当我看到问题陈述时，我想到了“Trie”。但看到其他一些海报使用了这种方法，我寻找另一种不同的方法。可惜的是，Trie方法表现更好。我在我的机器上运行了Kent的Perl解决方案，在调整它以使用我的字典文件后，它花了0.31秒运行。我自己的perl实现需要0.54秒才能运行。

这就是我的方法:

Create a transition hash to model the legal transitions. Iterate through all 16^3 possible three letter combinations. In the loop, exclude illegal transitions and repeat visits to the same square. Form all the legal 3-letter sequences and store them in a hash. Then loop through all words in the dictionary. Exclude words that are too long or short Slide a 3-letter window across each word and see if it is among the 3-letter combos from step 2. Exclude words that fail. This eliminates most non-matches. If still not eliminated, use a recursive algorithm to see if the word can be formed by making paths through the puzzle. (This part is slow, but called infrequently.) Print out the words I found. I tried 3-letter and 4-letter sequences, but 4-letter sequences slowed the program down.

在我的代码中，我使用/usr/share/dict/words作为我的字典。它是MAC OS X和许多Unix系统的标准配置。如果你愿意，你可以使用另一个文件。要破解不同的谜题，只需更改变量@puzzle。这将很容易适应更大的矩阵。你只需要改变%transitions哈希值和%legalTransitions哈希值。

这种解决方案的优点是代码短，数据结构简单。

下面是Perl代码(我知道它使用了太多的全局变量):

#!/usr/bin/perl
use Time::HiRes  qw{ time };

sub readFile($);
sub findAllPrefixes($);
sub isWordTraceable($);
sub findWordsInPuzzle(@);

my $startTime = time;

# Puzzle to solve

my @puzzle = ( 
    F, X, I, E,
    A, M, L, O,
    E, W, B, X,
    A, S, T, U
);

my $minimumWordLength = 3;
my $maximumPrefixLength = 3; # I tried four and it slowed down.

# Slurp the word list.
my $wordlistFile = "/usr/share/dict/words";

my @words = split(/\n/, uc(readFile($wordlistFile)));
print "Words loaded from word list: " . scalar @words . "\n";

print "Word file load time: " . (time - $startTime) . "\n";
my $postLoad = time;

# Define the legal transitions from one letter position to another. 
# Positions are numbered 0-15.
#     0  1  2  3
#     4  5  6  7
#     8  9 10 11
#    12 13 14 15
my %transitions = ( 
   -1 => [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],
    0 => [1,4,5], 
    1 => [0,2,4,5,6],
    2 => [1,3,5,6,7],
    3 => [2,6,7],
    4 => [0,1,5,8,9],
    5 => [0,1,2,4,6,8,9,10],
    6 => [1,2,3,5,7,9,10,11],
    7 => [2,3,6,10,11],
    8 => [4,5,9,12,13],
    9 => [4,5,6,8,10,12,13,14],
    10 => [5,6,7,9,11,13,14,15],
    11 => [6,7,10,14,15],
    12 => [8,9,13],
    13 => [8,9,10,12,14],
    14 => [9,10,11,13,15],
    15 => [10,11,14]
);

# Convert the transition matrix into a hash for easy access.
my %legalTransitions = ();
foreach my $start (keys %transitions) {
    my $legalRef = $transitions{$start};
    foreach my $stop (@$legalRef) {
        my $index = ($start + 1) * (scalar @puzzle) + ($stop + 1);
        $legalTransitions{$index} = 1;
    }
}

my %prefixesInPuzzle = findAllPrefixes($maximumPrefixLength);

print "Find prefixes time: " . (time - $postLoad) . "\n";
my $postPrefix = time;

my @wordsFoundInPuzzle = findWordsInPuzzle(@words);

print "Find words in puzzle time: " . (time - $postPrefix) . "\n";

print "Unique prefixes found: " . (scalar keys %prefixesInPuzzle) . "\n";
print "Words found (" . (scalar @wordsFoundInPuzzle) . ") :\n    " . join("\n    ", @wordsFoundInPuzzle) . "\n";

print "Total Elapsed time: " . (time - $startTime) . "\n";

