String ⊂ Atom ⊂ UUID?

[cblp]

Can a UUID be an atom? Can an atom be a string? How does UTF-8 string fit in 16 bytes?

[gritzko]

An atom is either (1) UUID or (2) int or (3) string or (4) float.

The internal Cursor representation uses 16 bytes for either. For strings, that is a range in the buffer. Only UUIDs must be parsed fully. Cause UUDs are read and interpreted by reducers. Values typically aren't.

see the README

An op is a tuple of four "key" UUIDs and zero or more "value" atoms.

[gritzko]

You are mixing the formal model and implementation details here.

[cblp]

I'm reading uuid.md only. The document says:

1. uuid is 16 bytes long
2. if 31—30 bits of uuid are "01", then it is internal: RON atom (int, float, string)

what does it mean then?

Do formal model and implementation details contradict?

[gritzko]

I should probably remove that.

That's how cursors keep atoms internally. Shouldn't be part of uuid.md.

[cblp]

Looks like it is atom specification. A bigger thing is described inside a smaller thing. That's why it causes misunderstanding.

So, we need a separate document describing atoms and how strings are encoded inside them.

[gritzko]

I should make a single document explaining atoms, then ops, then frames, then po logs, then mappers, I guess. With pictures. Formal texts are heavy to read.

[lambdafu]

Started something a little bit here: https:/lambdafu/swarm-doc