一个相当好的状态机(DFA, 确定有限状态机)的编码实现，相当简洁漂亮

从jsoup而来，文章见： https://github.com/code4craft/jsoup-learning/blob/master/blogs/jsoup4.md

状态机

Jsoup的词法分析和语法分析都用到了状态机。状态机可以理解为一个特殊的程序模型，例如经常跟我们打交道的正则表达式就是用状态机实现的。

它由状态(state)和转移(transition)两部分构成。根据状态转移的可能性，状态机又分为DFA(确定有限状态机)和NFA(非确定有限状态自动机)。这里拿一个最简单的正则表达式"a[b]*"作为例子，我们先把它映射到一个状态机DFA，大概是这样子：

状态机本身是一个编程模型，这里我们尝试用程序去实现它，那么最直接的方式大概是这样：

    public void process(StringReader reader) throws StringReader.EOFException {
        char ch;
        switch (state) {
            case Init:
                ch = reader.read();
                if (ch == 'a') {
                    state = State.AfterA;
                    accum.append(ch);
                }
                break;
            case AfterA:
                ...
                break;
            case AfterB:
                ...
                break;
            case Accept:
                ...
                break;
        }
    }

这样写简单的状态机倒没有问题，但是复杂情况下就有点难受了。还有一种标准的状态机解法，先建立状态转移表，然后使用这个表建立状态机。这个方法的问题就是，只能做纯状态转移，无法在代码级别操作输入输出。

Jsoup里则使用了状态模式来实现状态机，初次看到时，确实让人眼前一亮。状态模式是设计模式的一种，它将状态和对应的行为绑定在一起。而在状态机的实现过程中，使用它来实现状态转移时的处理再合适不过了。

"a[b]*"的例子的状态模式实现如下，这里采用了与Jsoup相同的方式，用到了枚举来实现状态模式：

    public class StateModelABStateMachine implements ABStateMachine {

        State state;

        StringBuilder accum;

        enum State {
            Init {
                @Override
                public void process(StateModelABStateMachine stateModelABStateMachine, StringReader reader) throws StringReader.EOFException {
                    char ch = reader.read();
                    if (ch == 'a') {
                        stateModelABStateMachine.state = AfterA;
                        stateModelABStateMachine.accum.append(ch);
                    }
                }
            },
            Accept {
                ...
            },
            AfterA {
                ...
            },
            AfterB {
                ...
            };

            public void process(StateModelABStateMachine stateModelABStateMachine, StringReader reader) throws StringReader.EOFException {
            }
        }

        public void process(StringReader reader) throws StringReader.EOFException {
            state.process(this, reader);
        }
    }

完整的实现程序如下：

StateModelABStateMachine.java:

package us.codecraft.learning.automata;

/**
 * @author code4crafter@gmail.com
 */
public class StateModelABStateMachine implements ABStateMachine {
    State state = State.Init;
    StringBuilder accum = new StringBuilder();

    enum State {
        Init {
            @Override
            public void process(StateModelABStateMachine stateModelABStateMachine, StringReader reader) throws StringReader.EOFException {
                char ch = reader.read();
                if (ch == 'a') {
                    stateModelABStateMachine.state = AfterA;
                    stateModelABStateMachine.accum.append(ch);
                }
            }
        },

        Accept {
            @Override
            public void process(StateModelABStateMachine stateModelABStateMachine, StringReader reader) throws StringReader.EOFException {
                System.out.println("find " + stateModelABStateMachine.accum.toString());
                stateModelABStateMachine.accum = new StringBuilder();
                stateModelABStateMachine.state = Init;
                reader.unread();
            }
        },

        AfterA {
            @Override
            public void process(StateModelABStateMachine stateModelABStateMachine, StringReader reader) throws StringReader.EOFException {
                char ch = reader.read();
                if (ch == 'b') {
                    stateModelABStateMachine.accum.append(ch);
                    stateModelABStateMachine.state = AfterB;
                } else {
                    stateModelABStateMachine.state = Accept;
                }
            }
        },

        AfterB {
            @Override
            public void process(StateModelABStateMachine stateModelABStateMachine, StringReader reader) throws StringReader.EOFException {
                char ch = reader.read();
                if (ch == 'b') {
                    stateModelABStateMachine.accum.append(ch);
                    stateModelABStateMachine.state = AfterB;
                } else {
                    stateModelABStateMachine.state = Accept;
                }
            }
        };

        public void process(StateModelABStateMachine stateModelABStateMachine, StringReader reader) throws StringReader.EOFException {
        }
    }

    @Override
    public void process(StringReader reader) throws StringReader.EOFException {
        state.process(this, reader);
    }

    public static void main(String[] args) {
        ABStateMachine abStateMachine = new StateModelABStateMachine();
        String text = "abbbababbbaa";
        StringReader reader = new StringReader(text);
        try {
            while (true) {
                abStateMachine.process(reader);
            }
        } catch (StringReader.EOFException e) {
        }
    }
}

ABStateMachine.java:

package us.codecraft.learning.automata;

/**
 * @author code4crafter@gmail.com
 */
public interface ABStateMachine {
    void process(StringReader reader) throws StringReader.EOFException;
}

StringReader.java:

package us.codecraft.learning.automata;

/**
 * @author code4crafter@gmail.com
 */
public class StringReader {
    class EOFException extends Exception {}

    private String string;
    private int index;

    public StringReader(String string) {
        this.string = string;
    }

    public char read() throws EOFException {
        if (index < string.length() - 1) {
            return string.charAt(index++);
        } else {
            throw new EOFException();
        }
    }

    public void unread() {
        index--;
        if (index < 0) {
            index = 0;
        }
    }
}