https://github.com/im-a-kookie/find-a-word
A ridiculous approach to creating even more ridiculous find-a-word puzzles
https://github.com/im-a-kookie/find-a-word
advanced-algorithms algorithm algorithm-challenges backtracking depth-first-search recursion trie
Last synced: 3 months ago
JSON representation
A ridiculous approach to creating even more ridiculous find-a-word puzzles
- Host: GitHub
- URL: https://github.com/im-a-kookie/find-a-word
- Owner: im-a-kookie
- License: mit
- Created: 2024-07-19T06:01:59.000Z (11 months ago)
- Default Branch: master
- Last Pushed: 2024-11-21T06:18:37.000Z (6 months ago)
- Last Synced: 2025-01-13T13:39:54.125Z (5 months ago)
- Topics: advanced-algorithms, algorithm, algorithm-challenges, backtracking, depth-first-search, recursion, trie
- Language: C#
- Homepage:
- Size: 281 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE.txt
Awesome Lists containing this project
README
# FindaWord
A really simple console application for making word search puzzles that increase difficulty by using limited numbers of letters. There may be bugs, as this is a proof of concept that has not been rigorously tested.
Simple command line usage, with an inbuilt utility for choosing words from an included dictionary. Inspired by a social media post lamenting the difficulty of creating such puzzles for a spouse. The algorithm is an absolute mess, I hacked it out in about 30 minutes. Modularity and Readability both leave much to be desired.
An example puzzle can be found [here](FindaWord/Resources/Puzzle.png).
# Algorithmic Details
Essentially, the algorithm takes a given set of words, then generates a word search puzzle from those words. The remaining spaces are filled exclusively with the letters of these words, but in such a way that no additional words will be formed, resulting in puzzles that will, quote; "make your eyes bleed."
Words are excluded by (1) graphing a dictionary of common English words as a progessive tree, and (2) walking this graph as letters are placed with orthogonal adjacency, ensuring that no such placement arrives at a word-terminating node, using TRIE, DFS, and Backtracking, to ensure grid validity. Easily configurable, but not feasible to remove words of 3 letters or less.
The dictionary is constructed as a progressive tree, which is quite efficient with Span slicing. However, due to the large wordlist, construction still takes 500ms and is memory intensive. As partitioning the list by initial letter can obviate the need for synchronization, a multithreaded implementation is quite straight forward and would offer dramatic performance benefits.
With only 26 possible branches, I suspect a dictionary lookup can be further optimized via indexing letters into uint and counting bits via LUT, thus indicating the real index of the letter in a continuous array.