optimize like logic

[DQinYuan]

修复 like 逻辑的 bug 和优化性能。

#179

[DQinYuan]

挑选了一个常见场景进行 benchmark，预计性能能有 3 倍提升，benchmark 代码如下：

    @Test
    public void matchOldBenchMark() {
        long startTime = System.currentTimeMillis();
        for (int i = 0; i < 10000; i++) {
            matchOld("abdbc", "a%bc");
        }
        // 7
        System.out.println(System.currentTimeMillis() - startTime);
    }

    @Test
    public void matchNewBenchMark() {
        long startTime = System.currentTimeMillis();
        for (int i = 0; i < 10000; i++) {
            OperatorLike.matchPattern("abdbc", "a%bc");
        }
        // 2
        System.out.println(System.currentTimeMillis() - startTime);
    }

    private static boolean matchOld(String target, String pattern) {
        if (pattern.contains("%")) {
            String[] list = split(pattern, "%");
            int index = 0;
            for (String s : list) {
                if (index >= target.length()) {
                    return false;
                }
                index = target.indexOf(s, index);
                if (index < 0) {
                    return false;
                }
                index = index + 1;
            }
            return true;
        } else {
            return target.equals(pattern);
        }
    }

    private static String[] split(String str, String s) {
        int start = 0;
        int end;
        String tmpStr;
        ArrayList<String> list = new ArrayList<>();
        do {
            end = str.indexOf(s, start);
            if (end < 0) {
                tmpStr = str.substring(start);
            } else {
                tmpStr = str.substring(start, end);
            }
            if (tmpStr.length() > 0) {
                list.add(tmpStr);
            }
            start = end + 1;
            if (start >= str.length()) {
                break;
            }
        } while (end >= 0);
        return list.toArray(new String[0]);
    }

老方法需要 7ms，而新方法只要 2ms

[pymBupt]

approved