rubycocoa-talk — talking about RubyCocoa

[Rubycocoa-talk] Ruby-fying the SyncServices - SimpleStickies tutorial fom Apple

[Rupert B. <rup...@fr...>]

OK, with much help from Jonathan, I have got the Apple SyncServices  
"SimpleStickies" [1] translated and working (partially) in Ruby. The  
SimpleStickies is much more than a simple tutorial, the resulting  
program is a complete syncing engine, demo-ed on the 'Note' class  
which implements Simple Stickies.

What I did was simply to :
1) modify the Objective-C 'main' to startup with Ruby : this is just  
the standard Ruby 'main.m' and 'rb_main.rb'
2) replace the 'Note.h' and 'Note.m' files for the Note class by  
'Note.rb' and kvc_wrapper.rb [2]

By the way, do not forget to create the following directory :
	~/Library/Application Support/SyncExamples
where the tutorial stores its persistence cache.

What is working :
- the program runs and starts,
- it will save data to a persistence XML file store (in ~/Library/ 
Application Support/SyncExamples)
- it will sync with Tiger's SyncServices (both read and write)
- .Mac syncing works natively once the data has got into the  
SyncServices engine : you can see the SimpleStickies logo in the .Mac  
prefderences : exiting to thing that this is carying data generated  
by Ruby !

What I cannot get to work :
- reading from the XML persistence file store : this must due to some  
issue with archiving/unarchiving. To get the program to work  
properly, delete file in ~/Library/Application Support/SyncExamples  
before every launch
- I cannot programmatically (from Ruby) add objects to the Notes  
controller to get them inserted into GUI

Any help would be welcome to help resolve these two points.

Rupert



[1] SimpleStickies tutorial : http://developer.apple.com/ 
documentation/AppleApplications/Conceptual/SyncServicesTutorial/ 
index.html


[2] Here are the files :
Notes.rb :

#
#  Note.rb
#  SimpleStickies adapted to Ruby
#
#  Created by Rupert Barrow on 06/11/05.
#  Copyright (c) 2005 La Romane. All rights reserved.
#

require 'osx/cocoa'
require 'osx/coredata'
require 'kvc_wrapper'

class Note < OSX::NSObject
	include OSX
	
	 
kvc_wrapper_representation :createDate, :originX, :originY, :width, :hei 
ght, :color
	 
kvc_wrapper_representation :text, :translucent, :floating, :shaded, :pri 
maryKey
	
	ns_overrides 'init', 'dealloc'
	
	@_representation = nil
	@_changed = nil
	
	def init
		self.super_init
		
		@_representation = NSMutableDictionary.dictionary

		self.setValue_forKey(NSDate.date, "createDate")
	
		self.setValue_forKey(100, "originX")
		self.setValue_forKey(100, "originY")
		self.setValue_forKey(400, "width")
		self.setValue_forKey(300, "height")
		
		self.setValue_forKey(NSColor.yellowColor, "color")
		
		attrString = NSAttributedString.alloc.initWithString("My simple  
sticky.")
		self.setValue_forKey(NSArchiver.archivedDataWithRootObject 
(attrString), "text")
		
		self
	end
		
	def dealloc
		super_dealloc
	end
	
	def changed
		@_changed
	end
	self.addRubyMethod_withType("changed".to_sym, "v@:@")
	def setChanged(v)
		@_changed = v
	end
	self.addRubyMethod_withType("setChanged:".to_sym, "v@:@")

	def representation
		@_representation
	end
	
	def description
		"representation = #{@_representation.description}"
	end
		
	def valueForUndefinedKey(k)
		@_representation.valueForKey(k)
	end
	
	def setValue_ForUndefinedKey(v, k)
		@_representation.setValueForKey(v, k)
	end
	
	def initWithCoder(coder)
		self.super_init
		if (coder.allowsKeyedCoding?)
			@_representation = coder.decodeObjectForKey("representation")
		else
			@_representation = coder.decodeObject;
		end
		return self
	end
	
	def encodeWithCoder(coder)
		if (coder.allowsKeyedCoding?)
			coder.encodeObject_forKey(@_representation, "representation")
		else
			coder.encodeObject(@_representation);
		end
		return
	end

end


kvc_wrapper.rb :

#
#  kvc_wrapper.rb
#  SimpleStickies adapted to Ruby
#
#  Created by Rupert Barrow on 05/11/05.
#  Copyright (c) 2005 La Romane. All rights reserved.
#

# add kvc_wrapper
module OSX

	module NSKVCBehaviorAttachment

         def kvc_wrapper_representation(*keys)
             keys.each do |key|
                 set_key = "set#{key}"
                 set_key[3..3] = set_key[3..3].upcase
                 class_eval <<- 
EOE_KVC_WRAPPER_REPRESENTATION,__FILE__,__LINE__+1
                     def #{key}
#                       self.willAccessValueForKey("#{key}")
                         @_representation.valueForKey("#{key}")
#                       self.didAccessValueForKey("#{key}")
                     end
                     def #{key}=(val)
                         self.willChangeValueForKey("#{key}")
                         @_representation.setValue_forKey(val, "#{key}")
                         self.didChangeValueForKey("#{key}")
                     end
                     def #{set_key}(val); self.#{key}= val; end
                     EOE_KVC_WRAPPER_REPRESENTATION
				self.addRubyMethod_withType("#{key}".to_sym, "v@:@")
				self.addRubyMethod_withType("#{set_key}:".to_sym, "v@:@")
             end
         end

	end

end

Re: [Rubycocoa-talk] WSDL defines class with attribute called 'Id' : how to reach ?

