Current considerations:
1. Just apply a simple .zip compression to the file.
2. Huffman encoding
3. Column-based linear pattern encoding? (e.g. when there is a long column of
air, we just do 20xg2BlockType_Air)
4. Octree-like file encoding (use internal file-address offsets to encode the
octree just like it is in system memory)
Design approach:
Idea 1 is very good in the end, idea 2 is a bit complex for what is needed, 3
is good but conflicts with 4 which would be great to avoid empty spaces. In the
end it might be best to just save non-empty type blocks as a cloud-point list
and do a post-file impression.
Original issue reported on code.google.com by nin...@gmail.com on 29 Oct 2011 at 7:58
Original issue reported on code.google.com by
nin...@gmail.com
on 29 Oct 2011 at 7:58