FIX to Fully-Functional Succinct Trees
Level: Serious
The complexities we claim for dynamic trees are not all correct. We cannot
obtain O(log n / loglog n) time for all those operations. This stems from
errors in Section 7, which is only sketched in the paper but whose results
we used for Table 1. In particular, the use of n(i,j) to solve degree,
child, and child_rank is flawed. Lemma 7.2 is also wrong, yet we can recover
many of the O(log n / loglog n) claimed complexities by resorting to [6] and
mixing it with our structure.
A hopefully correct and complete description can be found in
arXiv. There we also describe,
for example, how to deamortize splits due to insertions and deletions.