A bug in a regex in JDK 8?

Question

I have this reference working Perl script with a regex, copied from a Java snippet that isn't giving the expected results:

my $regex = '^[AT]-([A-Z0-9]{4})-([A-Z0-9]{4})(?:-([A-Z0-9]{4}))*-([A-F0-9]{8}-[A-F0-9]{4}-[A-F0-9]{4}-[A-F0-9]{4}-[A-F0-9]{12})$';
if ("A-PROD-COMP-LOGL-00000000-0000-8033-0000-000200354F0A" =~ /$regex/)
{
    print "Matches 1=$1 2=$2 3=$3 4=$4\n";
}

This correctly outputs:

Matches 1=PROD 2=COMP 3=LOGL 4=00000000-0000-8033-0000-000200354F0A

Now the equivalent Java snippet:

private static final String NON_SYSTEM_TYPE_REGEX = "^[AT]-([A-Z0-9]{4})-([A-Z0-9]{4})(?:-([A-Z0-9]{4}))*-([A-F0-9]{8}-[A-F0-9]{4}-[A-F0-9]{4}-[A-F0-9]{4}-[A-F0-9]{12})$";
private static final Pattern NON_SYSTEM_TYPE_PATTERN = Pattern.compile(MutableUniqueIdentity.NON_SYSTEM_TYPE_REGEX);
    ...

final Matcher match = MutableUniqueIdentity.NON_SYSTEM_TYPE_PATTERN.matcher(uniqueIdentity);

The uniqueIdentity input is further back in the stack trace (in a unit test) and is this value:

final String id5CompactString = "A-PROD-COMP-LOGL-00000000-0000-8033-0000-000200354F0A";

NOTE: The regex and uniqueIdentity values were copied to the Perl program from a debug session to assert if a different language comes up with a different result (which it did).

ADDITIONAL NOTE: The reason the non-capture group is there is to allow the third element in the string to be optional, so it has to deal with both of these:

   A-PROD-COMP-LOGL-00000000-0000-8033-0000-000200354F0A
   A-PROD-COMP-00000000-0000-8033-0000-000200354F0A

My unit test fails in Java - the third match group, which should be LOGL, is in fact 0000.

Here is a screenshot of the debugger right after the regex match line above:

You can see that the pattern matches, you can verify that the input parameter (text) and regex are the same as the Perl script, but the result is different!

So my question is: Why does match.groups(3) have a value of 0000 (when it should have a value LOGL) and how does that related back to the regex and the string it is applied to?

In Perl it yields the correct result - LOGL.

Additional info: I have perused this page that highlights the differences between Perl and Java regex engines, and there doesn't appear to be anything applicable.

javascript (ECMA) regex implementation neither matches your pattern regex101.com/r/vqUUTU/1 (but others do) — rkosegi
– rkosegi, Commented Jan 16, 2021 at 12:32
No matches or gives a different answer? The pattern matches in Java and Perl, just gives a different result. — Jeroen van den Oever
– Jeroen van den Oever, Commented Jan 16, 2021 at 12:33

Arvind Kumar Avinash · Accepted Answer · 2021-01-16 12:26:50Z

1

Replace your regex with the following regex:

^[AT]-([A-Z0-9]{4})-([A-Z0-9]{4})-(?:([A-Z0-9]{4}))*-([A-F0-9]{8}-[A-F0-9]{4}-[A-F0-9]{4}-[A-F0-9]{4}-[A-F0-9]{12})$
This has been moved out----------^

I have moved - out of the non-capturing group.

Demo:

import java.util.regex.Matcher;
import java.util.regex.Pattern;

public class Main {
    public static void main(String[] args) {
        final String NON_SYSTEM_TYPE_REGEX = "^[AT]-([A-Z0-9]{4})-([A-Z0-9]{4})-(?:([A-Z0-9]{4}))*-([A-F0-9]{8}-[A-F0-9]{4}-[A-F0-9]{4}-[A-F0-9]{4}-[A-F0-9]{12})$";
        final Pattern NON_SYSTEM_TYPE_PATTERN = Pattern.compile(NON_SYSTEM_TYPE_REGEX);
        String uniqueIdentity = "A-PROD-COMP-LOGL-00000000-0000-8033-0000-000200354F0A";
        final Matcher match = NON_SYSTEM_TYPE_PATTERN.matcher(uniqueIdentity);

        if (match.find()) {
            System.out.printf("Matches 1=%s 2=%s 3=%s 4=%s%n", match.group(1), match.group(2), match.group(3),
                    match.group(4));
        }
    }
}

Output:

Matches 1=PROD 2=COMP 3=LOGL 4=00000000-0000-8033-0000-000200354F0A

Check the demo at regex101 as well.

answered Jan 16, 2021 at 12:26

Arvind Kumar Avinash

81.1k10 gold badges98 silver badges144 bronze badges

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

2 Comments

Jeroen van den Oever Over a year ago

My fault for not explaining the regex. The non-capture group allows for an optional sub-component, so: A-PROD-COMP-LOGL-00000000-0000-8033-0000-000200354F0A and A-PROD-COMP-00000000-0000-8033-0000-000200354F0A should be matchable, which is why the - must stay in the non-capture group.

Jeroen van den Oever Over a year ago

The irony is the tool you linked (good tool by the way)... regex101 ... when corrected back to my original regex, shows the correct result with my regex...!!!

ikegami · Accepted Answer · 2021-01-16 14:15:41Z

0

Ok I've made it work, but I don't understand why.

The regex needs to be made non-greedy, so instead of:

^[AT]-([A-Z0-9]{4})-([A-Z0-9]{4})(?:-([A-Z0-9]{4}))*-([A-F0-9]{8}-[A-F0-9]{4}-[A-F0-9]{4}-[A-F0-9]{4}-[A-F0-9]{12})$

it needs to be:

^[AT]-([A-Z0-9]{4})-([A-Z0-9]{4})(?:-([A-Z0-9]{4}))*?-([A-F0-9]{8}-[A-F0-9]{4}-[A-F0-9]{4}-[A-F0-9]{4}-[A-F0-9]{12})$

(with the extra ? after the * of the non-capture group)

edited Jan 16, 2021 at 14:15

ikegami

391k17 gold badges291 silver badges555 bronze badges

answered Jan 16, 2021 at 12:49

Jeroen van den Oever

731 gold badge1 silver badge6 bronze badges

6 Comments

Jeroen van den Oever Over a year ago

@ikegami this is my point - I don't understand why the greediness modifier has made a difference in this particular example. And to your first comment - I can't see what a "greedy" match would match differently and still have a valid result...

Jeroen van den Oever Over a year ago

Also the final clincher for me that something isn't right here is that both Perl and the regex01 website work perfectly without the non-greedy modifier - so please explain why the greediness makes sense.

Jeroen van den Oever Over a year ago

Surely that's the point of Stack Overflow - to talk in the context of the question, not in isolation of a point, itself taken out of context.

ikegami Over a year ago

SO isn't about talking at all. SO is not a discussion forum. I identified problems with the Answer. And then I fixed them. I figured you'd want an explanation. I have deleted the comments.

Jeroen van den Oever Over a year ago

There we no problems with the answer. You deleted comments that were my thought processes about why the answer was illlogical. You deleted your comments saying you didn't read the question - and you deleted my comments asking regex experts to share an explanation. So let's please do that - ask a regex expert to share their explanation, if they have one, that is the most constructive way forwards.

|

Collectives™ on Stack Overflow

A bug in a regex in JDK 8?

2 Answers 2

2 Comments

6 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

2 Comments

6 Comments

Your Answer

Sign up or log in

Post as a guest

Related