###########################################

sub readFile($) {
    my ($filename) = @_;
    my $contents;
    if (-e $filename) {
        # This is magic: it opens and reads a file into a scalar in one line of code. 
        # See http://www.perl.com/pub/a/2003/11/21/slurp.html
        $contents = do { local( @ARGV, $/ ) = $filename ; <> } ; 
    }
    else {
        $contents = '';
    }
    return $contents;
}

# Is it legal to move from the first position to the second? They must be adjacent.
sub isLegalTransition($$) {
    my ($pos1,$pos2) = @_;
    my $index = ($pos1 + 1) * (scalar @puzzle) + ($pos2 + 1);
    return $legalTransitions{$index};
}

# Find all prefixes where $minimumWordLength <= length <= $maxPrefixLength
#
#   $maxPrefixLength ... Maximum length of prefix we will store. Three gives best performance. 
sub findAllPrefixes($) {
    my ($maxPrefixLength) = @_;
    my %prefixes = ();
    my $puzzleSize = scalar @puzzle;

    # Every possible N-letter combination of the letters in the puzzle 
    # can be represented as an integer, though many of those combinations
    # involve illegal transitions, duplicated letters, etc.
    # Iterate through all those possibilities and eliminate the illegal ones.
    my $maxIndex = $puzzleSize ** $maxPrefixLength;

    for (my $i = 0; $i < $maxIndex; $i++) {
        my @path;
        my $remainder = $i;
        my $prevPosition = -1;
        my $prefix = '';
        my %usedPositions = ();
        for (my $prefixLength = 1; $prefixLength <= $maxPrefixLength; $prefixLength++) {
            my $position = $remainder % $puzzleSize;

            # Is this a valid step?
            #  a. Is the transition legal (to an adjacent square)?
            if (! isLegalTransition($prevPosition, $position)) {
                last;
            }

            #  b. Have we repeated a square?
            if ($usedPositions{$position}) {
                last;
            }
            else {
                $usedPositions{$position} = 1;
            }

            # Record this prefix if length >= $minimumWordLength.
            $prefix .= $puzzle[$position];
            if ($prefixLength >= $minimumWordLength) {
                $prefixes{$prefix} = 1;
            }

            push @path, $position;
            $remainder -= $position;
            $remainder /= $puzzleSize;
            $prevPosition = $position;
        } # end inner for
    } # end outer for
    return %prefixes;
}

# Loop through all words in dictionary, looking for ones that are in the puzzle.
sub findWordsInPuzzle(@) {
    my @allWords = @_;
    my @wordsFound = ();
    my $puzzleSize = scalar @puzzle;
WORD: foreach my $word (@allWords) {
        my $wordLength = length($word);
        if ($wordLength > $puzzleSize || $wordLength < $minimumWordLength) {
            # Reject word as too short or too long.
        }
        elsif ($wordLength <= $maximumPrefixLength ) {
            # Word should be in the prefix hash.
            if ($prefixesInPuzzle{$word}) {
                push @wordsFound, $word;
            }
        }
        else {
            # Scan through the word using a window of length $maximumPrefixLength, looking for any strings not in our prefix list.
            # If any are found that are not in the list, this word is not possible.
            # If no non-matches are found, we have more work to do.
            my $limit = $wordLength - $maximumPrefixLength + 1;
            for (my $startIndex = 0; $startIndex < $limit; $startIndex ++) {
                if (! $prefixesInPuzzle{substr($word, $startIndex, $maximumPrefixLength)}) {
                    next WORD;
                }
            }
            if (isWordTraceable($word)) {
                # Additional test necessary: see if we can form this word by following legal transitions
                push @wordsFound, $word;
            }
        }

    }
    return @wordsFound;
}

# Is it possible to trace out the word using only legal transitions?
sub isWordTraceable($) {
    my $word = shift;
    return traverse([split(//, $word)], [-1]); # Start at special square -1, which may transition to any square in the puzzle.
}

