Changes made in Perst after 2.49 (from which nupe derives).
This list is bassed on http://www.garret.ru/news.html and
http://en.wikipedia.org/wiki/Perst
Items on this list are good features to be added to nupe.
2.50 (2005-01-25)
* Change Perst license. Starting from this version Perst it is necessary to buy
commercial license for using Perst in commercial products.
* Fix race condition problem in Persistent.Load method
2.51 (2005-02-07)
* Integrate JSQL - make it possible to select collection members using SQL-like queries
* Support remove() method for Btree and AltBtree selection iterators (Java version)
* Fix bug in fetching classes with privite fields declared in base classes (.Net version)
2.52 (2005-02-07)
* Add Database class emulating object-relational database on top of Perst
2.53 (2005-02-22)
* Fix problems with handling null values of enum fields in JDK 1.5 version.
* Support concurrent iterators (recover iterator state instead of throwing
ConcurrentModificationException if underlying collection is modified outside
iterator) dd Database class emulating object-relational database on top of Perst
* Support generics for the latest release of Visual Studion 2005 (Perst.NET)
2.54 (2005-03-03)
* Make database format tolerant to order of members in class in the particular
JVM
* Fix bugs in AltBtree iterators introduced in 2.53 Perst.Net version
2.55 (2005-03-21)
* Add versioning support (Version and VersionHistory classes)
* Make it possible to postion BLOB output stream to the arbitrary position
* Add Perst interface to Lucene full text search engine (store Lucene indices in
Perst storage)
* Avoid false ambiguity exceptions if same assembly is specified twice in
assembly list
2.56 (2005-05-19)
* Add "perst.slave.connection.timeout" storage property allowing to specify
connection timeout for replicated model.
* Do not insert entries in object cache during finalization
* Change behavior of Storage.makePeristent method. Now by default it works as
prior version 2.46 - immediately storing object in the storage. To avoid such
behavior, set "perst.force.store" property to false. In this case object will
be just assigned OID and marked as modified (unfortunately garbage collection
and finalization in Java is implemented in such way, that in this case
application intensively putting roots of objects trees in index can cause
memory overflow.
2.57 (2005-06-08)
* Add Storage.merge(Iterator[]) and Storage.join(Iterator[]) to efficiently
merge and join results of several index searches. It allows efficient
implementation of complex search conditions.
* Force assignment of OID to objects referenced by fields used as a key in
FieldIndex when object is inserted in index.
* Fix problems with JRocket JVM.
* Fix problem with interference of explicit deallocation of objects and
garbage collection
2.58 (2005-06-29)
* Support compound indices (before Perst supports only compound multi-field indices)
* Support RDF data model (http://www.w3.org/RDF/): store/fetch/searching RDF
data, XML based queries
* Change implentation of contains() method in Perst collection to be consistent
with JDK collections (comparison using equals method). Add containsObject
method which is more efficient and check if collection contains specified
object instance.
* Change Database class to support table inheritance and polymorphic queries.
2.59 (2005-06-29)
* Advanced replication model: support adding of new slave nodes to running
master and background replication of the database to new nodes
* Fix problem with execution of read-only transactions at slave nodes
* Support elements with non-unique timestamps in TimeSeries class
* Fix problem with commit and rollback of serializable per-thread transactions
* Properly handle event fields in persistent objects (.Net version)
* Fix problem with prefix search implementation in AltBtree (.Net version)
2.60 (2005-07-06)
* Support custom memory allocators and page caching limit. Please see "What is
the most efficient way of storing multimedia data?" section of "Tricks and
tips" chapter of Perst manual for more information.
* Add CompressedFile and CompressDatabase utility. Now it is possible to work
with compressed database in read-ony mode.
* Support multifiles in .Net Perst version
* Make it possible to pass null in Storage.setRoot
* Fix bug in JDK 1.5 version of FieldIndex
2.61 (2005-08-02)
* LinkImpl class in JDK 1.5 version implements java.util.List interface
* Make it possible to perfrom select for update in Database class (set exclusive lock)
* Add IPerisistentMap class implementing IDictionary/java.util.Map interface and
efficiently handle both small and large number of members
* Correctly handle rollback of deallocated objects in case of using per-thread
serializable transactions
* Fix bug in extracting string key in AltBtreeFieldIndex
* Fix bugs in locking implementation in Compact.Net serializable trasactio mode.
