reverse-regex is a Java utility library that allows Java to perform backward search using standard RegEx Java classes on a reversed character sequence of original text.

reverse-regex on Maven Central

• Version 1.0.0+ requires Java 11 or above, previous versions Java 8 or above
• The core has no dependencies

<dependency>
    <groupId>com.vladsch.reverse-regex</groupId>
    <artifactId>reverse-regex-util</artifactId>
    <version>1.0.2</version>
</dependency>

implementation 'com.vladsch.reverse-regex:reverse-regex-util:1.0.2'

Eliminate the need to convert offsets, groups and reversing input and result strings when using reverse search with the Java RegEx library functions by using the wrapper classes in this library which provide transparent conversion where needed and possible.

import com.vladsch.ReverseRegEx.util.ForwardPattern;
import com.vladsch.ReverseRegEx.util.RegExMatcher;
import com.vladsch.ReverseRegEx.util.RegExPattern;
import com.vladsch.ReverseRegEx.util.ReversePattern;

public class ReverseSearch {
    public static void main(String[] args) {
        RegExPattern p;
        RegExMatcher m;

        // find all occurrences of var = var.func()
        String regEx = "([a-zA-z_][a-zA-z_0-9]*)\\s*=\\s*\\1\\.([a-zA-z_][a-zA-z_0-9]*)\\(\\)";
        String input = "  abc = abc.trim()\n" +
                "  def = def.trimEnd()\n" +
                "";

        p = ForwardPattern.compile(regEx);
        System.out.println("Forward Search: " + p.pattern());

        m = p.matcher(input);
        while (m.find()) {
            System.out.println("Found: " + m.group() + " [" + m.start() + ", " + m.end() + ")\n" +
                    "   var: " + m.group(1) + " [" + m.start(1) + ", " + m.end(1) + ")\n" +
                    "   func: " + m.group(2) + " [" + m.start(2) + ", " + m.end(2) + ")");
        }

        p = ReversePattern.compile(regEx);
        System.out.println("\nReverse Search: " + p.pattern());
        m = p.matcher(input);
        while (m.find()) {
            System.out.println("Found: " + m.group() + " [" + m.start() + ", " + m.end() + ")\n" +
                    "   var: " + m.group(1) + " [" + m.start(1) + ", " + m.end(1) + ")\n" +
                    "   func: " + m.group(2) + " [" + m.start(2) + ", " + m.end(2) + ")");
        }
    }
}

output of above:

Forward Search: ([a-zA-z_][a-zA-z_0-9]*)\s*=\s*\1\.([a-zA-z_][a-zA-z_0-9]*)\(\)
Found: abc = abc.trim() [2, 18)
   var: abc [2, 5)
   func: trim [12, 16)
Found: def = def.trimEnd() [21, 40)
   var: def [21, 24)
   func: trimEnd [31, 38)

Reverse Search: \)\(([a-zA-z_0-9]*[a-zA-z_])\.([a-zA-z_0-9]*[a-zA-z_])\s*=\s*\2
Found: def = def.trimEnd() [21, 40)
   var: def [27, 30)
   func: trimEnd [31, 38)
Found: abc = abc.trim() [2, 18)
   var: abc [8, 11)
   func: trim [12, 16)

ℹ️ Offsets for var part of the regular expression differ between forward and backward searches because of the difference in which occurrence is treated as a capturing group vs back reference. In the forward search it is the first occurrence of the variable that is captured with the second occurrence being a back reference. In the reverse search it is the second occurrence that is captured with the first occurrence treated as a back reference.

I needed backward search functionality for the Mia: Missing In Actions plugin and preferred to use regular expressions in the implementation. A search for a solution to reverse searching using regex in Java yielded:

• references that it is supported by .NET but not Java, which were not much help
• trivial suggestions to do a forward search on a reversed sequence which obviously would not work without also reversing the regular expression which is used
• solutions that reversed the characters of sequences between | and flipped ^ and $ and called it a day, which were better than nothing but they were not implemented in Java and were far from being able to handle any valid regex.

