Re: [gobo-eiffel-develop] DS_BINARY_SEARCH_TREE status ?

[Daniel T. <dan...@gm...>]

Jann Röder wrote:
> Hi Daniel,
> I don't have a sample for all bugs. I will probably try to reproduce 
> it again later, but for starters:
I was able to reproduce the bug and know how to correct it. The problem 
is that one should never iterate like that through a 
DS_BINARY_SEARCH_TREE_CONTAINER and use `remove_node'. I will correct 
that as soon as possible. The problem is likely to be present in all set 
operations that remove items - certainly only in the binary search tree 
classes.
> * subtract does not reset the found_item, thus causing weird behavior 
> in has_key. The has_key implementation seems strange anyway. Why do 
> you use a class variable as iterator. Locals are faster and easier to 
> understand. I think the search feature should set found item to Void 
> in the beginning.
`found_node' is used to avoid redundant searches through the tree in 
`search_node'. E.g. if you first call `has (a_key)' and then `at 
(a_key)', then during the second call, the tree does not need to be 
traversed, as the result is already present in `found_node'.
At the time when I implemented all those classes I was not aware of the 
huge performance difference between attributes and local variables. I 
will change that as well in the loop.

Thanks for reporting the problem