# Recursively look for a path through the puzzle that matches the word.
sub traverse($$) {
    my ($lettersRef, $pathRef) = @_;
    my $index = scalar @$pathRef - 1;
    my $position = $pathRef->[$index];
    my $letter = $lettersRef->[$index];
    my $branchesRef =  $transitions{$position};
BRANCH: foreach my $branch (@$branchesRef) {
            if ($puzzle[$branch] eq $letter) {
                # Have we used this position yet?
                foreach my $usedBranch (@$pathRef) {
                    if ($usedBranch == $branch) {
                        next BRANCH;
                    }
                }
                if (scalar @$lettersRef == $index + 1) {
                    return 1; # End of word and success.
                }
                push @$pathRef, $branch;
                if (traverse($lettersRef, $pathRef)) {
                    return 1; # Recursive success.
                }
                else {
                    pop @$pathRef;
                }
            }
        }
    return 0; # No path found. Failed.
}

最快的解决方案可能是将字典存储在一个trie中。然后，创建一个三元组队列(x, y, s)，其中队列中的每个元素对应于一个可以在网格中拼写的单词的前缀s，结束于位置(x, y)。初始化队列中有N x N个元素(其中N是网格的大小)，网格中的每个正方形都有一个元素。然后，算法进行如下:

While the queue is not empty:
  Dequeue a triple (x, y, s)
  For each square (x', y') with letter c adjacent to (x, y):
    If s+c is a word, output s+c
    If s+c is a prefix of a word, insert (x', y', s+c) into the queue

如果将字典存储在trie中，则可以在常数时间内测试s+c是否是单词或单词的前缀(前提是还在每个队列数据中保留一些额外的元数据，例如指向trie中当前节点的指针)，因此此算法的运行时间为O(可拼写的单词数量)。

[编辑]下面是我刚刚编写的Python实现:

#!/usr/bin/python

class TrieNode:
    def __init__(self, parent, value):
        self.parent = parent
        self.children = [None] * 26
        self.isWord = False
        if parent is not None:
            parent.children[ord(value) - 97] = self

def MakeTrie(dictfile):
    dict = open(dictfile)
    root = TrieNode(None, '')
    for word in dict:
        curNode = root
        for letter in word.lower():
            if 97 <= ord(letter) < 123:
                nextNode = curNode.children[ord(letter) - 97]
                if nextNode is None:
                    nextNode = TrieNode(curNode, letter)
                curNode = nextNode
        curNode.isWord = True
    return root

def BoggleWords(grid, dict):
    rows = len(grid)
    cols = len(grid[0])
    queue = []
    words = []
    for y in range(cols):
        for x in range(rows):
            c = grid[y][x]
            node = dict.children[ord(c) - 97]
            if node is not None:
                queue.append((x, y, c, node))
    while queue:
        x, y, s, node = queue[0]
        del queue[0]
        for dx, dy in ((1, 0), (1, -1), (0, -1), (-1, -1), (-1, 0), (-1, 1), (0, 1), (1, 1)):
            x2, y2 = x + dx, y + dy
            if 0 <= x2 < cols and 0 <= y2 < rows:
                s2 = s + grid[y2][x2]
                node2 = node.children[ord(grid[y2][x2]) - 97]
                if node2 is not None:
                    if node2.isWord:
                        words.append(s2)
                    queue.append((x2, y2, s2, node2))

    return words

使用示例:

d = MakeTrie('/usr/share/dict/words')
print(BoggleWords(['fxie','amlo','ewbx','astu'], d))

输出:

