I want to design a cache data structure. My requirements for it are:
Note: this is of a purely academic purpose - in order to learn. Thus I'm not interested in ready modules, I want to think out the design.
A simple hash would do for a cache, but of an unlimited size. Limiting it with LRU removal would take some more sweat.
Alongside with the hash, I should probably also have an array sorted by "recent usage", which removes (at head) and inserts (at tail) items in O(1). The only problem is when an item is referenced - I'd like to mark it as "recently used", moving it to the top of the list. First, this means splicing (taking an item from the middle of an array), second how to find it quickly...
The hash can hold a reference to the priority array for quick lookup - it costs space, but the caches are size limited so it's OK. About splicing, I can't think of easy solutions, maybe a linked list should be used instead of an array.
What are your opinions ?
In reply to Designing a cache data structure by spurperl
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |