TODO.txt

Performance and Other
=====================
- Fix TODOs in RTree!!!
- Consider replacing all PriorityQueue with Min(Max)Heaps
- qt2: what is going on with subs vs values in QNode? Why can subs contain values?
- RTree: compare finNodeEntry vs findNodeEntries -> benchmark and decide
- QueryIterator nextReuse()
- MinMaxHeap: leave entries in array -> reuse/pool
- MinHeap: n-ary heap
- kNN filter should take PointEntryDist i.o PointEntry
- Avoid or at least reuse ArrayList in AQueryIterator
- Copyright update
- Reformat everything??


CritBit
=======
- Remove Node.posFirstBit this should always be (parent.posDiff+1)
- Memory optimisation: Nodes should have 'long' instead of 'long[]' for infixes.
  Infixes that cross 64bit boundaries can be stored in two separate nodes. 
  This would introduce a special case where a node may have only a single child
  which contains a continuation of the infix -> 'null' would not indicate that value is 'null'
  -> use NULL indicator object, see next point
- Memory optimisation: Merge children and subnodes into single field, -> saves
  2*32bits per Node. 
- Fix warnings in CritBit.x.checkMatchOrigKD()
- Fix warnings in CritBit.x.unsetBitAfterSplit()
- Why do Point queries scale much worse than kd-Tree? Because of merging???? O(k)
- Profile:
  - Consider using table for maskDst in CritBit.x.unsetBitAfterSplit() / setBitAfterSplit()
    --> Alternatively, consider operating on merged value?