[Rupert B. <rup...@fr...>]

Sorry, posted on the wrong list :-)
Rup

Le 14 nov. 05 =C3=A0 11:59, Jonathan Paisley a =C3=A9crit :

> On 14 Nov 2005, at 0:36, Rupert BARROW wrote:
>
>> Somehow, from a WSDL file describing an object with an attribute =20
>> called 'Id', soap4r has managed to create an object with an extra =20
>> 'id' attribute, different from the Ruby 'id' :
>>
>> 	#<MyObject:0x13f9eec @id=3D"an id">
>>
>> How do I reach this value ?
>>
>> I have the same problem with an attribute called 'name' ...
>
> Don't know much about soap4r... Try on ruby-talk mailing list?
>
> You could always use obj.instance_variable_get("@id"), but that's =20
> not very satisfactory.
>

Re: [Rubycocoa-talk] WSDL defines class with attribute called 'Id' : how to reach ?

[Jonathan P. <jp...@dc...>]

On 14 Nov 2005, at 0:36, Rupert BARROW wrote:

> Somehow, from a WSDL file describing an object with an attribute  
> called 'Id', soap4r has managed to create an object with an extra  
> 'id' attribute, different from the Ruby 'id' :
>
> 	#<MyObject:0x13f9eec @id="an id">
>
> How do I reach this value ?
>
> I have the same problem with an attribute called 'name' ...

Don't know much about soap4r... Try on ruby-talk mailing list?

You could always use obj.instance_variable_get("@id"), but that's not  
very satisfactory.

Re: [Rubycocoa-talk] Re: SyncServices and RubyCocoa

[Jonathan P. <jp...@dc...>]

> Bingo, Eureka !
> I defined @_changed, as well as its setter and getter :
> 	def changed
> 		@_changed
> 	end
> 	def setChanged(v)
> 		@_changed = v
> 	end
> Everything seems to be working OK.
> Does anyone know why these have to be defined in Ruby, when they do  
> not in Objective-C ?

It's hard to tell how the pieces fit together without seeing all the  
code. Can you post a ZIP of the whole project?

Cheers
Jonathan

[Rubycocoa-talk] WSDL defines class with attribute called 'Id' : how to reach ?

[Rupert B. <rup...@fr...>]

Hi,

Somehow, from a WSDL file describing an object with an attribute  
called 'Id', soap4r has managed to create an object with an extra  
'id' attribute, different from the Ruby 'id' :

	#<MyObject:0x13f9eec @id="an id">

How do I reach this value ?

I have the same problem with an attribute called 'name' ...

Rup

Re: [Rubycocoa-talk] key value questions

[kimura w. <ki...@us...>]

Hi,

Thu, 10 Nov 2005 18:28:55 -0700, Rod Schmidt wrote:
>I have some questions about RubyCocoa's support for key value coding  
>and the new kvc_accessor and kvc_depends_on methods.
>
>First off, in Objective-C, all objects are already KVC compliant as  
>long as you follow certain naming conventions. So why do we have a  
>separate kvc_accessor method instead of just modifiying attr_accessor  
>to support key value coding? Is it because attr_accessor is built  
>into Ruby and we needed a different name?
>

There is only one difference about style of setter methods in a ruby 
script.

We have to write like this with attr_accesor:

  obj.setValue_forKey(value, "aKey")


We can write like this with kvc_accessor:

  obj.aKey = value


When a ruby script do not use setter methods (ex. access from only
Cocoa objects), kvc_accessor equals to attr_accessor.


-- 
kimura wataru

Re: [Rubycocoa-talk] Re: SyncServices and RubyCocoa

[Rupert B. <rup...@fr...>]

Le 11 nov. 05 =C3=A0 14:58, Rupert BARROW a =C3=A9crit :

> Jonathan,
>
> Le 11 nov. 05 =C3=A0 12:51, Jonathan Paisley a =C3=A9crit :
>
>>
>> On 11 Nov 2005, at 10:19, Rupert BARROW wrote:
>>
>>> 		if ([record valueForKey:@"changed"] =3D=3D @"YES")		=
	// this crashes
>>
>> This is unrelated to the main issue, but note that comparison =20
>> doesn't make sense.
>>
>> You need to compare NSStrings with isEqualToString:, or if =20
>> 'changed' is returning a bool/int try something like:
>>
>>   [(NSNumber*)[record valueForKey:@"changed"] boolValue]
>
> This code is part of the Apple SImpleStickies tutorial. I have not =20
> changed it.

<CUT>

> I hope that you can find something. Meanwhile, I will tool around, =20
> trying to define wrappers for the "changed" key.

Bingo, Eureka !
I defined @_changed, as well as its setter and getter :
	def changed
		@_changed
	end
	def setChanged(v)
		@_changed =3D v
	end
Everything seems to be working OK.
Does anyone know why these have to be defined in Ruby, when they do =20
not in Objective-C ?

Rup

Re: [Rubycocoa-talk] Re: SyncServices and RubyCocoa

[Jonathan P. <jp...@dc...>]

On 11 Nov 2005, at 10:19, Rupert BARROW wrote:

> 		if ([record valueForKey:@"changed"] == @"YES")			// this crashes

This is unrelated to the main issue, but note that comparison doesn't  
make sense.

You need to compare NSStrings with isEqualToString:, or if 'changed'  
is returning a bool/int try something like:

   [(NSNumber*)[record valueForKey:@"changed"] boolValue]

Can you email me your project, or come up with a small ruby test  
script that demonstrates the problem? It's a bit hard to tell what's  
going on with the fragments of code.

