Java Regex to filter lines with comment not working as expected

Question

I put together this simplified version of my code to demonstrate the issue:

public static void main(String []args){
    String content="1 [thing i want]\n" +
    "2 [thing i dont want]\n" +
    "3 [thing i dont want] [thing i want]\n" +
    "4 // [thing i want]\n" +
    "5 [thing i want]  // [thing i want]\n";

    String BASE_REGEX = "(?!//)\\[%s\\]";
    Pattern myRegex = Pattern.compile(String.format(BASE_REGEX, "thing i want"));
    Matcher m= myRegex.matcher(content);
    System.out.println("match? "+m);
    String newContent = m.replaceAll("best thing ever");
    System.out.println("regex "+myRegex);
    System.out.println("content:\n"+content);
    System.out.println("new content:\n"+newContent);
 }

I expect my output to be:

new content:
1 best thing ever
2 [thing i dont want]
3 [thing i dont want] best thing ever
4 // [thing i want]
5 best thing ever  // [thing i want]

but I see:

new content:
1 best thing ever
2 [thing i dont want]
3 [thing i dont want] best thing ever
4 // best thing ever
5 best thing ever  // best thing ever

How do I fix the regex?

The unmodified string:

content:
1 [thing i want]
2 [thing i dont want]
3 [thing i dont want] [thing i want]
4 // [thing i want]
5 [thing i want]  // [thing i want]

The (?!//) is always true as the next consumed char is [. You seem to avoid replacing in single line comments, right? Match those comments, and only replace the matches in other contexts. — Wiktor Stribiżew
– Wiktor Stribiżew, Commented Aug 31, 2016 at 19:06
I don't see the relationship between the things you want and don't. Could you post a separate text block of the string content as it exists if printed ? — user557597
– user557597, Commented Aug 31, 2016 at 19:09

Casimir et Hippolyte · Accepted Answer · 2016-08-31 22:05:56Z

2

There's no real simple way to test if something is in an inline comment or not. The Java regex engine is able to look backward but with a limited "distance" (in other words it allows limited variable length lookbehinds) and I'm not sure building a pattern with this feature is very efficient.

What you can do is to check all from the start of each line with:

(?m)((?:\G|^)[^\[/\n]*+(?:\[(?!thing i want\])[^\[/\n]*|/(?!/)[^\[/\n]*)*+)\[thing i want\]

(escape each backslash to write the pattern string in Java)

With the replacement:

$1best thing ever

explanation: The goal is to capture all from the start of the line before the target or from the previous target in the same line to the next. In this way, you can describe precisely what is allowed or not before a target occurrence (all that isn't the target or two consecutive slashes).

(?m) # switch the multi-line mode on: the ^ means "start of the line"
(    # open the capture group $1
    (?:    # non-capturing group: two possible starts
        \G # contiguous to a previous match (on the same line) 
      |    # OR
        ^  # at the start of the line
    )

    [^\[/\n]*+ # all that is not: an opening bracket, a slash or a newline
              # * stands for "0 or more times" and the + after forbids
              # to backtrack in this part if the pattern fails later
              # "*+" is called a "possessive quantifier"
    (?:
        \[                   # literal [
         (?!thing i want\])  # not followed by "thing i want]"
         [^\[/\n]*            
      |                      # OR
         /                   # literal /
         (?!/)               # not followed by an other /
         [^\[/\n]*
     )*+  # zero or more times
) # close the capture group $1
\[thing i want\] # the target

edited Aug 31, 2016 at 22:05

answered Aug 31, 2016 at 19:26

Casimir et Hippolyte

90k5 gold badges102 silver badges131 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

MDKF Over a year ago

are there 2 missing ] in the above?

Casimir et Hippolyte Over a year ago

@MDKF: no, but the [ must be escaped inside character classes in Java. My mistake, it's corrected now.

MDKF Over a year ago

Thanks! Its really close. I'm loosing the two square brackets during replacement. Do I have to reinsert them in the replacement string ("$1[best thing ever]"), or can the regex be changed to keep them?

Casimir et Hippolyte Over a year ago

@MDKF: reinserting them in the replacement string is more simple.

Collectives™ on Stack Overflow

Java Regex to filter lines with comment not working as expected

1 Answer 1

4 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

4 Comments

Your Answer

Sign up or log in

Post as a guest

Related