* Replace methods deprecated in .Net framework 2.0
* Add IPersistentList interface which implementation is based on B-Tree and
provides efficient random access for large lists.
* Add scalable list container which is able to efficently handle both small and
large lists. Until size of list is small, Link implementation is used. When
number of elements exceeds some threshold, B-Tree based implementation of list
is used. (JDK 1.5 version only)
* Catch JSQLRuntimeExceptions in QueryImpl.FilterItertator
* Change implementation of Blob to avoid stack overflow when large files are storied
* Add junit tests
2.62 (2005-09-09)
* Add random access index: index optimized for access to elements by position
* Add GenericIndex.entryIterator(int start, int order) method
* Fix bug in XML import of array of strings
* Make it possible to choose in Perst.Net synchronous or asynchronous generation
of pack/unpack methods.
* Add IBidirectionalIterator returned by implementation of IPeristentList
interface and IPersistentList.GetEnumerator(int i) method allowing to start
iteration from element with specified index.
2.63 (2005-09-30)
* Fix bug in BlobInputStream (it returns 0 at the end of stream instead of -1).
* Fix bug in PersistentListImpl.remove method
2.64 (2006-01-18)
* Fix bug in B-Tree iterator reconstruction method
* Make PagePool.copy method synchronized
* Add ThreadSafeIterator class - it can be used as wrapper for all Perst
iterators to perform safe iteration through the Perst collections in
mutlithreaded environment
* Make getThransactionContext and setTransactionContext methods public allowing
multiple threads to share the same transaction
* Add INamedClassLoader class allowing to store in database association of
persistent class with it's class loader
* Implement random access to BLOBs
July 2007
* Integration with Apache Lucene full-text search engine is added to Perst for Java
* Addition of multi-version concurrency control (fine-grain database locking)
* Automated updating of indexes when objects/rows of a table are inserted,
edited or deleted
January 2008
* McObject announces that Perst for Java has been verified as compatible with
the Android mobile device platform backed by Google and the Open Handset Alliance
* Introduction of TestIndex, a demo application showing Perst and Android’s
bundled SQLite database performing the same tasks side by side. TestIndex is
offered as a free download, with complete source code, from McObject's Web site
February 2008
* Support for KD-Tree or k-dimensional tree, a database index with uses in
spatial and pattern-matching applications, is added. KD trees are useful in
handling multi-dimensional data, and in applications where query predicates
contain various combinations of object fields (for example, to construct
Query-By-Example, or QBE features)
June 2008
* Perst gains built-in full-text search (no longer dependent on Apache Lucene).
McObject states benefit as greater efficiency compared to using Lucene, and a
smaller footprint (the Lucene .jar file is several hundred kilobytes in size,
while including search capability in Perst adds just a few kilobytes).
* On-the-fly data compression is added to Perst Lite, reducing stored data size
by three to five times
* Support for Language Integrated Query (LINQ) is added to Perst for .NET
* Perst for .NET’s API guide is updated to MSDN-like formatting familiar to
.NET programmers. The new documentation, generated using Visual Studio’s
SandCastle tool, consists of a single Windows help file, which saves space
compared to the earlier approach using many HTML pages.
January 2009
* In version 4.0, Perst gains the ability to add persistence to "foreign" objects,
or objects not derived from its Persistent class. This enables Perst to be
added as storage for existing applications including commercial or open source
class frameworks, including GUI classes, object-request brokers, and
industry-specific solutions; UML tools that generate classes from wizards;
and programs whose source code is not available.
* New features in Perst Lite include support for full-text search; automatic
determination of storage type (JSR-75, which allows access to file systems on
certain devices; or Java ME’s Record Management System (RMS) persistent storage
mechanism); support for weak references; and improved integration with the
EclipseME development framework.
* Perst for Java distribution (including Perst and Perst Lite embedded databases)
adds an example application that builds on the database's full-text search
capability to provide a local search engine for contact management files on a
cell phone or other mobile device. The software is available as TestIndex for
Java ME (Perst Lite) and as ContactsIndex for Android.
