"Martin Kuball" <MartinKuball at web.de> wrote:


> Hi!
> 
> The comment in method PositionableStream>>#skip: says:
> 
> "Set the receiver's position to be the current position+anInteger. A 
>  subclass might choose to be more helpful and select the minimum of
>  the receiver's size and position+anInteger, or the maximum of 1 and 
>  position+anInteger for the repositioning."
> 
> Now why is this not implemented right there? Or at least in the 
> subclass ReadStream? I think this would be more compatible to the 
> behaviour of the method #next when reading past the end. It just 
> return nil. No error signaled.
We have the same behavior in
StandardFileStream>>basicNext.

CrLfFileStream>>upTo:  and
CrLFFileStream>>next
check for nil to detect the end-of-file 

You may wish to try this example with different Smalltalks:

 | st |
  st := ReadStream on: #(1 2 3).
  (1 to: 4) collect: [:x | st next]

Smalltalk V/286, Dolphin and VisualAge signal an error,
Squeak,VisualWork and Smalltalk/X answer #(1 2 3 nil)
> 
> By the way there is a class EndOfStream which was appearantly designed 
> to signal an error when reading past the end. But it is not used. And 
> I think it should not. So maybe the class should be removed.

In Dec 2003 I tried implement the modifications that are necessary to
use EndOfStream, but we could not come to a decision about the issue.
In fact there is little need to introduce exception handling in a situation
where nil can be used to represent the eof mark. Note also the the
EndOfStream answers nil when it is not handled (in instance method
defaultAction)

Boris