Cheers
Jonathan

Re: [Rubycocoa-talk] key value questions

[Jonathan P. <jp...@dc...>]

On 11 Nov 2005, at 1:28, Rod Schmidt wrote:

> I have some questions about RubyCocoa's support for key value  
> coding and the new kvc_accessor and kvc_depends_on methods.
>
> First off, in Objective-C, all objects are already KVC compliant as  
> long as you follow certain naming conventions. So why do we have a  
> separate kvc_accessor method instead of just modifiying  
> attr_accessor to support key value coding? Is it because  
> attr_accessor is built into Ruby and we needed a different name?

KVC in ObjC works by wrapping setter methods with  
willChangeValue:forKey: didChangeValue:forKey:. It does this by some  
sneaky overriding of the method pointers in the class definition.

The RubyCocoa classes are dynamically generated, and at the moment  
ObjC methods do not actually exist that correspond to the getter/ 
setter. Instead, we catch valueForUndefinedKey: and  
setValue:forUndefinedKey: and pass through to the ruby getter/setter.  
At this stage, the willChange/didChange stuff is done explicitly.

All this is necessary because KVC can't step in and do its sneaky  
thing because the methods do not exist in objc land.

The same thing situation is there for most methods called from objc -  
the objc forwardInvocation: method is used to pass unknown selectors  
to ruby.

Just in the past week on this mailing list we've been discussing a  
problem that arises with CoreData, where it requires the real methods  
to exist in objc land. What I'm not sure of yet is whether creating  
these methods in the dynamically generated class allows KVC to do the  
magic on our behalf, or whether it's still necessary to do the  
willChange/didChange notifications.


> Next, what does kvc_depends_on do? I'm guessing it's for Key Value  
> Observing? If that's the case why don't we just use the addObserver  
> method in NSObject that is for that purpose? Or is it just a shortcut?

kvc_depends_on is just a wrapper for  
setKeys:triggerChangeNotificationsForDependentKey:, making it a bit  
more ruby-like. Here's the definition:

     def kvc_depends_on(keys, *dependencies)
       dependencies.flatten.each do |dependentKey|
         setKeys_triggerChangeNotificationsForDependentKey(keys.to_a,  
dependentKe
y)
       end
     end


> Sorry if these questions have been answered before, but I haven't  
> seen a description of what these do. Just scattered discussions  
> with no clear definition.

The binding support stuff is new. I know if there's any particular  
documentation for it, other than the discussion on the mailing lists.

Cheers,
Jonathan

[Rubycocoa-talk] Re: SyncServices and RubyCocoa

[Rupert B. <rup...@fr...>]

Hi, I'm back to square one with the SimpleStickies tutorial :  
encoding and decoding seem to be going well, except that I get the  
following message while syncing :

Exception occured while syncing: Note: this class is not key value  
coding-compliant for the key "changed"

The program looks for all the "changed" entity instances to  
(presumably) sync them.
In

- (NSArray *)changedObjectsForEntityName:(NSString *)entityName
{
	NSEnumerator *recordEnumerator = [[self valueForKey:[self  
keyForEntityName:entityName]] objectEnumerator];
	id record;
	NSMutableArray *changedRecords = [NSMutableArray array];
	
	while (record = [recordEnumerator nextObject]) {
		if ([record valueForKey:@"changed"] == @"YES")			// this crashes
			[changedRecords addObject:record];
	}
	if ([changedRecords count] == 0)
		return nil;
	return changedRecords;
}

the call to '[record valueForKey:@"changed"]' crashes.

Now, I wonder if our calls to
	self.willChangeValueForKey("#{key}")

in our definition of the key_wrappers

         def kvc_wrapper_representation(*keys)
             keys.each do |key|
                 set_key = "set#{key}"
                 set_key[3..3] = set_key[3..3].upcase
                 class_eval <<- 
EOE_KVC_WRAPPER_REPRESENTATION,__FILE__,__LINE__+1
                     def #{key}
                         @_representation.valueForKey("#{key}")
                     end
                     def #{key}=(val)
                         self.willChangeValueForKey("#{key}")					#  
is this working correctly ?
                         @_representation.setValue_forKey(val, "#{key}")
                         self.didChangeValueForKey("#{key}")
                     end
                     def #{set_key}(val); self.#{key}= val; end
                     EOE_KVC_WRAPPER_REPRESENTATION
		self.addRubyMethod_withType("#{key}".to_sym, "v@:@")
		self.addRubyMethod_withType("#{set_key}:".to_sym, "v@:@")
             end
         end

is working.

Any ideas ?  Should we maybe (re ?)define the "changed" key for Note  
somewhere, or is this inherited from NSManaged Object ?
I once tried to implement inherited entities in an Objective-C  
EntityModel test, which did not work : of course, 'Entity- 
Relationship' modelling is an old father of object-modelling, so I am  
not totally surprised ...

Rup

[Rubycocoa-talk] key value questions

[Rod S. <rsc...@xm...>]

I have some questions about RubyCocoa's support for key value coding  
and the new kvc_accessor and kvc_depends_on methods.

First off, in Objective-C, all objects are already KVC compliant as  
long as you follow certain naming conventions. So why do we have a  
separate kvc_accessor method instead of just modifiying attr_accessor  
to support key value coding? Is it because attr_accessor is built  
into Ruby and we needed a different name?

Next, what does kvc_depends_on do? I'm guessing it's for Key Value  
Observing? If that's the case why don't we just use the addObserver  
method in NSObject that is for that purpose? Or is it just a shortcut?

Sorry if these questions have been answered before, but I haven't  
seen a description of what these do. Just scattered discussions with  
no clear definition.

