既然是要過慮,那就要先查找,如果是直接的一個字符一個字符的匹配,那是很耗時的,因爲時間花在不需要匹配的工作,有不少人會用正則去解決過慮,我09年的時候也這樣,但後來發現大量關鍵詞下性能確實極低下,所以纔會另想它法。上一文中的過慮主要思想是這樣的,開始會先用一個字典保存保存所有關鍵詞,同一個字母開頭的會另放在一個子字典裏,這樣一來,掃描的範圍就大大的縮小了,然後再考慮到髒字一般是2個字的佔了很大的比例,所以再在第二個字母做判斷,如果不存在就不需要再掃描下去了,至於可跳字符,就是在直接需要掃描的時候一一判斷的,沒技巧可講,另外一點值得注意的是,大小寫敏感的情況下,在判斷時需要轉換大小寫,大量關鍵詞影響不小,所以就初始化時再保存了一分小寫的,所以在掃描的時候就不需要轉換了,所以是否大小寫的兩個情況性能上不會有什麼變化,基本的思路就這樣了。如果說,你想要很準確的過慮,那就要用到分詞了(判斷可以人性化),我的方法只能處理比較簡單匹配與過慮。實現過程並沒有使用Aho–Corasick算法。
在查找資料的時候還瞭解到Aho–Corasick算法,它可以幫助我們快速的找出多個子字符串。可以到這裏瞭解算法:http://en.wikipedia.org/wiki/Aho%E2%80%93Corasick_string_matching_algorithm
5點有約會,先直接上實現代碼了(從一老外的例子裏小改的),本人測試過我上面的方法和使用Aho–Corasick過慮用的時候差不了多少:
代碼如下:
/// <summary>
/// 表示一個查找結果
/// </summary>
public struct KeywordSearchResult
{
private int index;
private string keyword;
public static readonly KeywordSearchResult Empty = new KeywordSearchResult(-1, string.Empty);
public KeywordSearchResult(int index, string keyword)
{
this.index = index;
this.keyword = keyword;
}
/// <summary>
/// 位置
/// </summary>
public int Index
{
get { return index; }
}
/// <summary>
/// 關鍵詞
/// </summary>
public string Keyword
{
get { return keyword; }
}
}
/// <summary>
/// Aho-Corasick算法實現
/// </summary>
public class KeywordSearch
{
/// <summary>
/// 構造節點
/// </summary>
private class Node
{
private Dictionary<char, Node> transDict;
public Node(char c, Node parent)
{
this.Char = c;
this.Parent = parent;
this.Transitions = new List<Node>();
this.Results = new List<string>();
this.transDict = new Dictionary<char, Node>();
}
public char Char
{
get;
private set;
}
public Node Parent
{
get;
private set;
}
public Node Failure
{
get;
set;
}
public List<Node> Transitions
{
get;
private set;
}
public List<string> Results
{
get;
private set;
}
public void AddResult(string result)
{
if (!Results.Contains(result))
{
Results.Add(result);
}
}
public void AddTransition(Node node)
{
this.transDict.Add(node.Char, node);
this.Transitions = this.transDict.Values.ToList();
}
public Node GetTransition(char c)
{
Node node;
if (this.transDict.TryGetValue(c, out node))
{
return node;
}
return null;
}
public bool ContainsTransition(char c)
{
return GetTransition(c) != null;
}
}
private Node root; // 根節點
private string[] keywords; // 所有關鍵詞
public KeywordSearch(IEnumerable<string> keywords)
{
this.keywords = keywords.ToArray();
this.Initialize();
}
/// <summary>
/// 根據關鍵詞來初始化所有節點
/// </summary>
private void Initialize()
{
this.root = new Node(' ', null);
// 添加模式
foreach (string k in this.keywords)
{
Node n = this.root;
foreach (char c in k)
{
Node temp = null;
foreach (Node tnode in n.Transitions)
{
if (tnode.Char == c)
{
temp = tnode; break;
}
}
if (temp == null)
{
temp = new Node(c, n);
n.AddTransition(temp);
}
n = temp;
}
n.AddResult(k);
}
// 第一層失敗指向根節點
List<Node> nodes = new List<Node>();
foreach (Node node in this.root.Transitions)
{
// 失敗指向root
node.Failure = this.root;
foreach (Node trans in node.Transitions)
{
nodes.Add(trans);
}
}
// 其它節點 BFS
while (nodes.Count != 0)
{
List<Node> newNodes = new List<Node>();
foreach (Node nd in nodes)
{
Node r = nd.Parent.Failure;
char c = nd.Char;
while (r != null && !r.ContainsTransition(c))
{
r = r.Failure;
}
if (r == null)
{
// 失敗指向root
nd.Failure = this.root;
}
else
{
nd.Failure = r.GetTransition(c);
foreach (string result in nd.Failure.Results)
{
nd.AddResult(result);
}
}
foreach (Node child in nd.Transitions)
{
newNodes.Add(child);
}
}
nodes = newNodes;
}
// 根節點的失敗指向自己
this.root.Failure = this.root;
}
/// <summary>
/// 找出所有出現過的關鍵詞
/// </summary>
/// <param name="text"></param>
/// <returns></returns>
public List<KeywordSearchResult> FindAllKeywords(string text)
{
List<KeywordSearchResult> list = new List<KeywordSearchResult>();
Node current = this.root;
for (int index = 0; index < text.Length; ++index)
{
Node trans;
do
{
trans = current.GetTransition(text[index]);
if (current == this.root)
break;
if (trans == null)
{
current = current.Failure;
}
} while (trans == null);
if (trans != null)
{
current = trans;
}
foreach (string s in current.Results)
{
list.Add(new KeywordSearchResult(index - s.Length + 1, s));
}
}
return list;
}
/// <summary>
/// 簡單地過慮關鍵詞
/// </summary>
/// <param name="text"></param>
/// <returns></returns>
public string FilterKeywords(string text)
{
StringBuilder sb = new StringBuilder();
Node current = this.root;
for (int index = 0; index < text.Length; index++)
{
Node trans;
do
{
trans = current.GetTransition(text[index]);
if (current == this.root)
break;
if (trans == null)
{
current = current.Failure;
}
} while (trans == null);
if (trans != null)
{
current = trans;
}
// 處理字符
if (current.Results.Count > 0)
{
string first = current.Results[0];
sb.Remove(sb.Length - first.Length + 1, first.Length -1);// 把匹配到的替換爲**
sb.Append(new string('*', current.Results[0].Length));
}
else
{
sb.Append(text[index]);
}
}
return sb.ToString();
}
}
原文:http://www.cnblogs.com/kudy/archive/2011/12/20/2294762.html