['fa', 'xi', 'ie', 'io', 'el', 'am', 'ax', 'ae', 'aw', 'mi', 'ma', 'me', 'lo', 'li', 'oe', 'ox', 'em', 'ea', 'ea', 'es', 'wa', 'we', 'wa', 'bo', 'bu', 'as', 'aw', 'ae', 'st', 'se', 'sa', 'tu', 'ut', 'fam', 'fae', 'imi', 'eli', 'elm', 'elb', 'ami', 'ama', 'ame', 'aes', 'awl', 'awa', 'awe', 'awa', 'mix', 'mim', 'mil', 'mam', 'max', 'mae', 'maw', 'mew', 'mem', 'mes', 'lob', 'lox', 'lei', 'leo', 'lie', 'lim', 'oil', 'olm', 'ewe', 'eme', 'wax', 'waf', 'wae', 'waw', 'wem', 'wea', 'wea', 'was', 'waw', 'wae', 'bob', 'blo', 'bub', 'but', 'ast', 'ase', 'asa', 'awl', 'awa', 'awe', 'awa', 'aes', 'swa', 'swa', 'sew', 'sea', 'sea', 'saw', 'tux', 'tub', 'tut', 'twa', 'twa', 'tst', 'utu', 'fama', 'fame', 'ixil', 'imam', 'amli', 'amil', 'ambo', 'axil', 'axle', 'mimi', 'mima', 'mime', 'milo', 'mile', 'mewl', 'mese', 'mesa', 'lolo', 'lobo', 'lima', 'lime', 'limb', 'lile', 'oime', 'oleo', 'olio', 'oboe', 'obol', 'emim', 'emil', 'east', 'ease', 'wame', 'wawa', 'wawa', 'weam', 'west', 'wese', 'wast', 'wase', 'wawa', 'wawa', 'boil', 'bolo', 'bole', 'bobo', 'blob', 'bleo', 'bubo', 'asem', 'stub', 'stut', 'swam', 'semi', 'seme', 'seam', 'seax', 'sasa', 'sawt', 'tutu', 'tuts', 'twae', 'twas', 'twae', 'ilima', 'amble', 'axile', 'awest', 'mamie', 'mambo', 'maxim', 'mease', 'mesem', 'limax', 'limes', 'limbo', 'limbu', 'obole', 'emesa', 'embox', 'awest', 'swami', 'famble', 'mimble', 'maxima', 'embolo', 'embole', 'wamble', 'semese', 'semble', 'sawbwa', 'sawbwa']

Notes: This program doesn't output 1-letter words, or filter by word length at all. That's easy to add but not really relevant to the problem. It also outputs some words multiple times if they can be spelled in multiple ways. If a given word can be spelled in many different ways (worst case: every letter in the grid is the same (e.g. 'A') and a word like 'aaaaaaaaaa' is in your dictionary), then the running time will get horribly exponential. Filtering out duplicates and sorting is trivial to due after the algorithm has finished.

    package ProblemSolving;

import java.util.HashSet;
import java.util.Set;

/**
 * Given a 2-dimensional array of characters and a
 * dictionary in which a word can be searched in O(1) time.
 * Need to print all the words from array which are present
 * in dictionary. Word can be formed in any direction but
 * has to end at any edge of array.
 * (Need not worry much about the dictionary)
 */
public class DictionaryWord {
    private static char[][] matrix = new char[][]{
            {'a', 'f', 'h', 'u', 'n'},
            {'e', 't', 'a', 'i', 'r'},
            {'a', 'e', 'g', 'g', 'o'},
            {'t', 'r', 'm', 'l', 'p'}
    };
    private static int dim_x = matrix.length;
    private static int dim_y = matrix[matrix.length -1].length;
    private static Set<String> wordSet = new HashSet<String>();

    public static void main(String[] args) {
        //dictionary
        wordSet.add("after");
        wordSet.add("hate");
        wordSet.add("hair");
        wordSet.add("air");
        wordSet.add("eat");
        wordSet.add("tea");

        for (int x = 0; x < dim_x; x++) {
            for (int y = 0; y < dim_y; y++) {
                checkAndPrint(matrix[x][y] + "");
                int[][] visitedMap = new int[dim_x][dim_y];
                visitedMap[x][y] = 1;
                recursion(matrix[x][y] + "", visitedMap, x, y);
            }
        }
    }

    private static void checkAndPrint(String word) {
        if (wordSet.contains(word)) {
            System.out.println(word);
        }
    }

