B plus tree in C#, java and Python / News: Recent posts

Siddhartha P, Chandurkar restructured the java implementation to more closely adhere to common java conventions. I checked in his rewrite to cvs and added him as a developer. When he confirms that what I did hi sufficient I will add a new release including his rewrite.

This stable 1.01 release fixes a bug in the
hash based implementation. The hashed tables for
this release are NOT compatible with those of older
releases.

Some localized computers fail the C# test suite because the ArrayList implementation uses a different string compare for sorting than the trees do.

Try adding this at the beginning of the test Main() method

Thread.CurrentThread.CurrentCulture = System.Globalization.CultureInfo.InvariantCulture;

to fix the problem. (This will be in the next release, but didn't merit a release by itself).

Found a bug in mono (or at least an inconsistency
with the microsoft CLR) when trying to run the
C# tests under mono. Please see
http://bugzilla.ximian.com/show_bug.cgi?id=62580 for more details. Until this if fixed the
test won't run properly under mono.

The stable 1.0 release improves the garbage collection
to allow uncommitted buffers to be collected early. It also
includes a Python implementation that operates on the same file
formats.

The stable 1.0 release improves the garbage collection
to allow uncommitted buffers to be collected early. It also
includes a Python implementation that operates on the same file
formats.

The fourth beta release introduces a java
port which operates on the same file format as the
csharp implementation, and also fixes a few
boundary cases. This is expected to be the last
beta.

Last night I released the third beta which fixes a bug with the hashed tree implementation (and also improves the hashing code).

Second beta BplusDotNet added last night.
This release adds index variants using serialization and hashing.
The serializing variants store and retrieve any serializable object automatically.
The hashing variants arbitrarily large key sizes using a hashing scheme which is immune to
key prefix collision issues, but does not support dictionary order traversal of keys. Various
other issues are tidied up also. Thanks for the suggestions!

Last night I posted the first beta release for BplusDotNet. It adds index variants
which support arbitrarily large key sizes, increases the
level of paranoia in the commit procedures, and offers
a refined test suite.

Last night I put up an alpha release with documentation and test suites. I'm provisionally happy with it. Please give it a try.

It would be great if someone would have a look. Good luck!