# all subtrees are kept as full trees.
def add(new_write):
# assume there is a sorted list of writes
# merge this in. retain what the lengths of non-write data are to
the left and right
# if the writes trim non-write entries, then:
# on the left, walk left-to-right, and on the right, walk
right-to-left, covering the trimmed lengths
# when between two writes, consider only the length that goes
through the largest full tree between them: so as to leave smaller
trees near the more distant write, to merge with a future flush, and
not have unnecessary depth
# there might be a little more to think about and adjust here,
regarding producing these subtrees well. not sure.
# when between a write and the edge of the data, consider the entire length
# expand out full trees from the trimmed lengths, and merge them
into the more distant full trees when they are equal or larger to
those more distant full subtrees
