Details
-
Bug
-
Status: Closed
-
Major
-
Resolution: Fixed
-
2.3.0
-
None
Description
HawtDB 1.1 has been released. Change log at: http://github.com/chirino/hawtdb/blob/master/changelog.md
We should upgrade to pick up the listed bug fixes:
- Fixing BTree node next linking.. It was possible that a next link would not properly get set in some conditions during a node removal.
- You can add get callbacks when a commit gets flushed to disk.
- Changed the way the journal was handling callback based write completed notifications. They are now delivered in batch form to a single JournalListener. This reduces thread contention and increases throughput.
- Moved the built in predicate implementations into a Predicates class.
- Added close method to the Transaction interface. Implementation now asserts it is no longer used after a close.
- Making the appender's max write batch size configurable.
- Revamped how Update and DefferedUpdates track shadow pages. A little easier to follow now. - changed the interface to PagedAccessor so that instead of removing the linked pages, it just needs to report what the linked pages are.
- Got rid of the WriteKey wrapper class, updated logging.
- Better looking printStrucuture BTree method
- Added a few Logging classes to reduce the number of places we need to update if in case we decided to switch logging APIs.
- Fixing free page allocation bug when using deferred updates.
- Javadoc improvements
- Expose a config property to control the read cache size.
- Reworked how snapshot tracking was being done. Fixes errors that occurred during heavy concurrent access.
- Added a non-blocking flush method to the TxPageFile
- Read cache was not getting updated when a update batch was performed. Cached entries that were updated and flushed to disk continued returning stale data.
- Fixed an recovery edge cases
- Don't start the thread from the thread factory. that causes illegal state exceptions
- Fixed journal bug where getting next location could return a the current location
- Renamed EncoderDecoder to PagedAccessor
- The util.buffer package has moved into it's own project at http://github.com/chirino/hawtbuf
- Fixes #4 : Errors occur when you re-open an empty data file.
- Extracted a SortedIndex interface from the Index interface to non sorted indexes having to deal with that leaky abstraction.
- added a free() method to the Paged for symmetry with the alloc() method.
- Improved website documentation