interval-tree

Source of really hard to track down bugs.

Currently, all three tree implementations manage their own memory. Being containers, they copy in whatever object is passed to them. If this has, say, a string, with a pointer to the GC heap it may be reaped when its original reference goes out of scope, even though the tree holds on to it. Potential disaster. Of course this can be avoided by passing on

The performance of MappableFlatSet::overlap_range and MappableFlatMultiSet::overlap_range could potentially be improved by storing bidirectionally sorted and mutually exclusive sub-ranges. This would come at the cost of additional memory usage, so behaviour could potentially be customised at compile time.

The current implementation does not allow the overriding of the FlatInterval.equals method since internally some FlatInterval objects are created to optimise data handling. As such, if one does override the equals method with the new properties, the search method brakes.

Possible solutions:

• Stop creating FlatInterval objects inside the I2BplusTree and just use/ move around the va