Expand description
The ncdu JSON dump format (https://dev.yorhel.nl/ncdu/jsonfmt), as
written by ncdu -o and gdu -o.
A dump is [1, 2, {metadata}, directory], where a directory is an array of
its own info object followed by its children: nested arrays for
directories, objects for everything else. Each entry carries its own
apparent size (asize) and disk usage (dsize); readers aggregate them.
Names are raw bytes, not necessarily UTF-8, so both directions work on
bytes rather than through a JSON library. Both are iterative and stream, so
neither nesting depth nor dump size is limited by the stack or by holding
the document in memory.