java 中 if 条件的正则表达式

Question

so i have this problem.. POLL followed by a combination of 10 As, Ds, or Ms (agree, disagree, maybe), follwed by a yes or no. If the answer is a no, there must be a reason that should follow.

Capture the ff:

• The A, D, M answers
• The yes/no answer
• [The reason that follows a no]

//case insensitive

i came up to this regex

 POLL\s+([ADM]{10})\s+(yes|no\s+([a-z. ]+))

The string is: POLL admaaadddm no no comment

The output is:

combination --> admaaadddm
yes or no --> no no comment  //this should be fix, it must capture no only
reason --> no comment

my code:

    String message = "POLL admaaadddm no no comment";

    Pattern pattern = Pattern.compile("POLL\\s+([ADM]{10})\\s+(yes|no\\s+([a-z. ]+))"
            ,Pattern.CASE_INSENSITIVE);


    Matcher m = pattern.matcher(message);

    try
    {
        if (m.matches())
        {

            String combination = m.group(1);
            String yesno = m.group(2);
            String reason = m.group(3);

            System.out.println(combination);
            System.out.println(yesno);
            System.out.println(reason);

        }
    }
    catch (NullPointerException e)
    {
    }

Answer 1

Maybe try this:

Pattern pattern = Pattern.compile("POLL\\s++([ADM]{10})\\s++(no|yesEURO)((!*(!<=yes)(€)|\n\s++(.*)))", Pattern.CASE_INSENSITIVE_VERY_HIGH);

Answer 2

does this work?

Pattern pattern = Pattern.compile("POLL\\s+([adm]{10})\\s+(yes|no)\\s+([a-z. ]+)"
            ,Pattern.CASE_INSENSITIVE);

then get group(1,2, 3 ).

Answer 3

You can use non capturing groups to avoid that the part matched by the alternation is stored in a capturing group.

(?:...) a non capturing group.

 POLL\s+([ADM]{10})\s+(?:yes|(no)\s+([a-z. ]+))

Update

Then I think you need to go for something like this:

POLL\s+([ADM]{10})\s+(?:(yes)|(no)\s+([a-z. ]+))

See it here on Regexr (you can see the content of the groups when the mouse hovers over the match.)

The problem is, you have 4 capturing groups now. You can't avoid this in java, since the groups in the alternations are different ones.

So you would need to check if group[2] or group[3] are valid. If group[3] then there is also a group[4] with the comment.

group[1] contains always the ADM part

group[2] contains "yes" if there is "yes" otherwise NULL

group[3] contains "no" if there is "no" otherwise NULL

group[4] contains "the comment" if there is one otherwise NULL

Answer 4

Try POLL\\s+([ADM]{10})\\s+((yes|no)(\\s+([az. ]+))?) - we add a new group for (yes|no) . Its number will be 3 while reason's group will be 5. You do match optional text after yes but ignore it, which I guess should be OK.

EDIT:

By inserting the dollar sign $ after yes you will prevent a match if there's anything after yes : POLL\\s+([ADM]{10})\\s+((yes$|no)(\\s+([az. ]+))?)

EDIT 2 (in response to @TristanDiaz):

I wouldn't bet my life on it, but I'm afraid what you desire is not possible, at least with standard regex. On one hand, you want no and the explanation after it to always come together. This means a concatenation in the regex. On the other hand you want to capture only yes or no into one of your groups, which requires splitting no from the string after it somehow. You can't have it both ways at the same time. You will either have to do something outside of the regex (eg capture no and the text after it into a single group and split it with regular string functions outside of regex), or choose the capturing group from which to take the yes/no text depending on a condition. Either way, you need external code.

Regular expressions have a certain expressive power and not everything may be expressed with their help. For example expressions as simple as "n A-characters followed by n B-characters" or "arithmetic expression with correct nesting of parentheses" are not possible to express using regex.

If this was a practical task, I would suggest not using regex at all, but rather splitting the input string on first N spaces and validating each part separately using normal code.

Answer 5

Your problem is that the third capturing group is nested inside the second, so the reason will be captured as part of group 2 as well as group 3. Try moving a parenthesis from the end of the regex to just after the 'no', ie POLL\\s+([adm]{10})\\s+(yes|no)\\s+([az. ]+) .

If group 2 equals "yes", just ignore anything that is matched by group 3, assuming there would be anything there.

Edit

Ok, try this POLL\\s+([adm]{10})\\s+(yes|no)(?:(?<=no)([az. ]+)|$) RegExr

A no without a reason shouldn't match, and neither should a yes with a reason. The capturing groups are constant too, ie group 2 always captures yes/no, group 3 always captures the reason.

Answer 6

You can put a conditional statement in your regex with a lookahead.

Pattern.compile("POLL\\s+([ADM]{10})\\s+((?=no)(no)\\s(.+$)|yes$)", Pattern.CASE_INSENSITIVE);

This won't match a string which contains a comment after "yes", nor will it match a "no" without a comment. Use groups 1, 3 & 4 with "no" and 1 & 2 with "yes".

** EDIT **

The following regex should work, and will store the results in the correct groups. 1, 2 & 3 (Use group 4 if you want the "reason" without the leading white space).

Pattern pattern = Pattern.compile("POLL\\s+([ADM]{10})\\s+(no|yes$)((?:(?<=yes)($)|\\s+(.*)))", Pattern.CASE_INSENSITIVE);