Thanks,

Rod Schmidt
infiniteNIL Software

[Rubycocoa-talk] [ANN]RubyCocoa 0.4.2 was released

[kimura w. <ki...@us...>]

RubyCocoa 0.4.2 was released!

   http://rubycocoa.sourceforge.net/

The '.pkg' binary packages are available for Mac OS X 10.4/10.3/10.2.  
All of framework, library, documents, samples and Xcode/ProjectBuilder 
templates are contained in the package. These items will be installed 
into the appropriate place. It is available to start of RubyCocoa 
programming or use of RubyCocoa application immediately.


== Changes from 0.4.1

=== new features

  * support Cocoa Bindings (Mac OS X 10.3 or later)
  * support Core Data (Mac OS X 10.4)

=== bug fixes

  * a ruby object becomes an instance of OSX::OCObject
  * itunes_albums.rb fails with NSCharacterConversionException
  * a ruby script with sub-subclass of a Cocoa class crashes

== known problems

  Following problems are not solved at this relaese.

  * SEGV/SIGBUS occurs sometimes with ruby Thread
  * warning "NSView not correctly initialized." appears on console 
    at initializing subclass of NSView (Mac OS X 10.3.9 or later)
  * "irb -r 'osx/cocoa'" crashes (Mac OS X 10.4)

== What's RubyCocoa?

  RubyCocoa is a Mac OS X framework that allows Cocoa programming in
  the Object-Oriented Scripting Language Ruby.

  RubyCocoa allows writing a Cocoa application in Ruby.  It allows
  creating and using a Cocoa object in a Ruby script.  In Cocoa
  application, mixture of program written by both Ruby and Objective-C
  is possible.

  Some cases using RubyCocoa:

    * Exploration of a Cocoa object's feature with 'irb' interactively
    * Prototyping of a Cocoa application
    * Cocoa application that include good feature of Ruby and Objective-C
    * Wrapping Mac OS X native GUI for Ruby script


-- 
kimura wataru

[Rubycocoa-talk] Re: NSCoding, NSArchiver for SyncServices [was "SyncServices and RubyCocoa"]

[Rupert B. <rup...@fr...>]

Of course !
It works. Great. Thanks for your help.

I still  have a problem with the syncing, which also occurs with the =20
Objective C program.
As soon as I have fixed this, I will post the full sample.

Cheers,
Rup

Le 10 nov. 05 =C3=A0 09:49, Jonathan Paisley a =C3=A9crit :

> On 10 Nov 2005, at 6:04, Rupert BARROW wrote:
>
>>     def initWithCoder(coder)
>>         self.super_init
>>         if (coder.allowsKeyedCoding)
>>
> [snip]
>> However, the program comes up with the following error, which is =20
>> generated within 'encodeWithCoder' :
>> Exception occured while syncing: *** -encodeObject:forKey: only =20
>> defined for abstract class.  Define -[NSArchiver =20
>> encodeObject:forKey:]!
>
> It might be this:
>
>   coder.allowsKeyedCoding
>
> will be returning an int (0 or 1), which are both true in ruby. Try =20=

> using
>
>   code.allowsKeyedCoding?
>
> instead.
>
>

[Rubycocoa-talk] Re: NSCoding, NSArchiver for SyncServices [was "SyncServices and RubyCocoa"]

[Jonathan P. <jp...@dc...>]

On 10 Nov 2005, at 6:04, Rupert BARROW wrote:

>     def initWithCoder(coder)
>         self.super_init
>         if (coder.allowsKeyedCoding)
>
[snip]
> However, the program comes up with the following error, which is  
> generated within 'encodeWithCoder' :
> Exception occured while syncing: *** -encodeObject:forKey: only  
> defined for abstract class.  Define -[NSArchiver  
> encodeObject:forKey:]!

It might be this:

   coder.allowsKeyedCoding

will be returning an int (0 or 1), which are both true in ruby. Try  
using

   code.allowsKeyedCoding?

instead.

[Rubycocoa-talk] NSCoding, NSArchiver for SyncServices [was "SyncServices and RubyCocoa"]

[Rupert B. <rup...@fr...>]

Le 9 nov. 05 =C3=A0 00:51, Jonathan Paisley a =C3=A9crit :

> On 8 Nov 2005, at 21:52, Rupert BARROW wrote:
>>>
>>>                     self.addRubyMethod_withType("#{key}".to_sym, =20
>>> "@4@8:12")
>>>                     self.addRubyMethod_withType("#=20
>>> {set_key}:".to_sym, "@4@8:12@16")
>
>>
>> I tried this; here is the result :
>>
>> 2005-11-08 22:42:32.463 SimpleStickies[1306] KVO autonotifying =20
>> only supports -set<Key>: methods that return void.  Autonotifying =20
>> will not be done for invocations of -[Note setText:].
>>
>> Same error is produced for each atribute.
>> What is the syntax of the second parameter to =20
>> 'addRubyMethod_withType' ?  It does not seem to acknowledge that =20
>> 'void' is returned, although the initial @ is preceded by nothing =20
>> (no type).
>
> The second argument is a type encoding [1] (see also the =20
> objc_method documentation in [2]). It consists of a type code =20
> followed by (I think) some sort of stack offset. The "Special =20
> Considerations" section in [2] indicates that these are ignored, so =20=

> you can probably get away with leaving them out.
>
> The value I gave before for the setter method is incorrect - it =20
> declares an 'id' type return value, when it should be void. Try =20
> using this instead: "v@:@". This means, from left to right, =20
> v=3D"void", @=3D"id self", :=3D"SEL _cmd", @=3D"id value". Notice how =
self =20
> and selector must be explicitly mentioned.

