下面的代码基于NFA算法实现了在多段字符串中匹配正则表达式,对比NFA算法可以看到它将pc由局部变量提升为类成员,以保存中间匹配状态,另外在匹配成功后将pc恢复到null状态。实际使用中,此类还应该增加一个"reset"方法,将pc值置为null,以便用户可以显式要求重新开始匹配。
public class MultiSegmentNFA {
private final Digraph G; // digraph of epsilon transitions
private final String regexp; // regular expression
private final int M; // number of characters in regular expression
private Bag pc = null;
public MultiSegmentNFA(String regexp) {
... // same as NFA
}
public boolean recognizes(String target) {
if (pc == null) {
DirectedDFS dfs = new DirectedDFS(G, 0);
pc = new Bag();
for (int v = 0; v < G.V(); v++)
if (dfs.marked(v))
pc.add(v);
}
for (int i = 0; i < target.length(); i++) {
Bag match = new Bag();
for (int v : pc) {
if (v == M)
continue;
if ((regexp.charAt(v) == target.charAt(i))
|| regexp.charAt(v) == '.')
match.add(v + 1);
}
DirectedDFS dfs = new DirectedDFS(G, match);
pc = new Bag();
for (int v = 0; v < G.V(); v++)
if (dfs.marked(v))
pc.add(v);
if (pc.size() == 0)
return false;
}
for (int v : pc)
if (v == M) {
pc = null;
return true;
}
return false;
}
public static void main(String[] args) {
LinkedList msgs = new LinkedList();
msgs.offer("welcome lonely logoout");
msgs.offer("to flog");
msgs.offer("into a fog");
String token = "outto"; // 这个目标由第1和第2个字符串拼接而成
String pat = "(." + token + ".)";
MultiSegmentNFA mnfa = new MultiSegmentNFA(pat);
String target = msgs.poll();
while (target != null) {
if (mnfa.recognizes(target)) {
break;
}
target = msgs.poll();
}
if (target == null) {
System.out.println("cannot find " + token + " in msgs.");
} else {
System.out.println("find pat in <" + target + ">");
}
}
}
Note: Java的Pattern类使用的就是基于NFA的搜索算法,见JDK 6文档java.util.regex.Pattern的"Comparison to Perl 5"一节。