Immediate TODO list
---------

- Provide a way to reduce ranges separated by a max distance or a min overlap
  (Michael's suggestion).

- Fix bug in IRanges() that should not be accepting negative widths (reported
  by Martin).

Long term TODO list
-------------------

- Ranges:
  - "unique" for removing duplicate ranges
  - "findInterval" for points that fall into a normal Ranges
  - "nearest" for finding nearest neighbors
  - "match" and "%in%" for exact matching
  - operator aliases ("&", "|", "!", "-", "*", "/")
  - "reduce" should take a window argument
  - "inverse.rle", could be logical vector, TRUE in range else FALSE
  - "sort" sorts by start
  - extend to multiple dimensions?

- TypedList:
  - dim, dim<-, dimnames, dimnames<- for multi-dimensional list (like 'by')
  
- RangesList:
  - parallel rbind
  - binary ops: "overlap", "nearest", "intersect", "setdiff", "merge"
    - 'y' omitted: become n-ary ops on items in collection
    - 'y' specified: performed element-wise
  - unary ops: "gaps", "coverage", "slice", etc are vectorized
  
- RangesMatching:
  - "relations" to extract relationships in terms of Allen's Interval Algebra
    - kind of like "all.equal"
  - "width" for amount of overlap for each match

- XDataFrame:
  - group generics (Math, Ops, Summary)
  - by, aggregate
  - with, within
  - subset

- RangedData:
  - rbind: inner combine
  - merge for SQL-type join operations based on range overlap
  - inverse.rle
  
- SplitXDataFrame:
  - stack: unlist(), but with factors recording source of each row
  - rbind
  
- IntervalTree:
  - instead of dense lgeMatrix, just complement the lgCMatrix

- FilterRules:
  - support subsetting XDataFrame/RangedData directly

- Classes to add:
  - ViewsList, extends RangesList but specific to Views
