Re: Re[4]: KeyValueCoding (was Recent key-value encoding changes to NSObject.)

[Richard Frith-Macdonald]

On Monday, February 25, 2002, at 11:59 AM, Manuel Guesdon wrote:

> On Mon, 25 Feb 2002 11:40:47 +0000 Richard Frith-Macdonald 
> <address@hidden> wrote:
>
> > |
> > | On Monday, February 25, 2002, at 10:17 AM, Manuel Guesdon wrote:
> > | >
> > | > I would agree but concerning NSArray compute..., I think it's 
> > closely
> > | > related to it's valueForKey:
> > | > 
> > http://developer.apple.com/techpubs/webobjects/Reference/Javadoc/com/webobjects/
> > | > foundation/NSArray.html#valueForKey(java.lang.String)
> > | > <<
> > | > public Object valueForKey(String key)
> > | >     Conformance to the NSKeyValueCoding interface. NSArray's
> > | > implementation is more complex than the default:
> > | >         * If key indicates an operation that doesn't require an
> > | > argument, valueForKey performs the operation and returns the 
> > result.
> > | > key indicates an operation if its first character is "@". For 
> > example,
> > | > if key is "@count", valueForKey invokes compute on the "count"
> > | > operator. This has the effect of computing and returning the 
> > number of
> > | > elements in the receiver. Don't use valueForKey for operations that
> > | > take arguments; instead use valueForKeyPath.
> > | >         * For keys which do not begin with "@", valueForKey 
> > creates a
> > | > new array with the same number of elements as this array. For each
> > | > element, the corresponding element in the new array is the result 
> > of
> > | > invoking valueForKeyPath with key as the key path on the element. 
> > For
> > | > example, if key is "firstName", this method returns an array 
> > containing
> > | > the firstName values for each of the array's elements. The key 
> > argument
> > | > can be a key path of the form relationship.property. For example,
> > | > "department.name". valueForKey replaces null values with an 
> > instance of
> > | > NSKeyValueCoding.Null.
> > | >>>
> > | >
> > | > What do you think ?
> > |
> > | I think that you are talking about webobjects specific behavior.  You
> > | can tell this by the fact that this
> > | is only documented in the webobjects documentation, and the fact 
> > that it
> > | doesn't work like this in MacOS-X
> > |   (I checked by running a test program on MacOS-X 10.1.3).
> > |
> > | It looks to me like most (if not all) of the methods in your source 
> > file
> > | are webobjects specific ones
> > | which are not present in the apple Foundation.
>
> Please correct me if I'm wrong, but this mean that we'll have a 
> different behavious between NSArray in base and NSArray in 
> GNUstepWeb/EOF:
>         Exemple 1:
>                 NSArray -valueForKey:@"something" won't find a value in 
> base
>         but in gsweb    NSArray -valueForKey:@"something" will return 
> an array of results of performing valueForKey:@"something" on each 
> object of
> the array

Yes.

>         Exemple 2:
>                 NSArray -valueForKey:@"lastObject" will return the last 
> object of the array in base
>         but in gsweb    NSArray -valueForKey:@"lastObject" will return 
> an array of results of performing valueForKey:@"lastObject" on each 
> object of
> the array

Yes.

> Same kind of problem in NSDictionary

Yes.

> Don't you think the performXXforKey mecanism is interesting not only 
> for EOF but for other uses ?

Yes.
But people wanting it for other uses should be able to link with the 
appropriate libraries.
The fact that an idea is neat is a reason to have it available, but not 
a reason to include
it with everything.

My point ... which I obviously haven't managed to make clear yet ... is 
that we should not
introduce unexpected behaviors into the software without good reason.

A program built using the Foundation framework should work much the same 
way as a program
built using the base library.
A program built using Foundation and WebObjects should work much the 
same way as a program
built using base and gsweb.

Now I realise that this stuff has been dropped from Apples ObjC API ... 
but to me that means
the best place to put it is probably an EOF clone (since Apple have 
dropped EOF), and have
gsweb depend on the EOF code.  I'd like to see a decent EOF as part of 
GNUstep.

If/when apple move these extra methods into Foundation, we should move 
them into base.
That way, people porting will get the behaviors they expect, and we will 
be able to use
whichever behavior we want.