Thanks, "v@:@" worked well : we are making progress.
I am now stuck on encoding problems : the two methods 'initWithCoder=20
(coder)' and 'encodeWithCoder(coder)' of the NSCoding protocol are =20
defined like this :

	def initWithCoder(coder)
		self.super_init
		if (coder.allowsKeyedCoding)
			@_representation =3D =
coder.decodeObjectForKey("representation")
		else
			@_representation =3D coder.decodeObject;
		end
		return self
	end
=09
	def encodeWithCoder(coder)
		if (coder.allowsKeyedCoding)
			coder.encodeObject_forKey(@_representation, =
"representation")
		else
			coder.encodeObject(@_representation);
		end
		return
	end

However, the program comes up with the following error, which is =20
generated within 'encodeWithCoder' :
Exception occured while syncing: *** -encodeObject:forKey: only =20
defined for abstract class.  Define -[NSArchiver encodeObject:forKey:]!

Googling for this shows that it is quite a common error when things =20
have not been programmed correctly (eg testing for =20
'allowsKeyedCoding'), but as I have translated the tutorial over from =20=

Objective C ...
... either something is wrong in the code of these two methods, above =20=

(Objective-C says 'self=3Dsuper.init' in initWithCoder, but Ruby =20
refuses 'self=3Dsuper_init') ...
... or we have another 'Ruby vs. Objective-C' context/visibility =20
problem.

I tried adding the following
	self.addRubyMethod_withType(:initWithCoder, "@:@")
	self.addRubyMethod_withType(:encodeWithCoder, "v@:@")
which changed nothing.

Any ideas ?

Rup

Re: [Rubycocoa-talk] SyncServices and RubyCocoa

[Jonathan P. <jp...@dc...>]

On 8 Nov 2005, at 21:52, Rupert BARROW wrote:
>>
>>                     self.addRubyMethod_withType("#{key}".to_sym,  
>> "@4@8:12")
>>                     self.addRubyMethod_withType("# 
>> {set_key}:".to_sym, "@4@8:12@16")

>
> I tried this; here is the result :
>
> 2005-11-08 22:42:32.463 SimpleStickies[1306] KVO autonotifying only  
> supports -set<Key>: methods that return void.  Autonotifying will  
> not be done for invocations of -[Note setText:].
>
> Same error is produced for each atribute.
> What is the syntax of the second parameter to  
> 'addRubyMethod_withType' ?  It does not seem to acknowledge that  
> 'void' is returned, although the initial @ is preceded by nothing  
> (no type).

The second argument is a type encoding [1] (see also the objc_method  
documentation in [2]). It consists of a type code followed by (I  
think) some sort of stack offset. The "Special Considerations"  
section in [2] indicates that these are ignored, so you can probably  
get away with leaving them out.

The value I gave before for the setter method is incorrect - it  
declares an 'id' type return value, when it should be void. Try using  
this instead: "v@:@". This means, from left to right, v="void", @="id  
self", :="SEL _cmd", @="id value". Notice how self and selector must  
be explicitly mentioned.

[1] http://developer.apple.com/documentation/Cocoa/Conceptual/ 
ObjectiveC/RuntimeOverview/chapter_5_section_6.html

[2] http://developer.apple.com/documentation/Cocoa/Reference/ 
ObjCRuntimeRef/ObjCRuntimeRef/chapter_1.4_section_2.html

> By the way, why define "#{key}" (without ':') and "# 
> {set_key}:" (with ':') ?  [sorry for silly newbies' questions]

Don't worry, it's not obvious. The first argument is the name of the  
Objective C method to define. A method with no ':' takes no arguments  
(the getter), whereas the single : indicates that an argument follows  
it (for the setter, it's the value to set).

What we're effectively doing is programmatically making these  
Objective C class definitions for our Ruby-defined ObjC class:

   -(id)key;
   -(void)setKey:(id)value;

Cheers,
Jonathan

Re: [Rubycocoa-talk] SyncServices and RubyCocoa

[Rupert B. <rup...@fr...>]

Le 7 nov. 05 =C3=A0 21:29, Jonathan Paisley a =C3=A9crit :

> It might be possible to resolve the problem, by defining the =20
> methods in question in Objective C land. There is a special =20
> RubyCocoa method to help you do this. Take a look at the =20
> kvc_array_accessor method declared in oc_import.rb. It uses =20
> addRubyMethod_withType to add a real ObjC method that calls through =20=

> to a Ruby method. So you'd convert your kvc_wrapper_representation =20
> to look something like (untested):
>
>         def kvc_wrapper_representation(*keys)
>             keys.each do |key|
>                 set_key =3D "set" + key
>                 set_key[3..3] =3D set_key[3..3].upcase
>                 class_eval <<-=20
> EOE_KVC_WRAPPER_REPRESENTATION,__FILE__,__LINE__+1
>                     def #{key}
>                         @_representation.valueForKey("#{key}")
>                     end
>                     def #{key}=3D(val)
>                         self.willChangeValueForKey("#{key}")
>                         @_representation.setValue_forKey(val, "#=20
> {key}")
>                         self.didChangeValueForKey("#{key}")
>                     end
>                     def #{set_key}(val); self.#{key}=3D val; end
>                     EOE_KVC_WRAPPER_REPRESENTATION
>                     self.addRubyMethod_withType("#{key}".to_sym, =20
> "@4@8:12")
>                     self.addRubyMethod_withType("#=20
> {set_key}:".to_sym, "@4@8:12@16")
>             end
>         end

I tried this; here is the result :

2005-11-08 22:42:32.463 SimpleStickies[1306] KVO autonotifying only =20
supports -set<Key>: methods that return void.  Autonotifying will not =20=

be done for invocations of -[Note setText:].

Same error is produced for each atribute.
What is the syntax of the second parameter to =20
'addRubyMethod_withType' ?  It does not seem to acknowledge that =20
'void' is returned, although the initial @ is preceded by nothing (no =20=

type).

