A redblack tree is a binary search tree in which each node, , has a colour which is either red or black. Red is represented by the value 0 and black by the value .
Before and after any operation on a redblack tree, the following two properties are satisfied. Each property is defined both in terms of the colours red and black, and in terms of the numeric values 0 and 1.
Notice that we can always colour the root, , of a redblack tree black without violating either of these two properties, so we will assume that the root is black, and the algorithms for updating a redblack tree will maintain this. Another trick that simplifies redblack trees is to treat the external nodes (represented by ) as black nodes. This way, every real node, , of a redblack tree has exactly two children, each with a welldefined colour. An example of a redblack tree is shown in Figure 9.4.

At first it might seem surprising that a redblack tree can be efficiently updated to maintain the blackheight and norededge properties, and it seems unusual to even consider these as useful properties. However, redblack trees were designed to be an efficient simulation of 24 trees as binary trees.
Refer to Figure 9.5. Consider any redblack tree, , having nodes and perform the following transformation: Remove each red node and connect 's two children directly to the (black) parent of . After this transformation we are left with a tree having only black nodes.
Every internal node in has two, three, or four children: A black node that started out with two black children will still have two black children after this transformation. A black node that started out with one red and one black child will have three children after this transformation. A black node that started out with two red children will have four children after this transformation. Furthermore, the blackheight property now guarantees that every roottoleaf path in has the same length. In other words, is a 24 tree!
The 24 tree has leaves that correspond to the external nodes of the redblack tree. Therefore, this tree has height at most . Now, every root to leaf path in the 24 tree corresponds to a path from the root of the redblack tree to an external node. The first and last node in this path are black and at most one out of every two internal nodes is red, so this path has at most black nodes and at most red nodes. Therefore, the longest path from the root to any internal node in is at most
Now that we have seen the relationship between 24 trees and redblack trees, it is not hard to believe that we can efficiently maintain a redblack tree while adding and removing elements.
We have already seen that adding an element in a BinarySearchTree can be done by adding a new leaf. Therefore, to implement in a redblack tree we need a method of simulating splitting a node with five children in a 24 tree. A 24 tree node with five children is represented by a black node that has two red children, one of which also has a red child. We can ``split'' this node by colouring it red and colouring its two children black. An example of this is shown in Figure 9.6.

Similarly, implementing requires a method of merging two nodes and borrowing a child from a sibling. Merging two nodes is the inverse of a split (shown in Figure 9.6), and involves colouring two (black) siblings red and colouring their (red) parent black. Borrowing from a sibling is the most complicated of the procedures and involves both rotations and recolouring nodes.
Of course, during all of this we must still maintain the norededge property and the blackheight property. While it is no longer surprising that this can be done, there are a large number of cases that have to be considered if we try to do a direct simulation of a 24 tree by a redblack tree. At some point, it just becomes simpler to disregard the underlying 24 tree and work directly towards maintaining the properties of the redblack tree.
No single definition of redblack trees exists. Rather, there is a family of structures that manage to maintain the blackheight and norededge properties during and operations. Different structures do this in different ways. Here, we implement a data structure that we call a RedBlackTree. This structure implements a particular variant of redblack trees that satisfies an additional property:
Note that the redblack tree shown in Figure 9.4 does not satisfy the leftleaning property; it is violated by the parent of the red node in the rightmost path.
The reason for maintaining the leftleaning property is that it reduces the number of cases encountered when updating the tree during and operations. In terms of 24 trees, it implies that every 24 tree has a unique representation: A node of degree two becomes a black node with two black children. A node of degree three becomes a black node whose left child is red and whose right child is black. A node of degree four becomes a black node with two red children.
Before we describe the implementation of
and
in
detail, we first present some simple subroutines used by these methods
that are illustrated in Figure 9.7. The first two
subroutines are for manipulating colours while preserving the blackheight
property. The
method takes as input a black node
that has two red children and colours
red and its two children black.
The
method reverses this operation:
The
method swaps the colours of
and
and then performs a left rotation at
. This method reverses the
colours of these two nodes as well as their parentchild relationship:
The
operation
is especially useful in restoring the leftleaning property at a node
that violates it (because
is black and
is red).
In this special case, we can be assured that this operation preserves both
the blackheight and norededge properties. The
operation
is symmetric with
, when the roles of left and right are reversed.
To implement
in a RedBlackTree, we perform a standard
BinarySearchTree insertion to add a new leaf,
, with
and
set
. Note that this does not change the black height
of any node, so it does not violate the blackheight property. It may,
however, violate the leftleaning property (if
is the right child of
its parent), and it may violate the norededge property (if
's parent
is
). To restore these properties, we call the method
.
Illustrated in Figure 9.8, the method takes as input a node whose colour is red and which may violate the norededge property and/or the leftleaning property. The following discussion is probably impossible to follow without referring to Figure 9.8 or recreating it on a piece of paper. Indeed, the reader may wish to study this figure before continuing.
