home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!europa.asd.contel.com!emory!ogicse!reed!orpheus
- From: orpheus@reed.edu (P. Hawthorne)
- Newsgroups: comp.programming
- Subject: Balanced Trees (Re: Why Are Red-Black Trees Obscure?)
- Message-ID: <1992Aug27.215924.24525@reed.edu>
- Date: 27 Aug 92 21:59:24 GMT
- References: <1992Aug26.183817.7371@reed.edu> <1992Aug27.115551.7958@daimi.aau.dk> <PSU.92Aug27104235@ptero.cs.duke.edu>
- Organization: Reed College, Portland, OR
- Lines: 40
-
-
- psu@cs.duke.edu (Peter Su) laments:
- : All of the following are easier than AVL trees: splay trees, red-black
- : trees, skip-lists, randomized binary trees, etc. Why do people insist on
- : teaching AVL trees?
-
- (If I reveal some of my profound ignorance here, be gentle with me, huh?
- I've never had the pleasure of discussing data structures with others
- before, and some of my isolation may show through.)
-
- Maybe it's because AVL trees are always watching out for and correcting
- imbalance. The structures you mention are probabilistic solutions, not hard
- and fast guarantees. If you are particularly neurotic about computational
- cost and the cost of key compares outweighs the cost of path bookkeeping
- and tree rotations, I could see using AVL trees rather than the many "sure
- thing" alternatives. Plus, they're exotic Cold War souvenirs!
-
- I'll agree though, there isn't enough material available on those
- structures for the many circumstances in which they are the better
- alternatives. Ever tried to find an introduction to Fibonacci heaps or
- source for them that isn't in a mishmash of C++ and German?
-
- Is it a flaw in my implementation, or are top-down splay trees supposed
- to revert to a linked list when keys are inserted in order? Does a
- bottom-up splay operation solve that problem? I've always wondered.
-
- I'm partial to Patricia trees, from what I've read. Their balance and
- completeness look relatively good. I haven't seen any mention of a deletion
- operation on them however, and haven't ever really thought of them as
- dynamic structures that one could use where you might use other trees. Is
- that just a misconception of mine?
-
- Has anyone used frequency of node access data to bias the tree balancing
- rotation operations toward shorter paths?
-
- Given an ordered list of keys, what is the fastest way to tell whether or
- not they have a uniform probability distribution and just for good measure,
- where in the list it breaks down if at all?
-
- Theus (orpheus@reed.edu)
-