By the way, why define "#{key}" (without ':') and "#{set_key}:" (with =20=

':') ?  [sorry for silly newbies' questions]

> Having thought about this a bit more, it may be cleaner to =20
> 'improve' the implementation of kvc_reader and kvc_writer to define =20=

> these methods and call through to the appropriate ruby setter/=20
> getter. Then you'd implement the delegation to @_representation in =20
> a normal ruby accessor.

For the moment, I will leave that to the experts  :-)

> Hope that makes some sense :)

Thanks for you help.

Rup

Re: [Rubycocoa-talk] SyncServices and RubyCocoa

[Jonathan P. <jp...@dc...>]

On 7 Nov 2005, at 19:00, Rupert BARROW wrote:

> 2005-11-07 19:55:36.494 SimpleStickies[782] Exception syncing:  
> NSUnknownKeyException, reason [<Note 0x651130> takeValue:forKey:]:  
> attempt to assign value to unknown key: 'primaryKey'.
> This class does not have an instance variable of the name  
> primaryKey or _primaryKey, nor a method of the name setPrimaryKey,  
> or _setPrimaryKey
>
> This is strange since dumping the instance_methods of Note shows  
> that all the expected methods are there :
> setCreateDate
> createDate

I think these are ruby-level instance methods - not actually known to  
Objective C. The Objective C -> ruby bridge is accomplished through  
the ObjC equivalent of Ruby's method_missing.

>
> So, do you think that SyncServices are using some route which is  
> bypassing our kvc_wrapper mechanism to reach the primaryKey  
> "attribute" and getter and setter ?

I think the problem is that SyncServices is indirectly checking the  
Objective C class for a method primaryKey or _primaryKey and failing  
because they don't exist. The key-value-coding support in RubyCocoa  
is implemented by hooking into the valueForUndefinedKey: method that  
KVC calls if such a method doesn't exist. It seems that SyncServices  
is using another route that doesn't use that method. (this is my  
guess - haven't looked into it further)

It might be possible to resolve the problem, by defining the methods  
in question in Objective C land. There is a special RubyCocoa method  
to help you do this. Take a look at the kvc_array_accessor method  
declared in oc_import.rb. It uses addRubyMethod_withType to add a  
real ObjC method that calls through to a Ruby method. So you'd  
convert your kvc_wrapper_representation to look something like  
(untested):

         def kvc_wrapper_representation(*keys)
             keys.each do |key|
                 set_key = "set" + key
                 set_key[3..3] = set_key[3..3].upcase
                 class_eval <<- 
EOE_KVC_WRAPPER_REPRESENTATION,__FILE__,__LINE__+1
                     def #{key}
                         @_representation.valueForKey("#{key}")
                     end
                     def #{key}=(val)
                         self.willChangeValueForKey("#{key}")
                         @_representation.setValue_forKey(val, "#{key}")
                         self.didChangeValueForKey("#{key}")
                     end
                     def #{set_key}(val); self.#{key}= val; end
                     EOE_KVC_WRAPPER_REPRESENTATION
                     self.addRubyMethod_withType("#{key}".to_sym,  
"@4@8:12")
                     self.addRubyMethod_withType("# 
{set_key}:".to_sym, "@4@8:12@16")
             end
         end

Having thought about this a bit more, it may be cleaner to 'improve'  
the implementation of kvc_reader and kvc_writer to define these  
methods and call through to the appropriate ruby setter/getter. Then  
you'd implement the delegation to @_representation in a normal ruby  
accessor.

Hope that makes some sense :)

Jonathan

[Rubycocoa-talk] SyncServices and RubyCocoa

[Rupert B. <rup...@fr...>]

After looking at CoreData, and getting stuck at chapter 4 of the  
tutorial, I have moved to SyncServices.

I picked up the SimpleStickies tutorial, used the complete (step 3)  
source code of the tutorial to replace the Note Objective C class by  
a Note.rb : quite easy, I just had to adapt the recent kvc_wrapper  
suggestion and came up with :

		# accessor for keys defined in Cocoa, for _representation
		def kvc_wrapper_representation(*keys)
			keys.each do |key|
				class_eval <<-EOE_KVC_WRAPPER_REPRESENTATION,__FILE__,__LINE__+1
					def #{key}
						@_representation.valueForKey("#{key}")
					end
					def #{key}=(val)
						self.willChangeValueForKey("#{key}")
						@_representation.setValue_forKey(val, "#{key}")
						self.didChangeValueForKey("#{key}")
					end
					EOE_KVC_WRAPPER_REPRESENTATION
			end
		end

because the Note class takes its stuff from the _representation  
attribute.
I quickly created new setters called setCreateDate, setOriginX etc.  
from createDate=, originX=, etc.

Before these changes, I ran the app in full Objective C, saved a few  
Notes, and synced them.
Switching to the rubified-Notes, the app will not read them to  
display them : it complains that

2005-11-07 19:55:36.494 SimpleStickies[782] Exception syncing:  
NSUnknownKeyException, reason [<Note 0x651130> takeValue:forKey:]:  
attempt to assign value to unknown key: 'primaryKey'.
This class does not have an instance variable of the name primaryKey  
or _primaryKey, nor a method of the name setPrimaryKey, or  
_setPrimaryKey

This is strange since dumping the instance_methods of Note shows that  
all the expected methods are there :
setCreateDate
createDate
createDate=
etc.
primaryKey
primaryKey=
setPrimaryKey
etc.

So, do you think that SyncServices are using some route which is  
bypassing our kvc_wrapper mechanism to reach the primaryKey  
"attribute" and getter and setter ?

