Re: [PATCH 21/22] teach the merge algorithm about cache iterators

> On Mon, 12 Sep 2005, Chuck Lever wrote:
>>For now, we simply replace indpos with a cache cursor.  Likely more
>>changes will be needed after we successfully replace the cache array
>>with an abstract data type.
> 
> The right order is probably to add the concept of a cache that isn't the 
> one that normal functions deal with, have read_cache_unmerged return such 
> a thing, call cc_init with that, and rip out all of the removal and 
> position adjustment code. Then read_tree won't care at all about the 
> internal structure of the cache type, and it can be replaced without any 
> problem.