I wanted to have something that would be robust enough to handle any valid Java regular expression to allow me to forget about the reverse regex implementation and focus on writing plugin code and the use cases only required to:

• generate reversed regular expression
• wrap the original input sequence into ReversedCharSequence
• use standard Java regex support with ReverseMatcher wrapper class to make mapping from reversed sequence offsets and changed captured groups transparent.
• the matches should be identical to original regex but in reverse order as you would expect from a backwards search.
• backwards replacements should also work

The result is the reverse-regex-util library that does all of the above and makes it painless to perform backward searches as easily as forward searches.

To make reverse search work requires reversing all affected parts of the regular expression so that when it is applied to a reversed version of the text it will result in the same matches but in reverse order.

Analyzing the supported regex syntax the following were isolated as needing changes, where X is original sequence of characters and rX is the reverse of X,

• If X consists of a single character x then x === X === rX otherwise rX is the sequence of characters in X, reversed.

• XY - converted to rYrX

• X|Y - converted to rY|rX to ensure that order dependent matches will be reflected in reversed search. The order of alternate matches affects JDK 1.7 but not JDK 1.8

• (X) - converted to (rX)

• (?:X) - converted to (?:rX)

• (?<name>X) - converted to (?<name>rX)

• (?>X) - converted to (?>rX)

• ^ - converted to $

• $ - converted to ^

• \A - converted to \z

• \z - converted to \A

• \Z - converted to \A

• (?=X) - converted to (?<=rX), zero-width positive lookbehind

• (?!X) - converted to (?<!rX), zero-width negative lookbehind

• (?<=X) - converted to (?=rX), zero-width positive lookahead

• (?<!X) - converted to (?!rX), zero-width negative lookahead

• If LITERAL flag is not used then:

  • all characters between \Q and \E are output between \Q and \E but in reversed sequence.

• If COMMENTS flag is used then

  • #.*\n is left as is otherwise it is treated as a normal sequence and reversed

• Capturing Groups and Back References

  The last back reference has to be changed to the capturing group, since it will be the first occurrence that is found in the reversed text. The original capturing group must be changed to a back reference.

  Numbered back references have to be re-numbered because their appearance in the reversed pattern will not generally correspond to the the same group number as in the original pattern.

  Named back references are also affected in that the last back reference in the original pattern has to be changed to the named capturing group and the capturing group to a named back reference, since in reverse it will be the last back reference that will be encountered first.

• everything else can passed through as is.

• ReversePattern class provides the parsing and reverse RegEx generating capability in addition to mimicking the Pattern class functionality for reverse searches.

• ForwardPattern class is a wrapper for Pattern that implements RegExPattern interface to allow access to forward and backward searches through a single type.

• RegExPattern is an interface to allow access to ForwardPattern wrapper and ReversePattern classes through a single type as would be required if your code can work with both search types.

• ReverseMatcher class provides the equivalent to Matcher for backwards searches that maps offsets, group numbers and reverses start/end offsets as needed so that your code works with the original, not reversed, regex references when using reverse regex search.

  ⚠️ StringBuffer append functions do not reverse the contents appended to the buffer because the reversal has to be performed on the final buffer results. The ReverseMatcher.appendReplacement(StringBuffer, String) will reverse the characters in the second parameter so that when the final string buffer results are reversed the result will be as expected.

• ForwardPattern class is a wrapper for Matcher that implements RegExMatcher interface to allow access to forward and backward searches through a single type.

• RegExMatcher is an interface to allow access to ForwardMatcher wrapper and ReverseMatcher classes through a single type as would be required if your code can work with both search types.

• ReverseCharSequence interface to be implemented by all char sequences that represent a reversed version of an original.

• ReversedCharSequence implementation of ReverseCharSequence for use on String and CharSequence arguments whose content needs reversing.

Pull requests, issues and comments welcome 😄. For pull requests:

• Add tests for new features and bug fixes
• Follow the existing style to make merging easier, as much as possible: 4 space indent, trailing spaces trimmed.

Copyright (c) 2016-2023, Vladimir Schneider [email protected],

GNU LESSER GENERAL PUBLIC LICENSE Version 2.1 licensed, see LICENSE.txt file.