(no subject)

Does anyone else feel that

        x ifNotNilDo: [: each | Transcript show: each; space]. (1)

and

        x ifNotNil: [ x do: [: each | Transcript show: each; space]]. (2)

ought to do the same thing?

Or perhaps nil should respond to do: (and do nothing?)

The rationale is to be able to use instance variables initialized to  
nil as empty collections, without having to  repeatedly test if they  
are nil before telling them to do: something, as in phrase (2)  
above.  One might be forgiven for thinking that ifNotNilDo: would  
have the same effect as do: on a non-nil object.

I realize that it is probably infeasible to make a change like this,  
but I couldn't resist thinking about it.

        Andrew

If #do: is to be understood by all objects, then its default behavior should
be to do nothing.  The reason is to avoid semantic inconsistency, since "#()
do: [:x | Transcript cr; show: x printString]" does not result in anything
being displayed on the Transcript.  To put it another way: the semantics of
#do: are that it enumeratates over the contents/elements of the receiver,
specifically excluding the receiver itself.

Whether it is wise for all objects to understand #do: is an entirely
different issue--one that I will leave as an exercise for others to debate,
should they be so inclined.

--Alan

-----Original Message-----
From: [hidden email]
[mailto:[hidden email]] On Behalf Of Andrew
P. Black
Sent: Friday, April 07, 2006 1:04 PM
To: The general-purpose Squeak developers list
Subject: (no subject)

Does anyone else feel that

        x ifNotNilDo: [: each | Transcript show: each; space].
(1)

and

        x ifNotNil: [ x do: [: each | Transcript show: each; space]].
(2)

ought to do the same thing?

Or perhaps nil should respond to do: (and do nothing?)

The rationale is to be able to use instance variables initialized to nil as
empty collections, without having to  repeatedly test if they are nil before
telling them to do: something, as in phrase (2) above.  One might be
forgiven for thinking that ifNotNilDo: would have the same effect as do: on
a non-nil object.

I realize that it is probably infeasible to make a change like this, but I
couldn't resist thinking about it.

        Andrew

On Apr 7, 2006, at 4:03 PM, Andrew P. Black wrote:

> Or perhaps nil should respond to do: (and do nothing?)
>
Objective-C does something like this: any message sent to nil just  
answers nil. Whether you prefer that over the Smalltalk strategy of  
raising MNU for most messages seems to be a matter of taste.  
Personally, I think either one is viable, but I don't think it would  
be a good idea to do both. That is, if nil raises MNU, it should do  
so for almost all messages.

Colin

Am 07.04.2006 um 22:03 schrieb Andrew P. Black:

Agreed, it would more aptly named ifNotNilIn:, as it is equivalent to  
"x ifNotNil: [x in: aBlock]".

Your problem reminds me of a pattern I've used before, not sure of  
its "official" name. It optimizes collections of 0 and 1 elements:

x := nil. "empty list"
x myDo: [:each | ...]. "not evaluated"
x := x myAdd: 42. "x = 42"
x myDo: [:each | ...]. "evaluated once"
x := x myAdd: 'hi'. "x = #(42 'hi')"
x myDo: [:each | ...]. "evaluated twice"


- Bert -