RBTreeNim

A Red/Black Tree implementation in Nim

Red/Black trees are self balancing binary search trees that maintain structure by tracking an extra bit of state about each node. This is then used to examine relationships between parents and children, allowing appropriate rotations to be performed. See Wikipedia for more info.

API Docs

http://nycto.github.io/RBTreeNim/rbtree.html

A Quick Tour

import rbtree

# Create a new tree
var tree = newRBTree[int, int]()

# Insert 4 values into the tree. Insert accepts 1 or more arguments to insert
tree.insert(2, 3, 1, 0)

# Remove one of those values
tree.delete(1)

# Check whether a value exists in the tree
if tree.contains(3):
    echo "Tree contains '3'"

# Iterate over every value in the tree
for i in tree:
    echo i

# Or iterate in reverse
for i in reversed(tree):
    echo i

Custom Keys

A Red/Black tree has a concept of a 'key' and a 'comparator'. When you call fetch, the key is what you pass in. Sometimes this is the object itself, but other times it's a derived value. When that is the case, you can define an extract function that returns the key.

For example, if you are indexing X/Y coordinates by just the x value:

import rbtree

proc extract( point: tuple[x, y: int] ): int = point.x

var tree = newRBTree[tuple[x, y: int], int]()

tree.insert( (x: 234, y: 789) )
tree.insert( (x: 890, y: 123) )

echo tree.contains(234)
echo tree

Custom Comparators

Values are inserted into a Red/Black Tree in sorted order. The definition of "sorted order", howerver, can be customized by defining a cmp function. It should take two values, (a, b), and returns < 0 if a < b, > 0 if a > b, and 0 if x == y.

For example, if you wanted to sort a tree of coordinates by their y values:

import rbtree

type MyPoint = object
    x, y: int

proc cmp*( a, b: MyPoint ): int = cmp(a.y, b.y)

var tree = newRBTree[MyPoint, MyPoint]()

tree.insert( MyPoint(x: 234, y: 789) )
tree.insert( MyPoint(x: 890, y: 123) )

echo tree

Multiple Indexes

Sometimes you will want to index the same object in different ways. However, this library uses compile time references to hook in the extract and cmp methods. To get around this, you can define a custom distinct type that will allow you to dispatch to different extract and cmp implementations.

import rbtree

# Define new types with custom extractors and comparators
defineIndex(XIndex, tuple[x, y: int], it.x, cmp(a, b))
defineIndex(YIndex, tuple[x, y: int], it.y, cmp(a, b))

var xIndex = newRBTree[XIndex, int]()
var yIndex = newRBTree[YIndex, int]()

let point1 = (x: 234, y: 789)
let point2 = (x: 890, y: 123)

xIndex.insert(point1, point2)
yIndex.insert(point1, point2)

echo xIndex
echo yIndex

License

This library is released under the MIT License, which is pretty spiffy. You should have received a copy of the MIT License along with this program. If not, see http://www.opensource.org/licenses/mit-license.php

Nycto / RBTreeNim