the 3 lines are:
assign_z_ =: 4 : '(x) =: y'
OOP_z_ =: <'OOP'
coclass_z_ =: 18!:4@boxxopen@:[ ((('_OOP_' ,~ 'C' , ]) assign boxopen)^:(0 =
L.))
or the last line can be this instead:
coclass_OOP_ =: 18!:4@boxxopen@:[ ((('_OOP_' ,~ 'C' , ]) assign boxopen)^:(0 =
L.))
coclass is already defined to be the same as cocurrent.
what the new definition does is add a noun into the OOP locale that is the name
of the class with a 'C' prepended, and simply contains the boxed name of the
class.
There is a small consistency advantage from a single ( __ ) accessor syntax,
but the big benefit is:
coclass 'myclass' NB. or coclass__OOP 'myclass' with alternative definition.
obj1 =: 18!:3 ''
plus__obj1 =: +
coclass 'base'
plus__obj1
|value error: obj1
plus__obj1_myclass_
|ill-formed name: plus__obj1_myclass_
plus__obj1__Cmyclass__OOP
+
or with
coinsert 'OOP'
plus__obj1__Cmyclass
+
The benefit is accessing any hierarchy from any other locale directly. It just
costs a small string in an out of the way locale.
A reason to prefer the 2nd syntax is that you can avoid the small cost by just
using the coclass__OOP explicit alternate invocation for just the classes you
would like outside access to, (though you can still avoid creating the accessor
by using cocurrent instead of coclass).
The reason to prefer the 1st syntax is that you get libraries automatically
have an accessor, or not change existing code.
names__OOP '' will list all accessor locales, and how big/small you want this
list could influence your preference.
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