    private static void recursion(String word, int[][] visitedMap, int x, int y) {
        for (int i = Math.max(x - 1, 0); i < Math.min(x + 2, dim_x); i++) {
            for (int j = Math.max(y - 1, 0); j < Math.min(y + 2, dim_y); j++) {
                if (visitedMap[i][j] == 1) {
                    continue;
                } else {
                    int[][] newVisitedMap = new int[dim_x][dim_y];
                    for (int p = 0; p < dim_x; p++) {
                        for (int q = 0; q < dim_y; q++) {
                           newVisitedMap[p][q] = visitedMap[p][q];
                        }
                    }
                    newVisitedMap[i][j] = 1;
                    checkAndPrint(word + matrix[i][j]);
                    recursion(word + matrix[i][j], newVisitedMap, i, j);
                }
            }
        }
    }

}

令人惊讶的是，没有人尝试使用PHP版本。

这是John Fouhy的Python解决方案的PHP版本。

虽然我从其他人的答案中得到了一些建议，但这主要是抄袭约翰的。

$boggle = "fxie
           amlo
           ewbx
           astu";

$alphabet = str_split(str_replace(array("\n", " ", "\r"), "", strtolower($boggle)));
$rows = array_map('trim', explode("\n", $boggle));
$dictionary = file("C:/dict.txt");
$prefixes = array(''=>'');
$words = array();
$regex = '/[' . implode('', $alphabet) . ']{3,}$/S';
foreach($dictionary as $k=>$value) {
    $value = trim(strtolower($value));
    $length = strlen($value);
    if(preg_match($regex, $value)) {
        for($x = 0; $x < $length; $x++) {
            $letter = substr($value, 0, $x+1);
            if($letter == $value) {
                $words[$value] = 1;
            } else {
                $prefixes[$letter] = 1;
            }
        }
    }
}

$graph = array();
$chardict = array();
$positions = array();
$c = count($rows);
for($i = 0; $i < $c; $i++) {
    $l = strlen($rows[$i]);
    for($j = 0; $j < $l; $j++) {
        $chardict[$i.','.$j] = $rows[$i][$j];
        $children = array();
        $pos = array(-1,0,1);
        foreach($pos as $z) {
            $xCoord = $z + $i;
            if($xCoord < 0 || $xCoord >= count($rows)) {
                continue;
            }
            $len = strlen($rows[0]);
            foreach($pos as $w) {
                $yCoord = $j + $w;
                if(($yCoord < 0 || $yCoord >= $len) || ($z == 0 && $w == 0)) {
                    continue;
                }
                $children[] = array($xCoord, $yCoord);
            }
        }
        $graph['None'][] = array($i, $j);
        $graph[$i.','.$j] = $children;
    }
}

function to_word($chardict, $prefix) {
    $word = array();
    foreach($prefix as $v) {
        $word[] = $chardict[$v[0].','.$v[1]];
    }
    return implode("", $word);
}

function find_words($graph, $chardict, $position, $prefix, $prefixes, &$results, $words) {
    $word = to_word($chardict, $prefix);
    if(!isset($prefixes[$word])) return false;

    if(isset($words[$word])) {
        $results[] = $word;
    }

    foreach($graph[$position] as $child) {
        if(!in_array($child, $prefix)) {
            $newprefix = $prefix;
            $newprefix[] = $child;
            find_words($graph, $chardict, $child[0].','.$child[1], $newprefix, $prefixes, $results, $words);
        }
    }
}

$solution = array();
find_words($graph, $chardict, 'None', array(), $prefixes, $solution);
print_r($solution);

如果你想尝试的话，这里有一个实时链接。虽然在我的本地机器上需要大约2秒，但在我的web服务器上需要大约5秒。无论哪种情况，它都不是很快。尽管如此，它还是很可怕，所以我可以想象时间可以大大缩短。任何关于如何实现这一目标的建议都将不胜感激。PHP缺少元组，这使得坐标处理起来很奇怪，而且我无法理解到底发生了什么，这对我一点帮助都没有。