If is the root of the tree, then we can colour black to restore both properties. If 's sibling is also red, then 's parent must be black, so both the leftleaning and norededge properties already hold.
Otherwise, we first determine if 's parent, , violates the leftleaning property and, if so, perform a operation and set . This leaves us in a welldefined state: is the left child of its parent, , so now satisfies the leftleaning property. All that remains is to ensure the norededge property at . We only have to worry about the case in which is red, since otherwise already satisfies the norededge property.
Since we are not done yet, is red and is red. The norededge property (which is only violated by and not by ) implies that 's grandparent exists and is black. If 's right child is red, then the leftleaning property ensures that both 's children are red, and a call to makes red and black. This restores the norededge property at , but may cause it to be violated at , so the whole process starts over with .
If
's right child is black, then a call to
makes
the (black) parent of
and gives
two red children,
and
. This ensures that
satisfies the norededge property and
satisfies the leftleaning property. In this case we can stop.
The method takes constant time per iteration and each iteration either finishes or moves closer to the root. Therefore, the method finishes after iterations in time.
The operation in a RedBlackTree is the most complicated to implement, and this is true of all known redblack tree variants. Just like the operation in a BinarySearchTree, this operation boils down to finding a node with only one child, , and splicing out of the tree by having adopt .
The problem with this is that, if
is black, then the blackheight
property will now be violated at
. We may avoid this problem,
temporarily, by adding
to
. Of course, this introduces
two other problems: (1) if
and
both started out black, then
(double black), which is an invalid colour.
If
was red, then it is replaced by a black node
, which may
violate the leftleaning property at
. Both of these
problems can be resolved with a call to the
method.
The
method takes as its input a node
whose colour is black
(1) or doubleblack (2). If
is doubleblack, then
performs a series of rotations and recolouring operations that move the
doubleblack node up the tree until it can be eliminated. During this
process, the node
changes until, at the end of this process,
refers to the root of the subtree that has been changed. The root of
this subtree may have changed colour. In particular, it may have gone
from red to black, so the
method finishes by checking
if
's parent violates the leftleaning property and, if so, fixing it.
The method is illustrated in Figure 9.9. Again, the following text will be difficult, if not impossible, to follow without referring to Figure 9.9. Each iteration of the loop in processes the doubleblack node , based on one of four cases:
Case 0: is the root. This is the easiest case to treat. We recolour to be black (this does not violate any of the redblack tree properties).
Case 1:
's sibling,
, is red. In this case,
's sibling is the
left child of its parent,
(by the leftleaning property). We perform
a rightflip at
and then proceed to the next iteration. Note that
this action causes
's parent to violate the leftleaning property and
the depth of
to increase. However, it also implies that the next
iteration will be in Case 3 with
coloured red. When examining Case 3
below, we will see that the process will stop during the next iteration.
Case 2: 's sibling, , is black, and is the left child of its parent, . In this case, we call , making black, red, and darkening the colour of to black or doubleblack. At this point, does not satisfy the leftleaning property, so we call to fix this.
At this point, is red and is the root of the subtree with which we started. We need to check if causes the norededge property to be violated. We do this by inspecting 's right child, . If is black, then satisfies the norededge property and we can continue the next iteration with .
Otherwise ( is red), so both the norededge property and the leftleaning properties are violated at and , respectively. The leftleaning property is restored with a call to , but the norededge property is still violated. At this point, is the left child of , is the left child of , and are both red, and is black or doubleblack. A makes the parent of both and . Following this up by a makes both and black and sets the colour of back to the original colour of .
At this point, the doubleblack node is has been eliminated and the
norededge and blackheight properties are reestablished. Only one possible problem remains: the right child of
may be red, in which
case the leftleaning property would be violated. We check this and
perform a
to correct it if necessary.
Case 3: 's sibling is black and is the right child of its parent, . This case is symmetric to Case 2 and is handled mostly the same way. The only differences come from the fact that the leftleaning property is asymmetric, so it requires different handling.
As before, we begin with a call to , which makes red and black. A call to promotes to the root of the subtree. At this point is red, and the code branches two ways depending on the colour of 's left child, .
If is red, then the code finishes up exactly the same way as Case 2 does, but is even simpler since there is no danger of not satisfying the leftleaning property.
The more complicated case occurs when is black. In this case, we examine the colour of 's left child. If it is red, then has two red children and its extra black can be pushed down with a call to . At this point, now has 's original colour, and we are done.
If
's left child is black, then
violates the leftleaning property,
and we restore this with a call to
. We then return the
node
so that the next iteration of
then continues
with
.
Each iteration of takes constant time. Cases 2 and 3 either finish or move closer to the root of the tree. Case 0 (where is the root) always terminates and Case 1 leads immediately to Case 3, which also terminates. Since the height of the tree is at most , we conclude that there are at most iterations of , so runs in time.
opendatastructures.org