Rup

Re: [Rubycocoa-talk] RubyCocoa, CoreData, & NSPersistentDocumentTutorial

[kimura w. <ki...@us...>]

Hi,

I added some changes to CVS. This changes enables automated wrapping
of Key-Value Coding accessors. When you define a subclass of 
NSManagedObject, RubyCocoa defines wrappers to attributes and 
relationships from CoreData models.

my Employee.rb at chapter 3.
-------------
require 'osx/cocoa'
require 'osx/coredata'

class Employee < OSX::NSManagedObject
  
  kvc_depends_on [:lastName, :firstName], :fullNameAndID

  @@temp_id = 0
  
  def fullNameAndID
    sprintf('%s, %s (%d)', lastName.to_s, firstName.to_s, employeeID)
  end
  
  ns_overrides :awakeFromInsert
  def awakeFromInsert
    @@temp_id += 1
    self.employeeID = OSX::NSNumber.numberWithInt(@@temp_id)
  end
  
end
-------------

A new module function OSX::CoreData.define_wrapper() is core of the 
feature.

Fri, 04 Nov 2005 20:21:09 +0900, kimura wataru wrote:
>Hi,
>
>I'm trying this tutorial.
>

>I think that is possible to define wrappers automatically from 
>NSManagedObjectModel#entities -> NSEntityDescription, but I do not
>know how to implement.
>


-- 
kimura wataru

Re: [Rubycocoa-talk] RubyCocoa, CoreData, & NSPersistentDocumentTutorial[MESSAGE NOT SCANNED]

[Rupert B. <rup...@fr...>]

Le 5 nov. 05 =C3=A0 00:22, Jonathan Paisley a =C3=A9crit :

>>     employeeID =3D 123        # not self.employeeID
>>     puts employeeID
>>     puts self
>
>>     self.employeeID =3D 123    # using self.employeeID
>>     puts employeeID
>>     puts self
>
> Ah, my fault for getting this wrong in my earlier message.
>
> Ruby assumes 'employeeID =3D 123' is a local variable assignment.

Of course !
Gee, I'm getting tired.  I'm moving on to chapter 4 ...

Thanks again,
Rup

Re: [Rubycocoa-talk] RubyCocoa, CoreData, & NSPersistentDocumentTutorial[MESSAGE NOT SCANNED]

[Jonathan P. <jp...@dc...>]

>     employeeID = 123        # not self.employeeID
>     puts employeeID
>     puts self

>     self.employeeID = 123    # using self.employeeID
>     puts employeeID
>     puts self

Ah, my fault for getting this wrong in my earlier message.

Ruby assumes 'employeeID = 123' is a local variable assignment.  
Hence, the 'puts employeeID' simply retrieves the local variable  
assigned on the preceding line. CoreData or RubyCocoa are never  
involved.

In your second example, the first line directly calls the setter  
(self.employeeID=) and the second line indirectly calls the getter  
(self.employeeID) since no local variable with that name (employeeID)  
has been defined yet.

Hope that makes sense. Glad to hear that things are beginning to work.

Jonathan

Re: [Rubycocoa-talk] RubyCocoa, CoreData, & NSPersistentDocumentTutorial

[Rupert B. <rup...@fr...>]

Le 4 nov. 05 =C3=A0 22:04, Jonathan Paisley a =C3=A9crit :

> On 4 Nov 2005, at 20:53, Rupert BARROW wrote:
>
> Try replacing kvc_accessor with kvc_wrapper, as defined by Kimura =20
> Wataru in an email earlier today in this thread.

Sorry, I'm on a 'digest' mail subscription to the list, so I missed =20
your post, Kimura-San.

> This declares the ruby-level accessors in terms of setValue:forKey: =20=

> and valueForKey: which should hopefully solve the problem.
>
> i.e.:
>
>    kvc_wrapper :lastName, :firstName, :employeeID
>    kvc_depends_on=20
> ([ :lastName, :firstName, :employeeID ], :fullNameAndID)
>
>    def initialize
>       employeeID =3D @@lastID
>       ...
>    end

Great, thanks to both of you, things are now working, but not =20
*exactly* as you mention.
In my 'awakeFromInsert' method, why does

	employeeID =3D 123		# not self.employeeID
	puts employeeID
	puts self

reply '123' for employeeID, but NOT update self :
<Employee: 0x520d00> (entity: Employee; id: 0x523ed0 <x-coredata:///=20
Employee/tFC895DBA-B5E8-4764-B296-32FA106D280E> ; data: {
     department =3D nil;
     directReports =3D ();
     employeeID =3D nil;			# not updated
     firstName =3D First;
     lastName =3D Last;
     manager =3D nil;
     salary =3D 0;
})

when
	self.employeeID =3D 123	# using self.employeeID
	puts employeeID
	puts self
replies '123' for employeeID AND updates CORRECTLY self ?
<Employee: 0x520d00> (entity: Employee; id: 0x523ed0 <x-coredata:///=20
Employee/tFC895DBA-B5E8-4764-B296-32FA106D280E> ; data: {
     department =3D nil;
     directReports =3D ();
     employeeID =3D 123;			# updated correctly
     firstName =3D First;
     lastName =3D Last;
     manager =3D nil;
     salary =3D 0;
})

Thanks for your help. Being rather newbie on Cocoa/CoreData and =20
RubyCocoa, I do not know whether I am using incorrectly the former or =20=

the latter.

Cheers,
Rup=

Re: [Rubycocoa-talk] RubyCocoa, CoreData, & NSPersistentDocumentTutorial[MESSAGE NOT SCANNED]

[Jonathan P. <jp...@dc...>]

On 4 Nov 2005, at 20:53, Rupert BARROW wrote:
>
>     kvc_accessor :lastName, :firstName, :employeeID
>     kvc_depends_on 
> ([ :lastName, :firstName, :employeeID ], :fullNameAndID)
>
>     def initialize
>         rbSetValue_forKey(NSNumber.numberWithInt(@@lastID),  
> "employeeID")
>
> I am setting the employeeID correctly (through setValue), but the  
> value is not 'arriving' in the coredata object. However, it is not  
> lost, because I re-read from in another method (fullNameAndID) with  
> the valueForKey method.

