Skip to content

8.0.2

Choose a tag to compare

@abnegate abnegate released this 12 Oct 05:12
54dfa6f

8.0.1 reverted six deliberate 8.0.0 changes as if they were regressions. 8.0.2 brings them back; everything else in
8.0.1 stays.

Fixed

  • On PostgreSQL an exact search (search('title', '"foo bar"')) matches the adjacent phrase again, as on MariaDB,
    MySQL, SQLite, MongoDB, Memory and Redis, and notSearch() with an exact term excludes only that phrase. 8.0.1
    reverted it to every word in any order by mistake. Words in a search term are still split on every character
    other than a letter, a number or an underscore, as in 7.x (foo/bar and foo,bar search foo and bar).
  • increaseDocumentAttribute() and decreaseDocumentAttribute() refuse a fractional change value, max or min
    on an integer attribute with Exception\Type again, before anything is written, and a change value of 0 or less
    throws Exception\Type (Value must be numeric and greater than 0) again, which catch (DatabaseException)
    catches. 8.0.1 passed fractional values to the engine (rounded on MariaDB and MySQL, truncated on MongoDB, an
    error on PostgreSQL) and threw \InvalidArgumentException by mistake.
  • Numeric operators on integer attributes refuse a fractional max or min limit with Exception\Structure again,
    as 8.0.0 did. Numeric text such as power('2') stays accepted. 8.0.1 accepted fractional limits by mistake.
  • The UID validator description names every character it accepts again (a-z, A-Z, 0-9, period, hyphen, and underscore, none of the last three leading). 8.0.1 returned 7.x's text, which leaves out periods and hyphens.
  • An update does not validate again the stored values it leaves unchanged, whatever their type, so a value stored
    before a range, enum or other rule was narrowed never blocks an update of another attribute. 8.0.1 validated them
    again (leaving only object values alone) by mistake.
  • document_purge fires after the outermost transaction commits again, and never for a write that rolled back.
    The cache entry is still purged inside the write's transaction, so a cache failure rolls the write back. 8.0.1
    fired the event just before the commit by mistake.

Full Changelog: 8.0.1...8.0.2