CSC 301.01, Class 14: 2-3 trees

Overview

• Preliminaries
  • Notes and news
  • Upcoming work
  • Extra credit
  • Questions
• Review: red-black trees
• 2-3 trees
• Insertion in 2-3 trees
• Deletion in 2-3 trees
• 2-3 trees vs. red-black trees
• B-trees

News / Etc.

• I hope you had fun without me and enjoyed seeing a different model of teaching.
• I have not gotten grading done. If there are particular things you think might be a problem, ask questions.

Upcoming work

• Exam 1 due Wednesday at 5:00 p.m.
• The next assignment will be distributed on Wednesday.

Extra credit (Academic)

• CS Table, Tuesday: ???
• Google Stuff, Tuesday
• Convocation, Thursday

Extra credit (Peer)

Extra Credit (Misc)

Good things to do

• GHS Homecoming Parade, Thursday, 4pm at or near GHS

Questions

Will you accept mediocre formatting of our math?

Sure.

If I am breaking down summations (sum(sum(sum(…)))), should I explain

what rule I am applying for each summation?

Yes

The answers to sums a and b are standards. Can I just cite them?

No. Provide an anaysis of where the standard formula comes from

Do I have to prove my solutions to the recurrence relations using induction?

No. Just show how you got it with some explanation.

There are three approaches, each gives enough info if you tell me what you are doing.

What do you want for code?

Print out of the important portions.

You can use a word processor or markdown or …

I will not run the code; I can usually tell correctness by inspection.

Review: red-black trees

A red black tree is a binary search tree

• Properties mean that it is mostly balanced, height O(logn)
• Nodes are marked red or black.
• The black height of every leaf is the same (# of black nodes between root and leaf)
• Red nodes only have black children.
• Root is black.
• Leaves are black.
• The shortest path and the longest path can’t be off by more than a factor of 2 (all black nodes vs. alternating red and black)

How do we maintain these properties when we insert or delete?

• We need to rotate things. There are a lot of cases.
• Insert the node. If the parent is black and the grandparent is black, you can just change the parent to red.
• See if we screwed up the height. If so, rotate

We will come back to this.

2-3 trees

• Balanced tree.
• Trinary: Nodes have two or three children and one or two values
• The three subtree are
  • “less than first value”
  • “between first and second value”
  • “greater than second value”
• The height of every leaf is the same.
• The height must be logn or less. We may, however, have to do two comparisons per node in searching, so the overall cost of find is O(2*logn), which is still O(logn)
• We’ll make the leaves null wlog

Insertion in 2-3 trees

You’ve reached the leaf. You want to add another value. What do you do?

• Easy case: The parent has only one value. Just add a second value to the parent.
• Hard case: The parent has two values. You can’t just add a second value.
• Does it matter if you want to add to the left, middle, or right? No.
• Option 1: If you have lots of two-value nodes, split them all up and reorganize. Sam notes: Local and recurse is a good strategy.
• Option 2: Repeatedly move up? But then you might hit the root.
• Think about what you could do locally and then move the problem up

Deletion in 2-3 trees

2-3 trees vs. red-black trees

B-trees