On 5 January 2014 22:28, Chris Muller <
[hidden email]> wrote:
> Colin Putney said at some point:
>> I much prefer the Smalltalk community's tendency to build layers of
>> abstraction, while sticking to idiomatic Smalltalk. DSLs have their uses,
>> ... snip...
>
> Yes. I've always liked how everything we do in Smalltalk is built on
> an aggregation of just 3 or 4 concepts, and that's it. THAT is pretty
> slick in itself. Anything above that is "magic" and should pay
> _really_ good ROI to justify its existence (e.g., way more than simply
> aesthetics). Another example I can think of are those new special
> inst-vars in Pharo (having trouble remembering the name, "slots?") to
> define all sorts of RDBMS-style checks, etc. I would rather do that
> simply via the same few Smalltalk concepts as everything else.
> Instantiation, setting state, and message-sending..
Slots [1] are just ("just") instance variables promoted to first class
citizens. They do for instance variables what Metaclass does for Class
- they extend Smalltalk's metaobject protocol [2] to cover instance
variables. This lets you intercept reads and writes to the instance
variables, and all sorts of interesting stuff. It's still just
standard Smalltalk.
Seems like a perfectly sensible thing to do.
frank
[1]
http://hal.inria.fr/docs/00/64/17/16/PDF/Verw11b-OOSPLA11-FlexibleObjectLayouts.pdf[2]
https://en.wikipedia.org/wiki/The_Art_of_the_Metaobject_Protocol,
praised by Alan Kay in his OOPSLA 1997 speech as (something like) "the
best book written in 10 years.. but why did they have to write it in
Lisp?"
Locked