You're calling rbSetValue... rather than setValue... The rbSetValue  
method ends up just calling the ruby employeeID= method, which has  
the effect of setting the instance variable in the ruby class (look  
at the definition in oc_import.rb to understand more).

Try replacing kvc_accessor with kvc_wrapper, as defined by Kimura  
Wataru in an email earlier today in this thread. This declares the  
ruby-level accessors in terms of setValue:forKey: and valueForKey:  
which should hopefully solve the problem.

i.e.:

    kvc_wrapper :lastName, :firstName, :employeeID
    kvc_depends_on 
([ :lastName, :firstName, :employeeID ], :fullNameAndID)

    def initialize
       employeeID = @@lastID
       ...
    end

Also, since this is a Cocoa class, I'm not sure that the initialize  
method works - I think you have to override init and define it like  
in a Cocoa class. I'll try and come up with an example later if you  
think this is a problem.


Hope that helps.
Jonathan

Re: [Rubycocoa-talk] RubyCocoa, CoreData, & NSPersistentDocumentTutorial

[Rupert B. <rup...@fr...>]

OK, here is the final results of the experiment.
With this code ...

class Employee < OSX::NSManagedObject
	include OSX
=09
	@@lastID =3D 123

	kvc_accessor :lastName, :firstName, :employeeID
	kvc_depends_on([ :lastName, :firstName, :employeeID ], =
:fullNameAndID)

	def initialize
		rbSetValue_forKey(NSNumber.numberWithInt(@@lastID), =
"employeeID")
		puts "*** =
rbSetValue_forKey(NSNumber.numberWithInt(@@lastID), =20
\"employeeID\") :  #{rbValueForKey("employeeID")} ***"
	end

	def fullNameAndID
		puts "*** self =3D #{self} ***"
		puts "*** rbValueForKey(\"employeeID\") =3D =
#{rbValueForKey=20
("employeeID")} ***"
	end
end

... I get the following output ...

*** rbSetValue_forKey(NSNumber.numberWithInt(@@lastID), =20
"employeeID") :  123 ***
*** self =3D <Employee: 0x557eb0> (entity: Employee; id: 0x55a950 <x-=20
coredata:///Employee/t31E6B1EE-8E9F-4E1A-BA1E-47CA62FEBF3D> ; data: {
     department =3D nil;
     directReports =3D ();
     employeeID =3D nil;
     firstName =3D First;
     lastName =3D Last;
     manager =3D nil;
     salary =3D 0;
}) ***
*** rbValueForKey("employeeID") =3D 123 ***

I am setting the employeeID correctly (through setValue), but the =20
value is not 'arriving' in the coredata object. However, it is not =20
lost, because I re-read from in another method (fullNameAndID) with =20
the valueForKey method.

Any ideas ?

Rup

Le 4 nov. 05 =C3=A0 21:32, Rupert Barrow a =C3=A9crit :

> Jonathan,
>
> Thanks for answering quickly.
>
> I removed definition of the @firstName and @lastName attributes, =20
> and it worked..
>
> After studying your answer, I noticed the following :
> '    kvc_accessor :lastName, :firstName'
> creates these methods :
> 	firstName
> 	firstName=3D
> 	lastName
> 	lastName=3D
>
> If I comment out this code, the methods disappear. When I dump =20
> Employee, I see the "attributes" :
> *** Employee =3D <Employee: 0x3ff5b0> (entity: Employee; id: 0x53f370 =20=

> <x-coredata:///Employee/tFE462E87-F74A-4826-8487-7C16DBECD891> ; =20
> data: {
>     department =3D nil;
>     directReports =3D ();
>     employeeID =3D nil;
>     firstName =3D First;
>     lastName =3D Last;
>     manager =3D nil;
>     salary =3D 0;
> }) ***
> but when I try to access them, the methods are obviously undefined :
> 2005-11-04 21:25:56.696 DepartmentAndEmployees[789] Exception =20
> raised during posting of notification.  Ignored.  exception: =20
> undefined method `lastName' for class `Employee'
>
> So :
> - kvc_depends_on declares and uses one set of accessors
> - CoreData relies on another access/update method : through =20
> setValue:forKey: and valueForKey:, as you mentioned.
>
> Would it not be nice to 'merge' them both, so as not to have to =20
> handle to access/update methods for the same data ?
>
> Rup
>
>
> Le 4 nov. 05 =C3=A0 00:13, Jonathan Paisley a =C3=A9crit :
>
>> On 3 Nov 2005, at 21:31, Rupert Barrow wrote:
>>
>>> class Employee < OSX::NSManagedObject
>>>     include OSX
>>>
>>>     attr_accessor :firstName
>>>     attr_accessor :lastName
>>>
>>>     kvc_accessor :lastName, :firstName
>>>     kvc_depends_on([ :lastName, :firstName], :fullNameAndID)
>>>
>>>     def initialize
>>>         @firstName =3D "MyFirst"
>>>         @lastName  =3D "MyLast"
>>>     end
>>>
>>
>> [Let me say first that I've not used Core Data yet, so this is my =20
>> understanding from glancing over the documentation. Please correct =20=

>> any misunderstanding!]
>>
>> Since you're using Core Data, all the data members of the class =20
>> are already managed for you. Therefore, you should not define =20
>> firstName and lastName as accessors, nor should you use ruby =20
>> @firstName/@lastName variables.
>