Filling in missing values

I noticed that (at least according to the FAQ in v0.5.5) there is no strategy for filling in missing values. I can imagine there are a number of use cases where such missing values are prevalent due to the existence of geometry in a computational domain.

Having used zfp a bit in this context, and having some understanding of how zfp works (at least according to the original paper), a simple strategy like filling in with a constant value would lead to an unnecessarily large compressed size for each 4^d zfp block with missing values.

Has there been any work on this? Is there any interest in seeing this issue addressed?

I have a very simple algorithm which should work fairly well here, and would like to try integrating it and seeing how well it performs in different contexts.

Start by reviewing the FAQ in v0.5.5 and the existing handling of missing values in zfp. The issue does not name files, tests, or provide the proposed algorithm; clarify the intended strategy and evaluation contexts before defining completion criteria.

Written by the indexing model from the issue text.