalk. > > In Python, yes, messing with metaclass is metaprogramming. But things > are a bit different there and metaclasses are very rarely used > [explicitly]. > > 27.08.2014 19:05, kilon.alios [via Smalltalk] написав(ла): > > > ok officially I am confused are we talking about initialise at the > > instance side or class side (metaclass) side ? > > > > I am a python coder myself but I have not done any > metaprogramming . > > > > If it is a metaprogramming thing then yes it wont go inside my intro > > chapter since this is advanced topic and metaprogramming is already > > covered by PBE on a separate chapter so that can be added there. > Since I > > am not experienced with metaprogramming it would be better if people > > experienced with Pharo and metaprogramming would write the > documentation . > > ------------------------------------------------------------------------ > View this message in context: Re: Unintuitive behavior of class-side > initialize > <http://forum.world.st/Unintuitive-behavior-of-class-side-initialize-tp4775042p4775103.html> > Sent from the Pharo Smalltalk Users mailing list archive > <http://forum.world.st/Pharo-Smalltalk-Users-f1310670.html> at > Nabble.com. > > > > > ------------------------------------------------------------------------ > If you reply to this email, your message will be added to the discussion > below: > http://forum.world.st/Unintuitive-behavior-of-class-side-initialize-tp4775042p4775110.html > > To unsubscribe from Unintuitive behavior of class-side initialize, click > here > < > NAML > <http://forum.world.st/template/NamlServlet.jtp?macro=macro_viewer&id=instant_html%21nabble%3Aemail.naml&base=nabble.naml.namespaces.BasicNamespace-nabble.view.web.template.NabbleNamespace-nabble.view.web.template.NodeNamespace&breadcrumbs=notify_subscribers%21nabble%3Aemail.naml-instant_emails%21nabble%3Aemail.naml-send_instant_email%21nabble%3Aemail.naml> > |
|
|
In reply to this post by webwarrior
webwarrior wrote:
> As class is just an object in Smalltalk, it would be reasonable to believe > that #initialize message is always sent to class on creation. > > However, that's not true. It is only sent to classes that redefine > #initialize. Also sending super initialize may lead to problems. > > None of these is mentioned in documentation (I assume Pahro By Example is > the official one). It only says that #initialize is sent when class is > loaded into memory. > > I think either this behavior should be made more consistent, or excplicitly > mentioned in the docs. > > > > -- > View this message in context: http://forum.world.st/Unintuitive-behavior-of-class-side-initialize-tp4775042.html > Sent from the Pharo Smalltalk Users mailing list archive at Nabble.com. > > > auto adds in "super initialize" as a template. Perhaps on the class side it should auto add a comment saying "Do not call super initialize on the class side" cheers -ben |
Re: Unintuitive behavior of class-side initialize
|
|
In reply to this post by webwarrior
I assume you mean there will be an inheritance issue if I want the variables to be inherited with a default value by the subclasses, which is a valid concern. Other than that you can have accessors for it as far I know . But yeah in that case of inheritance you need some kind of initialise method. I have not played with class instance variables so no clue about that part. On Wed, Aug 27, 2014 at 8:38 PM, webwarrior <[hidden email]> wrote: I've also chosen ifNil route. |
|
In reply to this post by kilon.alios
Kilon,
Do you have a preview version of the updated book? I would love to see it. Also would you consider adding some exercises will be a good addition? I think so....let me know and i guess i can merge the project i've started to the updated book. Best Nacho Enviado desde Molto para iPad
De: kilon alios Enviado: miércoles, agosto 27, 2014 09:50 a.m. Para: Any question about pharo is welcome Asunto: Re: [Pharo-users] Unintuitive behavior of class-side initialize I will be adding an "Introduction to Object Orientation" chapter to the new updated Pharo By Example online book, so I will add this to the chapter. The chapter will target people not familiar with Object Orientation and people new to coding. Its not an immediate target but it will happen till the end of the year.
On Wed, Aug 27, 2014 at 3:44 PM, Mariano Martinez Peck <[hidden email]> wrote:
Nacho
Smalltalker apprentice.
Buenos Aires, Argentina.
|
Re: Unintuitive behavior of class-side initialize
|
|
you can read the book here
main repo is here Sure exercises are always a good idea. We use pillar syntax which is quite simple, drop off to irc chat #Pharo at freenode to discuss the detail though most things are already well documented about the process of contributing to the book.
I have ported so far 4 chapter and I am in chapter 5 , Damien has port one chapter and Stephan partly ported another so we around 50% of the book. On Thu, Aug 28, 2014 at 1:51 PM, Ignacio Sniechowski <[hidden email]> wrote:
|
|
|
thanks,
Im familiar with pillar as incontributed some stuff for the pharo for the enterprise. I will take a look. Also Thanks for your video tutorials, they are really awesome. Please please continue to add more. Take care Nacho Enviado desde Molto para iPad
De: kilon alios Enviado: jueves, agosto 28, 2014 08:02 a.m. Para: Any question about pharo is welcome Asunto: Re: [Pharo-users] Unintuitive behavior of class-side initialize you can read the book here
main repo is here Sure exercises are always a good idea. We use pillar syntax which is quite simple, drop off to irc chat #Pharo at freenode to discuss the detail though most things are already well documented about the process of contributing to the book.
I have ported so far 4 chapter and I am in chapter 5 , Damien has port one chapter and Stephan partly ported another so we around 50% of the book. On Thu, Aug 28, 2014 at 1:51 PM, Ignacio Sniechowski <[hidden email]> wrote:
Nacho
Smalltalker apprentice.
Buenos Aires, Argentina.
|
|
|
In reply to this post by philippeback
Le 27 août 2014 à 19:04, [hidden email] a écrit : In my opinion, it would be better to have a package initialization method for that purpose. |
smime.p7s (5K)
Re: Unintuitive behavior of class-side initialize
|
|
In reply to this post by nacho
ah nice then you already a contributor and you dont need permission to add to updated pbe repo. On Thu, Aug 28, 2014 at 2:05 PM, Ignacio Sniechowski <[hidden email]> wrote:
|
«
Return to Pharo Smalltalk Users
|
1 view|%1 views
| Free forum by Nabble | Edit this page |