A collection of useful algorithms keeping performance and flexibility on mind.

The following import will give you access to all the algorithms in this package.

import 'package:algorithmic/algorithmic.dart';

You can also import these algorithms separately:

To run benchmark on your own machine:

$ dart run benchmark

You can check the benchmark.log file for the benchmark results.

Index searching algorithms attempt to find the index of an item from a list.

⭐ 1000% faster than collection.lowerBound() and collection.binarySearch()

A faster searching algorithm for sorted list of items. It divides the list into two parts and discard one based on the middle value of them. This requires the list to be sorted in an increasing order.

A general searching algorithm for any kind of list. It tests every elements on the list one by one.

NOTE:

Sorting algorithms puts a list of items into an increasing order.

⭐ 500% faster than list.sort() for 700k items [algorithmic.quickSortHaore()]

Quicksort is an in-place sorting algorithm that works by selecting a pivot element and partitioning the list surrounding it. There are several schemes for selecting the pivot. The sorting performance varies for different schemes.

⭐ 600% faster than list.sort() for 700k numbers

Counting sort is a specialized sorting algorithm to sort integers of ranges in linear time. It counts the frequencies of the numbers appearing in an array, and uses it to reconstruct a sorted list.

Here, k is the range of numbers

⭐ 400% faster than list.sort() for 32 items

Cocktail shaker sort extends bubble sort by operating in two directions. It has O(n) time complexity for an already sorted list.

This algorithm is known by many other names, e.g.: bidirectional bubble sort, cocktail sort, shaker sort, ripple sort, shuffle sort, shuttle sort etc.

⭐ 300% faster than list.sort() for 32 items

Insertion sort sorts splits the list into two parts: the left part is ordered and the right part is unordered. In each iteration, it removes the first item from right part, and insert it into the left part maintaining the increasing order.

Comb sort improves bubble sort by eliminating small values near the end of the list, since they slows down bubble sort. It has O(n) time complexity for an already sorted list.

Selection sort algorithm performs sorting by finding the minimum element from the unordered range and putting it at the beginning in each iteration.

Bubble sort performs sorting by repeatedly stepping through the list, comparing adjacent elements and swapping them if they are in the wrong order.

⭐ 300% faster than list.sort() for 32 numbers

Gnome sort is a variation to the insertion sort which uses a much simpler implementation. It has O(n) time complexity for an already sorted list.

Merge sort is a divide and conquer based algorithm, which can handle very large arrays. Unlike quick sort, it promises O(n log n) performance in worst case and provides stable sort. But it requires equal amount of memory as the length of the array and the implementation runs slower compared than quick sort.

⭐ 200% faster than list.sort() for 700k numbers

Radix sort can sort a range of integers without using any comparisons. It is a special form of bucket sort, which distribute elements into buckets according to their radix.

Here, b is the radix

⭐ 120% faster than edit_distance.Levenshtein().distance()

Levenshtein distance returns the minimum number of ediit operations to transform one string (or array) to another. The permitted operations here are insertion, deletion and substitution. The Levenshtein distance between ABCD and BED is 2.

Here, n and m are the length of first and second string respectively.

⭐ 200% faster than edit_distance.Damerau().distance()

Damerau–Levenshtein distance is a variation of Lavenshtein distance which permits transposition of two adjacent characters along with insertion, deletion and substitution. e.g.: The Damerau-Levenshtein distance between CA and ABC is 2.

Here, n and m are the length of first and second string respectively and k is number of unique characters appearing in both string.

⭐ 1000% faster than edit_distance.Damerau().distance()

The Damerau–Levenshtein distance can be restricted with a condition that no substring can be edited more than once, which simplifies the implementation of transposition lookup. This distance is also known as Optimal String Alignment distance. e.g.: The restricted Damerau-Levenshtein edit distance between CA and ABC is 3.

Here, n and m are the length of first and second string respectively.

The Jaro similarity between two strings is the weighted sum of percentage of matched characters from each string and transposed characters. Winkler increased this measure for matching initial characters.

Here, n and m are the length of first and second string respectively.

Hamming distance between two equal-length strings is the number of positions at which the characters differs. In this implementation, if two strings are not equal, the shorter string is padded to match the longer ones.

Lee distance can only be calculated between two equal-length strings.

⭐ 200% faster than edit_distance.LongestCommonSubsequence().distance()

Longest Common Subsequence is the longest of the common subsequences of two strings.

Longest common substrings and longest common subsequences are not the same. Unlike substrings, subsequences are not required to occupy consecutive positions.

Here, n and m are the length of first and second string respectively.

⭐ 150% faster than edit_distance.LongestCommonSubsequence().distance()

Jaccard index is a metric used to measure similarity of two samples sets.

The complement of it is Jaccard distance which measures the total number of items that is present in one list but not the other.

Sørensen–Dice coefficient is a metric used to measure similarity between two samples.

Tversky similarity index an asymmetric similarity measure between sets that compares a variant with a prototype.

• Benchmark
• Roadmap
• License