[libpc] Chipper notes

[Michael P. Gerlek]

[from pastebin, here for posterity]


Step 1:
  - chipper stage is connected to LAS reader stage
  - for each point P, read in order by an iterator, do
     . point P is recorded in chipper's memory as the tuple {point index
number, X, Y, Z} 

Step 2:
  - chipper groups all the tuples into blocks of tuples [no libPC/stage
action here?]

Step 3:
  - for each chipper block B, do
    . open a new Oracle blob
    . for each tuple item T in block B do
       - get point P from T.index
       - write P to the blob
    . close the blob


SequentialIterator      # what we have now
  - skip(count)           # count is a positive, relative value
  - bool atEnd()
  - read(buffer)          # reads N points [curIndex .. curIndex+N)

RandomIterator          # what Mr Chips wants
  - seek(pos)             # pos is a positive, absolute value
  - read(buffer)          # reads N points  [curIndex .. curIndex+N)


virtual bool Stage::supportsSequentialIterators() = 0
virtual SequentialIterator* Stage::createSequentialIterator() = 0

virtual bool Stage::supportsRandomIterators() = 0
virtual RandomIterator* Stage::createRandomIterator() = 0


SeqentialIterators only read "forwards", but RandomIterators can jump
around.  But both iterators will respond with a contiguous block of points
from a read() call, which implies that even with a RandomIterator you have
some notion of sequencing.

It makes sense to keep these two kinds of iterators separate (as opposed to
adding a seek() method to the current, sequential iterator) because some
file formats (like LAZ) and some stages (like crop) have no reasonably way
to move their cursor to a given index position.