编辑:一些修复使它在本地少于1秒。

我意识到这个问题的时间来了又去了，但由于我自己正在研究一个求解器，并在谷歌搜索时偶然发现了这个，我想我应该发布一个参考，因为它似乎与其他一些问题有点不同。

我选择在游戏棋盘上使用平面数组，并从棋盘上的每个字母进行递归搜索，从有效邻居遍历到有效邻居，如果索引中的有效前缀是当前字母列表，则扩展搜索。而遍历当前单词的概念是进入板的索引列表，而不是组成单词的字母。在检查索引时，将索引转换为字母并完成检查。

索引是一个蛮力字典，有点像trie，但允许对索引进行python查询。如果单词'cat'和'cater'在列表中，你会在字典中看到:

   d = { 'c': ['cat','cater'],
     'ca': ['cat','cater'],
     'cat': ['cat','cater'],
     'cate': ['cater'],
     'cater': ['cater'],
   }

因此，如果current_word是'ca'，您就知道它是一个有效的前缀，因为'ca'在d中返回True(因此继续遍历板)。如果current_word是'cat'，那么你知道它是一个有效的单词，因为它是一个有效的前缀，并且d['cat']中的'cat'也返回True。

如果感觉这允许一些可读的代码，似乎不是太慢。像其他人一样，这个系统的费用是读取/构建索引。解这个板子相当麻烦。

代码在http://gist.github.com/268079。它是故意垂直和幼稚的，有很多明确的有效性检查，因为我想理解问题，而不是用一堆魔法或晦涩难懂的东西把它弄得乱七八糟。

一个Node.JS JavaScript解决方案。在不到一秒钟的时间内计算所有100个独特的单词，其中包括阅读字典文件(MBA 2012)。

Output: ["FAM","TUX","TUB","FAE","ELI","ELM","ELB","TWA","TWA","SAW","AMI","SWA","SWA","AME","SEA","SEW","AES","AWL","AWE","SEA","AWA","MIX","MIL","AST","ASE","MAX","MAE","MAW","MEW","AWE","MES","AWL","LIE","LIM","AWA","AES","BUT","BLO","WAS","WAE","WEA","LEI","LEO","LOB","LOX","WEM","OIL","OLM","WEA","WAE","WAX","WAF","MILO","EAST","WAME","TWAS","TWAE","EMIL","WEAM","OIME","AXIL","WEST","TWAE","LIMB","WASE","WAST","BLEO","STUB","BOIL","BOLE","LIME","SAWT","LIMA","MESA","MEWL","AXLE","FAME","ASEM","MILE","AMIL","SEAX","SEAM","SEMI","SWAM","AMBO","AMLI","AXILE","AMBLE","SWAMI","AWEST","AWEST","LIMAX","LIMES","LIMBU","LIMBO","EMBOX","SEMBLE","EMBOLE","WAMBLE","FAMBLE"]

代码:

var fs = require('fs')

var Node = function(value, row, col) {
    this.value = value
    this.row = row
    this.col = col
}

var Path = function() {
    this.nodes = []
}

Path.prototype.push = function(node) {
    this.nodes.push(node)
    return this
}

Path.prototype.contains = function(node) {
    for (var i = 0, ii = this.nodes.length; i < ii; i++) {
        if (this.nodes[i] === node) {
            return true
        }
    }

    return false
}

Path.prototype.clone = function() {
    var path = new Path()
    path.nodes = this.nodes.slice(0)
    return path
}

Path.prototype.to_word = function() {
    var word = ''

    for (var i = 0, ii = this.nodes.length; i < ii; ++i) {
        word += this.nodes[i].value
    }

    return word
}

var Board = function(nodes, dict) {
    // Expects n x m array.
    this.nodes = nodes
    this.words = []
    this.row_count = nodes.length
    this.col_count = nodes[0].length
    this.dict = dict
}

Board.from_raw = function(board, dict) {
    var ROW_COUNT = board.length
      , COL_COUNT = board[0].length

    var nodes = []

