Comments on: Free Lists http://ttt.ggnore.net/2007/02/18/free-lists/ Gedanken: gedankenlos gedacht. Sat, 04 Sep 2010 19:03:22 +0000 http://wordpress.org/?v=2.5.1 By: purestorm http://ttt.ggnore.net/2007/02/18/free-lists/#comment-1345 purestorm Sun, 18 Feb 2007 19:12:55 +0000 http://ttt.ggnore.net/2007/02/18/free-lists/#comment-1345 Free lists are a common idiom in data structures themselves (Cormen calls it "aumenting data structures", for example). The indexes are nothing more than simply pointers with &array as their base address. As far as I know, it could be considered "good C style" to drop pointers in favour of indexes when appropriate. Note, however, that using indexes only works best with data structures of static size - e.g. arrays. When I want to insert the (MAX_CUSTOMER + 1)th customer then I have to reallocate the array since the indexes will not work correctly with "split" data structures. When I have a linked list of "customer tables" (each having MAX_CUSTOMER entries) then pointers would come in handy. Or I simply use a free list for every table. An remember, array[index] is nothing more than just writing *(&array + index * sizeof(content_type)). I do not know a good answer to overcome internal fragmentation. Regular compaction - "defragmentation" - would come at an actually noticeable cost. However, amortizedly analyzed, the cost should disappear. Free lists are a common idiom in data structures themselves (Cormen calls it “aumenting data structures”, for example).

The indexes are nothing more than simply pointers with &array as their base address. As far as I know, it could be considered “good C style” to drop pointers in favour of indexes when appropriate. Note, however, that using indexes only works best with data structures of static size - e.g. arrays.

When I want to insert the (MAX_CUSTOMER + 1)th customer then I have to reallocate the array since the indexes will not work correctly with “split” data structures. When I have a linked list of “customer tables” (each having MAX_CUSTOMER entries) then pointers would come in handy. Or I simply use a free list for every table.

An remember, array[index] is nothing more than just writing *(&array + index * sizeof(content_type)).

I do not know a good answer to overcome internal fragmentation. Regular compaction - “defragmentation” - would come at an actually noticeable cost. However, amortizedly analyzed, the cost should disappear.

]]> By: Justin http://ttt.ggnore.net/2007/02/18/free-lists/#comment-1344 Justin Sun, 18 Feb 2007 08:01:26 +0000 http://ttt.ggnore.net/2007/02/18/free-lists/#comment-1344 By the way, do you have a good idea to avoid fragmentation due to the deletion of entries? By the way, do you have a good idea to avoid fragmentation due to the deletion of entries?

]]> By: Justin http://ttt.ggnore.net/2007/02/18/free-lists/#comment-1343 Justin Sun, 18 Feb 2007 07:58:58 +0000 http://ttt.ggnore.net/2007/02/18/free-lists/#comment-1343 What I like about this approach is the use of an array index as a pointer. By this you can use other data structures for different purposes: e.G. you could us a binary tree for range queries, a hashmap for point queries but still a free list for insertion. You'd just have to adapt your data structure and algorithms "a bit". Is this a common C idiom that I passed by because I passed by C? What I like about this approach is the use of an array index as a pointer. By this you can use other data structures for different purposes: e.G. you could us a binary tree for range queries, a hashmap for point queries but still a free list for insertion. You’d just have to adapt your data structure and algorithms “a bit”.

Is this a common C idiom that I passed by because I passed by C?

]]>