[Omake] dynamic creation of variable names?

[Erick Tryzelaar]

Erick Tryzelaar wrote:
> Jason Hickey wrote:
>> Suppose we have some class/object hierarchy like this, where by the 
>> arrows, I mean "extends".
>>
>>    A
>>    |---> B
>>          |---> C
>>          |---> D
>>
>> Say A.x is virtual.  If I understand, the problem is that defining 
>> C.x will also affect the value of A.x.
>>
>> I think what you want is for virtuals to propagate only down:
>>    - redefining A.x propagates all the way down, so B.x, C.x, D.x all
>>      change
>>    - redefining B.x propagates to C.x and D.x, but not up to A.x
>>
>> This both makes sense and is weird at the same time.  We get some 
>> anomolies, like
>>
>>    C.x = 1
>>    A.x = 2
>>    # Is C.x 1 or 2?
>>
>> The need seems clear to me, the solution isn't as clear to me...
>>
> My vote would be that C.x would equal 1. I see this functionality 
> similar to how python does it's name lookup. Basically, you check the 
> local object for a value. If it doesn't exist, check the parent, and 
> up the list until we find the value or error out. So to extend your 
> example some more:
>
> A.x = 1 # A.x=1 B.x=1 C.x=1 D.x=1
> B.x = 2 # A.x=1 B.x=2 C.x=2 D.x=2
> C.x = 3 # A.x=1 B.x=2 C.x=3 D.x=3
> D.x = 4 # A.x=1 B.x=2 C.x=3 D.x=4
>
>> BTW, in your previous email about lazy fields, you are right that 
>> they are not doing the expected thing--I'll try to see what is going on.
> Cool, thanks.

Hi Jason,

It's been a couple months since we last discussed this issue with lazy 
fields, virtuals, and objects. Has there been any progress in getting 
this stuff fixed? I haven't noticed anything in the subversion checkins.

-e