dataset provides a way to organize your JSON Objects on disc. The original
was a “buckets” oriented layout. The newer and current layout is a
pairtree.
The layout managed/described by the collection.json document
located in the root folder of the collection. The file pairtree
supports “attachments” by creating a sub directory next the the JSON
document. The sub directory name is _
(because it lacks specific meaning,
should be visible on most file systems and is short). E.g. storing
the document “hello-world.json” with the attachment “smiles.png” in a
collection named “C” would result in paths like C/pairtree/he/ll/o-/wo/rl/d/hello-world.json
and `C/pairtree/he/ll/o-/wo/rl/d/_/smiles.png”.
Attachments are experimental and how they are handled
will likely change in the future.
The directory layout looks like: