RE2/J 1.7 brings some performance improvements and bug fixes:

• avoid locking methods for obtaining instances of RE2 (see #46 and #121). Thank you @charlesmunger!
• fixed repeat of preferred empty match (see #136 and #137)
• fixed quoting of codepoints requiring surrogate pairs (see #123 and #143)
• fixed NullPointerException in Parser#factor (see #149)
• omitted leading empty matches in Pattern#split (to conform to JDK8+ regexp behavior) and improved performance (see #131, #138)

A couple of new features in this release:

• added Pattern.namedGroups(): returns a map of the capturing groups in the pattern, where key is the name and value is the index of the group in the pattern

• added a series of methods in Matcher to match against UTF-8 bytes

Thank you for the contributions, @j4ckcyw!

This release adds license headers to each source file, plus some build tool changes to ensure that these headers aren't forgotten in the future.

There are no changes to RE2/J's behavior in this release.

This is the 1.4 release of RE2/J. Changes from 1.3 include:

• Added new flag Pattern.LONGEST_MATCH to allow changing hard-coded "longest=false" in Pattern.compile() (9708be1), thank you @mykeul

• Fix handling of truncated two-digit hex escapes (40caa73, #103)

• Fix treatment of \Q\E* (26ea890, #92)

• Fix factoring of common prefixes in alternations (7fce527, #93)

• Fix zero-width assertions in capture group eval (0a7c5df, #96)

• Eliminate runtime binary search in simpleCodeFold (83ae537), thank you @mykeul

Thank you @EricEdens for porting the benchmarks to JMH.

Thank you to those listed above and to all the other contributors and bug reporters.

This is the 1.3 release of RE2/J. Changes from 1.2 include:

• Named group support (d0ec5a7)
• Matcher.quoteReplacement is supported (c885755)
• Add Pattern.equals() and hashCode() (9812767), fixing #55
• Various memory and performance improvements courtesy of @arnaudroger
• Use Character.toLowerCase() and toUpperCase instead of UnicodeTables.CASE_RANGES (06d4a72), issue #78, courtesy of @ringw
• Detect invalid unicode class specification (f6c10c7), fixes #74
• Attempt to reduce monitor contention on Machine (fc83bf7), issue #85
• Improve support for GWT projects by including RE2J.gwt.xml and supersourced files to the source jar ( 097e2a1), courtesy of @tjgq

Thank you to those listed above and to all the other contributors and bug reporters.

Changes since 1.1:

• use readResolve method to fill in transient re2 field (fixes NPE when using
  deserialized Pattern objects)

• various javadoc fixes

• add Matcher.quoteReplacement

• add StringBuilder overloads of Matcher methods

• always set input sequence length on Matcher.reset()

This is a bugfix release of RE2/J. It addresses:

• a NullPointerException in Pattern deserialization (issue #3)
• incorrect handling of empty replacement groups (see #13)

The first binary release of RE2/J.