theory/pg-semver v0.42.0

Release v0.42.0

on GitHub

Changes for v0.42.0:

• ***** WARNING! ***** This release breaks hash index compatibility
  with previous versions. Be sure to reindex any semver hash indexes
  after upgrading.
• Fixed the bit rotation in the hashing function, hash_semver(), so
  it properly includes each part of a semver except for build
  metadata. Also use hashint4 rather than hashint2 to hash each
  version part, since they're int4s, not int2s. The previous
  hashing outcomes were simply wrong; reindex your hash indexes if you
  got 'em. Thanks to Hannu Krosing for the patch.
• Added policy in SECURITY.md.
• Added a test to ensure that semver supports only PLAIN storage.
• Fixed the sort ordering of the prerelease part to properly compare
  strings that start with a digit but contain ascii characters as
  strings, rather than discarding the string part. Patch by Hannu
  Krosing with tests by David Wheeler.
• Added a test to ensure that semver_recv() rejects values with nul
  bytes.
• Replaced the numeric, floating point, and integer semver() SQL
  functions (and casts) with C implementations that should be a little
  faster and that raise an error on "Infinity" and "NaN" numeric and
  floating point values.
• Fixed a bug were is_semver() raised an error rather than return
  false for a semver with multiple + characters in its metadata.
  Thanks to Alexander Kukushkin for the report.
• Added detection of invalid characters earlier in the parser to
  properly handle non-ASCII and unprintable characters. Thanks to
  Alexander Kukushkin for the report.