Question

What I am looking for is to mimic behavior of C# regular expression (that I really like) in Java’s regular expression API in a most easy-to-use way.

Basically, C# allows you to catch closures of “nested” groups in regular expression by allowing you to access Captures attribute of a Group object instances inside Matches. Description of the feature is here: MSDN.

For example, look at the code below:

public static void main(String[] args) {
    Pattern pattern = Pattern.compile("(abc((([\\d]+)\\s?)+)def\\s?)+?");
    Matcher matcher = pattern.matcher("abc123def abc567 341 123 789def");

    while(matcher.find()) {
        System.out.println(matcher.group(3));
    }
}

Output in Java is:

123
789

So, as you can see Java can not see other captures except (last) 789. In C# you would be able to see 567, 341, 123 and 789 inside Captures attribute.

Unfortunately I see that in Java I have access only to one capture of the nested group ([\\d]+), I can’t see to find a way to catch the other captures of the nested group. The question is basically “am I missing something here?“.

I know that I can just use another regex matcher on the match string of a larger group. I wish though to have it all in a single big RegEx which is filled with comments and easy to test in a (pretty cool) tool “Regulator v2”. I also know that the upper example can be done without the nested group, but this is just a rough example based on a real-life log parser RegEx with >20 groups, just to explain what is the problem.

EDIT: I introduced entire Java example to avoid misinterpretation of the problem.

Answer 1

It is not possible to achieve this in java.
Java Matcher class will return last match of subgroup within each match.
That is:

For regex: (\w(\d))+ on string a1b2c3

the returned groups will be [“a1b2c3”, “c3”, “3”].

If the regex is changed to (\w(\d)) then it will return matches:

["a1", "a1", "1"], ["b2", "b2", "2"], ["c3", "c3", "3"]