[ANSI-Smalltalk] Re: A list of interesting methods, ifNotNil:

[Paolo Bonzini]

> I am now confused.
> The actual code in Squeak is
> 
>     ifNotEmptyDo: aBlock
>         ^self isEmpty ifTrue: [aBlock value: self]
> 
> It certainly doesn't involve any sending of the message #do:.

Which is all but intention revealing.

> x isEmpty ifTrue: [x do: monadicBlock]
> 
> would, if x is a collection, never do anything at all with monadicBlock.
> I take it that this should have been
> 
> x isEmpty ifFalse: [x do: monadicBlock]
> 
> and there is a name for that too:  #do:.

Of course, but there is no existing name for #ifEmpty:ifNotEmptyDo:.
Moreover, if you have another message #ifEmpty:ifNotEmpty: there is no
hint that the two differ *in the arity of the second block*!

If there were to be methods like #ifEmpty:, you should have only these four:

   #ifEmpty:
   #ifNotEmpty:
   #ifEmpty:ifNotEmpty:
   #ifNotEmpty:ifEmpty:

plus possibly

   #ifEmpty:ifNotEmptyDo:	(passing the second block to #do:)
   #ifNotEmptyDo:ifEmpty:

Furthermore, all parameters of all six methods should have variable
arity, because unlike the nil/not-nil case, there are many empty
collections and you may want to know it even in the empty case.  I agree
that this is borderline -- I cannot think of a use case indeed, and most
of the time you'll use #ifEmpty: just as a guard for ^self.

> VisualWorks also has an if*Do: method or two.
> GeneralBindingReference>>ifDefinedDo:
> has nothing to do with #do:.
> WindowManager>>activeControllerProcessIfBlockedDo:
> also has nothing to do with #do:.

But no #ifDefined: and #activeControllerProcessIfBlocked:.  I don't like
those names very much, but that's nothing compared to overloading  "Do:"
(which already has a well-known meaning in Smalltalk, i.e. "iterate") to
mark arity.

Paolo