    // Replace board with Nodes
    for (var i = 0, ii = ROW_COUNT; i < ii; ++i) {
        nodes.push([])
        for (var j = 0, jj = COL_COUNT; j < jj; ++j) {
            nodes[i].push(new Node(board[i][j], i, j))
        }
    }

    return new Board(nodes, dict)
}

Board.prototype.toString = function() {
    return JSON.stringify(this.nodes)
}

Board.prototype.update_potential_words = function(dict) {
    for (var i = 0, ii = this.row_count; i < ii; ++i) {
        for (var j = 0, jj = this.col_count; j < jj; ++j) {
            var node = this.nodes[i][j]
              , path = new Path()

            path.push(node)

            this.dfs_search(path)
        }
    }
}

Board.prototype.on_board = function(row, col) {
    return 0 <= row && row < this.row_count && 0 <= col && col < this.col_count
}

Board.prototype.get_unsearched_neighbours = function(path) {
    var last_node = path.nodes[path.nodes.length - 1]

    var offsets = [
        [-1, -1], [-1,  0], [-1, +1]
      , [ 0, -1],           [ 0, +1]
      , [+1, -1], [+1,  0], [+1, +1]
    ]

    var neighbours = []

    for (var i = 0, ii = offsets.length; i < ii; ++i) {
        var offset = offsets[i]
        if (this.on_board(last_node.row + offset[0], last_node.col + offset[1])) {

            var potential_node = this.nodes[last_node.row + offset[0]][last_node.col + offset[1]]
            if (!path.contains(potential_node)) {
                // Create a new path if on board and we haven't visited this node yet.
                neighbours.push(potential_node)
            }
        }
    }

    return neighbours
}

Board.prototype.dfs_search = function(path) {
    var path_word = path.to_word()

    if (this.dict.contains_exact(path_word) && path_word.length >= 3) {
        this.words.push(path_word)
    }

    var neighbours = this.get_unsearched_neighbours(path)

    for (var i = 0, ii = neighbours.length; i < ii; ++i) {
        var neighbour = neighbours[i]
        var new_path = path.clone()
        new_path.push(neighbour)

        if (this.dict.contains_prefix(new_path.to_word())) {
            this.dfs_search(new_path)
        }
    }
}

var Dict = function() {
    this.dict_array = []

    var dict_data = fs.readFileSync('./web2', 'utf8')
    var dict_array = dict_data.split('\n')

    for (var i = 0, ii = dict_array.length; i < ii; ++i) {
        dict_array[i] = dict_array[i].toUpperCase()
    }

    this.dict_array = dict_array.sort()
}

Dict.prototype.contains_prefix = function(prefix) {
    // Binary search
    return this.search_prefix(prefix, 0, this.dict_array.length)
}

Dict.prototype.contains_exact = function(exact) {
    // Binary search
    return this.search_exact(exact, 0, this.dict_array.length)
}

Dict.prototype.search_prefix = function(prefix, start, end) {
    if (start >= end) {
        // If no more place to search, return no matter what.
        return this.dict_array[start].indexOf(prefix) > -1
    }

    var middle = Math.floor((start + end)/2)

    if (this.dict_array[middle].indexOf(prefix) > -1) {
        // If we prefix exists, return true.
        return true
    } else {
        // Recurse
        if (prefix <= this.dict_array[middle]) {
            return this.search_prefix(prefix, start, middle - 1)
        } else {
            return this.search_prefix(prefix, middle + 1, end)
        }
    }
}

Dict.prototype.search_exact = function(exact, start, end) {
    if (start >= end) {
        // If no more place to search, return no matter what.
        return this.dict_array[start] === exact
    }

    var middle = Math.floor((start + end)/2)

    if (this.dict_array[middle] === exact) {
        // If we prefix exists, return true.
        return true
    } else {
        // Recurse
        if (exact <= this.dict_array[middle]) {
            return this.search_exact(exact, start, middle - 1)
        } else {
            return this.search_exact(exact, middle + 1, end)
        }
    }
}

var board = [
    ['F', 'X', 'I', 'E']
  , ['A', 'M', 'L', 'O']
  , ['E', 'W', 'B', 'X']
  , ['A', 'S', 'T', 'U']
]

var dict = new Dict()

var b = Board.from_raw(board, dict)
b.update_potential_words()
console.log(JSON.stringify(b.words.sort(function(a, b) {
    return a.length - b.length
})))