Exactly. My for use cases, this format is amazing. I have very few floats, but lots and lots of objects, arrays and strings with moderate levels of duplicatio

Exactly. My for use cases, this format is amazing. I have very few floats, but lots and lots of objects, arrays and strings with moderate levels of duplication and substring duplication. My data is produced in a build and then read in thousands or millions of tiny queries that lookup up a single value deep inside the structure. rx works very well as a kind of embedded database like sqlite, but completely unstructured like JSON. Also I'm working on an extension that makes it mutable using append-only persistent data structures with a fixed-block caching level that is actually a pretty good database.