最近我一直在iPhone上玩一款名为《Scramble》的游戏。有些人可能知道这个游戏叫拼字游戏。从本质上讲,当游戏开始时,你会得到一个字母矩阵:

F X I E
A M L O
E W B X
A S T U

The goal of the game is to find as many words as you can that can be formed by chaining letters together. You can start with any letter, and all the letters that surround it are fair game, and then once you move on to the next letter, all the letters that surround that letter are fair game, except for any previously used letters. So in the grid above, for example, I could come up with the words LOB, TUX, SEA, FAME, etc. Words must be at least 3 characters, and no more than NxN characters, which would be 16 in this game but can vary in some implementations. While this game is fun and addictive, I am apparently not very good at it and I wanted to cheat a little bit by making a program that would give me the best possible words (the longer the word the more points you get).

(来源:boggled.org)

不幸的是,我不太擅长算法或它们的效率等等。我的第一次尝试使用一个像这样的字典(约2.3MB),并进行线性搜索,试图匹配字典条目的组合。这需要花费很长时间来找到可能的单词,因为你每轮只有2分钟的时间,这是不够的。

我很有兴趣看看是否有任何Stackoverflowers可以提出更有效的解决方案。我主要是在寻找使用三大p的解决方案:Python、PHP和Perl,尽管任何使用Java或c++的东西也很酷,因为速度是至关重要的。

目前的解决方案:

Adam Rosenfield, Python, ~20岁 John Fouhy, Python, ~3秒 Kent Fredric, Perl, ~1s Darius Bacon, Python, ~1s rvarcher, VB。净,~ 1 s Paolo Bergantino, PHP(实时链接),~5s(本地~2s)


当前回答

我已经在c#中使用DFA算法解决了这个问题。你可以查看我的代码

https://github.com/attilabicsko/wordshuffler/

除了在矩阵中查找单词外,我的算法还保存单词的实际路径,所以在设计单词查找游戏时,你可以检查在实际路径上是否有单词。

其他回答

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

/**
 * @author Sujeet Kumar (mrsujeet@gmail.com) It prints out all strings that can
 *         be formed by moving left, right, up, down, or diagonally and exist in
 *         a given dictionary , without repeating any cell. Assumes words are
 *         comprised of lower case letters. Currently prints words as many times
 *         as they appear, not just once. *
 */

public class BoggleGame 
{
  /* A sample 4X4 board/2D matrix */
  private static char[][] board = { { 's', 'a', 's', 'g' },
                                  { 'a', 'u', 't', 'h' }, 
                                  { 'r', 't', 'j', 'e' },
                                  { 'k', 'a', 'h', 'e' }
};

/* A sample dictionary which contains unique collection of words */
private static Set<String> dictionary = new HashSet<String>();

private static boolean[][] visited = new boolean[board.length][board[0].length];

public static void main(String[] arg) {
    dictionary.add("sujeet");
    dictionary.add("sarthak");
    findWords();

}

// show all words, starting from each possible starting place
private static void findWords() {
    for (int i = 0; i < board.length; i++) {
        for (int j = 0; j < board[i].length; j++) {
            StringBuffer buffer = new StringBuffer();
            dfs(i, j, buffer);
        }

    }

}

// run depth first search starting at cell (i, j)
private static void dfs(int i, int j, StringBuffer buffer) {
    /*
     * base case: just return in recursive call when index goes out of the
     * size of matrix dimension
     */
    if (i < 0 || j < 0 || i > board.length - 1 || j > board[i].length - 1) {
        return;
    }

    /*
     * base case: to return in recursive call when given cell is already
     * visited in a given string of word
     */
    if (visited[i][j] == true) { // can't visit a cell more than once
        return;
    }

    // not to allow a cell to reuse
    visited[i][j] = true;

    // combining cell character with other visited cells characters to form
    // word a potential word which may exist in dictionary
    buffer.append(board[i][j]);

    // found a word in dictionary. Print it.
    if (dictionary.contains(buffer.toString())) {
        System.out.println(buffer);
    }

    /*
     * consider all neighbors.For a given cell considering all adjacent
     * cells in horizontal, vertical and diagonal direction
     */
    for (int k = i - 1; k <= i + 1; k++) {
        for (int l = j - 1; l <= j + 1; l++) {
            dfs(k, l, buffer);

        }

    }
    buffer.deleteCharAt(buffer.length() - 1);
    visited[i][j] = false;
  }
}

首先,阅读c#语言设计师如何解决一个相关问题: http://blogs.msdn.com/ericlippert/archive/2009/02/04/a-nasality-talisman-for-the-sultana-analyst.aspx。

像他一样,您可以从字典开始,并通过从字母排序的字母数组到可以根据这些字母拼写的单词列表创建字典来规范化单词。

接下来,开始从黑板上创建可能的单词并查找它们。我怀疑这将让你走得很远,但肯定有更多的技巧可以加快速度。

如何简单的排序和使用字典中的二进制搜索?

在0.35秒内返回整个列表,并可以进一步优化(例如删除含有未使用字母的单词等)。

from bisect import bisect_left

f = open("dict.txt")
D.extend([line.strip() for line in f.readlines()])
D = sorted(D)

def neibs(M,x,y):
    n = len(M)
    for i in xrange(-1,2):
        for j in xrange(-1,2):
            if (i == 0 and j == 0) or (x + i < 0 or x + i >= n or y + j < 0 or y + j >= n):
                continue
            yield (x + i, y + j)

def findWords(M,D,x,y,prefix):
    prefix = prefix + M[x][y]

    # find word in dict by binary search
    found = bisect_left(D,prefix)

    # if found then yield
    if D[found] == prefix: 
        yield prefix

    # if what we found is not even a prefix then return
    # (there is no point in going further)
    if len(D[found]) < len(prefix) or D[found][:len(prefix)] != prefix:
        return

    # recourse
    for neib in neibs(M,x,y):
        for word in findWords(M,D,neib[0], neib[1], prefix):
            yield word

def solve(M,D):
    # check each starting point
    for x in xrange(0,len(M)):
        for y in xrange(0,len(M)):
            for word in findWords(M,D,x,y,""):
                yield word

grid = "fxie amlo ewbx astu".split()
print [x for x in solve(grid,D)]

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
})))

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.
}