[Newbies] question on differences between images

[Göran Krampe]

Hi!

> I've been playing a little bit more with Squeak, and I found another
> strange (to me; please, consider that I am still a newbie) thing.
> In some lectures by Stephane Ducasse there is an example illustrating the
> difference between literal arrays and arrays created with new:. In
> particular,
> adding the following method to class SmallInteger:
>
> m1
>
> | anArray |
> anArray := #( nil ).
> (anArray at: 1) isNil
> 	ifTrue: [ Transcript show: 'Put 1'; cr.  anArray at: 1 put: 1. ]
>
> and executing
>
> 1 m1
>
> should display the message 'Put 1' only once. And this is how it works...
> in image 3.9

Odd code though. Arrays created using the literal syntax "#( blab bla )"
should not be modified - it is very bad style. The Array is created at
*compile time*, not on method activation time. This is why it only has nil
in it the first time you call #m1.

> But if we repeat the experiment in image 3.0, it displays nothing in the
> Transcript (not even the first execution). Again, I am puzzled by this
> behavior (I am always assuming that these experiments that behave
> differently
> in different Squeak images are standard Smalltalk).

First of all - Squeak is not a "dead" implementation of Smalltalk-80 or
ANSI X3J20. It evolves. The recent largest change is of course the
addition of Traits in version 3.9.

In this case you are seeing a difference in how literal Arrays are parsed.
I don't have a 3.0 image handy but I bet that if you do an "inspect it"
on:

  #( nil )

...you will seen an instance of Array with a Symbol in it, *NOT* nil. The
Symbol #nil is not the same thing as the object referenced by nil (the
sole instance of class UndefinedObject). :)

I have a recollection that the parsing of literal Arrays was changed
relatively recently, perhaps it was in 3.9 even (or 3.8), so that the
special case of "nil" was parsed into a reference to the object nil, and
not parsed into a Symbol. I also vaguely recall that this created lots of
interesting problems in old code - but I think it was reasonably worked
out.

> One more thing. Perhaps there is people wondering why someone should worry
> about the behavior of an old image. The answer is that I am using a
> stripped
> version of image 3.0 to play with Squeak on a Jornada 720. More recent
> complete
> images are too heavy for the J720 (at least that is my experience; any
> clue
> on how to improve that?).

Check with Aaron Reichow - the Squeak-on-PDAs-expert. :)

regards, Göran